Neurobiology of motor control : fundamental concepts and new directions 9781118873397

Table of contents :
Content: List of Contributors xiii About the Cover xvii 1 Introduction 1Ansgar Buschges and Scott L. Hooper References 5 2 Electrophysiological Recording Techniques 7Scott L. Hooper and Joachim Schmidt 2.1 Introduction 7 2.2 Terminology 8 2.3 Intracellular and Patch Clamp Recording 9 2.3.1 Recording Electrodes 9 2.3.2 Current-Clamp:Measuring Transmembrane Potential 12 2.3.3 Voltage Clamp: Measuring Transmembrane Current 15 2.3.3.1 Voltage Clamp with Transmembrane Potential as Reference 15 2.3.3.2 Voltage Clamp with Preparation (Bath) Ground as Reference 16 2.4 Extracellular Recording and Stimulation 17 2.5 A Brief History of Electrophysiological Recording 21 2.6 Concepts Important to Understanding Neuron Recording Techniques 27 2.6.1 Membrane Properties 27 2.6.2 Intracellular Recording 29 2.6.3 Extracellular Recording 32 2.6.3.1 Intracellular Action Potential Shape 33 2.6.3.2 Axon Embedded in Uniform, Infinite Volume Conductor 33 2.6.3.3 Variations in Extracellular Action Potential Shape Induced by Non-Uniform, Non-Infinite Volume Conductors 42 2.6.3.4 Bipolar Recording 44 2.6.3.5 Extracellular Action Potential Summary 46 Acknowledgements 47 References 47 3 Multi-Unit Recording 55Arthur Leblois and Christophe Pouzat 3.1 Introduction 55 3.2 Chapter Organization and Expository Choices 56 3.3 Hardware 57 3.4 Spike Sorting Methods 60 Endnotes 69 References 70 4 The New Math of Neuroscience: Genetic Tools for Accessing and Electively Manipulating Neurons 75Andreas Schoofs,Michael J. Pankratz, and Martyn Goulding 4.1 Introduction 75 4.2 Restricting Gene Expression to Specific Neurons 76 4.2.1 Promoter Bashing, Enhancer Trapping: Binary Systems for Targeted Gene Expression 77 4.2.2 Intersectional Strategies 81 4.2.3 Temporally Inducible Systems 82 4.3 Tracing, Manipulating, and Monitoring Neurons 84 4.3.1 Tracing Neuronal Projections and Connections with Fluorescent Reporters 84 4.3.2 Viral Tracers for Mapping Neural Connections 85 4.3.3 Manipulating Neuronal Function 87 4.3.4 Monitoring Neuronal Activity 90 4.4 Case Studies 92 4.5 Future Perspective 98 References 98 5 Computer Simulation Power and Peril 107Astrid A. Prinz and Scott L. Hooper 5.1 Introduction 107 5.2 Why Model? 107 5.3 Modeling Approaches 110 5.4 Model Optimization and Validation 118 5.5 Beyond Purely ComputationalModels 120 5.6 Fundamental Concepts and Frequently Used Models in Motor Control 121 5.6.1 How to Predict the Future 121 5.6.2 Neuron Models 123 5.6.3 Synapse Models 127 5.6.4 Muscle Models 128 5.6.5 Biomechanical Models 128 5.7 The Future 129 Acknowledgements 130 References 130 6 Evolution of Motor Systems 135Paul S. Katz and Melina E. Hale 6.1 Introduction 135 6.2 Phylogenetics 136 6.3 Homology and Homoplasy 138 6.4 Levels of Biological Organization 139 6.5 Homologous Neurons 139 6.6 Deep Homology 142 6.7 Homoplasy 145 6.8 Convergence in Central Pattern Generators 150 6.9 Evolutionary Loss 152 6.10 Evolution of Novel Motor Behaviors 152 6.11 Three Scenarios for the Evolution of Novel Behavior 154 6.11.1 Generalist Neural Circuitry 154 6.11.2 Rewired Circuitry 157 6.11.3 Functional Rewiring with Neuromodulation 159 6.12 Motor System Evolvability 161 6.13 Neuron Duplication and Parcellation 162 6.14 Divergence of Neural Circuitry 164 6.15 Summary and Conclusions 165 Acknowledgements 165 References 165 7 Motor Pattern Selection 177 7.1 Introduction to Motor Pattern Selection in Vertebrates and Invertebrates 178Hans-Joachim Pfluger and Sten Grillner References 179 7.2 Selection of Action A Vertebrate Perspective 181 Sten Grillner and Brita Robertson 7.2.1 Introduction 181 7.2.2 Control of Locomotory Outputs 182 7.2.3 The Organization and Role of the Basal Ganglia 184 7.2.4 ConceptualModel of the Organization Underlying Selection of Behavior 187 7.2.5 The Organization of Motor Control From Cortex (Pallium in Lower Vertebrates) 189 7.2.6 The Relative Role of Different Forebrain Structures for Selection of Behavior 189 Acknowledgements 190 References 191 7.3 Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects 195Hans-Joachim Pfluger 7.3.1 Introduction 195 7.3.2 Organization Principles of Relevant Sensory Systems 196 7.3.3 Movement-Generating Neural Networks in Invertebrates 196 7.3.4 Motor Pattern Selection in Invertebrates 197 7.3.4.1 Probabilistic Selection : Intrinsically Variable CPGs in Mollusk Feeding 197 7.3.4.2 Selection via CPG Coordination 198 7.3.4.3 Selection by Neuromodulators or Neurohormones 198 7.3.4.4 Selection by Command Neurons Not in the Brain 201 7.3.4.5 The Brain is Crucial in the Motor Selection Process 202 7.3.5 Two Case Studies 207 7.3.6 Concluding Remarks on Invertebrates 213 7.3.7 Are There Common Themes between Motor Pattern Selection in Invertebrates and Vertebrates? 213 References 216 8 Neural Networks for the Generation of RhythmicMotor Behaviors 225Ronald M. Harris-Warrick and Jan-Marino Ramirez 8.1 Introduction 225 8.2 Concept of the Central Pattern Generator 225 8.3 Overall Organization of Rhythmic Motor Networks 227 8.4 Identification of CPG Neurons and Synapses: The Wiring Diagram 234 8.5 Cellular PropertiesThat Shape Network Output: Building Blocks for Network Operation 238 8.6 Combined Neural Mechanisms for Rhythmogenesis 240 8.7 Ionic Currents Shaping CPG Network Neuron Intrinsic Firing Properties 241 8.7.1 Role of Outward Currents in Regulating Pacemaker and Network Activity 241 8.7.2 Role of Inward Currents in the Generation of Pacemaker and Network Activity 243 8.7.3 Interaction of Inward and Outward Currents in the Generation of Pacemaker Activity 245 8.7.4 Homeostatic Plasticity and the Balance between Different Ion Channel Types 245 8.7.5 Rapid Changes in Extracellular Ion Concentrations during Rhythmic Network Function 246 8.8 Role of Network Synaptic Properties in Organizing Rhythmic Behaviors 246 8.9 Variable Output from Motor Networks 249 8.10 Conclusions 252 Acknowledgements 253 References 253 9 Sensory Feedback in the Control of Posture and Locomotion 263Donald H. Edwards and Boris I. Prilutsky 9.1 Introduction 263 9.2 History and Background of Feedback Control 264 9.3 Classical Control Theory 264 9.4 Nervous System Implementation in the Control of Posture and Limb Movements 267 9.5 Organization and Function in Arthropods 274 9.5.1 Locomotory System Gross Anatomy 274 9.5.2 Proprioceptors and Exteroceptors 274 9.5.3 Arthropod Nervous Systems 275 9.5.4 Postures and Movement Commands 275 9.5.5 Sensory Feedback in the Maintenance of Posture 275 9.5.6 Sensory Feedback in Movement andWalking 276 9.6 Organization and Function in Vertebrates 282 9.6.1 Sensory Feedback in the Maintenance of Posture 282 9.6.2 Sensory Feedback and its Integration with Motor Commands in Movement 285 9.7 Conclusions 293 Acknowledgements 294 Endnote 294 References 294 10 Coordination of Rhythmic Movements 305Jean-Patrick Le Gal, Rejean Dubuc, and Carmen Smarandache-Wellmann 10.1 Introduction 305 10.2 Overview of Invertebrate CPGs 306 10.2.1 Stomatogastric Nervous System: Feeding Circuits in Decapod Crustacea 308 10.2.2 Leech Locomotion 315 10.2.3 Crayfish Swimmeret System 317 10.2.4 Insect Locomotion 319 10.2.5 MultipleMechanisms Mediate Coordination in Invertebrate Systems 321 10.3 Overview of Vertebrate CPGs 321 10.3.1 General Characteristic of Vertebrate CPGs 322 10.3.1.1 Locomotor CPGs 322 10.3.1.2 Respiratory CPGs 323 10.3.1.3 Feeding CPGs 324 10.3.2 CPG Interactions within One Motor Function 324 10.3.2.1 Unit Generators in Limbless Swimming Vertebrates 324 10.3.2.2 Unit Generators in Mammalian Limbs 325 10.3.3 CPGs Interactions for Different Motor Functions 327 10.3.3.1 Coordination of Respiration and Swallowing 327 10.3.3.2 Coordination of Locomotion and Respiration 328 10.4 Conclusion 331 References 332 11 Prehensile Movements 341Till Bockemuhl 11.1 Introduction: Prehension as Goal-Directed Behavior 341 11.2 The Redundancy Problem in Motor Control 343 11.3 Redundancy Occurs on Multiple Levels of the Motor System 346 11.4 Overcoming the Redundancy Problem 349 11.4.1 InvariantMovement Features 350 11.4.2 Increasing the Number of Task Conditions 352 11.4.3 Reducing the Number of DOFs 357 References 361 12 Muscle, Biomechanics, and Implications for Neural Control 365Lena H. Ting and Hillel J. Chiel 12.1 Introduction 365 12.2 Behavioral Context Determines How Motorneuron Activity Is Transformed into Muscle Force and Power 366 12.2.1 The Neuromuscular Transform Is History-Dependent 367 12.2.1.1 Motorneurons Are Subject to Neuromodulation and History-Dependence That Can Significantly Alter Their Output 368 12.2.1.2 Presynaptic Neurotransmitter Release at the Neuromuscular Junction Is History-Dependent 368 12.2.1.3 Post-SynapticMuscle Excitation Is History-Dependent and Subject to Modulation 368 12.2.1.4 Contractile Dynamics of Cross-Bridge Interactions Are History Dependent 369 12.2.1.5 The Molecular Motors of Muscles Give Rise to the Functional and History-Dependent Properties of Muscle Force Generation 369 12.2.2 Muscle Power Depends on Behavioral Context 371 12.2.3 Muscle Specialization Reflects Behavioral Repertoire 373 12.3 Organismal Structures Transform Muscle Force into Behavior 374 12.3.1 Effects of Muscle Force Depend on the Properties of the Body and the Environment 375 12.3.1.1 The Relative Importance of Inertial, Viscous, and Spring-Like Forces Affect the Role of Muscle Force 375 12.3.1.2 Muscle Function Depends on Behavioral Context and Environmental Forces 377 12.3.1.3 Biomechanical Affordances and Constraints of Body Structures Affect Muscle Functions 377 12.3.2 Muscles Are Multi-Functional 381 12.3.3 Specialization of Biomechanical Structures Reflect Behavioral Repertoire 385 12.4 Biomechanics Defines Meaningful Patterns of Neural Activity 387 12.4.1 Organismal Structures Are Multi-Functional 389 12.4.2 Many Functionally-Equivalent Solutions Exist for Sensorimotor Tasks 392 12.4.3 Structure and Variability in Motor Patterns Reflect Biomechanics 394 12.4.4 Specialization of Neuromechanical Systems Reflect Behavioral Repertoire 399 12.5 Conclusions 401 Acknowledgements 402 References 402 13 Plasticity and Learning in Motor Control Networks 417John Simmers and Keith T. Sillar 13.1 Introduction 417 13.2 Homeostatic Motor Network Assembly 418 13.3 Short-Term Motor Learning Conferred by Sodium Pumps 420 13.3.1 Swimming CPG Network Plasticity in Xenopus Frog Tadpoles 420 13.3.2 Comparative Aspects of Na+ Pump Contribution to Neural Network Function 425 13.4 CPG Network Plasticity and Motor Learning Conferred by Operant Conditioning 426 13.5 Discussion and Conclusions 432 References 436 14 Bio-inspired Robot Locomotion 443Thomas Buschmann and Barry Trimmer 14.1 Introduction 443 14.2 Mechanical Engineering Background and a Biological Example 444 14.3 Legged Robots with Skeletal Structures 446 14.3.1 Mechanism Design, Sensing, and Actuation 446 14.3.2 Basic Dynamics of Legged Locomotion 447 14.3.3 Trajectory-OrientedWalking Control 448 14.3.4 Limit CycleWalkers 450 14.3.5 CPG-Based Control and Step-Phase Control 451 14.4 Soft Robots 452 14.4.1 Limitations and Advantages of Soft Materials 452 14.4.2 The Challenges 453 14.4.2.1 Actuators 453 14.4.2.2 Sensors 455 14.4.2.3 Control of Soft Robots 456 14.4.3 Bioinspired Locomotion in Soft Robots 459 14.5 Conclusion and Outlook 463 References 463 Index 473

Citation preview

Neurobiology of Motor Control

Neurobiology of Motor Control Fundamental Concepts and New Directions

Edited by Scott L. Hooper Department of Biological Sciences, Ohio University, USA

Ansgar Büschges Institut für Zoologie, Universität zu Köln, Germany

This edition first published 2017 © 2017 John Wiley & Sons, Inc. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions. The right of Scott L. Hooper and Ansgar Büschges to be identified as the author(s) of the editorial material in this work has been asserted in accordance with law. Registered Office John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA Editorial Office 111 River Street, Hoboken, NJ 07030, USA For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com. Wiley also publishes its books in a variety of electronic formats and by print-on-demand. Some content that appears in standard print versions of this book may not be available in other formats. Limit of Liability/Disclaimer of Warranty The publisher and the authors make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties; including without limitation any implied warranties of fitness for a particular purpose. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for every situation. In view of on-going research, equipment modifications, changes in governmental regulations, and the constant flow of information relating to the use of experimental reagents, equipment, and devices, the reader is urged to review and evaluate the information provided in the package insert or instructions for each chemical, piece of equipment, reagent, or device for, among other things, any changes in the instructions or indication of usage and for added warnings and precautions. The fact that an organization or website is referred to in this work as a citation and/or potential source of further information does not mean that the author or the publisher endorses the information the organization or website may provide or recommendations it may make. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this works was written and when it is read. No warranty may be created or extended by any promotional statements for this work. Neither the publisher nor the author shall be liable for any damages arising here from. Library of Congress Cataloguing-in-Publication Data Names: Hooper, Scott L., editor. | Büschges, Ansgar, editor. Title: Neurobiology of Motor Control : Fundamental Concepts and New Directions / [edited by] Scott L. Hooper, Ansgar Büschges. Description: Hoboken, New Jersey : John Wiley & Sons, Inc., 2017. | Includes bibliographical references and index. | Identifiers: LCCN 2017011654 (print) | LCCN 2017013127 (ebook) | ISBN 9781118873342 (pdf ) | ISBN 9781118873625 (epub) | ISBN 9781118873403 (cloth) Subjects: | MESH: Movement–physiology | Psychomotor Performance–physiology | Motor Skills–physiology | Biomechanical Phenomena Classification: LCC QP301 (ebook) | LCC QP301 (print) | NLM WE 103 | DDC 612/.04–dc23 LC record available at https://lccn.loc.gov/2017011654 Cover Design: Wiley Cover Images: (Background) © bgblue/Gettyimages; (Illustration) Designed by Scott L. Hooper, Sherylane Seeliger, and Susanne Schulze Set in 10/12pt WarnockPro by SPi Global, Chennai, India 10 9 8 7 6 5 4 3 2 1

v

Contents List of Contributors xiii About the Cover xvii 1

Introduction 1 Ansgar Büschges and Scott L. Hooper

References 5 2

Electrophysiological Recording Techniques 7 Scott L. Hooper and Joachim Schmidt

2.1 2.2 2.3 2.3.1 2.3.2 2.3.3 2.3.3.1 2.3.3.2 2.4 2.5 2.6 2.6.1 2.6.2 2.6.3 2.6.3.1 2.6.3.2 2.6.3.3

Introduction 7 Terminology 8 Intracellular and Patch Clamp Recording 9 Recording Electrodes 9 Current-Clamp: Measuring Transmembrane Potential 12 Voltage Clamp: Measuring Transmembrane Current 15 Voltage Clamp with Transmembrane Potential as Reference 15 Voltage Clamp with Preparation (Bath) Ground as Reference 16 Extracellular Recording and Stimulation 17 A Brief History of Electrophysiological Recording 21 Concepts Important to Understanding Neuron Recording Techniques 27 Membrane Properties 27 Intracellular Recording 29 Extracellular Recording 32 Intracellular Action Potential Shape 33 Axon Embedded in Uniform, Infinite Volume Conductor 33 Variations in Extracellular Action Potential Shape Induced by Non-Uniform, Non-Infinite Volume Conductors 42 Bipolar Recording 44 Extracellular Action Potential Summary 46 Acknowledgements 47 References 47

2.6.3.4 2.6.3.5

3

Multi-Unit Recording 55 Arthur Leblois and Christophe Pouzat

3.1 3.2

Introduction 55 Chapter Organization and Expository Choices 56

vi

Contents

3.3 3.4

Hardware 57 Spike Sorting Methods 60 Endnotes 69 References 70

4

The “New Math” of Neuroscience: Genetic Tools for Accessing and Electively Manipulating Neurons 75 Andreas Schoofs, Michael J. Pankratz, and Martyn Goulding

4.1 4.2 4.2.1

Introduction 75 Restricting Gene Expression to Specific Neurons 76 Promoter Bashing, Enhancer Trapping: Binary Systems for Targeted Gene Expression 77 Intersectional Strategies 81 Temporally Inducible Systems 82 Tracing, Manipulating, and Monitoring Neurons 84 Tracing Neuronal Projections and Connections with Fluorescent Reporters 84 Viral Tracers for Mapping Neural Connections 85 Manipulating Neuronal Function 87 Monitoring Neuronal Activity 90 Case Studies 92 Future Perspective 98 References 98

4.2.2 4.2.3 4.3 4.3.1 4.3.2 4.3.3 4.3.4 4.4 4.5

5

Computer Simulation—Power and Peril 107 Astrid A. Prinz and Scott L. Hooper

5.1 5.2 5.3 5.4 5.5 5.6 5.6.1 5.6.2 5.6.3 5.6.4 5.6.5 5.7

Introduction 107 Why Model? 107 Modeling Approaches 110 Model Optimization and Validation 118 Beyond Purely Computational Models 120 Fundamental Concepts and Frequently Used Models in Motor Control 121 How to Predict the Future 121 Neuron Models 123 Synapse Models 127 Muscle Models 128 Biomechanical Models 128 The Future 129 Acknowledgements 130 References 130

6

Evolution of Motor Systems 135 Paul S. Katz and Melina E. Hale

6.1 6.2 6.3 6.4

Introduction 135 Phylogenetics 136 Homology and Homoplasy 138 Levels of Biological Organization 139

Contents

6.5 6.6 6.7 6.8 6.9 6.10 6.11 6.11.1 6.11.2 6.11.3 6.12 6.13 6.14 6.15

Homologous Neurons 139 Deep Homology 142 Homoplasy 145 Convergence in Central Pattern Generators 150 Evolutionary Loss 152 Evolution of Novel Motor Behaviors 152 Three Scenarios for the Evolution of Novel Behavior 154 Generalist Neural Circuitry 154 Rewired Circuitry 157 Functional Rewiring with Neuromodulation 159 Motor System Evolvability 161 Neuron Duplication and Parcellation 162 Divergence of Neural Circuitry 164 Summary and Conclusions 165 Acknowledgements 165 References 165

7

Motor Pattern Selection 177

7.1

Introduction to Motor Pattern Selection in Vertebrates and Invertebrates 178 Hans-Joachim Pflüger and Sten Grillner

References 179 181

7.2

Selection of Action—A Vertebrate Perspective Sten Grillner and Brita Robertson

7.2.1 7.2.2 7.2.3 7.2.4

Introduction 181 Control of Locomotory Outputs 182 The Organization and Role of the Basal Ganglia 184 Conceptual Model of the Organization Underlying Selection of Behavior 187 The Organization of Motor Control From Cortex (Pallium in Lower Vertebrates) 189 The Relative Role of Different Forebrain Structures for Selection of Behavior 189 Acknowledgements 190 References 191

7.2.5 7.2.6

7.3

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects 195 Hans-Joachim Pflüger

7.3.1 7.3.2 7.3.3 7.3.4 7.3.4.1

Introduction 195 Organization Principles of Relevant Sensory Systems 196 Movement-Generating Neural Networks in Invertebrates 196 Motor Pattern Selection in Invertebrates 197 Probabilistic “Selection”: Intrinsically Variable CPGs in Mollusk Feeding 197 Selection via CPG Coordination 198

7.3.4.2

vii

viii

Contents

7.3.4.3 7.3.4.4 7.3.4.5 7.3.5 7.3.6 7.3.7

Selection by Neuromodulators or Neurohormones 198 Selection by Command Neurons Not in the Brain 201 The Brain is Crucial in the Motor Selection Process 202 Two Case Studies 207 Concluding Remarks on Invertebrates 213 Are There Common Themes between Motor Pattern Selection in Invertebrates and Vertebrates? 213 References 216

8

Neural Networks for the Generation of Rhythmic Motor Behaviors 225 Ronald M. Harris-Warrick and Jan-Marino Ramirez

8.1 8.2 8.3 8.4 8.5

Introduction 225 Concept of the Central Pattern Generator 225 Overall Organization of Rhythmic Motor Networks 227 Identification of CPG Neurons and Synapses: The “Wiring Diagram” 234 Cellular Properties That Shape Network Output: Building Blocks for Network Operation 238 Combined Neural Mechanisms for Rhythmogenesis 240 Ionic Currents Shaping CPG Network Neuron Intrinsic Firing Properties 241 Role of Outward Currents in Regulating Pacemaker and Network Activity 241 Role of Inward Currents in the Generation of Pacemaker and Network Activity 243 Interaction of Inward and Outward Currents in the Generation of Pacemaker Activity 245 Homeostatic Plasticity and the Balance between Different Ion Channel Types 245 Rapid Changes in Extracellular Ion Concentrations during Rhythmic Network Function 246 Role of Network Synaptic Properties in Organizing Rhythmic Behaviors 246 Variable Output from Motor Networks 249 Conclusions 252 Acknowledgements 253 References 253

8.6 8.7 8.7.1 8.7.2 8.7.3 8.7.4 8.7.5 8.8 8.9 8.10

263

9

Sensory Feedback in the Control of Posture and Locomotion Donald H. Edwards and Boris I. Prilutsky

9.1 9.2 9.3 9.4

Introduction 263 History and Background of Feedback Control 264 Classical Control Theory 264 Nervous System Implementation in the Control of Posture and Limb Movements 267 Organization and Function in Arthropods 274 Locomotory System Gross Anatomy 274 Proprioceptors and Exteroceptors 274

9.5 9.5.1 9.5.2

Contents

9.5.3 9.5.4 9.5.5 9.5.6 9.6 9.6.1 9.6.2 9.7

Arthropod Nervous Systems 275 Postures and Movement Commands 275 Sensory Feedback in the Maintenance of Posture 275 Sensory Feedback in Movement and Walking 276 Organization and Function in Vertebrates 282 Sensory Feedback in the Maintenance of Posture 282 Sensory Feedback and its Integration with Motor Commands in Movement 285 Conclusions 293 Acknowledgements 294 Endnote 294 References 294

10

Coordination of Rhythmic Movements 305 Jean-Patrick Le Gal, Réjean Dubuc, and Carmen Smarandache-Wellmann

10.1 10.2 10.2.1

Introduction 305 Overview of Invertebrate CPGs 306 Stomatogastric Nervous System: Feeding Circuits in Decapod Crustacea 308 Leech Locomotion 315 Crayfish Swimmeret System 317 Insect Locomotion 319 Multiple Mechanisms Mediate Coordination in Invertebrate Systems 321 Overview of Vertebrate CPGs 321 General Characteristic of Vertebrate CPGs 322 Locomotor CPGs 322 Respiratory CPGs 323 Feeding CPGs 324 CPG Interactions within One Motor Function 324 Unit Generators in Limbless Swimming Vertebrates 324 Unit Generators in Mammalian Limbs 325 CPGs Interactions for Different Motor Functions 327 Coordination of Respiration and Swallowing 327 Coordination of Locomotion and Respiration 328 Conclusion 331 References 332

10.2.2 10.2.3 10.2.4 10.2.5 10.3 10.3.1 10.3.1.1 10.3.1.2 10.3.1.3 10.3.2 10.3.2.1 10.3.2.2 10.3.3 10.3.3.1 10.3.3.2 10.4

11

Prehensile Movements 341 Till Bockemühl

11.1 11.2 11.3 11.4 11.4.1 11.4.2 11.4.3

Introduction: Prehension as Goal-Directed Behavior 341 The Redundancy Problem in Motor Control 343 Redundancy Occurs on Multiple Levels of the Motor System 346 Overcoming the Redundancy Problem 349 Invariant Movement Features 350 Increasing the Number of Task Conditions 352 Reducing the Number of DOFs 357 References 361

ix

x

Contents

12

Muscle, Biomechanics, and Implications for Neural Control 365 Lena H. Ting and Hillel J. Chiel

12.1 12.2

Introduction 365 Behavioral Context Determines How Motorneuron Activity Is Transformed into Muscle Force and Power 366 The Neuromuscular Transform Is History-Dependent 367 Motorneurons Are Subject to Neuromodulation and History-Dependence That Can Significantly Alter Their Output 368 Presynaptic Neurotransmitter Release at the Neuromuscular Junction Is History-Dependent 368 Post-Synaptic Muscle Excitation Is History-Dependent and Subject to Modulation 368 Contractile Dynamics of Cross-Bridge Interactions Are History Dependent 369 The Molecular Motors of Muscles Give Rise to the Functional and History-Dependent Properties of Muscle Force Generation 369 Muscle Power Depends on Behavioral Context 371 Muscle Specialization Reflects Behavioral Repertoire 373 Organismal Structures Transform Muscle Force into Behavior 374 Effects of Muscle Force Depend on the Properties of the Body and the Environment 375 The Relative Importance of Inertial, Viscous, and Spring-Like Forces Affect the Role of Muscle Force 375 Muscle Function Depends on Behavioral Context and Environmental Forces 377 Biomechanical Affordances and Constraints of Body Structures Affect Muscle Functions 377 Muscles Are Multi-Functional 381 Specialization of Biomechanical Structures Reflect Behavioral Repertoire 385 Biomechanics Defines Meaningful Patterns of Neural Activity 387 Organismal Structures Are Multi-Functional 389 Many Functionally-Equivalent Solutions Exist for Sensorimotor Tasks 392 Structure and Variability in Motor Patterns Reflect Biomechanics 394 Specialization of Neuromechanical Systems Reflect Behavioral Repertoire 399 Conclusions 401 Acknowledgements 402 References 402

12.2.1 12.2.1.1 12.2.1.2 12.2.1.3 12.2.1.4 12.2.1.5 12.2.2 12.2.3 12.3 12.3.1 12.3.1.1 12.3.1.2 12.3.1.3 12.3.2 12.3.3 12.4 12.4.1 12.4.2 12.4.3 12.4.4 12.5

13

Plasticity and Learning in Motor Control Networks 417 John Simmers and Keith T. Sillar

13.1 13.2 13.3 13.3.1

Introduction 417 Homeostatic Motor Network Assembly 418 Short-Term Motor Learning Conferred by Sodium Pumps 420 Swimming CPG Network Plasticity in Xenopus Frog Tadpoles 420

Contents

13.3.2 13.4 13.5

Comparative Aspects of Na+ Pump Contribution to Neural Network Function 425 CPG Network Plasticity and Motor Learning Conferred by Operant Conditioning 426 Discussion and Conclusions 432 References 436

14

Bio-inspired Robot Locomotion 443 Thomas Buschmann and Barry Trimmer

14.1 14.2 14.3 14.3.1 14.3.2 14.3.3 14.3.4 14.3.5 14.4 14.4.1 14.4.2 14.4.2.1 14.4.2.2 14.4.2.3 14.4.3 14.5

Introduction 443 Mechanical Engineering Background and a Biological Example Legged Robots with Skeletal Structures 446 Mechanism Design, Sensing, and Actuation 446 Basic Dynamics of Legged Locomotion 447 Trajectory-Oriented Walking Control 448 Limit Cycle Walkers 450 CPG-Based Control and Step-Phase Control 451 Soft Robots 452 Limitations and Advantages of Soft Materials 452 The Challenges 453 Actuators 453 Sensors 455 Control of Soft Robots 456 Bioinspired Locomotion in Soft Robots 459 Conclusion and Outlook 463 References 463 Index 473

444

xi

xiii

List of Contributors Till Bockemühl

and

Biozentrum Köln Institut für Zoologie Universität zu Köln Köln Germany

Groupe de Recherche en Activité Physique Adaptée Département des sciences de l’activité physique Université du Québec à Montréal Montréal QC Canada

Ansgar Büschges

Biozentrum Köln Institut für Zoologie Universität zu Köln Köln Germany Thomas Buschmann

Institute of Applied Mechanics Technische Universität München Garching Germany Hillel J. Chiel

Departments of Biology, Neurosciences, and Biomedical Engineering Case Western Reserve University Cleveland OH USA Réjean Dubuc

Groupe de Recherche sur le Système Nerveux Central Département de neurosciences Université de Montréal

Donald H. Edwards

Neuroscience Institute Georgia State University Atlanta GA USA Martyn Goulding

Salk Institute San Diego CA USA Sten Grillner

Department of Neuroscience Karolinska Institutet Stockholm Sweden Melina E. Hale

Department of Organismal Biology and Anatomy University of Chicago Chicago IL USA

xiv

List of Contributors

Ronald M. Harris-Warrick

Hans-Joachim Pflüger

Department of Neurobiology and Behavior Cornell University Ithaca NY USA

Institut für Biologie Neurobiologie Freie Universität Berlin Berlin Germany

Scott L. Hooper

Biozentrum Köln Institut für Zoologie Universität zu Köln Köln Germany

Neuroscience Program Department of Biological Sciences Ohio University Athens OH USA Paul S. Katz

Neuroscience Institute Georgia State University Atlanta GA USA Jean-Patrick Le Gal

Groupe de Recherche sur le Système Nerveux Central Département de neurosciences Université de Montréal Montréal QC Canada Arthur Leblois

Centre de Neurophysique Physiologie et Pathologie CNRS UMR 8119 Institut Neurosciences et Cognition Université Paris Descartes Paris France Michael J. Pankratz

Life & Medical Sciences Institute (LIMES) Molecular Brain Physiology and Behavior Bonn Germany

and

Christophe Pouzat

Mathématiques Appliquées à Paris 5 CNRS UMR 8145 Université Paris Descartes Paris France Boris I. Prilutsky

School of Biological Sciences Georgia Institute of Technology Atlanta GA USA Astrid A. Prinz

Department of Biology Emory University Atlanta GA USA Jan-Marino Ramirez

Department of Neurological Surgery University of Washington School of Medicine and Center for Integrative Brain Research Seattle Children’s Research Institute University of Washington Seattle WA USA

List of Contributors

Brita Robertson

Carmen Smarandache-Wellmann

Department of Neuroscience Karolinska Institutet Stockholm Sweden

Biozentrum Köln Institut für Zoologie Universität zu Köln Köln Germany

Joachim Schmidt

Biozentrum Köln Institut für Zoologie Universität zu Köln Köln Germany

Lena H. Ting

Andreas Schoofs

Life & Medical Sciences Institute (LIMES) Molecular Brain Physiology and Behavior Bonn Germany

Department of Rehabilitation Medicine Division of Physical Therapy Emory University Atlanta GA USA

Keith T. Sillar

Barry Trimmer

School of Psychology and Neuroscience University of St Andrews Fife Scotland UK

Department of Biology Tufts University Medford MA USA

John Simmers

Institut de Neurosciences Cognitives et Intégratives d’Aquitaine CNRS UMR 5287 Université de Bordeaux Bordeaux France

Department of Biomedical Engineering Emory University and Georgia Institute of Technology and

xv

xvii

About the Cover Moving from standing on both feet to standing on one. Higher centers and the basal ganglia “decide” to stand on the right foot, with the basal ganglia (blue shading in brain) playing a central role in suppressing expression of competing behaviors (standing on left foot, kneeling, jumping) (Chapter 7). Motor cortex (orange shading in brain) and brainstem (for facial movements, green shading in head) or spinal cord (for body movements) inter- and motor neuron networks (pink cell bodies in spinal cord) determine what pattern of motor neuron activity will produce the required movements (left hip flexion, changes in trunk and right leg posture to bring body center of mass over right foot) (Chapters 8, 10, 11). Motor neuron activity (upper red trace) induces muscle force production (bottom red trace) and, acting through limb moment arms, the joint torques (curved red arrow and equation) required to generate the movements (Chapter 12). Sensory feedback (green neuron) provides continuous updates of movement success and input necessary to maintain body stability (Chapter 9). Motor learning alters network properties to produce more fluid and effective movement (Chapter 13). Neuron activity and muscle activity and force can be monitored and modified by electrophysiological or molecular biology techniques (Chapters 2–4). Increased insight into the processes at work can be gained with computer simulation (Chapter 5) and evolutionary comparisons (Chapter 6), and applied to develop robots capable of producing more natural and robust movement (Chapter 14).

1

1 Introduction Ansgar Büschges 1 and Scott L. Hooper 2 1 2

Biozentrum Köln, Institute für Zoologie, Universität zu Köln, Köln, Germany Neuroscience Program, Department of Biological Sciences, Ohio University, Athens, OH, USA

It is de rigueur in a review or book on motor control to quote Sherrington’s (1924) statement that “To move things is all that mankind can do”. Although strictly true, this quotation discounts the central role in human experience of such actionless phenomena as ideation, emotion, and consciousness. However, it is nonetheless true that movement is an absolute requirement for animal survival and reproduction and, as the only observable output of the nervous system, is the defining basis of behavior. Movement is also self-defining, and hence allows analyzing nervous system function on the objective basis of its performance alone without reference to experimenter defined classifications. Disorders of movement also have great clinical importance, and production of functional and robust movement is a central problem in robotics. Because movements must be chosen among, and because almost all motor networks receive sensory input and information about internal state and “decide” how to alter their output in response, studying such networks may also provide insight into how the networks underlying “higher” abilities such as ideation function. Despite this, many researchers, as well as lay people, take the generation of motor behavior for granted, often rendering it as the outcome of simple and automatic neural processes that can be summarized with large arrows pointing “south” from an animal’s brain accompanied by the words “motor system”. Only when confronted with particularly outstanding motor performances, e.g., the graceful movements of a dancer or an acrobat, do we appreciate the complexity of generating motor output. This disparity was well captured more than 200 years ago in von Kleist’s (1810) essay Über das Marionetten Theater (On the Marionette Theater): “He asked me if indeed I hadn’t found some of the movements of the puppets…to be exceedingly graceful in the dances. I could not refute this observation”, a recognition that led Kleist to elaborate further on the potential mechanistic background of this observation. This text highlights how the ordinariness of movement can prevent us from appreciating how difficult it is to produce (something of which roboticists are well aware), and thus how extraordinary it is that nervous systems can do so. The last general textbook covering how nervous systems do so, at least with respect to locomotion, was Neural Control of Locomotion (Orlovsky et al. 1999). This exceptional book described the neural networks and mechanisms that generate locomotion Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

2

Neurobiology of Motor Control: Fundamental Concepts and New Directions

in mollusks, insects, anurans, lower vertebrates, mammals, and man. This book was the first comprehensive comparative account of how nervous systems generate locomotion. Such an overview had been lacking for decades and its detail and depth made and make it exceptional. However, the book’s concentration on locomotion meant that it, by design, did not cover the full range of movements animals produce. More importantly, dramatic advances in motor science have occurred since it was published. These advances represent a sea change in that motor research up to the 1990s primarily involved ever more elegant and detailed application of classical anatomical and single unit electrophysiological techniques. In the last two decades, alternatively, a much broader palette of methods has become available or practicable, including multi-unit recordings, molecular neurogenetics, computer simulation, and new approaches for studying how muscles and body anatomy transform motor neuron activity into movement. This broadening of experimental options has been exceptionally fruitful. However, it also means that researchers in motor control must be multi-competent, sufficiently informed and trained to be able to select from these multiple methodological options the optimal approach for the research question at hand. It is important to make this observation because human nature and the process by which researchers are typically trained (prolonged and intensely concentrated research on a narrowly-defined question in an individual mentor’s lab) work against achieving such multi-competence. Instead, as with a person with a hammer seeing every problem as a nail, it leads to researchers using the methods they know in preference to ones that might be better, but about which the researcher only peripherally knows. This is not a new observation, and conscious efforts are being made in training programs to train new researchers across fields. Nonetheless, in our experience barriers still exist between molecular biologists, electrophysiologists, muscle researchers, modelers, biomechanicists, and roboticists. It is a truism that reducing such barriers would serve all well. The question is, how to do so? This book, in part, is an attempt to contribute to this effort. Its intended audience is all workers in movement production, from molecular biologists to roboticists. Workers in each group will have most knowledge of fields nearest their own …thus an electrophysiologist from a biology program likely has greatest understanding of molecular biology, and perhaps least of robotics. A biomechanicist likely finds it easier to communicate with a roboticist than a molecular biologist. And in our experience, modelers, at least those whose training was in classical mathematics, always speak a foreign language. We therefore begin this book with four chapters covering basic knowledge on electrophysiological techniques, methods for large ensemble recordings, neurogenetic and molecular techniques, and computer simulation. These chapters are obviously not intended for experts in the field (although we hope they will be useful for beginning students in their labs, and the molecular biology and simulation chapters include case studies that will interest even experts in the fields). Rather, we hope that these chapters will allow workers outside each chapter’s field to better understand and critically assess the field’s literature, to understand the later chapters in the book, and encourage workers to reach outside their comfort zone and consider applying different methodological approaches to their research. We believe that writing these chapters, with their at least partially pedagogical nature, was likely a considerable change from the more purely research oriented reviews the authors would be typically asked to

Introduction

write. We are therefore particularly grateful to the highly distinguished colleagues in the field of motor control who were willing to take on this burden. Hooper and Schmidt cover classical (i.e., not multi-unit) electrophysiological recording techniques. The first sections of this chapter are practical, and provide the information necessary for readers to understand and interpret intracellular and extracellular recording in the contemporary literature without a detailed explanation of theory. It is very difficult for modern readers to appreciate just how difficult it was for these techniques to be developed. The authors therefore next provide a brief history of extracellular and intracellular recording. The authors end the chapter with a detailed explanation of the theory underlying both recording techniques, and potential pitfalls that can occur with them. Lebois and Pouzat cover multi-unit recordings…recordings in which electrodes that record the activity of multiple neurons are introduced into nervous tissue. The ability to do so strongly depends on proper electrode design and use, which the authors therefore first cover. Given that these electrodes record the activity of many neurons, advanced techniques are required to identify the individual activities of the many neurons being recorded from. The authors explain these techniques in detail in the chapter’s second part. Schoofs, Pankratz, and Goulding cover the use of molecular genetic tools to study neural network topology and function. They begin with a detailed explanation of the techniques available in invertebrates and vertebrates to observe and alter neuron activity. They then provide four cases studies, two in Drosophila and two in mice, in which these techniques were used to make novel findings in motor control that would have been presently impossible to achieve with other methods. Prinz and Hooper cover computational simulation. The authors first provide a relatively high level overview of both the great power, and also the potential pitfalls, of simulation, making use throughout of case studies relevant to motor control. Because computer simulation may not be a part of the training of many of the book’s intended audience, the authors then provide a detailed and basic explanation of how simulation is performed and how it is applied to neurons, synapses, muscles, and biomechanics. In planning this book we also aimed to reduce another set of barriers: those between workers in different experimental preparations, of which the greatest is between workers in invertebrates and vertebrates. Doing so is important on both historical and scientific grounds. First, many to perhaps most discoveries made in one of these groups have been later found to be also present in the other. Second, recent data suggest a deep homology between (bilaterian) invertebrate and vertebrate motor control structures. This observation suggests that the last common ancestor of these two groups (the urbilatarian) had a relatively complex nervous system from which both bilaterian invertebrate and vertebrate nervous systems developed. It would thus be expected that data from one group would often be relevant to the other. In the later chapters we therefore tried to team researchers in vertebrates and invertebrates, with the comparison between the groups being made implicitly or explicitly. In all cases the results of these across-group collaborations are excellent chapters whose synthesis, we believe, provide a depth and breadth of understanding and insight that could not otherwise have been achieved. We are very grateful to the many open-minded colleagues who were willing to accept this challenge to “work across the divide” in writing these chapters.

3

4

Neurobiology of Motor Control: Fundamental Concepts and New Directions

In choosing the topics for these chapters we strove to cover the full width of motor control research, areas which, in our opinion, are relevant to all workers, and particularly students and similar upcoming workers, in the field. These topics do not admit to an easy hierarchical ordering, but we tried to begin with the most general (evolution), then turned to the neural basis of movement generation, next to muscles and biomechanics, then to motor learning/plasticity, and finally to the application of these insights to robotics. Katz and Hale describe motor network evolution. Throughout they intermix explanation of evolutionary concepts and terminology with illustrative case study examples, including cautionary examples in which motor network similarity is solely through convergence. As one would expect, in this chapter the importance of the molecular biology advances discussed above in understanding the evolution of motor circuits is very apparent. Grillner, Robertson, and Pflüger in their chapters introduce the reader to the neural mechanisms that select which of the movement programs an animal has in its behavioral portfolio to produce. Unlike the other chapters in the book, these chapters are presented separately as vertebrate and invertebrate, with a general introduction. However, the chapters end with discussion of the possible deep homology between the movement selection centers in the two groups. Harris-Warrick and Ramirez cover the neural networks that generate rhythmic motor acts and identify general principles present across the animal kingdom. In doing so they describe both the importance of synaptic connectivity pattern and cell-specific properties, including the contributions made by specific ion conductances, in rhythm pattern generation. They also describe the effect of modulation in such networks, and the basis of their ability to produce multiple output patterns. Edwards and Prilutsky cover the role of sensory feedback in modifying and sculpting motor network activity. They begin with a description of control theory and then give case studies of how various types of sensory feedback and input function both in movement generation and the maintenance of posture, emphasizing the common problems and functional solution structures in vertebrates and invertebrates. Movements are often coordinated, e.g., breathing and locomotion in some gaits. Le Gal, Dubuc, and Smarandache-Wellmann describe the neural mechanisms that underlie these coordinations and the bases of their flexible expression using a wide variety of well-studied case studies. A key insight from this work is the frequent presence of multiple mechanisms subserving these coordinations. Not all movements are rhythmic. In particular, animals use appendages to reach out to objects in the environment, so-called prehensile movements. Bockemühl describes the theory of prehension with jointed limbs, particularly the redundancy problem (that a motor system can typically fulfill a given reaching task in a very large number of ways) and then both theoretical and neurobiological solutions to this problem. This chapter uses only vertebrate case studies, but the generality of its analysis makes it valuable to workers in all preparations. Ting and Chiel describe how neural and biomechanical systems interact to produce functional motor behaviors. Central points here are that several new types of redundancy exist on the muscle level, muscle response to neural input is history-dependent and non-linear, and the effect of muscle contraction (and hence motor neuron activity) depends on the contraction state of other muscles and the position of the structure to

Introduction

which the muscle attaches. This intimate interdependence of nervous system and body state clearly greatly complicates understanding how animals generate movement. Simmers and Sillar describe motor learning and plasticity in Xenopus tadpole swimming and Aplysia feeding. As such, their chapter does not cover higher level (motor cortex, cerebellum) involvement in learning complex motor patterns, e.g., to dance. This choice, however, allows them to focus on examples of motor learning in which the basis of the learning can be explained on a cellular level in relatively or very well described motor networks, and in which the learning occurs within these networks. In the final chapter Buschmann and Trimmer describe how neurobiological and biomechanical research can help design robots that can effectively move through unpredictable environments, covering both rigid limbed (analogous to human or insect limbs) and soft-bodied (analogous to caterpillars or octopus arms) robots. An important part of this chapter is that differences in biological and robotic sensors, force generators, and structural materials complicate applying biological principles to robot design. Nonetheless, this chapter is in a sense a measure of our understanding of biological movement; presumably when we understand the one, we can design the other. In closing, we want to again thank the authors for their work, and to express our hope that this work will help advance all of our efforts to understand how animals generate movement, particularly the efforts of those beginning their research in this field, who are its future.

References Orlovsky GN, Deliagina TG, Grillner S (1999) Neural Control of Locomotion. Oxford, UK: Oxford University Press. Sherrington CS (1924) Linacre Lecture, St John’s College, Cambridge. In Eccles JC, Gibson WC (eds). Sherrington: His Life and Thought. New York, NY: Springer-Verlag. von Kleist H (1810) Über das Marionetten Theater. Berlin Abendblätter, 4 installments Dec 12–15. Translation On the Marionette Theatre by TG Neumiller in The Drama Review: TDR (1972) 16:22–26.

5

7

2 Electrophysiological Recording Techniques Scott L. Hooper 1 and Joachim Schmidt 2 1 2

Neuroscience Program, Department of Biological Sciences, Ohio University, Athens, OH, USA Biozentrum Köln, Institute für Zoologie, Universität zu Köln, Köln, Germany

2.1 Introduction Selective ion transport across membranes, with the resulting development of transmembrane ion concentration and electrical potential differences, occurs in all cells, prokaryotic and eukaryotic. These differences are the bases of respiration and photosynthesis and are a form of energy storage used to power an enormous number of transmembrane transport systems. These differences also provide an environment in which ion (e.g., calcium), ligand, and voltage gated ionotropic channels can evolve. These channels evolved well before the origin of eukaryotes, with bacteria (Kubalski and Martinac 2005; Martinac et al. 2008) having ligand-gated ionotropic channels and voltage-gated Na+ and K+ channels, and producing action potential-like electrical spikes (Kralj et al. 2011). The bacterial channels are homologous to vertebrate voltage-gated Na+ and K+ channels and were indeed the source material for the crystal structure analysis of these proteins (Doyle et al. 1998; Jiang et al. 2003; Payandeh et al. 2011; Zhang et al. 2012). Transmembrane potentials, and a theoretical ability for more complex types of electrical activity, are thus present in all cells; that this ability is more than theoretical is shown by invertebrate and vertebrate oocytes (Hagiwara and Jaffe 1979) and even plant cells (Fromm and Lautner 2007) generating action potentials. This ability is most highly evolved in neuron and muscle, with both having excitable membranes, and motor network function and muscle force production critically depending on variation in transmembrane potential. Understanding the generation of motor behavior on the cellular level therefore requires measuring and manipulating neuron and muscle transmembrane potentials (Fig. 2.1). Our goal in this chapter is to provide readers with sufficient understanding of neuron and muscle intracellular and extracellular recording to read this literature with profit. Sections 2.3 and 2.4 are practically oriented and should be sufficient to understand most modern intracellular and extracellular recording work. Section 2.3 describes contemporary methods for measuring and manipulating neuron and muscle transmembrane potential and current. Section 2.4 discusses contemporary methods for recording electric field potentials generated by neurons, nerves, and muscles. In these recordings the electrodes are placed outside, but in close vicinity, of the cells or nerves, and hence Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

8

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Figure 2.1 Neuron schematic. Synaptic input generates graded synaptic potentials in the dendritic region. Information is coded in the axon by action potentials. voltage time

are called extracellular recordings. These electrodes can also be used to generate electric fields that elicit action potentials in neurons. Although the theory of the electric field generation (Section 2.6) is the same, extracellular recording from close packed assemblages of large numbers of neurons and axons (e.g., in brain), and analyzing the data so obtained, require specialized techniques. We do not cover these techniques here, which are instead presented in Chapter 3. Section 2.5 provides a brief history of electrophysiological recording. Section 2.6 covers the theory of cell transmembrane potential and action potential generation and intracellular and extracellular recording, and is provided for readers interested in a more detailed understanding of the issues presented in Sections 2.3 through 2.5. Readers not interested in Sections 2.5 and 2.6 can leap from the end of Section 2.4 to 2.6.3.5, Extracellular Action Potential Summary, without substantial loss of continuity.

2.2 Terminology Electrophysiology involves measuring electrical potentials and current flows inside, outside, and across the cell membrane. Substantial possibility for confusion exists unless one carefully distinguishes among these different potentials and current flows. This possibility is increased by a failure to distinguish between electrical potential (the work required to bring a test positive charge from infinity, at which the potential is zero, to the point at which the potential is being measured) and voltage (the difference in electrical potential between two points) in much neurobiological writing. For instance, the potential difference across a neuron membrane at rest is technically a voltage, but is very often referred to as the “resting potential”. Very often this difference is immaterial, as the extracellular medium is grounded and thus held at a potential of zero, and hence the intracellular electrical potential equals the transmembrane potential difference (the transmembrane voltage). However, in understanding extracellular recording, it is often important to distinguish between the transmembrane voltage and the electrical potential present at various points in the extracellular medium (see Section 2.6). To be both unambiguous and maintain typical usage, we use the following conventions. When writing of an electrical potential difference across the cell membrane, we

Electrophysiological Recording Techniques

use “transmembrane potential”. When referring to the potential of a point in space, or to the electrical potentials of the inside or outside of the neuron as individual entities (i.e., relative not to each other, but to ground or infinity) (Section 2.6), we use “potential”. Because they refer to specific, well-defined entities, we use “resting potential” and “action potential” unless it is necessary to specify (Section 2.6) whether it is transmembrane potential difference, or inside or outside potentials, that is being referred to. Current can flow across the neuron membrane, along the membrane’s inside and outside surfaces, and (theoretically) longitudinally inside the membrane itself. Lipid bilayers have very high resistance. Current flow in the last case is therefore negligible and can be ignored. We refer to current flowing across the membrane as “transmembrane current” and describe the direction and amplitude of inside and outside currents as necessary without using special terms.

2.3 Intracellular and Patch Clamp Recording Transmembrane potential and current flow are measured between two points. To measure these entities, we accordingly need pairs of electrodes connected to an appropriate measuring instrument: one electrode with access to the intracellular space and one placed in the extracellular space. The extracellular electrode, the reference electrode, is usually connected to ground, thus setting extracellular potential to 0 mV. An important concern that applies to all intracellular techniques is the cable properties of neurons, which become more pronounced as neuron geometry becomes more extended. For spherical neurons with no or limited processes (isopotential neurons), the transmembrane potential measured by the electrode is a good measure of the transmembrane potential throughout the neuron. For more typical neurons with extensive processes, electrodes measure (Section 2.3.2) or clamp (Section 2.3.3) the transmembrane potential only of the parts of the neuron electrically close to the electrode (see Fig. 2.3). 2.3.1 Recording Electrodes

Two approaches are commonly used to gain access to the intracellular side of cells. Intracellular Recordings with Sharp Glass Microelectrodes The cell is impaled with sharp intra-

cellular glass microelectrodes (0.01–0.1 μm tip diameter) that are typically filled with a highly concentrated electrolyte solution (e.g., 3M KCl) (Figs. 2.2, 2.3). Whole-Cell Patch Clamp Recordings This recording configuration is achieved in two steps.

First a relatively blunt (compared to sharp intracellular electrodes) (1–2 μm tip diameter) glass electrode is sealed by suction to the membrane, forming a very high (GΩ) resistance between the inside of the pipette and the external solution. At this point the activity of channels in the membrane under the electrode can be measured (see Fig. 2.8E and patch clamp part of Section 2.3.3.2). The membrane patch under the rim of the electrode tip is then ruptured (whole-cell patch-clamp configuration) providing relatively low resistance access to the cytoplasm (Fig. 2.4). Patch electrodes are filled with a solution that, ideally, is isotonic to the cytoplasm. However, it is impossible to match all the other soluble contents of the cytoplasm, and loss of these components to the relatively very large pipette volume (see below) is inevitable.

9

10

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

suction electrode

intracellular electrode

B suction electrode recording

intracellular recording

2 mV

20 ms

Figure 2.2 Extracellular and intracellular recording of a neuron in the abdominal ganglion of a crayfish. (A) Schematic of recording situation. A suction electrode (left) was placed on the surface of the ganglion where the primary dendrite of the neuron of interest was very close to the surface. The neuron was also impaled with an intracellular electrode (right). (B) The first trace shows two action potentials recorded with the suction electrode. The same action potentials were recorded by the intracellular electrode (second trace). Recording courtesy of C. Smarandache-Wellmann.

Comparison of Two Approaches Both methods are widely used and have advantages and

disadvantages. Sharp intracellular microelectrodes, especially when used in small cells, require small tip diameters to minimize cell damage when the electrode is inserted into the cell. Small tip diameters can create problems because tiny, high resistance electrode tips easily clog during the recording. This can lead to varying electrode resistance and unstable recordings and can cause problems for passing current. The electrodes can also be used to iontophorese charged dyes or indicators into the cells. However, again because of the small tip diameter, iontophoresis of the substances easily clogs the electrode tips. Another concern with almost all past sharp electrode work is that it has been performed with electrode fill solutions that do not match cytoplasm ionic makeup and typically have ionic strengths far greater than that of cytoplasm. Recent work has shown that high ionic strength electrode fill solutions diffuse even across the small tip diameters of sharp electrodes, and cause large changes in cell properties (Hooper et al. 2015) (see Section 2.6.2 for full discussion of this issue). Adequate sharp electrode current and voltage clamp recordings can be obtained using fill solutions whose ion concentrations

Electrophysiological Recording Techniques

A

B PD

10 mV

LP

5 mV

Ivn

50 μV 0.5 s

Figure 2.3 Intracellular and extracellular recordings in the lobster stomatogastric ganglion. (A) Schematic of recording situation. The PD and LP neurons were impaled with sharp electrodes. A pin electrode was placed close to a nerve (lvn) containing the axons of the LP, PD, and other neurons. The grey ring indicates the Vaseline well into which the pin electrode was placed. This electrical isolation of the pool from the bulk saline allowed the pin electrode to pick up the electric fields generated by action potentials in the nerve. (B) The PD neuron generated bursts of action potentials (first trace). LP neuron action potentials (second trace) evoked inhibitory postsynaptic potentials in the PD neuron. The extracellular recording shows action potentials from many neurons that can be discriminated by differences in amplitude. The largest spikes are LP neuron action potentials (note one-to-one relationship). Extracellular signals are much smaller than intracellular signals (compare scaling). The intracellularly recorded action potentials are not overshooting because they were recorded in the cell bodies, which are inexcitable; action potentials passively conduct to the cell bodies, thus decreasing their amplitude.

match that of cytoplasm (Hooper et al. 2015). This problem can thus be overcome in future work. However, it does pose a difficulty in interpreting the large amount of work using electrodes filled with high-ionic strength solutions that has been performed over the last 60 years. Since the recordings are not based on impaling the neuron, whole-cell patch electrodes have large diameter tips compared to sharp intracellular electrodes. They therefore have much lower resistances, which means that establishing adequate bridge balance and discontinuous current and voltage clamp is much easier, and a type of continuous single-electrode voltage clamp can even be obtained (see Section 2.3.3.2). One drawback is that establishing good seals requires that the electrode tip be very clean and an area of neuron membrane free of glial cells or other adhering material be available, which limits the experimental conditions in which good seals can be obtained. Another drawback is that, after establishing the whole-cell configuration, the solution in the patch pipette freely exchanges (to an extent much greater than with sharp electrodes) with solutes of the cytoplasm. It is important to appreciate the large difference between these two volumes—if the neuron were the size of a car, the pipette would be the size of the Eiffel tower (we thank C. Pouzat for this example). As such, electrode contents rapidly dominate because of the large volume ratio. While this can disturb intracellular signaling and normal cell function, it can also be used to deliberately change the intracellular solution and to inject the cell with dyes for single cell labeling, indicators for optical imaging, or pathway modulators to study intracellular signaling.

11

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B

20 mV whole cell patch-clamp 1s

C

Instantaneous frequency (Hz)

12

4

3

2

1 0

2

4 Time (s)

6

8

Figure 2.4 Current-clamp recording in the whole-cell patch-clamp configuration from a local interneuron of an insect antennal lobe. (A) Schematic of recording situation. (B) A train of action potentials evoked by a depolarizing current pulse. The spiking pattern shows strong spike frequency adaptation (SFA). (C) SFA shown as instantaneous frequency over time. Recording courtesy of J. Radermacher.

In experiments in which it is critical to insure the integrity of the intracellular components, perforated patch clamp recordings can be performed (see Fig. 2.5 for an example). In this configuration the solution in the patch electrode contains ionophores (e.g., amphotericin) that insert into the membrane after the electrode seals onto the cell. The ionophores used are typically permeable to monovalent ions but impermeable to divalent ions and large molecules. They thus provide electrical access to the intracellular side of the cell but prevent larger molecules from diffusing from the cell into the recording electrode. 2.3.2 Current-Clamp: Measuring Transmembrane Potential

In neurons information is computed, coded, and conducted by changes in transmembrane potential. Changes from resting potential can be graded (e.g., synaptic potentials) or stereotyped, threshold-dependent, all-or-nothing responses resulting from activation of cell active membrane processes (e.g., action potentials) (Fig. 2.1). The transmembrane potential, the potential difference between the extracellular and intracellular electrodes, is measured with a voltage measuring amplifier (insets to intracellular recordings, Figs. 2.2, 2.3, 2.4; see also Fig. 2.9).

Electrophysiological Recording Techniques

A –70 pA –190 pA

20 mV 500 ms

B –60 mV – 85 mV –100 mV –110 mV

200 pA 500 ms

Figure 2.5 Sag-potential and Ih currents in a dopaminergic neuron of the substantia nigra in a mouse brain slice. (A) Current-clamp recording in perforated patch-clamp configuration. Three pulses of increasing hyperpolarizing current evoked sag potentials of increasing amplitudes. (B) Voltage-clamp recording in perforated patch-clamp configuration. Pipette interior potential was clamped at a holding potential of –60 mV. Hyperpolarizing command potential steps of increasing amplitudes evoked slowly developing inward currents of increasing amplitude. These hyperpolarization activated currents (Ih ) generate the sag-potentials shown in A. Recording courtesy of U. Collienne.

In these treatments it is important to consider that the electrode forms a RC-circuit in series with the cell (see Fig. 2.9). For simple recording of transmembrane potential, the primary difficulty posed by this circuit is that it low pass filters, and thus limits, the ability to accurately measure rapid changes in transmembrane potential. With the commonly encountered RC values of sharp electrodes, this filtering can be great enough that signals with time-to-peak less than 5 ms (e.g., the action potential) would not be accurate measured (Halliwell and Whitaker 1987). This difficulty is overcome by electronic circuitry that “cancels” the electrode resistance and capacitance, thus reducing the electrode time constant.

13

14

Neurobiology of Motor Control: Fundamental Concepts and New Directions

To stimulate the neuron (e.g., to elicit action potentials) or analyze passive membrane properties (e.g., cell input resistance, cell capacitance) requires passing current into the cell. In this case the electrode resistance causes an additional difficulty because a voltage occurs across the electrode when passing current. Since the resistor is in series with it, this voltage adds to the transmembrane potential, and the electrometer thus measures the sum of the two voltages. The electrode voltage can be substantial, being, e.g., 25 mV if a 1 nA current is passed through a 25 MΩ electrode. The ideal method to avoid interference between transmembrane potential measurement and current injection is to impale the cell with two intracellular electrodes, one to measure transmembrane potential and the other to pass current (Two Electrode Current Clamp, TECC). However, many cells are too small to be impaled by two electrodes. In these cases, a single electrode must be used to measure both transmembrane potential and inject current. To compensate for the interference between these two tasks caused by electrode RC properties, special compensation circuitry (Bridge Balance), or a time sharing technique (discontinuous current-clamp, dSECC) (Brennecke and Lindemann 1974a,b; Finkel and Redman 1984; Halliwell and Whitaker 1987; Polder and Swandulla 2001; Pun 1988; Roelfsema et al. 2001; Wilson and Goldner 1975; references for discontinuous voltage clamp, but technology is the same), is used. Bridge Balance uses an electrical circuit to generate and subtract a signal from the voltage recording to compensate for the voltage across the resistance of the recording electrode. The primary difficulty with this technique is that the resistance of most sharp electrodes shows large variation as current sign (depolarizing, hyperpolarizing) and amplitude change (due to their lower electrode resistances, this problem is less important for current clamp recordings in whole cell, e.g., Fig. 2.4, or perforated patch, e.g., Fig. 2.5A, current clamp). Adequate bridge balance for sharp electrodes can therefore usually be achieved over only a small range of current injection values. In dSECC time-sharing amplifiers switch between transmembrane potential measurement and current injection to prevent interference between the tasks. Although this technique can be used with either sharp or whole cell current clamp recordings, because of the difficulties associated with their high resistances, it is most often used with sharp electrodes. This approach is possible because the RC time constant of most electrodes (of either type) is significantly smaller than those of most neurons. When current injection ceases, the electrode therefore discharges much faster than the cell membrane. Transmembrane potential uncontaminated by current flow through the electrode can thus be measured after electrode discharge is complete. It is critical both that the electrode is essentially completely discharged before transmembrane voltage measurement starts and that switching frequencies are high enough that substantial cell discharging does not occur during the transmembrane potential recording period. The difficulty of achieving this balance is increased because electrode capacitance is not constant during current injection, but instead changes due to induced redistribution of ions in the electrode (Finkel and Redman 1984; Roelfsema et al. 2001). Electrode discharge therefore has a rapid, RC-like initial component followed by a much slower late component; it is important not to mistake this flatter late component with full electrode discharge. Modern amplifiers have electronic circuitry that compensates for electrode capacitance to decrease electrode discharging time and thus increase the acceptable switching rate. As noted above, sharp electrode resistance varies with

Electrophysiological Recording Techniques

current injection amount and direction. Switching rates correct for one current injection level are thus not, in general, correct for others. 2.3.3 Voltage Clamp: Measuring Transmembrane Current

Voltage clamp recordings are used to measure the transmembrane currents that generate the transmembrane potential (Fig. 2.5). These currents are the product of ion channel conductance and the driving force on the ions the channel carries (see Section 2.6). When the channel’s reversal potential is known, these data can therefore be used to investigate how ion channel conductances depend on transmembrane potential. Voltage clamps are negative feedback systems that compare two potentials, one of which is user-set, and inject current to minimize the difference between them. The injected current is measured and equals the amount of current passing through the membrane at the command potential. They thus allow users to alter neuron transmembrane potentials and describe how transmembrane currents change in response. Two different techniques have been developed (see Section 2.6 for the electronic circuits that implement each type), which differ with respect to the definition of the voltage to which the user-set voltage is compared. We describe each technique separately. 2.3.3.1 Voltage Clamp with Transmembrane Potential as Reference

In the first technique neuron transmembrane potential is measured continuously, or so frequently that the measurement is effectively continuous, and (with correct settings) the voltage clamp circuit thus clamps a truly known transmembrane potential to the user-set potential. This goal can be achieved in two ways. Two Electrode Voltage Clamp (TEVC) Ideally, to avoid interference between measurement

of transmembrane potential and current injection, voltage clamp is performed with two intracellular electrodes, one that measures transmembrane potential and another that injects current into the cell. The feedback amplifier compares the measured transmembrane potential with the command potential and injects current through the current passing electrode to minimize the difference between the voltages (see Fig. 2.9). In this technique the only limitations are the amount of current the current-passing electrode can pass, and how large the gain of the feedback circuit, which determines how closely the neuron transmembrane potential tracks the user-set potential (Halliwell and Whitaker 1987), can be made without the electronic circuit oscillating. To achieve maximum possible gains, modern equipment has circuits that can be set to compensate for electrode resistance and capacitance. Discontinuous Single Electrode Voltage Clamp (dSEVC) Since most neurons are too small to

be impaled with two electrodes, an alternative voltage clamp method that requires only a single electrode has been developed. In dSEVC the same time sharing principle between transmembrane potential measurement and current injection as in dSECC is used to prevent interference between the two tasks. Again, high switching frequencies are important to optimize the clamp, but it is essential to ensure that no current passes through the electrode during transmembrane potential recording. dSEVC was originally used primarily with sharp microelectrodes but can be also be used with whole cell patch pipettes. In this case, given the smaller resistances of patch pipettes,

15

16

Neurobiology of Motor Control: Fundamental Concepts and New Directions

the difficulties of achieving adequate clamp are much reduced. As with dSECC, modern equipment has circuits that can be set to compensate for electrode resistance and capacitance so as to maximize switching frequency. 2.3.3.2 Voltage Clamp with Preparation (Bath) Ground as Reference

The other voltage clamp technique is patch clamp and its derivative, whole-cell patch clamp. Both these techniques are explained in detail in Ogden and Stanfield (1994). We first cover patch clamp. Patch Clamp In patch clamp recording the cell membrane under the pipette is kept

intact (it is the situation in the Whole Cell Patch Clamp Recordings part of Section 2.3.1 before membrane rupture). As such, absent information from intracellular electrodes, the experimenter has no direct knowledge of cell transmembrane potential. It is thus impossible to clamp cell transmembrane potential to a user-set potential. Instead, it is the potential of the interior of the pipette that is clamped to a user-set potential (see Fig. 2.9). In the absence of the patch-clamp apparatus, currents flowing through the membrane under the pipette would change the potential of the interior of the pipette. To maintain the potential of the pipette interior at the user-set value, the voltage clamp injects a current equal and opposite to the transmembrane currents. This injected current is measured as the output of the system. Because of the very high resistance between the pipette interior and the external solution, this clamping allows the current flowing through individual channels to be observed (see Fig. 2.8E). An important technical issue is that in patch clamp how one changes transmembrane potential is reversed. Consider, for example, the case in which one desires to bring the transmembrane potential of the membrane under the pipette to 0 mV. Using the frequent convention that current injected through an electrode is positive, to bring the transmembrane potential to 0 mV in two-electrode voltage clamp would entail injecting positive current to depolarize the cell interior from resting potential (say, –70 mV) to 0 mV. In patch clamp, alternatively, one controls only of the potential of the pipette interior. Thus, to bring the transmembrane potential to 0 mV, one must hyperpolarize the pipette interior to –70 mV, and thus would inject negative current. In the literature this “reversal” is typically taken into account, and the above procedure would be referred to as a transmembrane (or membrane) depolarization, not a pipette hyperpolarization. The fact that in patch clamp the pipette does not have access to the cell interior leads to a limitation of the technique. Without having another electrode in the cell, one does not know the cell’s transmembrane potential. As such, data from patch clamp experiments taken from intact cells (cell-attached patches) attempting, for instance, to describe the voltage-dependence of channel opening, suffer from one having to calculate the transmembrane potential using an assumed value of cell resting potential. This can be overcome by, at the end of the experiment, rupturing the membrane and switching to current clamp recording to measure the cell’s real transmembrane potential. However, this assumes that the cell’s transmembrane potential has been constant over the course of the experiment. A technique to overcome this difficulty is to perform experiments in which the membrane under the pipette is removed from the cell, and thus the potential on the external side of the membrane is the potential of the external solution, typically zero. Since the potential in the pipette is being clamped, in this case the transmembrane potential of

Electrophysiological Recording Techniques

the patch is unambiguously known. Such cell-detached patches can be created in two ways. In the first the seal is established and the pipette pulled away from the cell. The portion of membrane under the patch remains attached to the pipette, resulting in an “inside-out” patch, since the side of the membrane that was originally on the interior of the cell now faces the external solution. This also allows applying at will chemicals (e.g., Ca) to the inside side of the membrane. In the second way the membrane under the pipette is first ruptured and the pipette is then pulled away. The cell membrane that pulls away with the pipette then comes together and forms a new, continuous membrane across the pipette lumen, but in this case the membrane that faced the external solution when in the cell again faces the external solution in the detached patch (hence called an “outside-out” patch). Continuous Single Electrode Voltage Clamp (cSEVC) or Whole-Cell Patch Clamp In this technique

a patch clamp seal is achieved and then the cell membrane ruptured or perforated (Fig. 2.5) to access the cell interior. At this stage, if the electrode is attached to an electrometer, the cell’s transmembrane potential is measured as in the current-clamp section above (Figs. 2.4, 2.5A). If one instead switches to the patch clamp circuit explained above, the voltage clamp attempts to clamp the interior of the pipette to the user-set command potential. It therefore injects whatever currents are required to charge the electrode capacitance, establish the appropriate voltage across the electrode resistance, charge the cell membrane capacitance, and establish the appropriate voltage across the cell membrane resistance, including the changes in cell membrane resistance that occur as a result of changes in cell transmembrane potential, necessary to keep the pipette interior at the command potential. It should be clear from the discussions above about electrode resistance and capacitance that simultaneously fulfilling all these tasks is difficult (Ogden and Stanfield, 1994). For instance, if the voltage drop across the electrode was –10 mV, setting the command potential at –50 mV would result in clamping the cell transmembrane potential to only –40 mV. Rapidly changing command potential profiles are again low-pass filtered by the electrode RC properties. Finally, large cells have large capacitances, thus requiring large current injections, which both accentuate the voltage loss across the electrode (V = IR) and can strain the current delivery capacity of the voltage clamp circuit. The naturally low resistance of patch electrodes helps with the electrode voltage drop issue; whole cell patch is very often used in small neurons in which two-electrode voltage clamp is impossible, which helps with the cell capacitance charging issue; and compensatory circuits are used in all whole cell patch clamp apparatus to further reduce these errors. With proper care, high quality recordings of whole cell currents at well-controlled transmembrane potentials can thus be obtained (Fig. 2.5B).

2.4 Extracellular Recording and Stimulation When ion channels open in the cell membrane, currents flow between different parts of the cell and through the extracellular space. As current flows through the resistance of the extracellular space, a potential difference is produced that can be measured by extracellular electrodes (see Section 2.6). These electrodes are placed either close to the neurons or at some distance, e.g., on the skull or skin. Depending on placement,

17

18

Neurobiology of Motor Control: Fundamental Concepts and New Directions

the electrodes can record the sum of the action and synaptic potentials of only a few neurons or even one, or of large ensembles of neurons (field potentials). Field potentials are covered in Chapter 3. We therefore do not consider them here. Extracellular electrodes are often made of metal, typically steel or tungsten, insulated but for the tip. The difficulties described in Section 2.6 about electrode polarizability are in general not a concern because voltmeters (which do not depend on current flow through the circuit, and, ideally, have zero current flow through them), not galvanometers, are now used. Blunt glass pipettes filled with saline as an electrolyte can also be used, either by simple placement near or in the tissue of interest or, often with an attached plastic tube heat-pulled to the appropriate diameter, as suction electrodes. A reference electrode is always placed either close to the first electrode in contact with the same tissue, or somewhere in the body cavity or saline. Extracellular electrodes pick up action potentials from neurons in their vicinity. When recording from areas or nerves with relatively few active neurons (e.g., 1st trace, Fig. 2.2B; 3rd trace, Fig. 2.3B), individual action potentials can often be recorded. Depending on the recording method (see Section 2.6), such single action potentials may appear as mono-, bi- (also called “di-” in some articles), or triphasic (e.g., 1st trace, Fig. 2.2B) signals (Stein and Pearson 1971). In early work considerable effort was devoted to understanding the basis of extracellularly recorded action potential shape, and whether these differences could be used to gain insight into neuron and axon properties (Section 2.6). With special techniques, it is also possible to use extracellular electrodes to measure absolute transmembrane potential (Section 2.6). However, all extracellular recording techniques are highly susceptible to small changes in recording technique and extracellular amplifier filter settings. With the invention of intracellular recording, in modern work extracellular recording is therefore primarily used to detect action potential discharge patterns and measuring action potential frequencies. One important advantage of extracellular recording is that it is often possible to record the activity of multiple neurons simultaneously from single electrodes. Signal amplitude depends on the distance between the electrode and the active axon and how large the neuron or the axon diameter is. Large neurons or large diameter axons produce large electric fields, and hence large extracellular action potentials, because more current flows along and out of the axon. In situations where the activity of relatively small numbers of neurons are being recorded, these size differences sometimes allow action potentials from different neurons to be uniquely identified in recordings from single extracellular electrodes (Fig 2.3, 3rd trace). The use of multiple electrodes, and more advanced spike sorting techniques to identify the activity of individual neurons in more complex situations, are covered in Chapter 3. Extracellular electrodes are also used to stimulate single neurons or neuron ensembles, i.e., tracts or many axons in a nerve. Such stimulation leads to a special type of extracellular action potential, the compound action potential (CAP), which played a central role in understanding nerve axon make-up and physiology. CAPs are recorded by stimulation of a nerve at one position and recording the resulting activity at another site on the nerve. If the distance between the two electrodes is large enough, a complex, multi-peaked change in extracellular potential is observed (Fig. 2.6, trace with “Aα, Aβ”, “Aδ”, and “C”) (Erlanger and Gasser 1930). This complex shape arises from the presence of axons with varying diameters in the nerve, and action potentials in axons with larger diameters propagating with higher velocities (upper inset, Fig. 2.6) (Gasser and Grundfest 1939). As the action potentials propagate from the stimulation site, they therefore

Electrophysiological Recording Techniques

Aα, Aβ 100 Velocity (m/s)

6

Amplitude (mV)

5

80 60 40 20 0

4

0

2 4 6 8 10 12 Diameter (µm)

3

2

0

0.8

1.6 Time (ms)

Aδ

3.2

2.4

1 C 0 0

10

20

30

40 50 Time (ms)

60

70

80

90

Figure 2.6 The Compound Action Potential (CAP). The trace with the “Aα, Aβ”, “Aδ”, and “C” labels is a CAP recorded from a saphenous nerve with a distance of 37 mm between the stimulation and recording electrodes. Trace modified from Perl (2007) with permission. Uppermost inset: action potential velocity vs. axon diameter for multiple axons in cat saphenous nerve. Middle inset: Algebraic addition of action potentials from axons with different diameters gives rise to two-peaked CAP. Both insets modified from Gasser and Grundfest (1939) with permission.

sort by conduction velocity (axon diameter), and the sum of the separated classes of action potentials give rise to the multiple peaks of the CAP (middle insert, Fig. 2.6). The amplitude of the peaks depends on both the number of axons in each class, and the amplitude of the individual action potentials (with larger diameter axons, as noted above, having larger amplitudes). As such, the shape of a CAP tells nothing about the shape of any of the individual action potentials that comprise it. A large number of techniques can be used for extracellular recording and stimulation; we now summarize several commonly used methods. Neuron Recordings In some preparations the activity of single or a few neurons can be recorded with suction electrodes. These electrodes are saline-filled blunt micropipettes attached to a device that allows applying negative pressure (in which sense they are similar to patch-clamp electrodes). Suction electrodes are used to record activity from tracts in slice preparations or superficial neurons in invertebrate ganglia (e.g., Brown et al. 2006; Smarandache-Wellmann et al. 2009). Suction electrodes have also been used to record action potentials from single identified leech neurons by placing the electrode on the cell body (Cymbalyuk et al. 2002). Cymbalyuk et al. is a nice example of an advantage of extracellular over intracellular recording in that intracellular recording from this neuron introduces an additional leak current that markedly changes neuron firing.

19

20

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Nerve or Nerve Root Recordings The activity of axons in nerves or nerve roots can be

recorded with hook, suction, or pin electrodes. Hook electrodes are commonly used to record en passant from invertebrate nerves or connectives containing axons of interest. Hook electrodes are also used to record from vertebrate preparations, e.g., sciatic nerve. Such recordings can be made from nerves in intact, semi-intact, or isolated in vitro preparations. Often, the nerve is positioned on the hook-like ending of a single electrode while the reference electrode is positioned close by in the body cavity. Metals are used to fabricate the electrodes. It is critically important to isolate the electrode and recording site from the surrounding saline or body fluid to prevent a short circuit between the recording and reference electrodes. This is often done by lifting the hook and its overlain nerve into a layer of petroleum jelly (Vaseline) or viscous silicon oil. To achieve better isolation, particularly of fine nerves that are easily damaged, a device in which a plastic tube is situated above the hook and nerve, and from which highly viscous grease is extruded that forces the nerve firmly against the hook, and saline away from nerve and hook, can be used (Schmitz et al. 1988, 1991). For chronic in situ recordings cuff electrodes are often used (Foldes et al. 2011; Gruhn and Rathmayer 2002; Loeb and Peck 1996; Möhl 1977). Hook electrodes are often used to stimulate axons. A classic example is frog sciatic nerve. Generally, when stimulating a nerve extracellularly large diameter fibers are activated more readily than small diameter fibers. Large diameter fibers have a lower excitation threshold mainly because of their low longitudinal resistance. In vertebrates, selective activation of small axons can be achieved using a multi-contact electrode array placed along the nerve (Lertmanorat et al. 2006). Suction electrodes (Fig. 2.2) are used as an alternative to hook electrodes or when space limitations do not allow hook electrodes. Nerve activity can be recorded en passant or from a stump sucked into the electrode (e.g., Land 2001). Pin electrodes (Fig. 2.3) are used to record from invertebrate preparations with parts of the CNS placed in a petri dish lined with Sylgard, a silicon elastomere. A pin, typically steel, is pushed into the Sylgard immediately next to the nerve. The pin and nerve are then isolated from the surrounding medium by Vaseline. Alternatively, Vaseline is used to form a well separating a saline pool from the surrounding saline. The activity in a nerve running through this pool can be recorded by pushing a pin into the pool (the pin need not touch the nerve). A second pin electrode is placed anywhere in the saline as a reference. Recordings of Electrical Muscle Activity (EMG) We cover here recording the activity of skeletal

muscles only, as it is the activity of these muscles that generates behavior. In particular, we do not cover the specialized and extensive literature on extracellular recording of heart activity (electrocardiography). The electrical activity of skeletal muscles can be recorded extracellularly with surface electrodes attached to the skin or by intramuscular recordings. Here we consider only the latter because their placement is precise and measurements are not confined to only superficial muscles. For small animals like insects intramuscular recordings are the only way to record EMGs. Recordings are performed by inserting needles or wires insulated except for the tips into the muscles. In arthropods this is easily done because the electrodes can be pushed into a muscle through small holes made in the exoskeleton, with the wires then being glued to the exoskeleton for stabilization. Invertebrate muscles are often

Electrophysiological Recording Techniques

innervated by only a few motorneurons, allowing individual EMG potentials to be assigned in some cases to the action potentials of particular motorneurons. EMGs have been recorded from behaving adult mice by implanting a pair of twisted wires insulated except at the tips into the muscle (Akay 2014; Pearson et al. 2005). Decomposition of the signals can be used to identify individual motor unit action potentials (Merletti and Farina 2009). A concern in large muscles is that EMG electrodes may pick up signals from only a fraction of muscle fibers. This may lead to an over-simplified interpretation of the data if the signals are picked up from portions of the muscle containing primarily a single muscle fiber type (e.g., fast or slow fibers).

2.5 A Brief History of Electrophysiological Recording Our goals in this section are to provide a chronology of electrophysiological “firsts” and to help the reader understand how great the challenge recording from nerves and muscles was, something workers accustomed to modern equipment may not fully appreciate. We do not explain here the theoretical basis for the mono and diphasic action potential shapes seen in these figures; readers interested in this issue should see Section 2.6. Other descriptions of this history include Lenoir (1986); Nilius (2003); and Verkhratsky (2006). Although Newton (1713) hypothesized that electrical processes were involved in nervous system function, the first experimental evidence of this involvement was Galvani’s (1791) demonstration that electric shocks could induce muscle contraction. Further progress was stymied by the very small amplitude of axon and muscle transmembrane currents and potentials and the use, at the time, of galvanometers to measure nerve and muscle electrical activity. This use of a current-measuring device required inventing non-polarizable filter electrodes (ones in which, as in silver-chloride electrodes, a reversible redox reaction Ag + Cl− ⇌ AgCl + e− underlies the electrical interaction between the electrode and its environment) to prevent electrode polarization. du Bois-Reymond (1848) solved the small amplitude problem by using a coil-driven (to multiply the magnetic field induced by current flowing through the coil’s wire) galvanometer and the non-polarizable electrode problem by using filter paper electrodes. These allowed du Bois-Reymond to show that when electrodes were attached to the outside surface of a muscle and to its cut end, a current flowed through the galvanometer (the “injury current”, explicable today as arising from the interior of the muscle being hyperpolarized relative to its outside, see Section 2.6). He also showed that when the muscle’s nerve was tetanically stimulated, the injury current slowly decreased and then stabilized at a lesser value (with modern knowledge, due to the depolarization of the muscle interior during tetanus decreasing the potential difference between the inside and outside of the muscle). His galvanometer was much too slow to measure the time course of the change in injury current (the change in muscle transmembrane potential) that occurred at the beginning of the tetanus. That is, the slow decrease he observed at the beginning of the tetanus was the time course of the galvanometer response. At low stimulation frequencies, muscle contractions that occurred one-for-one with each nerve stimulation could nonetheless be observed. The question thus arose

21

22

Neurobiology of Motor Control: Fundamental Concepts and New Directions

whether the muscle’s one-for-one contractions were associated with one-for-one electrical responses in the muscle, or if the muscle’s electrical response was actually to move slowly and smoothly to a new steady-state condition as indicated by the galvanometer. That is, the galvonometer’s sustained response could be due either to the device actually measuring a sustained current, or to it slow-filtering a train of brief, individual electrical responses in the muscle. du Bois-Reymond resolved this issue by laying the nerve of one muscle (muscle B) across another muscle (muscle A). Stimulating muscle A’s nerve induced one-for-one contractions in both muscles. A continuous electrical change in muscle A, as reported by the galvanometer, would not be expected to cause one-for-one coupling of muscle A nerve shocks and muscle B contractions. du Bois-Reymond reasoned that this coupling instead occurred because rapid, unitary electrical events in muscle A induced by shocks of muscle A’s nerve were in turn shocking the nerve of muscle B lying over muscle A. du Bois-Reymond therefore concluded that the muscle response to nerve shocks was one-for-one electrical changes, and the sustained galvanometer response was due to the device slow-filtering these electrical changes. Despite these accomplishments, du Bois-Reymond’s hypothesis for the basis of the injury current was mistaken; he posited a battery-like construction of dipolar charged molecules, with changes in current occurring because of rotation of these molecules. A student of his, Julius Bernstein (1902, 1912), building on the work of Walther Nernst (1889) on the theory of electrolytes, subsequently proposed the correct explanation, that muscle (and axon) transmembrane potentials arise from diffusion potentials across semi-permeable cell membranes (see Section 2.6), and estimated that the axon interior was about 60 mV negative compared to its outside surface. Bernstein (1912) also showed that the temperature dependence of the injury current matched that predicted by the Nernst equation. He did not check the predicted dependence on extracellular K+ concentration, presumably because, as in all work of this era, the nerve was not bathed in an extracellular saline but instead suspended in air. Bernstein (1868, 1871) was also the first to measure the time course of a (compound, CAP) action potential (Fig. 2.7A, left). He accomplished this achievement not by designing a galvanometer that responded more rapidly to current changes, but instead by designing an ingenious mechanical device (Fig. 2.7A, right) that allowed measuring CAP amplitude at different times after its stimulation. He could thus repeatedly stimulate CAPs and, by altering when he measured the CAP amplitude after the stimulations,

Figure 2.7 A history of extracellular recording. (A) The first recording of a CAP (left) and the sampling machine that allowed it (right). “h” indicates resting injury current (bottom horizontal line is zero current), and thus the action potential overshoots. (B) Photograph of a capillary electrometer recording of a monophasic (muscle damaged under second electrode, see Section 2.6) muscle action potential (upper left) and calculated action potential (bottom left). The right panel shows calculated diphasic (muscle undamaged under the second electrode) action potential. (C) Photographs of capillary electrometer recordings (left) and calculated action potentials (right, “C” and “D” match in both) from a nerve that was undamaged between the recording electrodes (upper panels, diphasic action potentials) and when the nerve was killed between the electrodes (bottom panels, monophasic action potentials). (D) First cathode ray recording of a CAP induced by repeated nerve stimulations. (E) First cathode ray recording of three single (from individual axons) action potentials. A from Nilius (2003); B, Lucas (1909); C, Adrian (1926); D, Gasser and Elanger (1922); E, Blair and Erlanger (1933a), all with permission.

Electrophysiological Recording Techniques

A ϒ

m

ϑ

o

ϒ′

h t

τ1

t′

τ2

B +.04 +.03 +.02 +.01 STIM. 0 .03 –.01 –.02

.02

–.03 –.04

STIM. .005

SEC.

VOLT.

VOLT.

.01

.01

SEC.

.01

.02

C 10 gms. (diphasic)

D

Millivolts.

C

C

10 gms. (monophasic)

D

.02 .01

.01 SEC.

0

D

E

.02

.03

.04

.05

23

24

Neurobiology of Motor Control: Fundamental Concepts and New Directions

obtain enough data points to describe the CAP time course. He also measured the conduction velocity of the CAP, finding a value close to that already obtained by Helmholz (1850). An important result of this work was that during the action potential the interior of the axon did not just reach 0 mV, but became positive relative to the axon exterior. With modern knowledge, this overshoot implies that the membrane is permeable to two ions, one (K+ ) with a negative reversal potential and another (Na+ ) with a positive reversal potential, with the action potential arising because of a large increase in Na+ permeability. However, Bernstein was convinced that potassium alone created the transmembrane potential. He therefore discounted his own data showing overshooting behavior, arguing instead that the action potential resulted from an increased membrane permeability to all ions (which would result in transmembrane potential going only to zero, not overshooting). Overton (1902) corrected this error and showed that Na+ or Li+ are required for the muscle action potential. Thus, by the beginning of the 1900s, using what to modern minds seem almost inconceivably primitive equipment, it had been shown that axons and muscles had resting potentials due to charge separation across a semi-permeable membrane, that axons carried information in the form of brief, propagating electrical disturbances that involved the loss of this charge separation, and that muscle contractions were due to axon impulses inducing similar brief depolarizations of the muscle. Nonetheless, these techniques were insufficient to achieve the fundamental goal— investigating how nervous systems code information—which requires measuring nerve activity instant-by-instant as it occurs, as opposed to the method of du Bois-Reymond, in which a same event (the CAP) was induced over and over, and the time course of the event was reconstructed by measuring its amplitude at a different time in each repetition. The insurmountable difficulty was the slow time response of the galvanometers. This was first overcome by modification (Adrian 1926; Lucas 1908) of the capillary electrometer, first invented by Gabriel Lippmann (McKendrick 1883), in which voltage changes across a mercury column, by changing the mercury’s surface tension, change column height. Capillary electrometers respond rapidly to voltage changes which, when coupled with contemporary improvements in photography, allowed recording in real time the change in column height induced by an unamplified muscle action potential (Fig. 2.7B, upper left) (Lucas 1909). Transforming this change into transmembrane potential is straightforward; the bottom panel shows a monophasic action potential and the right panel a diphasic action potential. This use of a capillary electrometer was also the first direct measurement of voltage. The sensitivity of the capillary electrometer was too small to record unamplified axon action potentials. This limitation was overcome with the invention of the triode vacuum tube (“valve” in British English) amplifier by Lee De Forest in 1908. Edgar Adrian (1926) used a capillary electrometer to record valve-amplified sensory neuron action potentials triggered by muscle stretch (Fig. 2.7C; left, photographs of mercury column heights; right, transmembrane potentials derived from these data) and verified that these responses were action potentials by showing they were diphasic or monophasic depending on whether the nerve was injured between the two recording electrodes (see Section 2.6). Although this combination of a device with a rapid response characteristic (the capillary electrometer) and a valve amplifier could record both muscle and axon activity,

Electrophysiological Recording Techniques

transforming the column height variations into transmembrane potentials was tedious and could not be performed in real time. Around this same time another method to record transmembrane potentials, the cathode ray tube, was being developed. In this technique voltage is used to change the position of an electron beam on a fluorescent screen. In early work the electron beam intensity was too low to induce visible fluorescence with single beam excursions. However, when the nerve was repeatedly stimulated to produce identical CAPs, sufficient excitation was induced to see the CAP as a line of fluorescence (Fig. 2.7D) (Gasser and Erlanger 1922). Applying this technique to single action potentials required increasing electron beam intensity sufficiently to allow single sweeps to induce visible fluorescence. These changes were made relatively quickly, and by 1932 allowed recording individual axon action potentials (Fig. 2.7E) (Blair and Erlanger 1932, 1933a). Subsequent advances have resulted in more powerful amplifiers and ever progressing data storage technologies, but all the fundamental difficulties of recording device sensitivity and response rapidity required to record extracellular activity had thus been solved by 1932. With special recording methods (which, in retrospect, are similar to those used by Bernstein and other early workers), it is possible to use extracellular recordings to measure the action potential transmembrane potential, and, with modern knowledge and computing power, it is theoretically possible to calculate what transmembrane potential gives rise to a given extracellularly recorded action potential (Plonsey 1977). However, in the 1930s understanding was too little to use extracellular recordings to predict the transmembrane action potential, and in any case the shapes of extracellularly recorded action potentials highly depend on the details of the extracellular recording method used. Being able to directly measure transmembrane potential during an action potential was thus highly desirable. This goal was made possible by the discovery of the giant (as large as 1 mm diameter) squid axon (Young 1936). This axon is large enough to insert a glass micro-electrode into and thus directly measure the transmembrane action potential. Cole and Curtis (1939) and Hodgkin and Huxley (1939) published such data in the same year; the transmembrane potential was inside negative and the action potential overshooting (Fig. 2.8A from Hodgkin and Huxley). Although this work described the time course of the transmembrane action potential, it did not show the mechanism underlying these transmembrane potential changes. An important first step toward this goal was the demonstration that the giant axon action potential was associated with a large decrease in membrane resistance (Cole and Curtis 1939), thus proving Bernstein’s hypothesis that the action potential was associated with an increase in membrane permeability (although, interestingly, Cole and Curtis did not reference Bernstein’s work). With modern knowledge it seems clear that the next step would be to measure membrane resistance (and thus transmembrane current) under conditions in which transmembrane potential could be independently controlled. However, it is important to recognize that at this time the basis of transmembrane current was completely unknown, and in particular whether membrane resistance depended on transmembrane current flow, transmembrane potential, or even metabolic or other cellular processes. To investigate this issue, Cole and Marmont designed a feedback control apparatus in which the current flowing across the giant axon membrane could be controlled regardless of the transmembrane potential (design of the apparatus published under

25

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B +40

128 mV

1 0

64 mV

1 ms

mA/cm2

38 mV 27.5 mV

0 100 μs

–1

–70

C

D

120 100 80

20 mV

60

5 ms

40 Millivolts

26

E

20 0 –20

SubCh –40 500 ms

ACh

–60 –80

Carbachol

10 pA 0

10

20

30

m sec

Figure 2.8 A history of intracellular recording. (A) Intracellular recording obtained by inserting an electrode into a squid giant axon. (B) One of first voltage clamp recordings (see text). Numbers next to traces indicate how much the transmembrane potential was depolarized relative to rest. (C) First muscle action potential recorded with sharp electrode inserted through cell membrane; rest potential –80 mV. (D) First neuron action potential recorded with sharp electrode inserted through cell membrane. (E) First single channel recordings. A from Hodgkin and Huxley (1939); C, Nastuk and Hodgkin (1950); D, Woodbury (1952); E, Neher and Sakmann (1976). All with permission. B modified from Cole (1949).

Marmont’s name alone (1949)). Cole recognized that this apparatus could also be used to measure current flow under conditions of controlled transmembrane potential, but was unable to convince Marmont of the utility of this approach, and was allowed to perform only four such “voltage-clamp” experiments (Cole 1982). Around this time Hodgkin visited the Cole lab and upon his return home built his own feedback apparatus, which he then used as a voltage clamp to great success to explain how voltage-dependent conductances generate the action potential (Hodgkin and Huxley 1952a–d). Cole (1949) and Hodgkin et al. (1949) first published voltage-clamp data in the same year (in Hodgkin et al., with an incorrect hypothesis of the basis of the current

Electrophysiological Recording Techniques

flows); in light of the apparatus being invented in Cole’s laboratory, we present here Cole’s data (Fig. 2.8B). Being able to study only axons large enough to insert a glass microelectrode into was clearly limiting. Attempts were therefore soon made to insert small-tip, sharp electrodes through the cell membrane. The first such recording was from muscle cells, measuring a transmembrane potential of –80 mV (Ling and Gerard 1949). The first muscle cell (Fig. 2.8C) (Nastuk and Hodgkin 1950) and neuron (Fig. 2.8D) (Woodbury 1952) sharp microelectrode recordings of the action potential were obtained shortly thereafter. Up to 1970, although the electrical characteristics of very many membrane currents were extensively described, the molecular basis of the currents was not known, with arguments being made that DNA, cholesterol, or changes in lipid configuration were responsible for ion transfer across the membrane (Hille 2003). This difficulty was first resolved by using snake venom toxins to show that the nicotinic ACh receptor was a membrane-embedded protein (Changeux et al. 1970a,b; Cohen et al. 1972; Duguid and Raftery 1973; history in Changeux 2012). Work using tetrodotoxin as the binding element for purification soon after identified the fast voltage-dependent sodium channel as also a membrane-embedded protein (Agnew et al. 1978; Benzer and Raftery 1973; Henderson and Wank 1972; history in Agnew 1984). This work, coupled with the by then well-known fact that the action potential resulted from current flow across the membrane, suggested that current through single or a small number of these protein channels could be recorded by pressing a glass electrode against, but not through, the cell membrane, thus electrically isolating any channels under the electrode tip lumen from the surrounding bath. The small amplitude of single channel currents hampered this work, but it was finally achieved for muscle in 1976 (Fig. 2.8E) (Neher and Sakmann 1976). The use of extremely smooth and clean electrodes to achieve giga-ohm seals between the electrode and the cell membrane (Hamill et al. 1981; Sigworth and Neher 1980) allowed for very low noise recordings and cell-attached and detached patch clamping.

2.6 Concepts Important to Understanding Neuron Recording Techniques 2.6.1 Membrane Properties

The transmembrane potential is primarily due to transmembrane ion concentration gradients and membrane channels that show selective ion permeability. Consider a concentration gradient of only one ion, X + . If the membrane has channels permeable to X + , X ions will move across the membrane from the high to low concentration side. In addition to slightly lowering the X + concentration gradient, this movement changes the transmembrane potential by making the low concentration side positive compared to the high concentration side. Since X + is positively charged, this transmembrane potential opposes further X + movement across the membrane. This process continues until the transmembrane potential becomes great enough that no further net X + movement across the membrane occurs, the equilibrium potential (Ex ). For single ions the equilibrium potential is given by the Nernst equation [X]o RT , ln Ex = zX F [X]i

27

28

Neurobiology of Motor Control: Fundamental Concepts and New Directions

where R is the ideal gas constant, T is temperature in Kelvin, zx is ion valence, F is Faraday’s constant, and o and i refer to outside and inside the cell. The linear dependence of this equation on temperature was one check that Bernstein performed to test whether Nernst’s work applied to muscle transmembrane potentials, and thus whether the transmembrane potential was due to concentration differences across a semi-permeable membrane. Such a single ion situation cannot result in overshooting transmembrane potential excursions. That is, if Ex equals –80 mV and then pores open in the membrane that allow all ions to pass and the X concentration gradient thus collapses, transmembrane potential goes to zero, not positive. Bernstein, however, observed overshooting action potentials, and Overton showed that both K+ and Na+ (or Li+ ) were required for the action potential. These data prove that more than one ion concentration gradient exists. A consequence is that, unless the two ions have the same Ex , when the membrane is permeable to more than one ion there is no transmembrane potential at which both ions are at their equilibrium potential. When the transmembrane potential differs from an ion’s equilibrium potential, current due to that ion flows across ion channels permeable to it according to ix = gx (Vm − Ex ), where i is current, g is conductance (1/resistance), Vm is transmembrane potential, and negative i represent positive current flowing into the cell. If the cell’s transmembrane potential is such that the sum of the currents generated by all ions crossing the membrane does not equal zero, the transmembrane potential de- or hyperpolarizes because of this ongoing transfer of charge until it reaches a Vm at which the sum of the currents generated by all ion flows does equal zero. For a case in which the membrane is permeable to only two ions, this means that i1 = −i2 . Solving i1 = g1 (Vm − E1 ) and g E +g E i2 = g2 (Vm − E2 ) simultaneously for Vm gives Vm = 1 g1 +g2 2 . 1 2 Real neurons and muscles are, at rest, primarily permeable to potassium, sodium, and chloride, and applying exactly the same procedure shows that their resting transmemg E +gNa ENa +gCl ECl brane potential is given by Vm = K Kg +g ; note that couching this equation in K Na +gCl terms of gx ’s and Ex ’s means that the valencies of the ions are automatically accounted for in the Vm calculation. A more electrochemically precise formulation is in terms of channel permeability, which accounts for the effects of ion availability (if no ions are available to flow through a channel, it is immaterial if it opens or not); the equation in its permeability form can be seen in any neurobiology textbook. For our purposes the conductance-based form of the Goldman–Hodgkin–Katz (GHK) equation is sufficient. The ion concentration gradients themselves are established by membrane pump enzymes, the most important of which is the Na+ /K+ ATPase. These pumps do not always work in an electro-neutral fashion (e.g., the Na+ /K+ ATPase moves 3 Na+ out of the cell for every 2 K+ it moves in). This effect contributes only a few mV of the transmembrane potential, but in some cases can nonetheless play a substantial role in shaping neuron activity (Del Negro and Hays 2008; Zhang and Sillar 2012; see Chapters 8 and 13). In neurons and muscles the values of the gx and Ex at rest are such that Vm is –60 to –80 mV inside negative. It is this difference that gives rise to the injury current measured by du Bois-Reymond and Bernstein, in that one of their electrodes was located on the outside of the nerve or muscle and the other at the nerve’s or muscle’s cut end, and thus the two electrodes had different potentials (outside, 0 mV; inside of cut end, –60 to –80 mV). This equation also provides a partial explanation of the action potential and

Electrophysiological Recording Techniques

synaptic potentials, both of which involve gx changes. The GHK equation shows that Vm is a weighted function of the various gx , and thus as one gx increases, Vm drives closer to that ion’s equilibrium potential. This is precisely what occurs during the action potential, in which gNa becomes very large. The rates of change, and the steady-state values, of the gx ’s are generally functions of Vm and sometimes of intracellular ion concentrations (see Chapter 5); a major accomplishment of the voltage clamp was quantitatively describing these functions (Fig. 2.5). The reason the GHK equation only partially explains neuron and muscle activity is that it does not include membrane capacitance. Implicit in the above descriptions of charge separation is the fact that the membrane is a capacitor as well as a resistor. It therefore does not respond to step changes in membrane current with step changes in Vm . Biological membranes instead respond (even with fixed gx ) with a slow rise to a steady state, similar to the response of a resistor-capacitor (RC) circuit, V = VF (1 − e−t∕𝜏 ), where capacitor and resistor V ’s are originally zero, VF is their final V ’s, and 𝜏 = RC is the circuit’s time constant. This again is an oversimplification, as the extended geometry of most neurons makes their dynamic responses actually the sum of multiple exponential charging curves, again a topic not of concern here and one covered in detail in any good electrophysiology textbook. Neuron slow RC responses are critical to discontinuous current and voltage clamp. Note also that the presence of cell capacitance means that, when transmembrane potential changes are made in voltage clamp, the first, large currents injected are to charge the cell capacitor, with changes in currents through membrane channels only occurring subsequently after transmembrane potential (Vm ) has been changed ∑ by the charging of the dV cell capacitance. The neuron charging curve, Cm dtm = iinj − im , where Cm is membrane capacitance, iinj is current injected through intracellular electrodes, and im are the currents through the membrane channels, is also the basis of much neuron modeling (Chapter 5). 2.6.2 Intracellular Recording

Simplified versions of the electronic circuits used to achieve the various recording types described earlier are shown in Fig. 2.9. Simple transmembrane potential recording is straightforward, with the transmembrane potential being fed into a (typically unity gain) amplifier and then sent to a recording device (Fig. 2.9A). The only complication is that recording electrode resistance and capacitance low pass filters neuron transmembrane potential responses; this difficulty is overcome by electronic circuitry that “cancels” the electrode effects. Two-electrode voltage clamp (Fig. 2.9B) is again straightforward. Cell transmembrane potential is measured as in Fig. 2.9A (for simplicity, electrode resistances and capacitances are not shown). This voltage is fed into the inverting input of an amplifier whose output equals a gain factor times the difference in the cell transmembrane potential and the user-set command potential. This output is used to control current injection into the cell through a second electrode. This negative feedback drives the cell transmembrane potential to the command potential; the current necessary to do so is the output and equals the current across the cell membrane at the command potential. Patch clamp (Fig. 2.9C) instead clamps the interior of the pipette to a user-set command potential. It does so by connecting the electrode (in this case a cell-attached electrode, but the circuit is the same for all patch clamp, including whole cell) to the

29

30

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

+

B

Vm

+

im

+

command voltage

Vm

C ip

Rf

iRf i-

+

Vout = –ip·Rf + Vcom

Vcom

Figure 2.9 Electronic circuits for measuring and controlling transmembrane potential (A), two-electrode voltage clamp (B), and patch clamp (C). See text for explanation.

inverting input of an amplifier and the non-inverting input to the command potential. The output of the amplifier is then fed back to the amplifier’s inverting input. How this arrangement results in clamping of the inverting input voltage is not intuitively obvious, and we therefore provide the mathematical explanation (Ogden and Stanfield 1994). From current conservation, the currents going through the node marked with the open circle is ip = iRf + i− , where ip is current through the patch, iRf is the current through the feedback resistor, and i− is the current through the inverting input of the amplifier. The voltage across the feedback resistor is the voltage at the inverting input of the amplifier minus the voltage of the amplifier output (V− − Vout ), and therefore ip − i− =

V− − Vout . Rf

Electrophysiological Recording Techniques

The feedback is such that the voltage drop across Rf minimizes the difference between the two inputs of the amplifier, and for a perfect operational amplifier they are equal, V− = V+ = Vcom (the user set command potential), and i− is zero. Substituting gives ip =

Vcom − Vout . Rf

Rearranging gives Vout = −ip Rf + Vcom . Thus, this circuit clamps the interior of the pipette to the command potential, and the output of the amplifier, when the command potential is subtracted, measures the current entering the electrode from the patch of membrane. Note that the electrode resistance and capacitance is “before” the node, and thus the circuit is clamping the voltage after the voltage drop across the electrode and with the time limitations imposed by the electrode RC properties (see Continuous Single Electrode Voltage Clamp (cSEVC) or Whole-Cell Patch Clamp part of Section 2.3.3.2). With respect to the electrodes used, two important technical issues are present. The first is that sharp electrodes, which basically rip their way into the cell membrane, introduce a leak. The cell membrane slowly seals around the electrode, thus reducing this leak, and in many (but not all (Cymbalyuk et al. 2002)) cases the decrease in cell resistance is small enough to introduce only small changes in cell activity. For patch electrodes, which seal onto the cell membrane, this is not a concern. The second is the danger of electrode contents diffusing into the cell and changing cytoplasm ion and metabolite concentrations. This danger has been long recognized by the patch clamp community and it is standard in it to use electrode fill solutions that ionically match cytoplasmic composition, and also sometimes contain other molecules (e.g., ATP) in a further attempt to match cytoplasm contents. However, as noted above, there are limits to how well this can be achieved, and exacting work is therefore presumably best performed with perforated patch technology. Sharp electrode work has from its inception typically used electrode fill solutions with much higher ionic strengths than neuron and muscle cytoplasm. This was done (Nastuk and Hodgkin 1950) because (1) of a belief that high ionic strength fills would give lower electrode resistances and thus electrode noise and (2) it obviates the liquid junctional potential problem. With respect to the first reason, electrode resistance decreases much less than linearly with fill ionic strength (a 10-fold decrease in ionic strength increases electrode resistance only 4.4 fold) (Brown and Flaming 1986). Moreover, in a well-shielded set-up, electrode noise appears to be primarily due to net flow of ions across the tip of the electrode (Brown and Flaming 1986), which will be reduced by using fill solutions whose ionic composition matches cytoplasmic values. Brown and Flaming (1986) report that electrode noise indeed decreases as fill ionic strength is lowered, and both they and Hooper et al. (2015) show that electrodes with low ionic strength fills give high-quality recordings with electrode noise at least as low as that obtained with electrodes filled with high ionic strength solutions. With respect to the second reason, liquid junctional potentials are a potential that exists at the interface of two solutions composed of ions with different mobilities. The liquid junctional potential is part of the reason for the potential seen when an electrode is first placed in the preparation bath, which is then “zeroed out” as a first step

31

32

Neurobiology of Motor Control: Fundamental Concepts and New Directions

before cell penetration. The problem is that liquid junctional potentials depend in part on the concentration gradient between the two liquids. Hence, zeroing against saline does not zero against the different ion concentrations present in cytoplasm. However, the difference between the fill:saline and fill:cytoplasm gradients is much smaller the greater the fill solution ionic concentration. Using high ionic strength electrode fill solutions thus reduces the error in Vm measurement caused by an uncorrected liquid junctional potential between electrode fill and cell cytoplasm. In patch clamp work Vm measurements are corrected by explicit calculation of what the liquid junctional potential should be. The sharp electrode community has instead assumed that the difference between the fill:saline and fill:cytoplasm liquid junctional potentials will be negligible. This assumption is not completely correct, as the uncorrected liquid junctional potential with standard electrode fills can be as great as 10 mV (Hooper et al. 2015). Exacting work, even with high-ionic strength sharp electrode fills, should thus either measure or calculate liquid junctional potential correction. With respect to the more substantial concern that high ionic strength fills may leak into the cell and alter cell properties, Hooper et al. (2015) showed that this indeed occurs, changing conductance levels many-fold for some currents and activation time constant for at least one current two-fold, changes large enough to alter neuron spiking activity. The extent to which these changes occur depend on the electrode fill used and how much current is injected into the cell over the recording session. It is thus difficult to assess how much of the sharp-electrode literature is affected by these data. Nonetheless, they do make it clear that in general sharp electrode recording, just as patch electrode recording, should be made as much as possible with cytoplasm-matched electrode fill solutions. Note that this does not pose a difficulty for continued use of silver–silver chloride electrodes, as cytoplasm always contains sufficient chloride to interact with these electrodes. 2.6.3 Extracellular Recording

Reading the nerve extracellular recording literature can be confusing, with its mono-, bi-, and triphasic action potentials, and the multiple types of recording techniques (cut end with and without killing the nerve between the two electrodes, mono and bipolar electrodes, cuffs, suction electrodes) used, each of which affects the observed shape of the action potential. The goal of this section is to provide sufficient background that interested readers can access this literature with reduced difficulty. For a more detailed but still generally accessible review of extracellular recording theory and techniques see Stys and Kocsis (1995); for mathematically rigorous presentations see Plonsey and Barr (1988) and Malmivuo and Plonsey (1995). We cover here primarily recordings from axons and nerves, not the much more complicated situation encountered with extracellular recordings from tissues containing large numbers of intermixed neuron cell bodies and processes (e.g., cortex) (see Chapter 3). We also cover only single, not compound action potentials, since, as summed events, CAPs contain no information about the shapes of the single action potentials that comprise them, and no general statements as to what waveform CAPs should have, or what effect different recording techniques or volume conductivities should have on this waveform, can be made. Much of the work described below was motivated not only to understand the basis of extracellular waveform, but also by the hope that better understanding of

Electrophysiological Recording Techniques

extracellular recording could give the level of insight into neuronal physiology that, for instance, intracellular recording has provided. Unfortunately, this is a hope largely unfulfilled. The difficulty is that extracellular action potential shape highly depends on the properties of the volume surrounding the nerve being recorded from. Each kind of electrode introduces different changes in this volume, and even with a single kind of electrode, how the nerve interacts with it will vary from experiment to experiment. Repeated recordings, even from the same nerve with the same type of electrode, therefore typically give differently shaped action potentials in different individuals. In practice, shape discrimination in nerve extracellular recordings is consequently typically limited to distinguishing among units on the basis of repeatable differences in unit amplitude (due to some axons always being larger than others across animals, e.g., Fig. 2.3) or similar shape-independent measures of identity. Importantly, these concerns do not hinder the ability to distinguish between action potentials arising from different neurons when recording extracellularly from large neuron assemblages (e.g., brain, large ganglia). Moreover, this experimental situation best reproduces the case, extracellular recording in a continuous, uniform conductor, that is theoretically best understood. For workers in field recordings, the section below (2.6.3.2) dealing with this case may thus have practical value, and for all readers should be useful in understanding Chapter 3, which deals with recording from neuron ensembles. An important issue when recording from such tissues that does not arise in nerve recordings is that a spiking neuron is a dipole in which the somata where the spikes arise is depolarized in comparison to the dendrites. This dipole adds to the fields being produced by propagating action potentials, further complicating interpretation of extracellularly recorded field potentials (Fatt 1957; Graziane and Dong 2016). 2.6.3.1 Intracellular Action Potential Shape

The squid giant axon intracellular (transmembrane) action potential is biphasic, first depolarizing above rest and then hyperpolarizing below rest (undershooting) (Fig. 2.8A) (Hodgkin and Huxley 1939). This undershoot arises because the axon’s resting transmembrane potential is depolarized relative to EK , and potassium current increases late in the action potential. This biphasic axon action potential appears to be unusual, with axonal action potentials (from rest) in crayfish (Watanabe and Grundfest 1961), lobster (Ballo and Bucher 2009; Nusbaum et al. 1992), Xenopus (Chiu et al. 1979; Frankenhaeuser and Huxley 1964), and rabbit (Chiu et al. 1979) all lacking a late undershoot and hence being monophasic. Based on these data, all calculations of extracellular action potential shape of which we are aware, and the presentation below, assume a monophasic transmembrane action potential (see, e.g., Fig. 2.11A). In cases where this assumption is not true, the logic of the following does not change, with its application simply resulting in action potentials of greater complexity (i.e., more phases) than in the monophasic transmembrane action potential case. 2.6.3.2 Axon Embedded in Uniform, Infinite Volume Conductor

Extracellular recordings of nerves and axons are almost always performed under conditions (suction electrodes, pin electrodes, etc.) in which the nerve or axon are not embedded in a uniform, infinite volume conductor. However, it is for this condition that the theory is best developed, and, as the simplest case, it is necessary to understand this condition before considering more complex ones. Even understanding this simplest of cases is daunting, and multiple explanations of how the potential field around an axon

33

34

Neurobiology of Motor Control: Fundamental Concepts and New Directions

or nerve in a uniform, infinite volume arises are available, in each of which the conflict between being intuitively understandable and correct is resolved to varying degrees. Because reading many different approaches to the same problem can be useful, and because in different literatures different explanations are more or less commonly used, we summarize them all here. Solid Geometry Approach A qualitative, intuitive understanding of many aspects of

extracellular action potential shape in continuous, infinite volume conductor can be attained by simple geometric arguments (Splinter 2010) that can be rigorously formalized (Plonsey 1965, 1977). Consider the axon segment and point P in Fig. 2.10A. The resting transmembrane potential (pre-action potential) portion is that with negative inside oriented dipoles. The “action potential” propagates right and, in this case, consists of a single depolarization front with no repolarization—all the membrane behind the front remains inside positive. The angle Ω subtends the region where the depolarizing front occurs. The situation is actually three dimensional, and thus Ω is a solid—three-dimensional—angle, not a simple one. However, the logic is the same in the simplified two-dimensional case, and we therefore discuss it in two dimensions. Before considering the situation with an action potential, imagine that the depolarized membrane segment was not there. In this case, from the perspective of P, the “top” and “bottom” edges of the axon in Ω (and in any other angle) would contain oppositely-oriented dipoles (positive poles of the dipoles would be closest to P in the “top” edge, but furthest from P in the “bottom” edge). Because of the spread of the angle going from the top to bottom edge, in the two-dimensional case shown Fig. 2.10A there would be slightly more subtended dipoles in the bottom edge than the top. However, if one considers the situation in three dimensions and assumes that membrane thickness is extremely small, the contributions from the “front” and “back” sides cancel and thus produce no potential at P (Splinter 2010). As such, a quiescent axon produces no potential in the extracellular volume. When the action potential is present the situation completely changes. When the depolarizing front is very distant from P, because potential declines with distance, it makes an unmeasurable change at P. As the front approaches it induces an increasing potential at P because, at the front, the balance between the dipoles on the near and far side of the axon is destroyed. Specifically, for the position shown in Fig. 2.10A, in Ω the

Figure 2.10 (A) A depolarization front (propagating right, all membrane behind it remains inside positive) induces a diphasic potential that is first positive and grows as the front approaches point P, then decreases to zero directly under P, then becomes negative as the front propagates further right of P, and finally declines again to zero as the front becomes increasingly distant from P. (B) A rectangular action potential (“Depolarized membrane”) induces two potentials at P, one due to the depolarization front (solid angle ΩD ) and another due to the repolarization front (solid angle ΩR ). Note that in ΩD the positive poles of the dipoles on the two sides of the axon are closest to P, whereas in ΩR the negative poles of the dipoles on the two sides are closest to P. (C) Addition of the potentials from the two fronts gives rise to a summed triphasic potential at point P. 1st , 2nd traces: Potentials induced at P by each front as action potential propagates right. Because of the changing solid angles and distances between P and the action potential fronts as the action potential propagates, these potentials are bipolar and out of phase; because of the opposite orientation of each front’s dipoles, the signs of the potentials are reversed. 3rd trace: Summed potential at P is triphasic because of offset and reversed polarity of the biphasic front potentials. Figure new, but inspired by Splinter (2010).

Electrophysiological Recording Techniques

P

Ω

A + + + + + + + + + + + + + + + +

+ + + + + + + + + + + + + + + + + + + + + + + + + + Resting membrane Depolarized membrane P

ΩR

ΩD

B + + + + + + + +

+ + + + + + + + + + + + + + + + + + +

+ + + + + + + + Repolarized membrane

Depolarized membrane

C Depolarization front at P (biphasic) + 0 – +

* ** Repolarization front at P (biphasic)

0 – + Summed recording at P (triphasic) 0 –

+ + + + + + + Resting membrane

35

36

Neurobiology of Motor Control: Fundamental Concepts and New Directions

positive sides of the dipoles on both the near and the far sides of the axon are nearest to P. By convention, they therefore induce a positive potential at P. As the front moves right toward P, two things occur. The first is that the distance to P continually decreases, increasing the potential at P. The second is that Ω continually decreases, decreasing the potential at P, reaching zero when the front is directly under P. The combined effect of these two processes is that the potential at P first grows, reaches a maximum (single asterisk in first trace in Fig. 2.10C), and then decreases, reaching zero when the front is directly under P (double asterisk in first trace in Fig. 2.10C). When the front is to the right of P, the dipole imbalance becomes opposite, with the negative poles of the dipoles on both the near and far sides of the axon being nearest to P. Potential at P therefore turns negative, first growing and then decreasing as the front becomes ever more distant (right half of first trace in Fig. 2.10C). An action potential does not consist of a permanent depolarization of the inside of the axon, but instead of a discrete length of depolarization (Fig. 2.10B). In this case both the depolarization and repolarization fronts induce potentials at P. Because of the opposite signs of the two fronts, the shapes of the potentials they induce at P as they move under it are opposite: depolarization–hyperpolarization for the depolarizing front (1st trace, Fig. 2.10C) but hyperpolarization–depolarization for the repolarizing front (2nd trace, Fig. 2.10C). The repolarization front’s potential wave is also shifted back in time because of the action potential width (compare 1st and 2nd traces, Fig. 2.10C). The potential at P (3rd trace, Fig. 2.10C) is the sum of the potentials produced by the two fronts. This is a triphasic positive–negative–positive wave, exactly the sequence predicted by quantitative approaches (see below). It is worth stressing how profoundly the changes in Ω and distance that occur as the action potential propagates along the axon alter transmembrane action potential shape, transforming a monophasic, always positive, discontinuously-changing square wave into a smoothly-varying triphasic wave with both positive and negative components. Laplacian Calculation of External Potential Field from Action Potential Transmembrane Potential

The geometric approach outlined above is not quantitative. Once the transmembrane potential profile of an action potential (as, for example, in Fig. 2.11A) has been experimentally described, electrostatic theory, alternatively, should allow quantitative prediction of the potential field it induces. Axons are cylinders, and the equation for solving electrostatic potential fields for an infinite cylinder is called Laplace’s equation

Figure 2.11 (A) Outer membrane surface potential calculated directly from (a real) action potential transmembrane potential. (B) Potential field induced by transmembrane potential in A. General form is similar to that in A, but amplitude drops with distance from axon. Current arising at axial distances distal to the two positive peaks of the extracellular action potential (most distal dashed lines) would, in flowing from high to low potential regions, flow away (skew) from the center of the action potential before ultimately curving back and re-entering axon in central region. (C) Comparison of transmembrane current per unit length calculated from data in A and the 2nd derivative of the transmembrane potential. Note close similarity. Dashed lines help identify commonalities in three plots: positions of positive peaks of extracellular action potential match positions of maximum outward current, position of negative peak of extracellular action potential matches position of maximum inward current, positions of zero crossings of extracellular action potential match positions of zero transmembrane current. Modified with permission from Clark and Plonsey (1968).

SURFACE POTENTIAL DISTRIBUTION - μV

60

40

80

20

40

0

.2

.4

.6

.8

1.0 1.2 AXIAL DISTANCE (cm)

z

–20 Outer membrane surface potential –40

–60

B

660

–0+

Radial distance (μm)

+0–

560

C 10.0

10 5

5

3

3

10 5

2

5

360

30

260

20

160

30

8

40

.3

.4

50

.5

.6

2

.7

.8

.9

1.0 1.1 Axial distance (cm)

2nd derivative of transmembrane potential

10 Transmembrane current per unit length

5.0

0

10 20

460

60 0.2

Transmembrane current per unit length (im) (µA/cm)

120

Transmembrane potential

.2

.4

.6

.8

5

1.0 1.2 Axial distance (cm)

0

–5.0

–5

–10.0

–10

2nd derivative of transmembrane potential (V/cm2)

A

TRANSMEMBRANE POTENTIAL DISTRIBUTION - mv

Electrophysiological Recording Techniques

37

38

Neurobiology of Motor Control: Fundamental Concepts and New Directions

in cylindrical coordinates. Applying this equation to predict the extracellular potential field is formidable, involving taking the Fourier transform of the transmembrane action potential and the use of Bessel functions (solutions to the Laplace equation) (Clark and Plonsey 1966, 1968; Geselowitz 1966; Plonsey 1964, 1977). We do not attempt to explain this procedure here. For a given transmembrane action potential profile, this equation can be numerically evaluated (Fig. 2.11A) (Clark and Plonsey 1968). In agreement with the simple geometric approach above, a monophasic transmembrane action potential profile gives rise, at the surface of the axon, to a triphasic (positive–negative–positive) extracellular potential field. As such, when the action potential moves under an electrode located at the surface of the axon, this triphasic potential variation would be seen vs. time. If the electrode is placed further from the axon, the amplitudes of the measured potentials decrease, but the triphasic potential profile is maintained (Fig. 2.11B). Calculation of the transmembrane current flow during the action potential (Fig. 2.11C) shows that current flows out of the axon during the positive portions of the extracellular action potential. In the negative portion of the extracellular action potential, current flows into the axon. Using Current Flow to Derive a Particularly Simple, Approximate Predictive Relationship Another

approach is to make use of current flow in the axon interior and across its membrane to derive a simple relationship between the transmembrane potential (Vm ) and the extracellular potential field (𝜙e ). Detailed explanations are again formidable, involving vector calculus (Malmivuo and Plonsey 1995; Plonsey 1977; Plonsey and Barr 1988), but in this case a relatively simple derivation can also be given (Plonsey and Barr (1988); we thank C. Pouzat for this derivation). Understanding this procedure requires, for many readers, a change of point of view, as people often think of potential differences being the cause of current flows. However, just as a charge creates an electric potential in the volume surrounding it, so does a current source embedded in a conductor (Malmivuo and Plonsey 1995). Indeed, the fundamental equation underlying this approach, 𝜙e =

1 I0 , 4𝜋𝜎e r

(2.1)

where 𝜙e is the electrical potential, 𝜎e is the conductivity of the extracellular medium, I0 is the current source intensity, and r is the distance from the source, is exactly the same equation as that for the potential created by a single charge if 𝜎e is replaced with the permittivity 𝜀 and I0 is replaced with charge magnitude (Malmivuo and Plonsey 1995). Axons are not point sources of current, but instead cylinders across whose surface current flow varies with position. Calculating the electric field produced by such an object requires integrating current density, im (x), position by position over the object’s entire extent: im (x) 1 𝜙e = dx. (2.2) 4𝜋𝜎e ∫ r(x) Note that im is expressed as only a function of x, not time, even though as the action potential propagates it moves to different positions on the axon. However, the potential at a fixed point P generated by a moving action potential is the same as the potential observed if the point P is moved, in the opposite direction, at the action potential propagation speed, through the potential induced by a stationary action potential

Electrophysiological Recording Techniques

(Offner 1954). As such, we need only to solve for the potential produced by a stationary action potential, that is, as a function of x. The next step is to calculate im as a function of the intracellular potential along the axon. If one considers a small segment of axon with a radius a, an intracellular conductivity 𝜎i , and a length Δx, if the intracellular potential at one end of the segment, 𝜙i (x), does not equal the intracellular potential at the other end, 𝜙i (x + Δx), then from Ohm’s law the following axial current, Ii (x) flows in the interior of the segment: 𝜙i (x + Δx) − 𝜙i (x) , (2.3) Δx where current is positive if it flows in the direction of increasing x. In the limit as Δx goes to zero, this equation becomes Ii (x) = −𝜋a2 𝜎i

𝜕𝜙i (x) . (2.4) 𝜕x If equal current does not flow through both ends of the segment, charge conservation requires that the difference flows through the axon membrane. Keeping with convention, when this transmembrane current (im ) is outward, its sign is positive. In equation form, Ii (x) = −𝜋a2 𝜎i

Ii (x + Δx) − Ii (x) = −Δxim (x),

(2.5)

or, again in the limit as Δx goes to zero, 𝜕Ii (x) = −im (x). 𝜕x Differentiating equation 2.4 and substituting equation 2.6 into the result gives 𝜕 2 𝜙i (x) . 𝜕x2 Substituting equation 2.7 into equation 2.2 gives im = 𝜋a2 𝜎i

𝜙e =

2 a2 𝜎i 1 𝜕 𝜙i (x) dx. 4𝜎e ∫ r(x) 𝜕x2

(2.6)

(2.7)

(2.8)

Equation 2.8 gives extracellular potential as a function of intracellular potential. Experimental work, however, does not measure intracellular potential, but instead transmembrane potential, V m = 𝜙 e − 𝜙i .

(2.9)

This difficulty can be resolved by noting that the infinite (ideal case) and very large (real cases) extracellular volume is much larger than the very small intracellular volume. Extracellular resistance is therefore much smaller than intracellular resistance. 𝜙e is therefore much smaller than 𝜙i , and thus to a good approximation Vm = −𝜙i (Clark and Plonsey 1968; Plonsey and Barr 1988). Using this approximation allows expressing 𝜙e as a function of Vm : 𝜙e = −

2 a2 𝜎i 1 𝜕 Vm (x) dx. 4𝜎e ∫ r(x) 𝜕x2

(2.10)

We have not found a comparison of the 𝜙e ’s obtained by this and the Laplacian approach when applied to identical Vm ’s (that is, identical action potentials). This

39

40

Neurobiology of Motor Control: Fundamental Concepts and New Directions

comparison has, however, been made for the im ’s (equation 2.7 multiplied by –1 and with 𝜙i replaced with Vm ) (Fig. 2.11C) (Clark and Plonsey 1968). The im ’s calculated from the Laplacian method (“transmembrane current per unit length”) and that in equation 2.7 (“2nd derivative of transmembrane potential”) are very similar. Note also that, when applied to real data, equation 2.10 gives, as expected, a triphasic waveform (Plonsey and Barr 1988). Explanation Based on Directions of Longitudinal Current Flow The final explanation of the

extracellular action potential in a uniform, infinite volume uses geometry and extracellular current flow (Fig. 2.12A) (Offner 1954). Early work (Lorente de Nó 1947) showed that the transmembrane current flow associated with the action potential is a central region of current inflow flanked by two regions of current outflow (Fig. 2.11C). Charge conservation requires that the sums of these flows be zero, and thus that current flows from the flanking regions to the center. With respect to how current flows in the volume surrounding the axon, current always flows from higher to lower potential, in a direction orthogonal to the isopotential lines. Considering the potential field in Fig. 2.11B shows that currents arising distal to the two positive peaks of the extracellular action potential therefore slant away (skew) from the center of the action potential (Clark and Plonsey 1968). For instance, transmembrane current leaving the axon at an axial distance of 9.5 flows up and to the right (moving from higher to lower potential and orthogonal to the iso-potential lines) before eventually curving back (region not shown) and flowing left and down back into the axon. Figure 2.12A shows a schematic of these current flows. This figure was made before the Laplacian work in Fig. 2.11, and thus shows the current skewing away from the center of the action potential even at positions proximal to the positive peaks of the extracellular action potential. However, this does not affect the logic of the following arguments. The filled arrows show the direction of total current flow; the spacing of the lines indicates current flow density (see Fig. 2.11C for current flow density at the surface of the axon). Because of the skewing of these lines, the direction of longitudinal current flow varies at different distances from the center of the action potential. For instance, if an electrode is moved along the line marked by the open arrows, at position 7 the longitudinal current flows right, at position 6 longitudinal current flow is zero, and at position 5 longitudinal current flows left. The voltage between the recording electrode and a reference electrode is the integral of the current density taken on any line connecting the two electrodes. For convenience, if the reference electrode is located the same radial distance from the axon, this line can be the line defined by the open arrows. In this case the integral relationship means that the slope of the voltage change between the two electrodes (dV ∕dt) as the recording electrode is moved (mimicking the movement of an action potential under a fixed electrode) will depend on the longitudinal current, with the sign of the slope depending on the direction of this current relative to the reference. This analysis has the usual annoying problem of defining negative and positive directions of current flow, dV ∕dt, and action potential propagation. Choosing the following definitions maintains the normal convention that the extracellular action potential is positive–negative–positive. For cases in which both the recording and reference electrode are distal to the action potential, for action potentials propagating toward the electrodes (which they must to be recorded), dV ∕dt is positive if the longitudinal

Electrophysiological Recording Techniques

A

1

B

3

2

1

2

5

4

3

4

5

6

7

6

Figure 2.12 (A) Current flow in a volume conductor during an action potential (B) Illustrative example of effect of changing volume conductor properties on lines of current flow. Filled arrowheads show direction of total current; open arrowheads direction of longitudinal current; open circles are where longitudinal current amplitude is zero. A and B modified with permission from Offner (1954). In both panels the drawings have been altered to reflect modern conventions: the potential traces have been flipped vertically to give positive–negative–positive triphasic, and negative monophasic, action potentials; the directions of current flow have been changed to show how positive current would flow, as opposed to the direction of electron flow shown in the originals.

current at the recording electrode points toward the reference electrode. For cases in which the action potential is between the two electrodes (in which case, to be recorded, it must propagate toward the recording electrode), dV ∕dt is negative if the longitudinal current at the recording electrode points toward the reference electrode. See Stys and Kocsis (1995) for an equivalent circuit explanation of these sign issues, although this treatment deals only with voltage, not voltage slope. With these definitions, the extracellular action potential can be explained as follows. Take the case that both electrodes are to the left of the action potential (reference much

41

42

Neurobiology of Motor Control: Fundamental Concepts and New Directions

further left than recording) and the action potential propagates left (which we mimic here by moving the recording electrode right). When the recording electrode is far to the left of the action potential, there is essentially no current density between the recording and reference electrodes, and recorded voltage is thus zero. As the recording electrode is moved right, it enters the region between points 1 and 2 (Fig. 2.12A). In this region the longitudinal current always points toward the reference, and thus dV ∕dt is positive. The amplitude of dV ∕dt depends on the interplay of the current amplitude at each point (increasing from pt. 1 to pt. 2, see Fig. 2.11A and C) and the slopes of the total current flow lines, which become increasingly vertical, and thus have continually decreasing longitudinal components, as the electrode moves from pt. 1 to pt. 2. This interplay explains the changing slope of the voltage trace, first increasing and then decreasing, in this region. At point 2 total current is maximum, but its direction is vertical. The longitudinal component is thus zero and dV ∕dt is therefore zero. From pt. 2 to pt. 4 the longitudinal component of the current flow is away from the reference electrode. dV ∕dt is therefore negative and extracellular action potential amplitude continually decreases, going through zero and becoming negative, with the observed slopes depending on the same interplay of total current amplitude and slope as described for the interval between pts. 1 and 2. At pt. 4 (the maximum of the negative portion of the extracellular action potential), total current amplitude is (negative) maximum but its direction is again vertical and thus dV ∕dt again also zero. As the recording electrode is moved further right the process exactly repeats, with the longitudinal component pointing originally toward the reference electrode and dV ∕dt being positive (interval between pts. 4 and 6), the longitudinal component becoming zero where the total current amplitude is at a local maximum (pt. 6), the longitudinal component then switching to point away from the reference electrode, dV ∕dt being negative (interval between pts. 6 and 7), and then finally going to zero as the recording electrode is moved further from the action potential. Summary of Extracellular Recording in Uniform, Infinite Volume Conductor Although com-

plicated to describe rigorously mathematically, the combination of geometric and mathematical work summarized above provides a good conceptual image of current flow and potential generated by an action potential in a cylindrical axon embedded in a uniform, infinite volume conductor. In all cases a triphasic extracellular action potential (positive–negative–positive) and pattern of transmembrane current flow (outward–inward–outward) is predicted. An important issue is that the highly ordered current flows and potential variations in the uniform surrounding volume that give rise to these triphasic waveforms would be expected to be disturbed if the volume were not homogenous (Clark and Plonsey 1966). We next describe some of the effects of such non-uniformity. 2.6.3.3 Variations in Extracellular Action Potential Shape Induced by Non-Uniform, Non-Infinite Volume Conductors Monophasic Extracellular Action Potentials To avoid confusion it is important to understand

that monophasic extracellular action potentials can arise in several different ways. One such case is “effectively intracellular” extracellular recording methods, which we cover in Section 2.6.3.4 below. Another case is in heart electrophysiology (Franz 1999). The basis of this activity is controversial (Nesterenko et al. 2005), but one explanation is

Electrophysiological Recording Techniques

that it is similar to “effectively intracellular” extracellular recording methods. Regardless, as heart physiology is not part of this book, we do not cover this issue further here. Still another case is CAPs, in which the summing of the (presumably) triphasic action potentials results in a waveform that is only positive-going. Given that this summing destroys any information about the shape of single action potentials (see Fig. 2.6), we again do not cover this issue. The final case in which monophasic action potentials have been observed is in recording conditions in which the nerve being recorded from is surrounded by only a thin film of adhering saline which is itself surrounded by a large volume of insulating medium (e.g., hook electrodes in which the nerve is lifted into air or oil). In much classical work under these conditions, extracellular recordings of single cells were observed to have a monophasic waveform (e.g., from muscle, the bottom left panel of Fig. 2.7B; from nerve, Fig. 2.7E). One explanation for this shape is that, when the nerve is surrounded by only a thin film of conductive medium, current flow is restricted to near the nerve surface (Offner 1954) (Fig. 2.12B). This alteration removes almost all the portions of the current flows in which longitudinal current flows away from the central zone, and the extracellular recording therefore changes from triphasic to monophasic. A difficulty with this explanation is that modern extracellular recording with pin electrodes where the pin is situated next to the nerve and Vaseline used insulate the nerve/pin from the saline, or more complicated devices are used to accomplish the same goal (Schmitz et al. 1988, 1991), should also result in the nerve being surrounded by only a thin film of conducting medium. However, these recordings routinely give multiphasic extracellular action potentials. In an attempt to resolve this discrepancy, we have recorded from stomatogastric and locust nerves raised into the air. This work showed that the action potential waveform highly depended on the extent to which the nerve wrapped around the pin, and that it was also important to remove high-pass frequency filtering, which could add additional artifactual phases to the waveform. However, we were at best able to obtain only biphasic, not monophasic, action potentials (data not shown). We are unable to explain why we cannot repeat the old work, but note that, dating from the introduction of valve amplifiers and cathode ray recording, this difficulty has been reported by other workers (Blair and Erlanger 1933b; Cole and Curtis 1939). Other Effects of Non-Uniform Embedding Medium Extracellular recordings are often obtained

by building a small well around the nerve with Vaseline and putting the recording electrode in the well. Other recording techniques use tubes or cuffs placed around the nerve, or suck the cut end of a nerve, or a U-shaped portion of the nerve along its length (en passant recording), into a plastic or glass capillary tube. Because these techniques record current and voltage from a length of nerve instead of at a point, and because they alter the extracellular current flow and potential field, they would be expected to alter measured action potential shape. What these effects will be depends strongly on the details of the particular situation, and thus no general predictions can be made. In one case in which a detailed examination was performed, for a cuff electrode on an unmyelinated nerve with the cuff in a non-conductive medium, triphasic activity was predicted and observed (Stein and Pearson 1971). Myelination and the resulting restriction to action potentials to the nodes of Ranvier add further complications. With tubular recording electrodes, myelination results in

43

44

Neurobiology of Motor Control: Fundamental Concepts and New Directions

monophasic action potential recordings for tubes containing only a single node, with the sign of the action potential switching depending on where the node is located in the tube, and biphasic recordings for tubes containing multiple nodes (Marks and Loeb 1976). An additional concern for in situ or in vivo recordings is that the medium surrounding a nerve is seldom uniform. For instance, human bodies are a complex mélange of different tissues, each typically densely packed with cells and their processes. Paths to any recording electrode from an active cell are thus extremely non-uniform. Furthermore, low conductance paths often exist to even relatively distant locations, and thus achieving a truly indifferent reference electrode is typically very difficult (Rutkove 2007). 2.6.3.4 Bipolar Recording

The above sections have implicitly assumed that the reference electrode is not in contact with the nerve. However, extracellular action potentials are also measured with two more or less closely spaced electrodes in contact with the nerve. The advantage of bipolar recording is that, with proper electrode placement, the apparent amplitude of the action potential is substantially increased because (due to the fact that the outputs of the two electrodes are subtracted from each other) the negative peak of the action potential recorded on one electrode becomes a positive peak on the second (Fig. 2.13). The difficulty is that this subtraction completely distorts the shape of the underlying action potential (Fig. 2.13 B, C), including when the electrodes are placed to maximize recorded action potential amplitude (Fig. 2.13D; in the bipolar recording the final positive peak is larger than the initial positive peak, whereas in the actual potential field surrounding the axon, the initial positive peak is larger than the final). Furthermore, because of the dependence of action potential velocity on axon diameter, an inter-electrode spacing that optimizes recorded spike amplitude for one unit will not necessarily do so for others. As such, absent compelling justification, in most work it is best practice to use monopolar recordings. One such compelling justification is that, if the extracellular resistance between the two electrodes is made very large and the transmembrane potential under the second

Figure 2.13 Effect of bipolar electrode placement on recorded action potential shape. The extracellular action potential shown in Figure 2.11B (“outer membrane surface potential” trace) is shown approaching a pair of bipolar electrodes placed at decreasing inter-electrode differences (action potential propagates left). Left sides in each panel show recording set up. Right sides show recorded action potential, which equals the potential recorded at the right electrode minus the potential recorded at the left electrode. Scale bars on both sides are accurate for an action potential with a 1 m/sec propagation velocity. (A) Electrodes far enough apart (1.35 mm) that the action potential propagates completely under the right electrode before reaching the left electrode. The action potential is therefore recorded separately at each electrode. Because the potentials at the two electrodes are subtracted from each other, the potential recorded by the left electrode is -1 times the potential recorded by the right electrode (the left recording is a “vertically reflected” mirror image of the right recording). (B, C) As inter-electrode distance is decreased (B, 0.75 mm; C, 0.38 mm), the potentials recorded at the two electrodes begin to interact, resulting in complex shapes with increased total (most depolarized minus most hyperpolarized) voltage amplitude. D) At an inter-electrode distance of 0.19 mm total voltage amplitude is maximized, and has an almost triphasic shape. Note, however, that this is not the same shape as that recorded by a monopolar electrode (compare to action potential shape in left side of panel).

Electrophysiological Recording Techniques

A

left electrode recording

right electrode recording –+

B –+

C -+

D -+

10 μV 0.5 mm

10 μV 0.5 ms

electrode is made to be zero (by destroying this portion of the axon or bathing it in isotonic KCl) bipolar extracellular recording can measure transmembrane potential—that is, achieve the same measurements as do intracellular recordings (Berger and Barr 1969; Cleeman and Suenson 1984; Hoyle 1987; Huxley and Stämpfli 1951; Kocsis and Waxman 1983; Merrem et al. 1968; Stämpfli 1954; Stys et al. 1991; Tasaki and Frank 1955). This type of recording played an important role in measuring transmembrane potential and action potential shape in a variety of tissues (Burnstock and Straub 1958; Julian

45

46

Neurobiology of Motor Control: Fundamental Concepts and New Directions

et al. 1962; König 1962; Ritchie and Straub 1957; Rougier et al. 1968), but advances in intracellular techniques have reduced its use in neurobiology. These considerations are also important for reading the pre 1930s literature. This work was done without exception with two recording electrodes on the nerve, one on its side and the other on its cut end, and the nerve suspended in air. These recordings resulted in biphasic action potentials when the nerve under both electrodes was alive and monophasic when the nerve under the cut end was killed (Fig. 2.7B, bottom left panel vs. right panel; Fig. 2.7C, “C” traces vs. “D” traces). Given the expectation of modern workers that monopolar recordings should give triphasic action potentials (Figs. 2.2B, 2.10C, 2.11A, 2.12A), and bipolar recordings at least three phases (Fig. 2.13), these early data demand explanation. The basis for these early observations vary depending on the experimental situation, with multiple possible explanations being present. First, many of these recordings were of compound action potentials with the distance between the stimulation and recording electrodes being insufficient to allow separation of different axon diameter classes. In this case CAPs will be monophasic when recorded with a monopolar electrode (nerve killed under the second electrode), and thus, from the summation shown in Fig. 2.13, diphasic with bipolar electrodes (nerve alive under both). With respect to single axon action potentials, our inability, and that of others (Blair and Erlanger 1933b; Cole and Curtis 1939), to obtain with modern equipment monophasic action potentials with monopolar electrodes in air suggests that one explanation for monophasic action potentials in the older work is that earlier recording equipment filtered or otherwise missed the smaller amplitude portions of the action potentials. Another explanation is that of Offer (1954) (Fig. 2.12B). Because the older work was always with the nerves in air, Offer’s analysis argues that single action potentials recorded by each electrode would be monophasic. From Fig. 2.13, the recorded action potentials would thus again be biphasic when the nerve was alive under both electrodes and monophasic when the cut end of the nerve under the second electrode was killed. Another explanation of monophasic single axon action potentials in the data with the cut end of the nerve killed is that the effectively intracellular recording situation explained above occurred. Because the transmembrane action potential is monophasic (Fig. 2.11A), this would result in monophasic action potentials. With the cut end of the nerve alive, a similar situation exists except that the membrane potential under the cut end electrode is not zero, but instead the intracellular membrane potential, and the cut end electrode would record the action potential as it invaded the cut end of the nerve. From Fig. 2.13, this would again result in biphasic action potentials for the cut end alive situation. This explanation is also consistent with the injury currents measured in the earlier work, which are also present in effectively intracellular recording, and result from the voltage difference between the two electrodes (with the potential under the cut end electrode being zero when the nerve under it is killed, and the axon intracellular potential when the nerve under it is alive). 2.6.3.5 Extracellular Action Potential Summary

The most general conclusion is that extracellular action potential waveform shows great variation depending upon details of nerve anatomy, the characteristics of the medium or tissue in which the nerve is embedded, and the techniques used to make the recordings. This does not mean that, with proper care, analysis of extracellular action potential

Electrophysiological Recording Techniques

waveforms cannot give valuable data (e.g., Pearson (1970) and Schmitz et al. (1991) show how such analysis can be used to measure axon diameter). Nonetheless, in general, detailed analysis of extracellular action potential shape is of limited use in modern neurophysiology (Johnston and Wu 1995), particularly given the great advances in intracellular and patch recording technology that have occurred since the early days of neurophysiology. The primary uses of extracellular techniques in modern electrophysiology are therefore to stimulate axons and simply to measure spiking activity, not the detailed shape of the spikes themselves. With respect to action potential stimulation, the current flow considerations presented above explain why large axons fire at lowest stimulation amplitudes. When a voltage difference is applied across the two stimulation electrodes, current enters the axon under the positive electrode, flows along the cytoplasm, and leaves the axon under the negative electrode, where it depolarizes the membrane, which may therefore reach spike threshold. Longitudinal resistance decreases as diameter squared whereas membrane resistance decreases only as the 3/2 power of diameter. As a result, in larger diameter fibers a greater percentage of the applied voltage difference occurs across the membrane. Larger diameter axons therefore reach spike threshold at lower stimulation voltages. With respect to action potential recording, these techniques fall into two general classes. The first are neuron ensemble recordings from tissues containing both cell bodies and axons (cortex and the like) (see Chapter 3). The second are nerve recordings. In vertebrate nerves, with their large numbers of axons and lack, in general, of neurons that can be identified across individuals, extracellular recording is most typically used to measure population activities. Invertebrates, alternatively, often have nerves with relatively few axons and neurons that can be identified across individuals (e.g., Fig. 2.3). In these preparations the activity of individual units can often be identified on the basis of spike height or other spike sorting techniques, and thus extracellular recordings alone can provide detailed information about the activity of central neural networks.

Acknowledgements We thank C. Pouzat for help and editorial comments on Section 2.6.3, in particular the approximate solution to the exterior potential field calculation from transmembrane current; M. Dübbet and J. Sydow for help interpreting the circuits in Fig. 2.9; and P. Kloppenburg for discussion and improvement of all parts of the chapter.

References Adrian ED (1926) The impulses produced by sensory nerve endings. Part. 1. J Physiol (Lond) 61:49–72. Agnew WS (1984) Voltage-regulated sodium channel molecules. Annu Rev Physiol 46:571–530. Agnew WS, Levinson SR, Brabson JS, Raftery MA (1978) Purification of the tetrodotoxin-binding component associated with the voltage-sensitive sodium channel

47

48

Neurobiology of Motor Control: Fundamental Concepts and New Directions

from Electrophorus electricus electroplax membranes. Proc Natl Acad Sci USA 75:2606–2611. Akay T (2014) Long-term measurement of muscle denervation and locomotor behavior in individual wild-type and ALS model mice. J Neurophysiol 111:694–703. Ballo AW, Bucher D (2009) Complex intrinsic membrane properties and dopamine shape spiking activity in a motor axon. J Neurosci 29:5062–5074. Benzer TI, Raftery MA (1973) Solubilization and partial characterization of the tetrodotoxin binding component from nerve axons. Biochem Biophys Res Commun 51:939–944. Berger W, Barr L (1969) Use of rubber membranes to improve sucrose-gap and other electrical recording techniques. J Appl Physiol 26:378–382. Bernstein J (1868) Über den zeitlichen Verlauf der negativen Schwankung der Nervenstroms. Pflüg Arch 1:173–207. Bernstein J (1871) Untersuchungen über den Erregungsvorgang im Nerven- and Muskelsystem. Heidelberg. Germany: Winters Universitätbuchhandlung. Bernstein J (1902) Untersuchungen zur Thermodynamik der bioelektrischen Ströme. Pflüg Arch 92:512–562. Bernstein J (1912) Elektrobiologie: die Lehre von den elektrischen Vorgängen im Organismus auf moderner Grundlage dargestellt. Braunschweig, Germany: Vieweg und Sohn. Blair EA, Erlanger J (1932) Responses of axons to brief shocks. Exp Biol Med 29:926–927. Blair EA, Erlanger J (1933a) A comparison of properties of individual axons in the frog. Exp Biol Med 30:728–729. Blair EA, Erlanger J (1933b) A comparison of the characteristics of axons through their individual electrical responses. Am J Physiol 106:524–564. Brennecke R, Lindemann B (1974a) Design of a fast voltage clamp for biological membranes, using discontinuous feedback. Rev Sci Instr 45:656–661. Brennecke R, Lindemann B (1974b) Theory of a membrane-voltage clamp with discontinuous feedback through a pulsed current clamp. Rev Sci Instr 45:184–188. Brown KT, Flaming DG (1986) Advanced Micropipette Techniques for Cell Physiology. Chichester, UK: John Wiley & Sons. Brown TM, Banks JR, Piggins HD (2006) A novel suction electrode recording technique for monitoring circadian rhythms in single and multiunit discharge from brain slices. J Neurosci Meth 156:173–181. Burnstock G, Straub RW (1958) A method for studying the effects of ions and drugs on the resting and action potentials of smooth muscle with external electrodes. J Physiol (Lond) 140:156–167. Changeux J-P (2012) The nicotinic acetlycholine receptor: the founding father of the pentameric ligand–gated ion channel superfamily. J Biol Chem 287:40207–40215. Changeux J-P, Kasai M, Huchet M, Meunier JC (1970a) Extraction à partir du tissu électrique de gymnote d’une protéine présentant plusieurs propriétés caractéristiques du récepteur physiologique de l’acétylcholine. C R Acad Sci, Paris 270D:2864–2867. Changeux J-P, Kasai M, Lee C-Y (1970b) Use of a snake venom toxin to characterize the cholinergic receptor protein. Proc Natl Acad Sci USA 67:1241–1247. Chiu SY, Ritchi JM, Rogart RB, Stagg D (1979) A quantitative description of membrane currents in rabbit myelinated nerve. J Physiol (Lond) 292:149–166.

Electrophysiological Recording Techniques

Clark J, Plonsey R (1966) A mathematical evalulation of the core conductor model. Biophys J 6:95–112. Clark J, Plonsey R (1968) The extracellular potential field of the single active nerve fiber in a volume conductor. Biophys J 8:842–864. Cleeman L, Suenson M (1984) Reduction of the sucrose–saline interdiffusion in the sucrose gap technique by controlled compression of the extracellular space in myocardial preparations. Acta Physiol Scand 120:417–427. Cohen JB, Weber M, Huchet M, Changeux J-P (1972) Purification from Torpedo marmorata electric tissue of membrane fragments particularly rich in cholinergic receptor protein. FEBS Lett 26:43–47. Cole KS (1949) Dynamic electrical characteristics of the squid axon membrane. Arch Sci Physiol 3:253–258. Cole KS (1982) Squid axon membrane: impedance decrease to voltage clamp. Annu Rev Neurosci 5:302–323. Cole KS, Curtis HJ (1939) Electrical impedance of the squid giant axon during activity. J Gen Physiol 22:649–670. Cymbalyuk GS, Gaudry Q, Masino MA, Calabrese RL (2002) Bursting in leech heart interneurons: cell-autonomous and network-based mechanisms. J Neurosci 22:10580–10592. Del Negro CA, Hayes JA (2008) A “group pacemaker” mechanism for respiratory rhythm generation. J Physiol 586:2245–2246. Doyle DA, Cabral JM, Pfuetzner RA, Juo A, et al. (1998) The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science 280:69–77. du Bois-Reymond E (1848) Untersuchung über tierische Elektricität [1848–1884 (2 Bande)]. Berlin, Germany: Reimer. Duguid JR, Raftery MA (1973) Fractionation and partial characterization of membrane particles from Torpedo californica electroplax. Biochemistry 12:3593–3597. Erlanger J, Gasser HS (1930) The action potential in fibers of slow conduction in spinal roots and somatic nerves. Am J Physiol 92:43–82. Fatt P (1957) Electric potentials occurring around a neurone during its antidromic activation. J Neurophysiol 20:27–60. Finkel AS, Redman S (1984) Theory and operation of a single microelectrode voltage clamp. J Neurosci Meth 11:101–127. Foldes EL, Ackermann DM, Bhadra N, Kilgore KL, Bhadra N (2011) Design, fabrication and evaluation of a conforming circumpolar peripheral nerve cuff electrode for acute experimental use. J Neurosci Meth 196:31–37. Frankenhaeuser B, Huxley AF (1964) The action potential in the myelinated nerve fibre of Xenopus laevis as computed on the basis of voltage clamp data. J Physiol (Lond) 171:302–315. Franz MR (1999) Current status of monophasic action potential recording: theories, measurements and interpretations. Cardiovasc Res 41:25–40. Fromm J, Lautner S (2007) Electrical signals and their physiological significance in plants. Plant, Cell Environ 30:249–257. Galvani L (1791) De viribus electricitatis in motu musculari commentarius. Bon Sci Art Inst Acad Comm 7:363–418. Gasser HS, Erlanger J (1922) A study of the action currents of nerve with the cathode ray oscillograph. Am J Physiol 62:496–524.

49

50

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Gasser HS, Grundfest H (1939) Axon diameters in relation to the spike dimension and the conduction velocity in mammalian A fibres. Am J Physiol 127:393–414. Geselowitz D (1966) Comment on the core conductor model. Biophys J 6:691–692. Graziane N, Dong Y (2016) Electrical theory. In Electrophysiological Analysis of Synaptic Transmission. Neuromethods, vol 112. New York, NY: Springer. Gruhn M, Rathmayer W (2002) An implantable electrode design for both chronic in vivo nerve recording and axon stimulation in freely behaving crayfish. J Neurosci Meth 118:33–40. Hagiwara S, Jaffe LA (1979) Electrical properties of egg membranes. Annu Rev Biophys Bioeng 8:385–416. Halliwell JV, Whitaker MJ (1987) Using microelectrodes. In Standen NB, Gray PTA, Whitaker MJ (eds). Microelectrode Techniques. The Plymouth Workshop Handbook. 1st edn. Cambridge, UK: The Company of Biologists. Hamill OP, Marty A, Neher E, Sakmann B, Sigworth FJ (1981) Improved patch-clamp techniques for high–resolution current recording from cells and cell-free membrane patches. Pflüg Arch 391:85–100. Helmholz H (1850) Messungen über den zeitlichen Verlauf der Zuckung animalischer Muskeln und die Fortpflanzungsgeschwindigkeit der Reizung in den Nerven. Archiv für Anatomie, Physiologie und wissenschaftliche Medicin (1850):276–364. Henderson R, Wank JH (1972) Solubilization of a specific tetrodotoxin-binding component from garfish olfactory nerve membrane. Biochemistry 11:4565–4569. Hille B (2003) Ion Channels of Excitable Membranes. Sunderland, MA: Sinauer Associates. Hodgkin AL, Huxley AF (1939) Action potentials recorded from inside a nerve fibre. Nature 144:710–711. Hodgkin AL, Huxley AF (1952a) The components of membrane conductance in the giant axon of Loligo. J Physiol 116:473–496. Hodgkin AL, Huxley AF (1952b) Currents carried by sodium and potassium ions through the membrane of the giant axon of Loligo. J Physiol 116:449–472. Hodgkin AL, Huxley AF (1952c) The dual effect of membrane potential on sodium conductance in the giant axon of Loligo. J Physiol 116:497–506. Hodgkin AL, Huxley AF (1952d) A quantitative description of membrane current and its application to conduction and excitation in nerve. J Physiol 117:500–544. Hodgkin AL, Huxley AF, Katz B (1949) Ionic currents underlying activity in the giant axon of the squid. Arch Sci Physiol 3:129–150. Hooper SL, Thuma JB, Guschlbauer C, Schmidt J, Büschges A (2015) Cell dialysis by sharp electrodes can cause non–physiological changes in neuron properties. J Neurophysiol 114:1255–1271. Hoyle CHV (1987) A modified sucrose gap. J Pharm Methods 18:219–226. Huxley AF, Stämpfli R (1951) Direct determination of membrane resting potential and action potential in single myelinated nerve fibres. J Physiol (Lond) 112:476–495. Jiang Y, Lee A, Chen J, Ruta V, et al. (2003) X-ray structure of a voltage-dependent K+ channel. Nature 423:33–41. Johnston D, Wu SM (1995) Extracellular field recordings. In Foundations of Cellular Neurophysiology. Cambridge, MA: MIT Press. Julian FJ, Moore JW, Goldman DE (1962) Membrane potentials of the lobster giant axon obtained by use of the sucrose-gap technique. J Gen Physiol 45:1195–1216.

Electrophysiological Recording Techniques

Kocsis JD, Waxman SG (1983) Long-term regenerated nerve fibers retain sensitivity to potassium channel blocking agents. Nature 304:640–642. König K (1962) Membranepotenialmessungen am Skeletmuskel mit der “saccharose-trennward” Methode. Pflüg Arch 275:452–460. Kralj JM, Hochbaum DR, Douglass AD, Cohen AE (2011) Electrical spiking in Escherichia coli probed with a fluorescent voltage-indicating protein. Science 333:345–348. Kubalski A, Martinac Be (2005) Bacterial Ion Channels and their Eukaryotic Homologs. Washington, DC: ASM Press. Land BR, Wyttenbach RA, Johnson BR (2001) Tools for physiology labs: an inexpensive high–performance amplifier and electrode for extracellular recording. J Neurosci Meth 106:47–55. Lenoir T (1986) Models and instrument in the development of electrophysiology, 1845–1912. Hist Stud Phys Biol 17:1–54. Lertmanorat Z, Gustafson KJ, Durand DM (2006) Electrode array for reversing the recruitment order of peripheral nerve stimulation: experimental studies. Ann Biomed Eng 31:152–160. Ling G, Gerard RW (1949) The normal membrane potential of frog sartorius fibers. J Cell Physiol 34:383–396. Loeb GE, Peck RA (1996) Cuff electrodes for chronic stimulation and recording of peripheral nerve activity. J Neurosci Meth 64:95–103. Lorente de Nó R (1947) Analysis of the distribution of action currents in volume conductors. Stud Rockefeller Inst Med Res Repr 132:384–477. Lucas K (1908) A method of drawing tubes for the capillary electrometer. J Physiol (Lond) 37 (supplementary issue):xxviii–xxx. Lucas K (1909) On the relation between the electric disturbance in muscle and the propagation of the excited state. J Physiol (Lond) 39:207–227. Malmivuo J, Plonsey R (1995) Bioelectromagnetism. Principles and Applications of Bioelectric and Biomagnetic Fields. Oxford, UK: Oxford University Press. Marks WB, Loeb GE (1976) Action currents, internodal potentials, and extracellular records of myelinated mammalian nerve fibers derived from node potentials. Biophys J 16:655–668. Marmont G (1949) Studies on the axon membrane. I. A new method. J Cell Comp Physiol 34:351–382. Martinac B, Saimi Y, Kung C (2008) Ion channels in microbes. Physiol Rev 88:1449–1490. McKendrick JG (1883) Note on a simple form of Lippmann’s capillary electrometer useful to physiologists. J Anat 17:345–348. Merletti R, Farina D (2009) Analysis of intramuscular electromyogram signals. Philos T Roy Soc A 367:357–368. Merrem B, Kuchler G, Isenberg G (1968) On the deduction of bioelectric potentials of single skeletal muscle fibres by sucrose gap method. I. Modification of the specimen by experimental conditions. Acta Biol Med Ger 21:51–65. Möhl B (1977) High-frequency discharge of the locust wing hinge stretch receptor during flight. Naturwissenschaften 66:158–159. Nastuk WI, Hodgkin AL (1950) The electrical activity of single muscle fibers. J Cell Physiol 35:39–73. Neher E, Sakmann B (1976) Single-channel currents recorded from membrane of denervated frog muscle fibres. Nature 260:799–802.

51

52

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Nernst W (1889) Elektromotorische Wirksamkeit der Ionen. Z Phys Chem 4:129–181. Nesterenko VV, Kondo M, Antzelevitch C (2005) Biophysical basis for monophasic action potential. Cardiovasc Res 65:940–944. Newton I (1713) Principia Mathematica (2nd edn) English translation by Andrew Motte, Sir Isaac Newton’s Mathematical Principles of Natural Philosophy and his System of the World (1729, reprinted 1934 by Univ Calif Press, Oakland, CA). Nilius B (2003) Plügers Archiv and the advent of modern electrophysiology. Pflügers Arch/Eur J Physiol 447:267–271. Nusbaum MP, Weimann JM, Golowasch J, Marder E (1992) Presynaptic control of modulatory fibers by their neural network targets. J Neurosci 12:2706–2714. Offner FF (1954) The triphasic action potential. Electroencephalogr Clin Neurophysiol 6:507–508. Ogden D, Stanfield P (1994) Patch clamp techniques for single channel and whole-cell recording. In Odgen D (ed). Microelectrode Techniques: the Plymouth Workshop Handbook. 2nd edn. Cambridge, MA: The Company of Biologists. Overton E (1902) Beiträge zur allgemeinen Muskel- und Nervenphysiologie. II. Mittheilung. Ueber die Unentbehrlichkeit von Natrium- (oder Lithium-) lonen für den Contractionsact des Muskels. Pflüg Arch 92:346–386. Payandeh J, Scheuer T, Zheng N, Catterall WA (2011) The crystal structure of voltage–gated sodium channel. Nature 475:353–357. Pearson KG, Acharya H, Fouad K (2005) A new electrode configuration for recording electromyographic activity in behaving mice. J Neurosci Meth 148:36–42. Pearson KG, Stein RB, Malhotra SK (1970) Properties of action potentials from insect motor nerve fibres. J Exp Biol 53:299–316. Perl ER (2007) Ideas about pain, a historical view. Nat Rev Neurosci 8:71–80. Plonsey R (1964) Volume conductor fields of action currents. Biophys J 4:317–328. Plonsey R (1965) An extension of the solid angle potential formulation for an active cell. Biophys J 5:663–667. Plonsey R (1977) Action potential sources and their volume conductor fields. Proc IEEE 65:601–611. Plonsey R, Barr RC (1988) Bioelectricity. A Quantitative Approach. New York: Plenum Press. Polder HR, Swandulla D (2001) The use of control theory for the design of voltage clamp systems: a simple and standardized procedure for evaluating system parameters. J Neurosci Meth 109:97–109. Pun RYK (1988) Voltage clamping with single microelectrodes: comparison of the discontinuous mode and continuous mode using the Axoclamp 2A amplifier. Mol Cell Biochem 80:109–120. Ritchie JM, Straub RW (1957) The hyperpolarization which follows activity in mammalian non–medullated fibers. J Physiol (Lond) 136:80–97. Roelfsema MRG, Steinmeyer R, Hedrich R (2001) Discontinous single electrode voltage–clamp measurements: assessment of clamp accuracy in Vici faba guard cells. J Exp Bot 52:1933–1939. Rougier O, Vassort G, Stämpfli R (1968) Voltage clamp experiments on frog atrial heart muscle fibres with the sucrose gap technique. Pflügers Arch/Eur J Physiol 301:91–108. Rutkove SB (2007) Introduction to volume conduction. In Blum AS, Rutkove SB (eds). Clinical Neurophysiology Primer. Totowa, NJ: Humana Press.

Electrophysiological Recording Techniques

Schmitz J, Büschges A, Delcomyn F (1988) An improved electrode design for en passant recording from small nerves. Comp Biochem Physiol 91A:769–772. Schmitz J, Delcomyn F, Büschges A (1991) Oil and hook electrodes for en passant recording from small nerves. In Conn PM (ed). Methods in Neurosciences. Volume 4: Electrophysiogy and Microinjection. San Diego, CA: Academic Press. Sigworth FJ, Neher E (1980) Single Na+ channel currents observed in cultured rat muscle cells. Nature 287:447–449. Smarandache-Wellmann C, Hall WM, Mulloney B (2009) Coordination of rhythmic motor activity by gradients of synaptic strength in a neural circuit that couples modular neural oscillators. J Neurosci 29:9351–9360. Splinter R (2010) Action potential transmission and volume conduction. In Splinter R (ed). Handbook of Physics in Medicine and Biology. Boca Raton, FL: CRC Press Taylor & Francis Group. Stämpfli R (1954) A new method for measuring membrane potentials with external electrodes. Experientia 10:508–509. Stein RB, Pearson KG (1971) Predicted amplitude and form of action potentials recorded from unmyelinated nerve fibres. J Theor Biol 32:539–558. Stys PK, Kocsis JD (1995) Electrophysiological approaches to the study of axons. In Waxman SG, Kocsis JD, Stys PK (eds). The Axon: Structure, Function, and Pathophysiology. Oxford, UK: Oxford University Press. Stys PK, Ransom BR, Waxman SG (1991) Compound action potential of nerve recorded by suction electrode: a theoretical and experimental analysis. Brain Res 546:18–32. Tasaki I, Frank K (1955) Measurement of the action potential of myelinated nerve fiber. Am J Physiol 182:572–578. Verkhratsky A, Krishtal OA, Petersen OH (2006) From Galvani to patch clamp: the development of electrophysiology. Pflügers Arch/Eur J Physiol 453:233–247. Watanabe A, Grundfest H (1961) Impulse propagation at the septal and commissural junctions of crayfish lateral giant axons. J Gen Physiol 45:267–308. Wilson WA, Goldner MM (1975) Voltage clamping with a single electrode. J Neurobiol 6:411–422. Woodbury JW (1952) Direct membrane resting and action potentials from single myelinated nerve fibers. J Cell Physiol 39:323–339. Young JZ (1936) The structure of nerve fibers in cephalopods and crustacea. P Roy Soc B–Biol Sci 121:319–337. Zhang H-Y, Sillar KT (2012) Short-term memory of motor network performance via activity–dependent potentiation of Na+ /K+ pump function. Curr Biol 22:526–531. Zhang X, Ren W, DeCaen P, Yan C, et al. (2012) Crystal structure of an orthologue of the NaChBac voltage–gated sodium channel. Nature 486:130–134.

53

55

3 Multi-Unit Recording Arthur Leblois 1 and Christophe Pouzat 2 1 Centre de Neurophysique, Physiologie et Pathologie, CNRS UMR 8119, Institut Neurosciences et Cognition, Université Paris Descartes, Paris, France 2 Mathématiques Appliquées à Paris 5, CNRS UMR 8145, Université Paris Descartes, Paris, France

3.1 Introduction During reaching, and ballistic arm movements in general, the firing rate of large numbers of neurons in the arm-related area of the motor cortex changes in a manner that depends on movement direction, with movement direction being well-predicted by the weighted vector sum of the individual neuron activities (see Chapter 11). Moreover, changes in activity of single neurons in motor cortex or brainstem are neither necessary nor sufficient to induce movement. These data are from monkey, but are believed to apply to all mammalian, and likely vertebrate, systems. Movement direction is thus collectively encoded across a population of motor cortex neurons, and motor control results from the concerted activity of large numbers of neurons (Georgopoulos et al. 2014; Houk et al. 1993). Understanding how neurons encode movement therefore requires measuring the activity of large numbers of neurons. In principle, neurons could be sampled one by one, as was performed in early movement studies. However, instantaneous correlations among the activity of different neurons are an important part of movement encoding and decoding by the nervous system (Pesaran 2010; Shadlen and Newsome 1998). Similarly, hippocampal pyramidal cells during rest and sleep produce strongly coherent ensemble bursts believed to be critical in transferring information to the neocortex (Buzáki 2004). These patterns of across-neuron coordinated activity reveal the existence of dynamically-defined functional neuron assemblies that come into being, and whose neuronal make-up shifts, according to behavioral demand. Sequential single-cell recording cannot reveal such cooperative patterns. Understanding how these assemblies form and change, and how they generate and process information—that is, understanding the neural code underlying motor behavior—thus requires the ability to simultaneously record from multiple neurons during movement execution.

Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

56

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Recording and decoding the activity of multiple neurons during movement, however, poses multiple challenges. First, the recordings must be performed in vivo. Movements that can be made by head-restrained animals permit acute neural recordings using large recording apparatuses (Humphrey 1970; Reitbock et al. 1981). However, many natural motor behaviors, such as exploration and locomotion (O’Keefe and Dostrovsky 1971), song vocalization (McCasland 1987), and social interactions, are difficult or impossible for head-restrained animals to make. In these cases, a small recording apparatus is chronically implanted on the animal’s head, thereby permitting it to move in the environment with relatively little constraint while signals from individual neurons or clusters of neurons are recorded (McNaughton et al. 1983; Nordhausen et al. 1996). These recordings are almost always extracellular (but see Long et al. 2010; Hamaguchi et al. 2014), and thus action potentials from multiple neurons are recorded on single channels. These action potentials must be segregated on the basis of amplitude and wave-form to assign them to their proper source (neuron or class of neurons). The timing of action potentials from each neuron of the cell assembly can then be analyzed in relation to the motor behavior and the firing activity of other recorded neurons to extract the relevant motor code. Neuron ensemble recording is a rapidly evolving field. Currently, wire and silicon electrode arrays can simultaneously record from large numbers of neurons and thus monitor local neural circuits at work. These techniques can be used in a wide range of species, including humans, non-human primates, rodents (mainly rats and mice), birds, and invertebrates (e.g., fly, cockroach). We review here this technology, and provide general guidance on analysis of the large data sets these techniques generate.

3.2 Chapter Organization and Expository Choices This chapter is split into two sections: (1) hardware and data acquisition and (2) spike sorting methods. Most of its content applies to any kind of extracellular recording but particular attention is given to methods relevant to motor systems. As an illustrative example of multi-unit recording hardware, we describe chronic extracellular recording in songbirds using a custom-designed motorized microdrive. This choice is particularly appropriate because the species used are relatively small and songbirds will not sing spontaneously when stressed. Work in this system has therefore particularly strongly required developing recording systems that maintain signal quality while minimizing disruption of animal behavior. This requirement was fulfilled by developing very light (100 80 60 40 20

Figure 7.2.8 Stimulation of the lamprey pallium/cortex evokes coordinated movements. Threshold maps for different types of movements. (A) Dorsal schematic view of a lamprey brain showing the reconstructed distribution of electrically evoked motor responses. (B–E) Contour plots representing the threshold current needed to evoke different motor patterns. Note that the excitability of the LPal decreases from the caudal to the rostral pole. Scale bar (B) represents 500 μm. Abbreviations: Hb, habenula; LPal, lateral pallium; MPal, medial pallium; OB, olfactory bulb; OT, optic tectum; Pi, pineal gland; Rh, rhombencephalon. From Ocaña et al. (2015) with permission.

Decorticate mammals can thus display a complicated goal-directed behavioral repertoire. In contrast, decerebrate animals with lesions lower down, in front of the midbrain, can still be made to generate individual coordinated movements, but do so out of context. They can be made to locomote, swallow, and chew, and display a variety of reflexes, but the goal-directed aspect of the movement is completely lost. Goal-directed behavior thus requires the forebrain below neocortex. Although goal-directed movements can be generated without neocortex, normally the cortical input to the basal ganglia and concomitant projections directly to the brainstem appear to be complementary and act in concert to maintain precise control over motor pattern production. The basic forebrain circuitry responsible for initiating, maintaining, and terminating motor program activity appears to have originated early in vertebrate evolution and to have been maintained throughout phylogeny (Ocana et al. 2015; Stephenson-Jones et al. 2011, 2013).

Acknowledgements We thank Dr. Peter Wallén for valuable comments on the manuscript. This work was supported by grants VR-M-K2013-62X-03026 and VR-NT 621-2013-4613 from the Swedish Research Council and EU/FP7 no 604102 from the Human Brain Project.

Selection of Action—A Vertebrate Perspective electrical microstimulation

Birds Mammals Amphibians Teleosts

Reptiles

Sharks

Lampreys Hb

Optic tectum DLR

Spinal cord

Reticulospinal

MLR

Str

LPal (cortex)

STN Basal ganglia

SNc

Lamprey motor infrastructure

Figure 7.2.9 “Cortical” motor projections are conserved from lamprey to mammals. Schematic illustrating a phylogenetic tree with black areas indicating species in which electrical microstimulation of pallium/cortex have been shown to elicit movement. Below is a sagittal view of the lamprey brain indicating the descending pallial pathways to the basal ganglia, motor output regions (optic tectum, DLR, MLR), reticulospinal nuclei, and spinal cord. From Ocaña et al. (2015) with permission.

References Beloozerova IN, Sirota MG, Orlovsky GN, Deliagina TG (2005) Activity of pyramidal tract neurons in the cat during postural corrections. J Neurophysiol 93:1831–1844. Beloozerova IN, Zelenin PV, Popova LB, Orlovsky GN, et al. (2003) Postural control in the rabbit maintaining balance on the tilting platform. J Neurophysiol 90:3783–3793. Bjursten LM, Norrsell K, Norrsell U (1976) Behavioural repertory of cats without cerebral cortex from infancy. Exp Brain Res 25:115–130. Büschges A (2012) Lessons for circuit function from large insects: towards understanding the neural basis of motor flexibility. Curr Opin Neurobiol 22:602–608. Büschges A, Scholz H, El Manira A (2011) New moves in motor control. Curr Biol 21:R513–524. Carlsson A (1959) The occurrence, distribution and physiological role of catecholamines in the nervous system. Pharmacol Rev 11:490–493. Carlsson A (1964) Evidence for a role of dopamine in extrapyramidal functions. Acta Neuroveg (Wien) 26:484–493. Cui G, Jun SB, Jin X, Pham MD, et al. (2013) Concurrent activation of striatal direct and indirect pathways during action initiation. Nature 494:238–242.

191

192

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Deliagina TG, Beloozerova IN, Orlovsky GN, Zelenin PV (2014) Contribution of supraspinal systems to generation of automatic postural responses. Front Integr Neurosci 8:76. Deliagina TG, Fagerstedt P (2000) Responses of reticulospinal neurons in intact lamprey to vestibular and visual inputs. J Neurophysiol 83:864–878. Deliagina TG, Orlovsky GN, Grillner S, Wallén P (1992) Vestibular control of swimming in lamprey. II. Characteristics of spatial sensitivity of reticulospinal neurons. Exp Brain Res 90:489–498. DeLong MR (1990) Primate models of movement disorders of basal ganglia origin. Trends Neurosci 13:281–285. Deng Y, Lanciego J, Goff LK, Coulon P, et al. (2015) Differential organization of cortical inputs to striatal projection neurons of the matrix compartment in rats. Front Syst Neurosci 9:51. Doig NM, Moss J, Bolam JP (2010) Cortical and thalamic innervation of direct and indirect pathway medium-sized spiny neurons in mouse striatum. J Neurosci 30:14610–14618. Drew T, Prentice S, Schepens B (2004) Cortical and brainstem control of locomotion. Prog Brain Res 143:251–261. Dum RP, Strick PL (2002) Motor areas in the frontal lobe of the primate. Physiol Behav 77:677–682. Duysens J, Pearson KG (1976) The role of cutaneous afferents from the distal hindlimb in the regulation of the step cycle of thalamic cats. Exp Brain Res 24:245–255. El Manira A (2014) Dynamics and plasticity of spinal locomotor circuits. Curr Opin Neurobiol 29:133–141. Ellender TJ, Huerta-Ocampo I, Deisseroth K, Capogna M, Bolam JP (2011) Differential modulation of excitatory and inhibitory striatal synaptic transmission by histamine. J Neurosci 31:15340–15351. Ericsson J, Silberberg G, Robertson B, Wikstrom MA, Grillner S (2011) Striatal cellular properties conserved from lampreys to mammals. J Physiol 589:2979–2992. Ericsson J, Stephenson-Jones M, Kardamakis A, Robertson B, et al. (2013a) Evolutionarily conserved differences in pallial and thalamic short-term synaptic plasticity in striatum. J Physiol 591:859–874. Ericsson J, Stephenson-Jones M, Perez-Fernandez J, Robertson B, et al. (2013b) Dopamine differentially modulates the excitability of striatal neurons of the direct and indirect pathways in lamprey. J Neurosci 33:8045–8054. Forssberg H, Grillner S, Rossignol S (1975) Phase dependent reflex reversal during walking in chronic spinal cats. Brain Res 85:103–107. Freeze BS, Kravitz AV, Hammack N, Berke JD, Kreitzer AC (2013) Control of basal ganglia output by direct and indirect pathway projection neurons. J Neurosci 33:18531–18539. Georgopoulos AP, Grillner S (1989) Visuomotor coordination in reaching and locomotion. Science 245:1209–1210. Gerfen CR (1992) The neostriatal mosaic: multiple levels of compartmental organization. Trends Neurosci 15:133–139. Gerfen CR, Surmeier DJ (2011) Modulation of striatal projection systems by dopamine. Annu Rev Neurosci 34:441–466. Graybiel AM, Ragsdale Jr CW (1978) Histochemically distinct compartments in the striatum of human, monkeys, and cat demonstrated by acetylthiocholinesterase staining. Proc Natl Acad Sci USA 75:5723–5726.

Selection of Action—A Vertebrate Perspective

Grillner S (1985) Neurobiological bases of rhythmic motor acts in vertebrates. Science 228:143–149. Grillner S (2003) The motor infrastructure: from ion channels to neuronal networks. Nat Rev Neurosci 4:573–586. Grillner S (2006) Biological pattern generation: the cellular and computational logic of networks in motion. Neuron 52:751–766. Grillner S (2015) Action: the role of motor cortex challenged. Curr Biol 25:R508–R511. Grillner S, Robertson B, Stephenson-Jones M (2013) The evolutionary origin of the vertebrate basal ganglia and its role in action selection. J Physiol 591:5425–5431. Grillner S, Rossignol S (1978) Contralateral reflex reversal controlled by limb position in the acute spinal cat injected with clonidine i.v. Brain Res 144:411–414. Hikosaka O (2010) The habenula: from stress evasion to value-based decision-making. Nat Rev Neurosci 11:503–513. Hikosaka O, Wurtz RH (1983) Effects on eye movements of a GABA agonist and antagonist injected into monkey superior colliculus. Brain Res 272:368–372. Jordan LM, Brownstone RM, Noga BR (1992) Control of functional systems in the brainstem and spinal cord. Curr Opin Neurobiol 2:794–801. Kiehn O (2006) Locomotor circuits in the mammalian spinal cord. Annu Rev Neurosci 29:279–306. Kozlov AK, Kardamakis AA, Hellgren Kotaleski J, Grillner S (2014) Gating of steering signals through phasic modulation of reticulospinal neurons during locomotion. Proc Natl Acad Sci USA 111:3591–3596. Kumar S, Hedges SB (1998) A molecular timescale for vertebrate evolution. Nature 392:917–920. Lemon RN (2008) Descending pathways in motor control. Annu Rev Neurosci 31:195–218. Mathur BN, Capik NA, Alvarez VA, Lovinger DM (2011) Serotonin induces long-term depression at corticostriatal synapses. J Neurosci 31:7402–7411. Ménard A, Grillner S (2008) Diencephalic locomotor region in the lamprey—afferents and efferent control. J Neurophysiol 100:1343–1353. Mink JW (2003) The basal ganglia and involuntary movements: impaired inhibition of competing motor patterns. Arch Neurol 60:1365–1368. Newman H, Liu FC, Graybiel AM (2015) Dynamic ordering of early generated striatal cells destined to form the striosomal compartment of the striatum. J Comp Neurol 523:943–962. Ocaña FM, Suryanarayana SM, Saitoh K, Kardamakis AA, et al. (2015) The lamprey pallium provides a blueprint of the mammalian motor projections from cortex. Curr Biol 25:413–423. Orlovsky GN, Deliagina TG, Grillner S (1999) Neuronal Control of Locomotion: from Mollusc to Man. New York, NY: Oxford University Press. Parkinson J (1817) An essay on the shaking palsy. London: Sherwood. Pérez-Fernández J, Stephenson-Jones M, Suryanarayana SM, Robertson B, Grillner S (2014) Evolutionarily conserved organization of the dopaminergic system in lamprey: SNc/VTA afferent and efferent connectivity and D2 receptor expression. J Comp Neurol 522:3775–3794. Pombal MA, El Manira A, Grillner S (1997) Organization of the lamprey striatum—transmitters and projections. Brain Res 766:249–254.

193

194

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Pombal MA, Marin O, Gonzalez A (2001) Distribution of choline acetyltransferase-immunoreactive structures in the lamprey brain. J Comp Neurol 431:105–126. Reiner A, Jiao Y, Del Mar N, Laverghetta AV, Lei WL (2003) Differential morphology of pyramidal tract-type and intratelencephalically projecting–type corticostriatal neurons and their intrastriatal terminals in rats. J Comp Neurol 457:420–440. Rossignol S, Dubuc R, Gossard JP (2006) Dynamic sensorimotor interactions in locomotion. Physiol Rev 86:89–154. Ryczko D, Dubuc R (2013) The multifunctional mesencephalic locomotor region. Curr Pharm Des 19:4448–4470. Saitoh K, Ménard A, Grillner S (2007) Tectal control of locomotion, steering, and eye movements in lamprey. J Neurophysiol 97:3093–3108. Schultz W (2007) Multiple dopamine functions at different time courses. Annu Rev Neurosci 30:259–288. Shen W, Tian X, Day M, Ulrich S, et al. (2007) Cholinergic modulation of Kir2 channels selectively elevates dendritic excitability in striatopallidal neurons. Nat Neurosci 10:1458–1466. Stephenson–Jones M, Ericsson J, Robertson B, Grillner S (2012a) Evolution of the basal ganglia: dual-output pathways conserved throughout vertebrate phylogeny. J Comp Neurol 520:2957–2973. Stephenson-Jones M, Floros O, Robertson B, Grillner S (2012b) Evolutionary conservation of the habenular nuclei and their circuitry controlling the dopamine and 5-hydroxytryptophan (5-HT) systems. Proc Natl Acad Sci USA 109:E164–173. Stephenson-Jones M, Kardamakis AA, Robertson B, Grillner S (2013) Independent circuits in the basal ganglia for the evaluation and selection of actions. Proc Natl Acad Sci USA 110:E3670–3679. Stephenson-Jones M, Samuelsson E, Ericsson J, Robertson B, Grillner S (2011) Evolutionary conservation of the basal ganglia as a common vertebrate mechanism for action selection. Curr Biol 21:1081–1091. Surmeier DJ, Ding J, Day M, Wang Z, Shen W (2007) D1 and D2 dopamine-receptor modulation of striatal glutamatergic signaling in striatal medium spiny neurons. Trends Neurosci 30:228–235. Takakusaki K (2008) Forebrain control of locomotor behaviors. Brain Res Rev 57:192–198. Thompson RH, Ménard A, Pombal M, Grillner S (2008) Forebrain dopamine depletion impairs motor behavior in lamprey. Eur J Neurosci 27:1452–1460. Ullén NF, Deliagina T, Orlovsky G, Grillner S (1995) Spatial orientation in the lamprey. II. Visual influence on orientation during locomotion and in the attached state. J Exp Biol 198:675–681. Wilson CJ (1993) The generation of natural firing patterns in neostriatal neurons. Prog Brain Res 99:277–297. Wurtz RH, Hikosaka O (1986) Role of the basal ganglia in the initiation of saccadic eye movements. Prog Brain Res 64:175–190. Zelenin PV, Pavlova EL, Grillner S, Orlovsky GN, Deliagina TG (2003) Comparison of the motor effects of individual vestibulo- and reticulospinal neurons on dorsal and ventral myotomes in lamprey. J Neurophysiol 90:3161–3167.

195

7.3 Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects Hans-Joachim Pflüger Institut für Biologie, Neurobiologie; Freie Universität Berlin, Berlin, Germany and Biozentrum Köln, Institut für Zoologie, Universität zu Köln, Köln, Germany

7.3.1 Introduction In this chapter I review mechanisms of motor pattern selection and initiation. Before motor pattern selection occurs, however, a higher-level motivation decision (shall I eat or shall I mate), must be made, after which one of possibly many motor patterns appropriate to fulfilling the motivational goal are chosen for action. Motivational decision making is beyond the scope of this chapter. It is nonetheless useful to provide an example of this process in a particularly well understood case, as motivational state and motor pattern selection are clearly intimately intertwined, and many of the same issues arise in choosing among the many motor patterns that are appropriate for a given motivational state. In all animals situations occur where motivational decisions to fight or flee have to be made. In a detailed study of aggression between crickets, Stevenson and Rillich (2012) proposed that a cumulative assessment model (Payne 1998) best explains cricket’s actions. Crickets persist in fighting until the sum of the stimuli generated by their opponent’s actions accumulates to an “escape-threshold”, at which point they withdraw. Octopamine promotes aggression in insects whereas serotonin, nitric oxide, and several neuropeptides promote the motivation to flee. The pathways in the cricket nervous system where these decisions are made remain unknown. This model, however, implies that sensory stimuli from external and internal receptors, internal state as indicated by neuromodulator or neurohormone concentrations, and intrinsic neuronal activity in the brain and nervous system converge to determine whether the motivational state is aggressive or submissive. The tipping point between these two states will vary across individuals depending on genetics, experience, learning, age, and developmental state. As such, the process determining which motivational state exists must receive and integrate multimodal sensory information and stored information from learning processes and memories. Which motivational state is active will determine which set of motor behaviors will be selected among. Similar threshold mechanisms and multi-modal integration play a role in selection among motor patterns as well. Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

196

Neurobiology of Motor Control: Fundamental Concepts and New Directions

7.3.2 Organization Principles of Relevant Sensory Systems Topographic central representations of sensory fields are the rule for most sensory modalities. In vertebrates these topographic central structures occur in the mid or forebrain. In all known cases these central sensory representations are not “size calibrated” but “functionally calibrated”, meaning that functionally important parts are overrepresented in the central fields, even if the actual size of the region of the body from which the sensory input arises is small (e.g., central mechanosensory representation of the lips in humans is larger than that for the entirety of the back). Indeed, these size-mismatches are a good indication of the importance of sensory input from different body regions in the “life style” or “ecological niche” of the animal. In insects, as in other invertebrates, sensory pathways and their terminal fields are much more distributed and less centralized. However, topographic organization of sensory modalities does occur in the sensory neuropiles of ventral nerve cord ganglia (for insects see Bräunig et al. 1981; Hustert et al. 1981; Newland 1991; Pflüger et al. 1981, 1988, 1994). For example, the terminal axonal fibers of mechanosensory receptor cells of a particular body segment are organized strictly topographically and reveal a somatotopy (Burrows and Newland 1993; Grillenzoni et al. 1998; Lüdke and Lakes-Harlan 2008; Newland 1991; Newland et al. 2000). Because some mechanosensory primary afferent fibers (e.g., chordotonal organ fibers) project to many segmental ganglia, this means that the topographic organization is either (depending on the afferent) segment specific and confined to one ganglion or spreads over multiple ganglia, for example over all three thoracic ganglia (Bräunig et al. 1981). Moreover, similar to the vertebrate cases mentioned above, studies on the locust metathoracic segment, which bears a hind leg specialized for jumping, suggest that this leg is over-represented in the somatotopic organization of the respective sensory neuropil. The sense organs of the insect head are similarly topographically represented within their respective brain neuropils (e.g., antennal Johnston organ, Ai et al. 2007). Whether, as in vertebrates, there are also topographic representations of the whole body (homunculi) within structures in insect brain is unknown.

7.3.3 Movement-Generating Neural Networks in Invertebrates In insects, goal directed motor patterns can be initiated in segmental ventral cord ganglia to a limited extent but usually the selection of a behaviorally relevant, well-coordinated, and adaptive motor pattern requires involvement of the brain and its integrative capacity as in vertebrates. To carry out coordinated movements, the motor pattern must be initiated, maintained, and ultimately stopped. Reflex-Generated Movements To some extent, stimulus-oriented (reflex) motor patterns can be decided upon and carried out at the level of the ventral nerve cord and spinal cord alone. For example, mechanical stimulation of a locust wing leads to precise targeting-movements of the hind leg to remove a presumed hindering object (Matheson 1997, 1998; Page and Matheson 2004). Similar targeted grooming movements to many other body parts can be elicited in isolated locust ventral nerve cord (grooming, Berkowitz and Laurent 1996), proving that insects possess a topographic

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

representation of their body somewhere within the central nervous system, although this topographic representation is most likely more or less segmentally organized rather than centralized in the brain as in vertebrates. In frogs or turtles, precise targeting movements of appendages are possible at the level of the spinal cord (scratch reflexes; Hao et al. 2011; Stein 2010), and thus in vertebrates as well non-brain topographic body representations must exist. Rhythmic Motor Pattern Generation: CPGs The basic rhythmicity and pattern of neuron fir-

ing of most rhythmic motor patterns are produced by central pattern generator (CPG) neural networks that can be active in the absence of sensory feedback or patterned descending input (see Chapter 8). The first to recognize such CPGs were Hughes and Wiersma (1960) for the swimmeret system in crayfish, Wilson (1961, 1966) for locust flight, and Stent et al. (1979) for leech heart-beat. A general rule is that the more stereotyped a motor pattern is, the more important the CPG contribution. Some CPGs exhibit an inherent flexibility in their pattern (see below). Additionally, all CPGs receive sensory input that alters their output to compensate for unexpected variations from the intended output, either from insufficient motor response or environmental perturbations (see Chapter 9). Motor learning also occurs (Möhl 1988, 2000) (see also Chapter 13), with many insects learning to fly straight by producing asymmetric wing motor patterns when, for instance, large parts of wings are missing. In addition, insects can immediately change their inter-leg coordination so as to continue when two of the six legs are removed, and thus CPG coordination must also be flexible and adaptive (see Chapter 10). Motor patterns must also be expressed multiple ways, being fast in escape behavior but slow, adaptive, and flexible in exploratory behavior. In producing this wide range of adaptively appropriate motor behaviors, CPG alterations by neuromodulators, neurohormones, and other markers of internal state (e.g., hunger, stress), and sensory input and feedback, all play important roles.

7.3.4 Motor Pattern Selection in Invertebrates 7.3.4.1 Probabilistic “Selection”: Intrinsically Variable CPGs in Mollusk Feeding

In gastropod mollusks feeding patterns consist of movements towards food sources (appetitive movements or orientation) followed by patterns that underlie grazing, biting, swallowing, and also regurgitation and whose component neurons overlap between the respective circuits or CPGs (Kupfermann and Weiss 2001). A distinguishing feature of the mollusk feeding motor pattern both in vivo and in vitro is the high variability of its output. The continuing presence of this variability in isolated nervous system preparations indicates that this variability is inherent to the CPG network itself. In vitro, depending on different combinations of component neurons, even opposing patterns such as biting/ingestion and egestion could be generated and these respective neurons could act as switches between two patterns. The essential source of this variability is network neurons that alter post-inhibitory rebound, regenerative plateauing mechanisms, and endogenous oscillator properties in other network neurons and thus globally alter burst generation and autonomous network function, including blocking motor pattern production if “key component neurons” are inhibited. Various network neurons show

197

198

Neurobiology of Motor Control: Fundamental Concepts and New Directions

large variations in ion channels and electrical properties and are thought to be responsible for the observed flexibility. Network activity can be biased by learning or other forms of plasticity (Nargeot and Simmers 2012; see also Selverston 2010). This example shows that CPGs can possess an inherent mechanism for variability. Variability is observed at many levels during behavior in all animals and may possess an adaptive evolutionary value. This means that in addition to motor patterns being selected for, motor pattern “selection” can also occur because a probabilistic component may be present in many networks that introduces a random component into how the network functions or whether a response (behavior) occurs as, for example, has been demonstrated for an olfactory circuit in C. elegans (Gordus et al. 2015). 7.3.4.2 Selection via CPG Coordination

In insects, as in vertebrates, each joint is believed to be controlled by its own CPG-like element, a unit burst generator (Büschges 2005; Grillner 1981). The decision to walk, and what gait to produce, thus requires that a coordination among these elements be chosen. Decerebrate insects, i.e., those without a brain (supraoesophageal ganglion), walk more or less uninhibitedly in a coordinated fashion on horizontal surfaces (Huber 1960). Thus, suboesophageal (SOG, now called gnathal, Ito et al. 2014) ganglion input and local sensory feedback alone are sufficient to maintain an upright walking posture and proper leg and leg joint coordination. It is also necessary, as, if the connectives between the SOG and the rest of the ventral nerve cord are cut, animals no longer walk in a coordinated fashion. Individual legs can, however, execute single steps when stimulated by touch. This shows that the leg joints of individual legs can still move in a coordinated fashion, but coordination of all six legs requires signals from the SOG. These experiments may also suggest that the brain exerts a predominantly inhibitory influence on the SOG, and brain and SOG may therefore function antagonistically (Huber 1960). Interestingly, fruit flies without mushroom bodies no longer regulate locomotor actions and, thus have much longer walking bouts than wildtype flies (Martin et al. 1998). 7.3.4.3 Selection by Neuromodulators or Neurohormones

As mentioned above, rhythmic motor patterns are typically produced by CPGs. Isolated insect ventral nerve cords can produce properly coordinated motor neuron output activity (“fictive” motor patterns). These CPGs not only function, but can also develop, in the complete absence of sensory input (Selverston 2010; Suster and Bate 2002). In most cases these CPGs reside in one or several ventral cord ganglia. The general theme is that, depending on which neuromodulator is applied, a particular fictive motor pattern is induced. These data imply that these neuromodulators, or related substances that can interact with the same receptor, are released from projection neurons descending from the brain or anterior ganglia. The best studied motor system in this respect is the crustacean stomatogastric ganglion, which controls the different muscles of the stomach (Marder and Bucher 2001, 2007; Marder et al. 2005; Nusbaum et al. 2001; Selverston et al. 1976). The activities of the different stomatogastric neural networks can be dramatically altered by activity of neuromodulatory neurons projecting from more anterior ganglia, circulating neurohormones, or sensory feedback. These neuromodulators/neurohormones act on particular

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

component neurons of the networks and change their synaptic und electrical properties, which in turn leads to different output rhythms. Similar studies have been made in isolated ventral nerve cords of mollusks, leech, crayfish, and insects. Application of pilocarpine, a muscarinic acetylcholine receptor agonist, to an isolated crayfish ventral nerve cord produces motor rhythms and fictive locomotion (Chrachri and Clarac 1987, 1989, 1990). In insects, pilocarpine releases segment specific motor rhythms: if applied to a locust suboesophageal ganglion, fictive feeding is elicited (Rast and Bräunig 1997, 2001) but if applied to a metathoracic ganglion, a “fictive walking pattern” is induced (Ryckebusch and Laurent 1993). If more than one thoracic ganglion was isolated, rhythmicity in the meso- and prothoracic ganglia was generally less strong than that in the metathoracic ganglion, and coordination between the ganglia was weak and varying (Ryckebusch and Laurent 1994). This suggests that usually the CPG of one ganglion, here the metathoracic, plays a leading role, and that sensory information is required for inter-segment coordination. Since then pilocarpine has been found to induce fictive walking when applied to isolated ventral nerve cords in multiple insects (stick insect (Carausius morosus), Büschges 1995; Büschges et al. 1995; cockroach (Periplaneta americana), Fuchs et al. 2011; adult tobacco hawkmoth (Manduca sexta), Johnston and Levine 1996, 2002). In larval Manduca pilocarpine induces fictive crawling which, when applied to the whole ventral nerve cord except the brain, is well coordinated (Johnston and Levine 1996). In isolated thoracic nervous systems of adult Manduca it was later shown that pilocarpine not only induces fictive walking (Johnston and Levine 2002) but also fictive flight (Vierk et al. 2009). This apparent paradox was resolved by Rillich et al. (2013), who showed that, when applied to isolated locust thoracic nervous systems (Fig. 7.3.1A) at low concentrations, pilocarpine induced almost exclusively fictive walking, whereas at high concentrations it induced simultaneous fictive walking and flying, with the only interaction between the two motor patterns being that the fast motor units in leg levator motor neuron bursts occurred at the fictive flight frequency (Fig. 7.3.1B). This suggests that there is no, or at most very little, overlap between the walking and flight CPGs in locust thoracic ganglia.

Figure 7.3.1 (A) Schematic drawing showing a standard locust isolated meso (T2)-meta (T3) thoracic ganglion preparation with intact nerves to a wing elevator (WE) and wing depressor (WD) muscle and severed nerves to leg muscles: N3B, which carries trochanteral levator (TL) and slow extensor tibiae motoneuron (SETi) motor neurons, and N5A, which carries the common inhibitor (CI) neuron and slow and fast trochanteral depressor (TD) motor neurons. The WE and WD electromyograms show fictive flight induced by bath application of the muscarinic agonist pilocarpine. The N3B and N5A nerve recordings show fictive walking evoked by pilocarpine. The TL units in N3B identify swing phase and the TD units in N5A stance. (B–D) Simultaneous extracellular N3B and N5A nerve recordings and WE and WD electromyograms from preparation shown in A. Expanded time views of the patterns in the “i” panels are shown in the right column (ii). Pilocarpine (B, 1 mM) induced both fictive walking and flight. Arrows in Bii show the precise coupling in pilocarpine of TD motor neuron firing during TD bursts to the fictive flight rhythm. Octopamine (C, 100 mM) elicited a fictive flight pattern alone to which the leg motoneurons were recruited and again tightly coupled (arrows). Tyramine (D, 100 mM) induced both fictive walking and flight without coupling between TD firing and the flight rhythm. The black bars in the NSB and N5A traces in Bi and Di mark presumed swing and stance phases. All panels from Rillich et al. (2013) with permission.

199

200

Neurobiology of Motor Control: Fundamental Concepts and New Directions WD-WE latency

fictive flight

A

WE WD-WD cycle WD T2

WE WD 60ms fictive walking levators (TL)

T3

levators (TL)

SETi

N3B depressors (TD) Cl

N5A

2s

step cycle swing

stance

#

Bii pilocarpine detail

Bi pilocarpine (swing) N3B

TD

(stance) N5A

WE WD 5s

Ci octopamine

100ms

Cii octopamine detail

N3B TL

TD N5A WE WD 5s

100ms

Di tyramine

Dii tyramine detail

(swing) N3B TD

(stance) N5A

WE WD 5s

100ms

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

In contrast to pilocarpine, octopamine or its receptor agonist chlordimeform predominantly induced fictive flying in both locusts and moths and only very rarely, if ever, fictive walking. Interestingly, during octopamine or chlordimeform mediated fictive flight, leg muscle motor neurons fired at the flight frequency as long as the flight sequence lasted (Fig. 7.3.1C). Tyramine, the precursor of octopamine, like pilocarpine, induced fictive walking at low concentrations and fictive walking and flight at high concentrations (Fig. 7.3.1D). Agonists of nicotinic acetyl choline receptors such as carbachol (Buhl et al. 2008) and a blocker of GABA-receptor mediated chloride channels, picrotoxin, can also induce fictive flight (Vierk et al. 2009). Taken together, these data suggest it is not only the nature of transmitters/modulators, but also their concentration, that matters in releasing particular motor patterns. Somewhat different results are obtained when various biogenic amines (octopamine, serotonin, dopamine) are injected into the hemolymph of decapitated fruit flies (Yellman et al. 1997). Octopamine, for example, induces slow circular walking with occasional wing movements, whereas dopamine elicits grooming behavior. In fruit flies decapitation also removes the suboesophageal ganglion. Thus, flies, unlike the other insects that have been examined, seem to be able to produce a walking pattern at the level of the pterothoracic and abdominal ventral nerve cord. Another example of neurohormonal motor pattern selection is molting, the complex motor pattern in holometabolous insects in which the old cuticle is shed and replaced by a new one. For this motor behavior the muscular system of larvae has to be coordinated in a precise way. Usually a pre-ecdysis pattern is followed by the true ecdysis pattern after which a post-ecdysis pattern occurs. Ecdysis is under hormonal control and ecdysis-triggering-hormone (ETH) acts directly on neurons possessing ETH receptors. Work in Drosophila shows that FMRFamide-neurons are active during pre-ecdysis, that neurons releasing eclosion hormone (EH), crustacean cardioactive peptide (CCAP), and CCAP/myoinhibitory peptide (CCAP/MIP) are active prior to and during ecdysis, and CCAP/MIP/bursicon neurons are active during post-ecdysis. Kim et al. (2006) described ETH as a command chemical as ablation of the ETH-releasing neurons (the Inka cells) leads to lethal ecdysis deficiencies. The Inka cells are themselves regulated by corazonin, whose concentration in hemolymph is increased by systemic release from the corpora cardiac/corpora allata (Kim et al. 2004). 7.3.4.4 Selection by Command Neurons Not in the Brain

Activation of one or a small number of neurons is sometimes sufficient to activate a motor behavior. By definition, stimulation of such “command neurons” releases a particular complex behavior with great reliability (Kupfermann and Weiss 1978). The first to identify such neurons were Wiersma and Ikeda (1964) in the crayfish swimmeret system; since then such neurons have been identified in multiple preparations. Command Neurons for Escape Typical command neurons, often called giant fibers due to

their large diameter axons, have been identified in escape systems in a wide range of animals including annelids, cephalopods, insect, and fish, for example the Mauthner neuron. In all escape systems the common necessities are reliability and speed. One of the best studied escape behaviors in invertebrates is crayfish tail flip, where stimulation of particular giant fibers leads to fast complex movements including a somersault

201

202

Neurobiology of Motor Control: Fundamental Concepts and New Directions

(Edwards et al. 1999). Once the giant fiber is activated, the behavior proceeds in an “all or none” fashion. The giant fibers are activated by sensory stimuli that possess a threatening value such as a fast approaching object from the front or a sudden touch or water-jet from behind. In crayfish, a second system exists that also produces fast and complex swimming movements but, in contrast to the giant fiber escape system, can be adapted to requirements other than pure escape and thus is more flexible in its output (Edwards et al. 1999). Command Neurons for Food Intake vs. Locomotion Recent data from Drosophila larvae show that 20 neuropeptide hugin-expressing neurons in the SOG influence the decision between food intake or locomotion (Schoofs et al. 2014) (see also Chapter 4). When these neurons were stimulated food intake was depressed but wandering behavior, a special form of locomotion, was increased. Locomotion speed could be controlled by just 4 of the 20 hugin-expressing neurons. These experiments show that motor programs can be selected serially, in parallel, or exclusively, and again a limited number of neurons seems to serve this task. 7.3.4.5 The Brain is Crucial in the Motor Selection Process

As in vertebrates, the invertebrate brain clearly is often required for appropriate motor pattern selection. Our knowledge of which parts of the brain are involved in this process is still incomplete and comes from stimulation experiments or electrophysiological recordings from individual neurons or multi-electrode recordings from neuron populations (ensembles, see Chapter 3) in annelids (leech) and insects (stick insect, locust, cockroach, moth, fly). Recent progress largely is the result of studies on genetically tractable organisms such as nematodes (C. elegans) and insects (Drosophila). Brain Stimulation A method widely used by Huber (1960) was observing the behavior

of a cricket tethered on a ball when different brain areas of the brain were stimulated. Stimulation of the mushroom bodies usually inhibited locomotor movement whereas stimulation of the central complex or more lateral parts never did. In addition, electrical stimulation of the mushroom body and central complex released stridulation. Although the locations of the electrodes were determined histologically, the nature of the experiments and their resolution did not allow precise identification of neurons or fiber pathways. Command Neurons for Stridulation Stridulation is a component of the courtship behavior

of some insects and has been studied in grasshoppers, which rub their legs against the forewing, and crickets, which rub the two wings against each other. A command neuron in the grasshopper brain (B-DC-3 interneuron) was identified by Hedwig (1994) and described as “sufficient and necessary for the generation of the stridulatory leg movements”. Depolarization of this neuron induced stridulation and hyperpolarization stopped it. A similar neuron was also described in the cricket which, when depolarized, initiated stridulation (Hedwig 1996). Although there is evidence for a stridulation CPG (Schöneich and Hedwig 2011), how the command neurons activate the CPG is unknown.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

Command Neurons for Courtship Large advances in understanding the decision for

courtship behavior and its motor patterns have been recently achieved in Drosophila (von Philipsborn et al. 2011). In Drosophila a region of the dorsal posterior male brain is required to initiate singing whereas regions in the mesothoracic ganglion are required to execute correct song structure (Hall 1979; von Schlichet and Hall 1979). Thus, in Drosophila, similar to what is known from other insects, a thoracic central pattern generator is controlled by descending interneurons (perhaps command neurons) from the brain. The gene fruitless (fru) is predominantly involved in song production in males and a causal link has been established between the activity of fru-expressing neurons and song production. Von Philipsborn et al. (2011) identified two brain neurons (P1, or pMP4, and plP10) that can initiate an authentic song. plP10 may receive inputs from P1 and descends from the brain with axonal terminals within the mesothoracic ganglion. In the mesothoracic ganglion three other types of neurons (dPR1, vPR6, and vMS11) appear to be members of the CPG for a particular part of the male song and are most likely controlled by the two brain neurons. The Role of the Subesophageal Ganglion and Brain in Insect Walking and Flight In general, studies

where neurons in the connectives or the brain were recorded from in tethered locusts or crickets walking on a stationary Styrofoam ball (Böhm and Schildberger 1992; Kien 1990a,b; Staudacher 2001; Staudacher and Schildberger 1998), sets of neurons were found that fire synchronously with different aspects of the step patterns or which when stimulated change the direction or speed of walking, the form of a step, or the coupling between sides (Kien and Altman 1984). In addition, in some units firing frequency increased before stepping began (Kien 1990a) whereas in others firing was more clearly correlated with stepping itself (Kien 1990b). None of the fibers was stained, however, and the identity of neurons remains unclear. The location of these descending interneurons (named DINs) is either the brain, with many of them sending their axons through the dorsal deutocerebrum, or the SOG. Böhm and Schildberger (1992) identified one descending interneuron in the protocerebrum (IDIN, ipsilateral descending brain interneuron) which influenced cricket walking. Injection of current into this neuron to increase its firing rate above a certain threshold initiated walking whereas hyperpolarizing it stopped walking. Camera lucida drawings of this neuron indicate that its dendritic branches contact neither the mushroom bodies nor the central complex. The decision to fly involves deciding to jump, in that the front and middle legs lose tarsal contact first while the hind legs are still pushing the body forward before the body finally becomes airborne, only after which the wings open. For locust flight some prominent brain interneurons involved in various aspects of flight control have been identified, among them the Descending Contralateral Movement Detector (DCMD), later also described as being important in avoiding collisions during flight (Rind and Simmons 1992; Simmons and Rind 1992), and the Tritocerebral Commissure Giant Interneuron (TCG), involved in conveying information from wind sensitive head hairs to thoracic neurons. Several other multimodal brain interneurons are also involved in flight control (Simmons 1980a,b). An interesting interneuron is the A4I1 neuron (Pflüger 1984), which has its cell body in abdominal ganglion 4 and its main input zone in the prothoracic ganglion. A4I1 detects wind stimuli from the locust front and sends signals to the

203

204

Neurobiology of Motor Control: Fundamental Concepts and New Directions

thoracic ganglia and the brain, thus exhibiting corollary discharge properties. Descending interneurons in the SOG that contribute to various aspects of flight control have also been identified (Ramirez 1988). However, these neurons mainly convey sensory information from the front or elsewhere on the body to particular sets of flight motor neurons and thus control cycle-by-cycle aspects of flight. It remains unclear how much they contribute to flight initiation. Important new contributions have come from studies on Drosophila. Bidaye et al. (2014) identified neurons involved in switching between forward and backward walking. Usually flies are “reluctant” to walk backwards and do so only if other behaviors such as turning cannot be performed or if very strong stimuli are applied to the antennae. In an elegant study, Bidaye et al. (2014) characterized a strain of backward walking fruit flies they called “moonwalker”. They identified one type of descending brain interneuron, the moonwalker descending neuron (MDN), as both sufficient and necessary for initiating backward walking. There are two MDN per hemisphere, with their cell soma in the medial posterior protocerebrum, bilateral dendritic arborizations in the medial ventral protocerebrum and SOG, and axons descending to the contralateral thoracic ganglia. A second pair of neurons with functions in sustained backward walking was the moonwalker ascending neuron (MAN), with a soma and dendritic region in the metathoracic ganglion and an ascending axon to the brain. Interestingly, the MDN resemble in structure neurons previously identified in crickets (DBNc1-2 or DBNc2-2, Schöneich et al. 2011) and cockroaches (DMIa-1, Burdohan and Comer 1996). Neurons sufficient and necessary for forward walking were not identified. The authors suggest that forward walking may be the default state, but this still means that unidentified descending neurons for controlling the various parameters of forward locomotion (speed, gait, turning) should exist. The Central Complex of the Insect Brain is an Important Motor Area Elegant studies of wildtype

and mutant fruit flies have identified several brain areas as important for movements and higher locomotor control. One of these structures is the Central Complex (CC). Four well-defined areas can be distinguished: the fan shaped body, the ellipsoid body, the proto-cerebral bridge, and the paired noduli (Fig. 7.3.2A). The conspicuously regular pattern of fibers comes from columnar interneurons within these neuropilar structures. Flies with altered CC structures walk slowly, react more slowly to changing stimuli during flight, and show altered orientation behavior toward landmarks. They are also less active or quickly cease activity, or fail to start walking or flying under circumstances in which wild-type flies would readily do so (Ilius et al. 1994; Strauss et al. 1992; Strauss and Heisenberg 1993, reviewed in Heisenberg 1995). Based on detailed work on Drosophila mutants, Strauss (2002) reported seven distinct classes of aberrant walking: predominantly affected was the temporal pattern of the swing phase, spatial placement of the leg, correct initiation and maintenance of step length, across-body symmetry of step-length, range of stepping frequencies, swing phase duration, and leg swing speed. Of the 230 locomotor-defective lines, 30 exhibited CC structural changes. Interestingly, ablation or blocking of mushroom body instead of CC activity increased walking activity to above normal levels (Martin et al. 1998). Strauss (2002) concluded that “comparatively easy-to-conceive technical functions such as the balancing of step length on both sides of the body, along with seemingly complex decision functions such as whether to initiate walking voluntarily, are located in the CC”.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

New work recording from cockroach CC neurons while the animals walk on a slippery surface to some extent confirms the fruit fly data. CC neural activity strongly correlated with walking bouts, step frequency and sometimes could be predicted from the neural activity (Fig. 7.3.2B). Lesions within the CC and adjacent areas mainly affected the ability to make correct turns, suggesting the CC plays a role in integrating sensory inputs from the head into commands for turning (Ridge et al. 2007). In addition, electrical stimulation of the units also affected walking (Bender et al. 2010). The CC is also involved in more complex tasks, such as surmounting barriers or gaps, that require visual and mechanosensory antennal input, as lesions within CC disrupt these abilities (Fig. 7.3.2C; Bläsing and Cruse 2004a,b; Harley et al. 2009; Harley and Ritzman 2010; Pick and Strauss 2005). The CC, in particular the ellipsoid body, an area where inputs from various sensory modalities converge (Martin-Pena et al. 2014), was also involved in the special locomotor turns initiated exclusively by aversive odors, which are more straight and pronounced than turns to attractive odors (Gao et al. 2013). In Drosophila, neurons of the CC proto-cerebral bridge were identified in the tay bridge mutant that reduced walking speed and activity but did not affect compensation for rotatory stimuli during walking (Poeck et al. 2008). Fruit flies can also remember objects that are obstacles in their normal movement trajectories and, like cats (McVea and Pearson 2007), remember them even when the objects had been removed. Ring neurons of the ellipsoid body of the CC expressing protein kinase S6KII are necessary for this memory (Neuser et al. 2008). The importance of the CC and also the SOG, for initiation (“motivation”) and maintenance of coordinated walking is also evident from studies on how the predatory jewel wasp (Ampulex compressa) manipulates motor behavior of cockroach (Periplaneta americana) and drags it into a burrow to deposit an egg on the lethargic insect (Libersat and Gal 2013, 2104; reviewed in Gal and Libersat 2010). The venom, a cocktail of various substances including GABA and dopamine (Moore et al. 2006), is applied to the SOG and the brain, in particular areas near the CC and MBs, through a prolonged sting by the wasp, and most likely influences the octopaminergic and opioid systems (Gal et al. 2005; Gavra and Libersat 2011; Rosenberg et al. 2006, 2007). The venom prevents voluntary and sensory induced walking or escape movements with the exception that if the wasp pulls the cockroach at one of the previously shortened antennae, the cockroach will follow step by step to the wasp burrow. The venom thus does not interfere with the networks that are necessary for correct walking, but instead with the “motivation to walk” and the “threshold for maintained walking”. Other Insect Brain Areas in Insects: The Lateral Accessory Lobe and the Mushroom Bodies Another

interesting area of sensory to motor overlap is the Lateral Accessory Lobe (LAL) of olfactory orienting moths (Iwano et al. 2010; Kanzaki et al. 2013) which has been suggested to be the region where steering command information is generated (Kanzaki et al. 1991; Mishima and Kanzaki 1999; Namiki et al. 2014; Wada and Kanzaki 2005). Both the LAL and the adjacent ventral protocerebrum (VPC) contain projection neurons linking brain areas with those of the thoracic ganglia. Most of the neurons seem to be involved in the characteristic zig-zag locomotor pattern typical for pheromone-mediated orientation. A second area of insect brain important for locomotion, but less well studied than the CC in this respect, are the mushroom bodies. Mizunami et al. (1998a–c) recorded from various neuronal components during cockroach free walking. In recordings from the

205

206

Neurobiology of Motor Control: Fundamental Concepts and New Directions Central Complex (CC)

A MB PB FB EB

LAL

B

Figure 7.3.2 Neural responses in the cockroach Central Complex correlate with walking. (A) Diagram of the cockroach brain and its major neuropiles: MB, mushroom bodies; LAL lateral accessory lobes; AL, glomeruli of the antennal lobe. The Central Complex (CC) with its components Protocerebral Bridge (PB), Fan-shaped Body (FB), and Ellipsoid Body (EB) are shown within the rectangle. (B) Raw voltage traces from single electrodes in two tetrode bundles plotted during a bout of walking. Swing and stance phases of middle leg steps were extracted from high-speed video and the spike times of single units (expansion at bottom of panel) were isolated from the tetrode recordings. Individual neural units identified by animal, tetrode, and unit number (e.g., “unit 1, 2, 3” is animal 1, tetrode 2, unit 3). (C) Summary of motor task impairments caused by particular lesions. FB and LAL lesions caused abnormalities for all obstacles. PB lesions caused abnormalities only for shelf and turn obstacles. EB lesions caused abnormalities for shelf, turn, and wall obstacles, but not in reflexes or block behavior. Lesions outside the CC caused fewer abnormalities overall than lesions within the CC. This suggests that the CC is involved in complex behaviors and that different sub-regions of the CC have different levels of involvement in these behaviors. A from Ritzman et al. (2012), B Bender et al. (2010), C Harley and Ritzman (2010), all with permission.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

C

Figure 7.3.2 (Continued)

pedunculus and 𝛼- and 𝛽-lobes, units were found that fire during motor actions, with some of them predicting locomotion or turns (Fig. 7.3.3).

7.3.5 Two Case Studies I now present case studies in two model systems, C. elegans and leech. Motor Pattern Selection in C. elegans Touch-induced (avoidance) movements in C. elegans are mediated by the six touch receptors, five pairs of interneurons, and 69 motor neurons (Chalfie et al. 1985). Three principal pathways for touch-induced movements exist: two pathways for avoidance movements induced by anterior touch, mediated by two interneuron pairs (AVD and AVB), and one pathway for avoidance due to posterior touch mediated by one interneuron pair (PVC). The functions of these interneurons were inferred from the movement deficits laser ablation of the neurons induced. Current understanding is that locomotion is controlled by two sets of command neurons, forward locomotion by PVC/AVB and backward by AVD/AVA. Further analysis showed that aminergic signaling is most important for C. elegans movement. Activation of a chloride channel (LGC-55, Donnelly et al. 2013) by tyramine inhibits head movements and forward locomotion. For normal locomotion GABAergic neurons as well as cholinergic motor neurons play a role. Donnelly et al. (2013) “hypothesize that the temporally coordinated activation of ionotropic and metabotropic receptors may be a common signaling motif employed across organisms to orchestrate behavioral responses”. As for switching between crawling and swimming, biogenic amines are important in inducing each behavior, with dopamine being necessary for the swim-to-crawl transition in water and serotonin for the crawl-to-swim transition (Vidal-Gadea et al. 2011). Motor Pattern Selection (“Decision Making”) in the Leech An interesting hypothesis was

advanced by Esch and Kristan (2002), who studied motor decision making for crawling,

207

208

Neurobiology of Motor Control: Fundamental Concepts and New Directions 2

A a

5

b

c

4 3 1

3 cm

2 3

B

1

a

b

1

Cleaning of a foreleg 4 2 3

Walking c

1

Running

5

2

3

1 sec

Running

Figure 7.3.3 Recording from neural units in cockroach mushroom body. Increasingly vigorous responses of mushroom body units to self-stimulation, i.e., cleaning of a foreleg (A), walking (B), and running (C). Video frames (numbers) in these episodes (a–c in A) are indicated above the relevant electrophysiological records (a–c in B). Each group of recordings shows unfiltered responses of about six units above filtered data selecting two units. From Mizunami et al. (1998c) with permission.

swimming, or body shortening. They defined interneurons which, when stimulated, generated one of these behaviors, as “decision neurons” rather than command neurons (Fig. 7.3.4). The location and quality of sensory stimulation seems to matter in the decision to swim, crawl, or shorten: mechanical stimulation of the front end leads to shortening and of the rear to either swimming or crawling. Swimming occurs when the posterior sucker is not attached to the substrate and crawling when the leech is exposed to air. Therefore, additional sensory input, for example, from load sensors, may be crucial. With respect to the decision neurons, they found that they are involved in selecting more than one motor behavior and they therefore proposed that a combinatorial code of active decision neurons selects for particular motor patterns. In general, this work

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

A

B

C

D

E

F

G

209

210

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Figure 7.3.4 Neuronal basis of three leech behaviors. (A) Preparations used. Intact animals (top) are used to characterize movement kinematics, semi-intact preparations (middle) to relate muscle kinematics and motor neuron activity, isolated nerve cords (bottom) for more complex electrophysiological recording. (B) Front end touch induces whole-body shortening. The electrophysiological traces show dorsal and ventral longitudinal muscle activity. Recordings from the exposed part of a semi-intact preparation; a burst of electrical stimuli were delivered during shaded part of the recordings. (C) Twelve successive frames showing a complete swim cycle of a leech swimming left. The up-and-down undulatory waves can be made apparent by following the crests and troughs of the body wave across the frames. The electrophysiological recordings show that the dorsal and ventral longitudinal motor neurons produce alternating bursts at about 1 Hz, the swim cycle period. (D) A left-to-right crawl step. The leech is initially fully contracted with both suckers attached to the substrate. The front sucker then releases and the leech elongates (E) due to an anterior-to-posterior wave of contractions of the circular muscles in each segment. When the leech is fully elongated, it attaches its front sucker and shortens (contracts, C) due to an anterior-to-posterior wave of longitudinal muscle contractions. The tail sucker releases while the leech is contracting. When the leech is fully contracted the tail sucker re-attaches, completing one step of crawling. The electrophysiological traces show longitudinal (contraction, top and bottom, bars marked “C”) and circular (extension, middle, bars marked “E”) motor neuron activity. Longitudinal motor neuron bursts alternate with circular muscle motor neuron bursts in a segment (Gang. 10 recordings). Contraction bursts in more anterior segments occur earlier in more posterior segments (compare Gang. 10 and Gang. 13 recordings). The crawling cycle period is about ten times longer than that of swimming. Recordings in C and D obtained from isolated nerve cords. (E–G) Current injection (gray bars above each set of traces) into neuron R3b1 can elicit crawling or swimming. R3b1 activity monitored in the extracellular connective (conn 5/6). (E) Schematic of the isolated nerve cord preparation. (F) Intracellular recording of a CV neuron and extracellular recording of a DP nerve show a crawling motor pattern. Black bars below DP trace indicate fictive contractions. (G) R3b1 stimulation in a different preparation induced approximately equal R3b1 firing but elicited the swim motor pattern (note short cycle period bursts in time-expansion of DP 11 recording). In this case, swimming continued several cycles after current injection ceased. A–D from Esch and Kristan (2002), E–G from Esch et al. (2002), both with permission.

suggests that sensory feedback from external receptors about the condition or quality of the substrate, for example rough versus soft, water versus air or land, neutral or “smelly” or acid, and internal receptors carrying information about the physiological or motivational state of the animal, for example hungry or satiated, in reproductive state or pre- or post-mating, in an aggressive, dominant or a submissive, timid state, are crucial for the decision of which locomotor pattern is selected (Mesce and Pierce-Shimomura 2010). Further work identified several brain neurons involved in the decision to swim or crawl. One neuron, R3b-1, elicited swimming when the leech was immersed in deep water but crawling when the animal was in shallow water or exposed to air (Fig. 7.3.5, Esch et al. 2002). Thus, the command to locomote was generated by R3b-1 but sensory input selected the appropriate motor behavior in a context dependent manner. Even hybrid swim-crawl motor patterns could be evoked under certain environmental conditions (Esch et al. 2002; Mesce and Pierce-Shimomura 2010). R3b-1 was also active during more naturally occurring behaviors induced by sensory stimulation (Fig. 7.3.6). However, R3b-1 input is not required for swimming, as swimming can be elicited without the brain (cephalic ganglion) (Puhl and Mesce 2010), for example, by tactile stimulus activating skin mechanosensory neurons (Carlton and McVean 1995; Debski and Friesen 1987).

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

A

B

C

D

Figure 7.3.5 Effect of R3b1 electrical stimulation depends on the saline level in the chamber. Current was injected into R3b1 for the times indicated by gray bars in B–D; R3b1 spike frequency shown in each panel. Behaviors were observed directly and also recorded in segment 3 DP nerve (DP 3) activity. Elongation (E) and contraction (C) are indicated below the traces. (A) Schematic drawing of semi-intact preparation. (B) In low saline, current injection elicited crawling. In this example, cell 3 bursts occur in DP 3 during elongation, probably to assist in raising the head. (C) After the saline level was raised, R3b1 stimulation elicited swimming in the same preparation. (D) In intermediate saline levels, R3b1 stimulation elicited a hybrid behavior in which the leech swam (dots below DP trace) during elongation. In C and D, a portion of the traces has been expanded to show details of the swim bursts. From Esch et al. (2002) with permission.

The most important transmitters with respect to motor selection are biogenic amines. Dopamine elicits crawling and inhibits swimming behavior (Crisp and Mesce 2004; Puhl et al. 2012; Puhl and Mesce 2008, 2010) and may also be involved in feeding and biting behavior (O’Gara et al. 1991). Serotonin, in contrast, elicits swimming (Friesen and Kristan Jr 2007; Willard 1981), but is also the transmitter which, during feeding, inhibits sensory gating input that would elicit swimming, crawling or other body movements, and thus prevents these behaviors (Gaudry and Kristan Jr 2009, 2012). This apparent discrepancy was resolved by finding that it matters where serotonin is active: if serotonin is focally delivered to the brain, swimming is inhibited, whereas delivery to the ventral nerve cord elicits swimming. Correspondingly, some swim-gating neurons involved

211

212

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A.1

A.2

B

C

D

Figure 7.3.6 R3b1 is active during naturally elicited crawling and swimming. In A–C, the posterior end of the leech was briefly prodded (arrows) with a wooden applicator to elicit behaviors. (A.1) Touch-elicited crawling. R3b1 was activated and fired (raster above traces) while the leech elongated. E and C below traces indicate elongation and contraction. Stimulus-associated noise in the intracellular recording precluded spike identification during the time indicated by the gray box. (A.2) Mean (black) of traces (gray) centered on spikes marked in A.1 shows that intracellular spikes (top) align with spikes in the connective (bottom). To simplify the figure, the connective recording is not shown in A.1. (B) Touch-elicited swimming. In this preparation, an oscillation at the swim frequency occurred about a sustained depolarization. (C) Weak stimulation elicited a small R3b1 depolarization, but no overt behavior. (D) R3b1 depolarized and spiked at an increased frequency throughout a spontaneous swim. From Esch et al. (2002) with permission.

in initiating swimming through mechanosensory input from the skin are serotonergic (Nusbaum and Kristan Jr 1986). The same sensory stimuli also activate octopaminergic neurons (Gilchrist and Mesce 1997) and indeed octopamine delivered to the ventral nerve cord also elicits swimming. Cocktails of serotonin and octopamine, however, strongly inhibited swimming (Mesce et al. 2001, reviewed in Mesce 2002). In contrast, during wash-out the cocktail induced robust swimming that was much stronger than that induced by applying any amine by itself. These effects only occurred if the brain was attached to the ventral nerve cord and subsequent experiments showed that focal application of the 5-HT/octopamine cocktail inhibited swimming and its removal initiated robust swimming. Serotonin applied exclusively to the brain inhibited swimming whereas octopamine applied in the same way initiated swimming (Mesce 2002). This is interesting because the two amines act cooperatively at the level of the ventral nerve cord, with both initiating swimming, but act oppositely in the brain. Mixtures of dopamine

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

and serotonin also block swimming (Mesce and Pierce-Shimomura 2010) and this may be the reason why feeding with food containing high levels of dopamine prevents swimming.

7.3.6 Concluding Remarks on Invertebrates Two common themes in invertebrate locomotion are that CPGs in the segmental ganglia generate the basic rhythmicity and phase relationships of each motor pattern and, in the absence of descending input from the head ganglia (brain and/or suboesophageal ganglion), motor patterns can be initiated at the level of the ventral nerve cord with sufficient sensory input. Consequently, application of neurotransmitters or neuromodulators onto whole or parts of isolated ventral nerve cords can activate fictive motor patterns. Under natural conditions with the brain and suboesophageal ganglion attached to the ventral nerve cord, activation of single or small numbers of command neurons is responsible for motor pattern selection. The decision of which motor pattern is selected depends on input signaling (1) from many external sensory receptors such as the antennae with their compliment of chemoreceptors and mechanoreceptors or other parts of the body such as the thorax, for example leg load sensors, or the abdominal cerci; (2) from visual signals coming from the compound eyes or ocelli; and (3) signaling from internal receptors measuring physiological state, for example, whether the animal is hungry, satiated, or ready to reproduce. Internal state will also affect the release of neurohormones and/or neuromodulators that act on the descending command, decision making, or projection neurons, or neurons presynaptic to them. These substances play an important dual role because they not act not only on the central and peripheral nervous system but also control many physiological and metabolic processes that act in concert with, and may be a characteristic feature of, a particular behavior. In further investigation of the brain areas involved in motor pattern selection, studies on the so-called “genetic model organisms” such as nematode and fruit fly will play increasingly important roles. The work in Drosophila has already shown that the central complex plays a central role in motor pattern selection. For some specialized behaviors such as backward-walking only a few neurons may be responsible. New promising insights may come from optogenetic and thermogenetic techniques (see Chapter 4) that allow imaging brain activity (Grover et al. 2016) or controlling sets of neurons by activating them (Bath et al. 2014) in freely walking Drosophila. Similar experiments are being performed in a vertebrate system, the zebrafish, that will allow functional comparison (Dunn et al. 2016). However, understanding the exact role particular neurons or subset of neurons play will also require experiments on larger organisms in which electrophysiological recordings from neurons in intact or semi-intact organisms are possible. Experiments on leech and large insect species will therefore still be essential.

7.3.7 Are There Common Themes between Motor Pattern Selection in Invertebrates and Vertebrates? This and the preceding chapter showed that the most important areas for motor pattern selection in insects are the various subdivisions of the central complex and in

213

214

Neurobiology of Motor Control: Fundamental Concepts and New Directions

vertebrates the different nuclei and parts of the basal ganglia. Although it is clear that both groups depend on CPGs to generate the motor patterns themselves, whether common mechanisms with respect to motor pattern selection has been less clear. However, there is accumulating evidence that before the split between vertebrates (Deuterostomia) and bilaterian invertebrates (Protosomia), an “urbilaterian” brain existed. This is based upon results from molecular genetics that the basic body plan of bilateral invertebrates and vertebrates is similar but inverted (Arendt and Nübler-Jung 1994; Reichert and Simeone 2001). These data argue that, as has long been hypothesized (Dohrn 1875; Hanström 1928; Zawarzin 1925), the dorsal spinal cord of vertebrates and the ventral nerve cord of bilaterian invertebrates are homologous, not analogous. In addition, many genes involved in patterning the early embryonic brain are homologous between vertebrates and bilaterian invertebrates and, thus many of the structures may have common design principles (Arendt 2005; Hirth et al. 2003; Kammermeier

A

Striatum ground pattern

B

C

Matrisome (modality, sensory representation)

Striosome (modulation, memory)

D

E

Fan-shaped body ground pattern

F Module (modality-sensory space)

M

Stratum (modulation, memory)

Figure 7.3.7 Similar modular Baupläne of mammalian striatum and insect fan-shaped body. (A–C) Striatum primarily consists of closely associated striosomes (dark gray in A) and matrisomes (as marked in B and C). Striosomes receive inputs from hippocampus and amygdala about internal state (modulation, memory). Matrisomes receive cortical input about sensory space and modalities. (C) Local interneurons (arrows internal to slab) mediate interconnections between matrisomes and between matrisomes and striosomes and provide integrated information to GABAergic output pathways (vertical arrows). (D–F) The insect fan-shaped body (D) consists of tangential stratifications intersected by columnar modules (bracketed arrows marked M in D and E). Columnar modules receive input about sensory modality and space. Tangential elements are relays primarily from protocerebrum coding for physiological state, learned sensory associations, and higher-order sensory cues. (F) As in striatum, local interneurons (arrows internal to slab) mediate interconnections between columnar modules and columnar modules and planar tangentials and provide integrated information to outgoing GABAergic pathways (vertical arrows). From Strausfeld and Hirth (2013) with permission.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

Figure 7.3.8 Proposed correspondences between mammalian basal ganglia (left) and insect central complex (right) regions and connections. Dopaminergic pathways are black, inhibitory pathways white, other pathways (excitatory or modulatory) gray. Output pathways shown by arrows at bottom of figure. Mirror-image regions about figure center correspond in mammal and insect (e.g., mammalian striatum (ST) corresponds to the insect fan-shaped body (FB) and protocerebral bridge (PB)). Note strong similarities in regions, pathway connections among regions, and neurotransmitter use. Other abbreviations: external and internal globus pallidus, GPe, GPi; ellipsoid body, EB; hippocampus, HI; mushroom bodies, MB; subthalamic nucleus, STN; substantia nigra pars compacta (SNc); amygdala, AM; intermediate and inferior lateral protocerebra, IMP, ILP; superior medial protocerebrum, SMP; lateral accessory lobes, LAL; thalamus, TH; inferior and ventrolateral protocerebra, ILP, VLP. Modified from Strausfeld and Hirth (2013) with permission.

and Reichert 2001; Lichtneckert and Reichert 2005; Sprecher and Reichert 2003). Increasingly, comparative studies of molecular genetics reveal commonalities of design even for specific brain areas. With respect to this issue, a recent review by Strausfeld and Hirth (2013) is of particular interest. They suggest that the arthropod central complex and vertebrate basal ganglia both result from an “evolutionarily conserved genetic program leading to interconnected neuropils and nuclei that populate the midline of the forebrain-midbrain boundary region”. Similar inhibitory and modulatory transmitters such as GABA and dopamine seem to play decisive roles in both areas, and the authors argue that “the

215

216

Neurobiology of Motor Control: Fundamental Concepts and New Directions

observed multitude of similarities suggest deep homology of arthropod central complex and vertebrate basal ganglia circuits underlying the selection and maintenance of behavioral actions”. Thus far the similarities between the insect central complex and the vertebrate basal ganglia are homologous genes involved in formation, patterning, and specification of subdivisions as well as similar transmitters and receptor molecules including inhibitory and modulatory connection between subsystems. In addition, both the central complex, in particular the fan-shaped body, and the striatum of the basal ganglia (striosomes and matrisomes) are modularly organized (Fig. 7.3.7). Both the central complex and the basal ganglia also function similarly in that interactions between inhibitory, excitatory, and modulatory inputs lead to motor pattern selection (Fig. 7.3.8). In addition, based on its integration of gustatory inputs and information on many other aspects of feeding with locomotion, and its role of connecting the ventral nerve cord to the brain, the insect SOG (gnathal ganglion) has been compared to vertebrate brain stem, which also connects the brain hemispheres to the spinal cord (Schoofs et al. 2014). The urbilaterian common ancestor also had to select which motor programs to execute, and taken together, the above data suggest that, just as with the use of CPGs, the neural basis of this process may have been preserved through phylogeny. With the recent advances in molecular biology and the explosion of work applying molecular biological techniques to neurobiological problems, we now have the tools to investigate this question experimentally on a level of detail never before possible. These are thus truly exciting times in motor control science, and it is possible that combined studies in invertebrates and vertebrates are on the verge of revealing common design principles of motor selection and decision making in both groups.

References Ai H, Nishino H, Itoh T (2007) Topographic organization of sensory afferents of Johnston’s organ in the honeybee brain. J Comp Neurol 502:1030–1046. Arendt D (2005) Genes and homology in nervous system evolution: comparing gene function, expression patterns, and cell type molecular fingerprints. Theory Biosci 124:185–197. Arendt D, Nübler-Jung K (1994) Inversion of dorsoventral axis? Nature 371:26. Bath DE, Stowers JR, Hörmann D, Poehlmann A, et al. (2014) flyMAD: rapid thermogenetic control of neuronal activity in freely walking Drosophila. Nature Methods 11:756–762 Bender J, Pollack A, Ritzmann R (2010) Neural activity in the central complex of the insect brain is linked to locomotor changes. Curr Biol 20:921–926. Berkowitz A, Laurent G (1996) Central generation of grooming motor patterns and interlimb coordination in locusts. J Neurosci 16:8079–8091. Bidaye S, Machacek C, Wu Y, Dickson B (2014) Neuronal control of Drosophila walking direction. Science 344:97–101. Bläsing B, Cruse H (2004a) Mechanisms of stick insect locomotion in a gap-crossing paradigm. J Comp Physiol A 190:173–183. Bläsing B, Cruse H (2004b) Stick insect locomotion in a complex environment: climbing over large gaps. J Exp Biol 207:1273–1286.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

Böhm H, Schildberger K (1992) Brain neurones involved in the control of walking in the cricket Gryllus bimaculatus. J Exp Biol 166:113–130. Bräunig P, Hustert R, Pflüger H (1981) Distribution and specific central projections of mechanoreceptors in the thorax and proximal leg joints of locusts. I. Morphology, location and innervation of internal proprioceptors of pro- and metathorax and their central projections. Cell Tissue Res 216:57–77. Burdohan JA, Comer CM (1996) Cellular organization of an antennal mechanosensory pathway in the cockroach, Periplaneta americana. J Neurosci 16:5830–5843. Burrows M, Newland P (1993) Correlation between the receptive fields of locust interneurons, their dendritic morphology, and the central projections of mechanosensory neurons. J Comp Neurol 329:412–426. Büschges A (1995) Role of local nonspiking interneurons in the generation of rhythmic motor activity in the stick insect. J Neurobiol 27:488–512. Büschges A (2005) Sensory control and organization of neural networks mediating coordination of multisegmental organs for locomotion. J Neurophysiol 93:1127–1135. Büschges A, Schmitz J, Bässler U (1995) Rhythmic patterns in the thoracic nerve cord of the stick insect induced by pilocarpine. J Exp Biol 198:435–456. Carlton T, McVean A (1995) The role of touch, pressure and nociceptive mechanoreceptors of the leech in unrestrained behavior. J Comp Physiol A 177:781–791. Chalfie M, Sulston J, White J, Southgate E, et al. (1985) The neural circuit for touch sensitivity in Caenorhabditis elegans. J Neurosci 5:956–964. Chrachri A, Clarac F (1987) Induction of rhythmic activity in motoneurons of crayfish thoracic ganglia by cholinergic agonists. Neurosci Lett 77:49–54. Chrachri A, Clarac F (1989) Synaptic connections between motor neurons and interneurons in the fourth thoracic ganglion of the crayfish, Procambarus clarkii. J Neurophysiol 62:1237–1250. Chrachri A, Clarac F (1990) Fictive locomotion in the fourth thoracic ganglion of the crayfish, Procambarus clarkii. J Neurosci 10:707–719. Crisp K, Mesce K (2004) A cephalic projection neuron involved in locomotion is dye coupled to the dopaminergic neural network in the medicinal leech. J Exp Biol 207:4535–4542. Debski E, Friesen W (1987) Intracellular stimulation of sensory cells elicits swimming activity in the medicinal leech. J Comp Physiol A 160. Dohrn A (1875) Der Ursprung der Wirbelthiere und das princip des Functionswechsels. Leipzig, Germany: Wilhelm Engelmann. Donnelly J, Clark C, Leifer A, Pirri J, et al. (2013) Monoaminergic orchestration of motor programs in a complex C. elegans behavior. PLoS Biol 11:e1001529. Dunn TW, Mu Y, Narayan S, Randlett O, et al. (2016) Brain-wide mapping of neural activity controlling zebrafish exploratory locomotion. eLife doi: 10.7554/eLife.12741. Edwards DH, Heitler WJ, Krasne FB (1999) Fifty years of a command neuron: the neurobiology of escape behavior in the crayfish. Trends Neurosci 22:153–161. Esch T, Kristan WJ (2002) Decision-making in the leech nervous system. Integr Comp Biol 42:716–724. Esch T, Mesce KA, Kristan Jr, WB (2002) Evidence for sequential decision making in the medicinal leech. J Neurosci 22:11045–11054. Friesen W, Kristan Jr, W (2007) Leech locomotion: swimming, crawling, and decisions. Curr Opin Neurobiol 17:704–711.

217

218

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Fuchs E, Holmes P, Kiemel T, Ayali A (2011) Intersegmental coordination of cockroach locomotion: adaptive control of centrally coupled pattern generator circuits. Front Neural Circuit 4:125. Gal R, Libersat F (2010) A wasp manipulates neuronal activity in the sub–esophageal ganglion to decrease the drive for walking in its cockroach prey. PLoS ONE 5:e10019. Gal R, Rosenberg L, Libersat F (2005) Parasitoid wasp uses a venom cocktail injected into the brain to manipulate the behavior and metabolism of its cockroach prey. Arch Insect Biochem Physiol 60:198–208. Gao X, Potter C, Gohl D, Silie SM, et al. (2013) Specific kinematics and motor-related neurons for aversive chemotaxis in Drosophila. Curr Biol 23:1163–1172. Gaudry Q, Kristan Jr, W (2009) Behavioral choice by presynaptic inhibition of tactile sensory terminals. Nat Neurosci 12:1450–1457. Gaudry Q, Kristan Jr, W (2012) Decision points: the factors influencing the decision to feed in the medicinal leech. Front Neurosci 6:101. Gavra T, Libersat F (2011) Involvement of the opioid system in the hypokinetic state induced in cockroaches by a parasitoid wasp. J Comp Physiol A 197:2279–2991. Gilchrist L, Mesce K (1997) Coactivation of putative octopamine- and serotonin-containing interneurons in the medicinal leech. J Neurophysiol 78:2108–2115. Gordus A, Pokala N, Levy S, Flavell S, Bargmann C (2015) Feedback from network states generates variability in a probabilistic olfactory circuit. Cell 161:215–227. Grover D, Katsuki T, Greenspan RJ (2016) Flyception: imaging brain activity in freely walking fruit flies. Nature Methods 13:569–572. Grillenzoni N, van Helden J, Dambly-Chaudiere C, Ghysen A (1998) The iroquois complex controls the somatotopy of Drosophila notum mechanosensory projections. Development 125:3563–3569. Grillner S (1981) Control of locomotion in bipeds, tetrapods and fish. In Handbook of Physiology. The Nervous System. Motor Control. Brooks V (ed). Bethesda, MD: Am Physiol Soc. Hall J (1979) Control of male reproductive behavior by the central nervous system of Drosophila: dissection of a courtship pathway by genetic mosaics. Genetics 92:437–457. Hanström B (1928) Vergleichende Anatomie des Nervensystems der Wirbellosen Tiere (republished by Amsterday: Asher). New York, NY: Springer. Hao Z, Spardy L, Nguyen E, Rubin J, Berkowitz A (2011) Strong interactions between spinal cord networks for locomotion and scratching. J Neurophysiol 106:1766–1781. Harley C, English B, Ritzmann R (2009) Characterization of obstacle negotiation behaviors in the cockroach, Blaberus discoidalis. J Exp Biol 212:1463–1476. Harley C, Ritzman R (2010) Electrolytic lesions within central complex neuropils of the cockroach brain affect negotiation of barriers. J Exp Biol 213:2851–2864. Hedwig B (1994) A cephalothoracic command system controls stridulation in the acridid grasshopper Omocestus viridulus L. J Neurophysiol 72:2015–2025. Hedwig B (1996) A descending brain neuron elicits stridulation in the cricket Gryllus bimaculatus (de Geer). Naturwissenschaften 83:428–429. Heisenberg M (1995) Pattern recognition in insects. Curr Opin Neurobiol 5:475–481. Hirth F, Kammermeier L, Frei E, Walldorf U, et al. (2003) An urbilaterian origin of the tripartite brain: developmental genetic insights from Drosophila. Development 130:2365–2373.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

Huber F (1960) Untersuchungen über die Funktion des Zentralnervensystems und insbesondere des Gehirnes bei der Fortbewegung und der Lauterzeugung der Grillen. Z Vergl Physiol 44:60–132. Hughes G, Wiersma C (1960) Neuronal pathways and synaptic connexions in the abdominal cord of the crayfish. J Exp Biol 37:291–307. Hustert R, Pflüger H, Braunig P (1981) Distribution and specific central projections of mechanoreceptors in the thorax and proximal leg joints of locusts. Cell Tissue Res 216:97–111. Ilius M, Wolf R, Heisenberg M (1994) The central complex of Drosophila melanogaster is involved in flight control: studies on mutants and mosaics of the gene ellipsoid body open. J Neurogenet 9:189–206. Ito K, Shinomiya K, Ito M, Armstrong J, et al., Insect Brain Name Working Group (2014) A systematic nomenclature for the insect brain. Neuron 81:755–765. Iwano M, Hill E, Mori A, Mishima T, et al. (2010) Neurons associated with the flip–flop activity in the lateral accessory lobe and ventral protocerebrum of the silkworm moth brain. J Comp Neurol 18:366–388. Johnston R, Levine R (1996) Crawling motor patterns induced by pilocarpine in isolated larval nerve cords of Manduca sexta. J Neurophysiol 76:3178–3195. Johnston R, Levine R (2002) Thoracic leg motoneurons in the isolated CNS of adult Manduca produce patterned activity in response to pilocarpine, which is distinct from that produced in larvae. Invertebr Neurosci 4:175–192. Kammermeier L, Reichert H (2001) Common developmental genetic mechanisms for patterning invertebrate and vertebrate brains. Brain Res Bull 55:675–682. Kanzaki R, Arbas E, Hildebrand J (1991) Physiology and morphology of descending neurons in phermone-processing olfactory pathways in the male moth Manduca sexta. J Comp Physiol A 169:1–14. Kanzaki R, Minegishi R, Namiki S, Ando N (2013) Insect-machine hybrid system for understanding and evaluating sensory-motor control by sex pheromone in Bombyx mori. J Comp Physiol A 199:1037–1052. Kien J (1990a) Neuronal activity during spontaneous walking. I. Starting and stopping. Comp Biochem Physiol 95:607–621. Kien J (1990b) Neuronal activity during spontaneous walking. II. Correlation with stepping. Comp Biochem Physiol 95:623–638. Kien J, Altman J (1984) Descending interneurones from the brain and suboesophageal ganglia and their role in the control of locust behavior. J Insect Physiol 30:59–72. Kim Y, Spalovská-Valachová I, Cho K, Zitnanova I, et al. (2004) Corazonin receptor signaling in ecdysis initiation. Proc Natl Acad Sci USA 101:6704–6709. Kim Y, Zitnan D, Galizia C, Cho K, Adams M (2006) A command chemical triggers an innate behavior by sequential activation of multiple peptidergic ensembles. Curr Biol 16:1395–1407. Kupfermann I, Weiss KR (1978) The command neuron concept. Behav Brain Sci 1:3–10. Kupfermann I, Weiss KR (2001) Motor program selections in simple model systems. Curr Opin Neurobiol 11:673–677. Libersat F, Gal R (2013) What can parasitoid wasps teach us about decision-making in insect? J Exp Biol 216:47–55. Libersat F, Gal R (2014) Wasp voodoo rituals, venom-cocktails, and the zombification of cockroach hosts. Integr Comp Biol 54:129–142.

219

220

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Lichtneckert R, Reichert H (2005) Insights into the urbilaterian brain: conserved genetic patterning mechanisms in insect and vertebrate brain development. Heredity 94:465–477. Lüdke J, Lakes-Harlan R (2008) Regeneration of the tibia and somatotopy of regenerated hair sensilla in Schistocerca gregaria (Forskal). Arthropod Struct Dev 37:210–220. Marder E, Bucher D (2001) Central pattern generators and the control of rhythmic movements. Curr Biol 11:R986–R996. Marder E, Bucher D (2007) Understanding circuit dynamics using the stomatogastric nervous system of lobsters and crabs. Annu Rev Physiol 69:291–316. Marder E, Bucher D, Schluz DJ, Taylor AL (2005) Invertebrate central pattern generation moves along. Curr Biol 15:R685–R699. Martin J, Ernst R, Heisenberg M (1998) Mushroom bodies suppress locomotor activity in Drosophila melanogaster. Learn Memory 5:179–191. Martin-Pena A, Acebes A, Rodriguez J, Chevalier V, et al. (2014) Cell types and coincident synapses in the ellipsoid body of Drosophila. Eur J Neurosci 39:1586–1601. Matheson T (1997) Hindleg targeting during scratching in the locust. J Exp Biol 200:93–100. Matheson T (1998) Contralateral coordination and retargeting of limb movements during scratching in the locust. J Exp Biol 201:2021–2032. McVea D, Pearson K (2007) Contextual learning and obstacle memory in the walking cat. Integr Comp Biol 47:457–464. Mesce KA (2002) Metamodulation of the biogenic amines: second order modulation by steroid hormones and amine cocktails. Brain Behav Evol 60:339–349. Mesce KA, Crisp KM, Gilchrist LS (2001) Mixtures of octopamine and serotonin have nonadditive effects on CNS of the medicinal leech. J Neurophysiol 85:2039–2046. Mesce KA, Pierce-Shimomura J (2010) Shared strategies for behavioral switching: understanding how locomotor patterns are turned on and off. Front Behav Neurosci 4:49. Mishima T, Kanzaki R (1999) Coordination of flipflopping neural signals and head turning during pheromone-mediated walking in a male silkworm moth Bombyx mori. J Comp Physiol A 183:273–282. Mizunami M, Iwasaki M, Okada R, Nishikawa M (1998a) Topography of four classes of Kenyon cells in the mushroom bodies of the cockroach. J Comp Neurol 399:162–175. Mizunami M, Iwasaki M, Okada R, Nishikawa M (1998b) Topography of modular subunits in the mushroom bodies of the cockroach. J Comp Neurol 399:153–161. Mizunami M, Okada R, Li Y, Strausfeld N (1998c) Mushroom bodies of the cockroach: activity and identities of neurons recorded in freely moving animals. J Comp Neurol 402:501–519. Möhl B (1988) Short-term learning during flight control in Locusta migratoria. J Comp Physiol A 163:803–812. Möhl B (2000) Sensorimotor mechanism and learning in the locust flight system. In Studies in Cognitive Systems, Prerational Intelligence: Adaptive Behavior and Intelligent Systems Without Symbols and Logic. Ritter H, Cruse H, Dean J (eds). New York, NY: Springer Science. Moore E, Haspel G, Libersat F, Adams M (2006) Parasitoid wasp sting: a cocktail of GABA, taurine, and beta-alanine opens chloride channels for central synaptic block and transient paralysis of cockroach host. J Neurobiol 66:811–820.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

Namiki S, Iwabuchi S, PP K, Kanzaki R (2014) Information flow through neural circuits for pheromone orientation. Nat Commun 5:5919. Nargeot R, Simmers J (2012) Functional organization and adaptability of a decision-making network in Aplysia. Front Neurosci 6:113. doi: 10.3389/fnins.2012.00113. Neuser K, Triphan T, Mronz M, Poeck B, Strauss R (2008) Analysis of a spatial orientation memory in Drosophila. Nature 453:1244–1244. Newland P (1991) Morphology and somatotopic organisation of the central projections of afferents from tactile hairs on the hind leg of the locust. J Comp Neurol 312:493–508. Newland P, Rogers S, Gaaboub I, Matheson T (2000) Parallel somatotopic maps of gustatory and mechanosensory neurons in the central nervous system of an insect. J Comp Neurol 425:82–96. Nusbaum MP, Blitz DM, Swensen AM, Wood D, Marder E (2001) The roles of co-transmission in neural network modulation. Trends Neurosci 24:146–154. Nusbaum MP, Kristan Jr, W (1986) Swim initiation in the leech by serotonin–containing interneurones, cells 21 and 61. J Exp Biol 122:277–302. O’Gara B, Chae H, Latham L, Friesen W (1991) Modification of leech behavior patterns by reserpine-induced amine depletion. J Neurosci 11:96–110. Page K, Matheson T (2004) Wing hair sensilla underlying aimed hindleg scratching of the locust. J Exp Biol 207:2691–2703. Payne R (1998) Gradually escalating fights and displays: the cumulative assessment model. Anim Behav 56:651–662. Pflüger H (1984) The large fourth abdominal intersegmental interneuron: a new type of wind-sensitive ventral cord interneuron in locusts. J Comp Neurol 222:343–357. Pflüger H, Braunig P, Hustert R (1981) Distribution and specific central projections of mechanoreceptors in the thorax and proximal leg joints of locusts. II. The external mechanoreceptors: hair plates and tactile hairs. Cell Tissue Res 216:79–96. Pflüger H, Bräunig P, Hustert R (1988) The organization of mechanosensory neuropiles in locust thoracic ganglia. Philos T Roy Soc B 321:1–16. Pflüger H, Hurdelbrink S, Czjzek A, Burrows M (1994) Activity-dependent structural dynamics of insect sensory fibers. J Neurosci 14:6946–6955. Pick S, Strauss R (2005) Goal-driven behavioral adaptations in gap-climbing Drosophila. Curr Biol 15:1473–1478. Poeck B, Triphan T, Neuser K, Strauss R (2008) Locomotor control by the central complex in Drosophila—an analysis of the tay bridge mutant. Dev Neurobiol 68:1046–1058. Puhl JG, Masino MA, Mesce KA (2012) Necessary, sufficient and permissive: a single locomotor command neuron important for intersegmental coordination. J Neurosci 32:17646–17657. Puhl JG, Mesce KA (2008) Dopamine activates the motor pattern for crawling in the medicinal leech. J Neurosci 28:4192–4200. Puhl JG, Mesce KA (2010) Keeping it together: mechanisms of intersegmental coordination for a flexible locomotor behavior. J Neurosci 30:2372–2383. Ramirez J-M (1988) Interneurons in the suboesophageal ganglion of the locust associated with flight initiation. J Comp Physiol A 162:669–685. Rast G, Bräunig P (1997) Pilocarpine-induced motor rhythms in the isolated locust suboesophageal ganglion. J Exp Biol 200:2197–2207. Rast G, Bräunig P (2001) Insect mouthpart motor patterns: central circuits modified for highly derived appendages? Neuroscience 108:167–176.

221

222

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Reichert H, Simeone A (2001) Developmental genetic evidence for a monophyletic origin of the bilaterian brain. Philos T Roy Soc B 356:1533–1544. Ridge A, Alexander B, Ritzmann R (2007) Descending control of turning behavior in the cockroach, Blaberus discoidalis. J Comp Physiol A 193:385–402. Rillich J, Stevenson P, Pflüger H (2013) Flight and walking in locusts—cholinergic co-activation, temporal coupling and its modulation by biogenic amines. PLoS ONE 8:e62899. Rind F, Simmons P (1992) Orthopteran DCMD neuron: a reevaluation of responses to moving objects. I. Selective responses to approaching objects. J Neurophysiol 68:1654–1666. Rosenberg L, Glusman J, Libersat F (2007) Octopamine partially restores walking in hypokinetic cockroaches stung by the parasitoid wasp Ampulex compressa. J Exp Biol 210:4411–4417. Rosenberg L, Pflüger H, Wegener G, Libersat F (2006) Wasp venom injected into the prey’s brain modulates thoracic identified monoaminergic neurons. J Neurobiol 66:155–168. Ryckebusch S, Laurent G (1993) Rhythmic patterns evoked in locust leg motor neurons by the muscarinic agonist pilocarpine. J Neurophysiol 69:1583–1595. Ryckebusch S, Laurent G (1994) Interactions between segmental leg central pattern generators during fictive rhythms in the locust. J Neurophysiol 72:2771–2785. Schöneich S, Hedwig B (2011) Neural basis of singing in crickets: central pattern generation in abdominal ganglia. Naturwissenschaften 98:1069–1073. Schöneich S, Schildberger K, Stevenson P (2011) Neuronal organization of a fast-mediating cephalothoracic pathway for antennal-tactile information in the cricket (Gryllus bimaculatus DeGeer). J Comp Neurol 519:1677–1690. Schoofs A, Hückesfeld S, Schlegel P, Miroschnikow A, et al. (2014) Selection of motor programs for suppressing food intake and inducing locomotion in the Drosophila brain. PLoS Biol 12:e1001893. Selverston A (2010) Invertebrate central pattern generator circuits. Philos T Roy Soc B 365:2329–2345. Selverston AI, Russell DF, Miller JP, King DG (1976) The stomatogastric nervous system: structure and function of a small neural network. Prog Neurobiol 7:215–290. Simmons P (1980a) Connexions between a movement-detecting visual interneurone and flight motoneurones of a locust. J Exp Biol 86:87–98. Simmons P (1980b) A locust wind and ocellar brain neuron. J Exp Biol 85:281–294. Simmons P, Rind F (1992) Orthopteran DCMD neuron: a reevaluation of responses to moving objects. II. Critical cues for detecting approaching objects. J Neurophysiol 68:1667–1682. Sprecher S, Reichert H (2003) The urbilaterian brain: developmental insights into the evolutionary origin of the brain in insects and invertebrates. Arthropod Struct Dev 32:141–156. Staudacher E (2001) Sensory responses of descending brain neurons in the walking cricket, Gryllus bimaculatus. J Comp Physiol A 187:1–17. Staudacher E, Schildberger K (1998) Gating of sensory responses of descending brain neurones during walking in crickets. J Exp Biol 201:559–572. Stein P (2010) Alternation of agonists and antagonists during turtle hindlimb motor rhythms. Ann N Y Acad Sci 1198:105–118.

Motor Pattern Selection and Initiation in Invertebrates with an Emphasis on Insects

Stent GS, Thompson WJ, Calabrese RL (1979) Neural control of heartbeat in the leech and in some other invertebrates. Physiol Rev 59:101–136. Stevenson P, Rillich J (2012) The decision to fight or flee—insights into underlying mechanism in crickets. Front Neurosci 6:1–12. Strausfeld N, Hirth F (2013) Deep homology of arthropod central complex and vertebrate basal ganglia. Science 340:157–161. Strauss R (2002) The central complex and the genetic dissection of locomotor behavior. Curr Opin Neurobiol 12:633–638. Strauss R, Hanesch U, Kinkelin M, Wolf R, Heisenberg M (1992) No-bridge of Drosophila melanogaster: portrait of a structural brain mutant of the central complex. J Neurogenet 8:125–155. Strauss R, Heisenberg M (1993) A higher control center of locomotor behavior in the Drosophila brain. J Neurosci 13:1852–61. Suster ML, Bate M (2002) Embryonic assembly of a central pattern generator without sensory input. Nature 416:174–178. Vidal-Gadea A, Topper S, Young L, Crisp A, et al. (2011) Caenorhabditis elegans selects distinct crawling and swimming gaits via dopamine and serotonin. Proc Natl Acad Sci USA 108:17504–17509. Vierk R, Pflüger H, Duch C (2009) Differential effects of octopamine and tyramine on the central pattern generator for Manduca flight. J Comp Physiol A 195:265–277. von Philipsborn A, Liu T, Yu J, Masser C, et al. (2011) Neuronal control of Drosophila courtship song. Neuron 69:509–522. von Schlichet F, Hall J (1979) Neural topography of courtship song in sex mosaics of Drosophila melanogaster. J Comp Physiol A 129:85–95. Wada S, Kanzaki R (2005) Neural control mechanisms of the pheromone-triggered programmed behavior in male silkmoths revealed by double-labeling of descending interneurons and a motor neuron. J Comp Neurol 484:168–182. Wiersma CAG, Ikeda K (1964) Interneurons commanding swimmeret movements in the crayfish, Procambarus clarkia (Girard). Comp Biochem Physiol 12:509–525. Willard AL (1981) Effects of serotonin on the generation of the motor program for swimming by the medicinal leech. J Neurosci 1:936–944. Wilson DM (1961) The central nervous control of locust flight. J Exp Biol 38:471–490. Wilson DM (1966) Central nervous mechanisms for the generation of rhythmic behavior in arthropods. Symp Soc Exp Biol 20:199–228. Yellman C, Tao H, He B, Hirsh J (1997) Conserved and sexually dimorphic behavioral responses to biogenic amines in decapitated Drosophila. Proc Natl Acad Sci USA 94:4131–4136. Zawarzin A (1925) Der Parallelismus der Strukturen als ein Grundprinzip der Morphologie. Z wiss Zool 124:118–212.

223

225

8 Neural Networks for the Generation of Rhythmic Motor Behaviors Ronald M. Harris-Warrick 1 and Jan-Marino Ramirez 2 1

Department of Neurobiology and Behavior, Cornell University, Ithaca, NY, USA Department of Neurological Surgery, University of Washington School of Medicine and Center for Integrative Brain Research, Seattle Children’s Research Institute, University of Washington, Seattle, WA, USA 2

8.1 Introduction Rhythmic behaviors provide unique benefits for the study of the neural mechanisms underlying movement. By their very nature, rhythmic movements repeat in time, making it much easier to identify neurons whose activity is related to the behavior. As a result, our understanding of the neural networks driving locomotion, respiration, mastication, and rhythmic autonomic behaviors such as digestion are typically more advanced than for other, non-repetitive behaviors. In this chapter, we discuss these neural networks with a focus on general principles of function that may be applicable to a broad spectrum of rhythmic behaviors.

8.2 Concept of the Central Pattern Generator It was once thought (Sherrington 1906) that rhythmic behaviors require sensory feedback, with each step in the behavior eliciting the next step by a chain of reflexes. This model was disproved by experiments showing that deafferented cat spinal cords could generate the basic alternating rhythms of locomotion (Graham-Brown 1911) (Fig. 8.1). This led to the concept of the “Central Pattern Generator” (CPG), defined as a network of CNS neurons that organize and drive the timing, phasing, and intensity of rhythmic motorneuron activity without sensory feedback or rhythmic descending drive from other neural centers (Wilson 1961). It is generally demonstrated by evoking a “fictive” form of the basic rhythmic motor pattern in an isolated region of the nervous system (for example, isolated invertebrate nerve cord, mammalian spinal cord, or hindbrain slices (Smith et al. 1991; Zhong et al. 2010). CPGs have been found throughout the animal kingdom (Delcomyn 1980; Marder and Calabrese 1996), and are a foundational concept for understanding how rhythmic behaviors are generated. However, this does not mean that sensory feedback plays no role in rhythmic motor pattern generation. Indeed, in many systems, ranging from stick insect walking to mammalian locomotion, sensory feedback normally plays a decisive role in triggering some phase transitions in Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

226

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A Central Pattern Generator (CPG)

B Half Center Concept

F–E

+ F G

Flexor

+

=d g E =D E–F

Extensor

5s

Figure 8.1 The “central pattern generator” and “half center” concepts. Graham Brown demonstrated that cat locomotor activity does not require sensory feedback by studying locomotion after cutting the dorsal roots. (A) Illustrations from experiments of decerebrate cats walking on treadmills. (B) The first demonstration of rhythmic and alternating stepping movements in spinalized and deafferented cats. The recordings shows alternating movements of the tendons of an ankle flexor (upper trace) and an ankle extensor (lower trace), which inspired his proposal that a network of reciprocally inhibitory groups of neurons gives rise to alternating stepping movements (right). Drawing of half-center concept from Brown’s MD thesis (1912). All photographs and figures modified from Jones et al. (2011) with permission.

Neural Networks for the Generation of Rhythmic Motor Behaviors

the behavior (e.g., initiating limb swing to stance transitions) (Büschges and El Manira 1998; Pearson 2008; Büschges 2012). In addition, sensory feedback plays crucial roles in moment-to-moment motor pattern corrections to adjust to uneven or unpredictable surfaces (see Chapter 9). Understanding CPG organization and function requires answering several questions. First, what is the overall organization of the networks, including those that generate complex, multi-component behaviors? Second, what network and cellular mechanisms generate the basic rhythm of the motor pattern? Third, how do intrinsic and synaptic mechanisms interact to shape the patterns of muscle contractions for the behavior? Fourth, what mechanisms allow flexible and variable behaviors rather than robotic, invariant ones?

8.3 Overall Organization of Rhythmic Motor Networks Much has been learned by studying the organization of rhythmic motor networks in particularly experimentally advantageous invertebrate and vertebrate preparations. Among the best understood rhythm-generating networks are those that control crustacean foregut movements (Marder and Bucher 2007; Harris-Warrick 2014b), leech heartbeat (Lamb and Calabrese 2011), stick insect walking (Büschges 2012), locust flight (Ramirez and Pearson 1993), mammalian ventilation (Garcia et al. 2011; Ramirez et al. 2012), and swimming in lamprey (Grillner 2003), leech (Kristan et al. 2005), Tritonia (Sakurai and Katz 2009), tadpole (Combes et al. 2012), and zebrafish (Fetcho et al. 2008). The ability in these preparations to identify networks within a specific region or ganglion, and to link the network activity to the behavior even following isolation, has allowed definitive identification of neurons that participate in network function. This analysis has identified several organizational principles that govern CPGs in general. CPGs and Network Reconfiguration Numerous CPGs have been studied in isolation,

but the details of the underlying neural network structures are best understood in the numerically simpler invertebrate nervous systems. Most prominent is a rhythm generating network located in the stomatogastric ganglion (Marder and Bucher 2007; Harris-Warrick and Johnson 2009). This network is composed of only 26–30 neurons and generates the rhythmic foregut movements in Crustacea. The stomatogastric network can be configured to generate two distinct rhythmic activities, the pyloric and gastric rhythms, each of which can assume many forms and interact with each other in multiple ways (Fig. 8.2A). Although detailed network mechanisms are much less understood in mammals, from an organizational perspective there are many similarities. For example, the Pre-Bötzinger Complex (preBötC) in the brainstem contains a CPG that drives the mammalian inspiratory breathing rhythm (Smith et al. 1991; Ramirez et al. 2012). Much like the stomatogastric network, the preBötC reconfigures to give rise to different forms of inspiratory rhythms, such as sighing, normal inspiration, and gasping (Fig. 8.2B) (Lieske et al. 2000; Lieske and Ramirez 2006; Tryba et al. 2008). There is increasing evidence that this network is also involved in generating a variety of other rhythmic oral behaviors (Fig. 8.2C) (Moore et al. 2013), further supporting the notion that a core network can generate more than one rhythm.

227

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

Pyloric LP

AB

PY 5

IC

LG

GM 4

MG

Gastric

LPG 2

Gastro-pyloric PD 2

VD

Int 1

DG

AM

100%

B X

50%

XII SP5 NA O IO

PBC

0

preBötC 5s

Eupnea

Sigh

Gasp

Ambiguus (airway)

Tractus solitarius

Pre-Bötzinger

Hypoglossal IRt

W

h

is

Vibrissa IRt kin

g

Trigeminal IRt

Suc klin g

Bötzinger

Licking

Ventral respiratory group

Sw all ow

g in

Brea thin g

C

ing ew Ch

228

Neural Networks for the Generation of Rhythmic Motor Behaviors

CPGs and Neuromodulation While these “core CPGs” can generate rhythmicity and

reconfigure to generate different rhythmic outputs in isolation, they rarely function in isolation in the intact nervous system. An organizational principle is that all rhythm generating networks are controlled by, and probably dependent on, neuromodulatory inputs (Marder 2012). This has been well-documented in the stomatogastric ganglion (Harris-Warrick 2014b) and mammalian respiratory system (Tryba et al. 2006; Garcia et al. 2011) (Fig. 8.3). In many systems, neuromodulatory inputs are required for rhythm generation to occur at all. All known CPGs receive multiple neuromodulatory inputs from projection neurons in other areas of the nervous system (Fig. 8.3). CPG neurons themselves can also release neuromodulators onto other CPG neurons (intrinsic neuromodulation), thus shaping the behavior as it arises (Katz 1995). Coupling of Multiple CPGs and Left–Right Coordination Another organizational principle is

that complex behaviors require interactions between CPGs (see also Chapter 10). For example, it is often necessary to control movements that are symmetrical along an axis: examples include locomotion, respiration, and heartbeat. For these and similar motor patterns, a modular design has been proposed. To coordinate bilateral movements such as fish swimming, vertebrate locomotion, and leech heartbeat, the networks are bilaterally symmetric, with similar organization and network composition on the right and left sides of the nervous system. Commissural interneurons coordinate the activities of the two sides. The leech heartbeat CPG has several sets of neurons with crossed inhibitory synapses to contralateral homologs, ensuring the sides are normally active out of phase (Fig. 8.4A) (Kristan et al. 2005). In the rodent spinal cord, three commissural pathways regulate left–right hindlimb phasing to ensure alternation at low or high speeds, or synchronization when the alternating pathways are silenced (Crone et al. 2009; Talpalar et al. 2013; Shevtsova et al. 2015). In the lamprey swim CPG, commissural pathways again ensure right-left alternation (Fig. 8.4B) (Mentel et al. 2008). These symmetrical networks do not have to operate in a tightly coupled manner; in an extreme case, humans can hop with only one limb moving rhythmically. In the stick insect, locomotion can occur with very variable coordination of the six limbs, and single limbs can generate rhythmic movements in the absence of the others (Gabriel and Büschges 2007). Both intact and spinal-transected cats walking on split treadmills can adopt different left and right limbs frequencies, showing that the coupling between the

Figure 8.2 Reconfiguration of central pattern generators. (A) Schematic of the crab stomatogastric network, depicting the interactions of pyloric, gastric, and gastro-pyloric neurons in the configuration when the network generates the gastric and pyloric motor patterns. (B, C) Reconfiguration and integration in the pre-Bötzinger complex. (B) An isolated transverse slice of the preBötC can reconfigure to generate normal (eupneic) inspiration, sighing, and gasping. The activity maps represent the distribution of population activity during a given behavior; all are generated in the same area of the preBötC. [In this grayscale figure, 0% and 100% activity are both black. The white arrows in the bottom portion of the panel indicate the regions of the pre-Bötzinger complex that were strongly (close to 100%, central black spots) rhythmically active with the respiratory rhythm. The surrounding gray areas show decreasing rhythmic respiratory activity. The rest of the slice (black) showed no significant rhythmic activity in phase with the respiratory rhythm.] (C) The preBötC (middle) has been implicated in driving the rhythmic activities of a variety of premotor nuclei (light gray). The nucleus tractus solitarius provides resetting afferent input (dark gray). A redrawn from Nusbaum and Beenhakker (2002). B taken from Lieske et al. (2000), C from Moore et al. (2013), both with permission.

229

230

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A α1NE NK1 5-HT2 M3ACh P2Y1 TRH mGluRl Gq

Synaptic

Modulation βNE α2NE 5-HT1A 5-HT4 5-HT7 A1 D1 SST GABAB mGluRll mGluRlll Gs

Gi GLY-R

NMDA-R

Ca2+ AMPA-R

NALCN

Cl–

Intracellular

GABAA-R Cl–

Na+

Ca++

Na+

ATP

K+ KATP

TASK + TREK K TWIK

Na+

Na+ Ca2+

Na+ TRPC3/7 TRPM4/5 ICAN

INaP

Na+ K+ GIRK

Ih

K+ BK SK

Ca2+ T-Ca2+ P-Ca2+ L-Ca2+

Nav

Intrinsic

B

Neuromodulatory neurons

STG

ACh GABA NO DA HA Oct

AST-A (1) AST-B (8) Buc (?) CabTRP (2) CCK (?) FMRFamide-like (10) HIGSLYRamide (1) Myo (?) OMTR (1) Orc (9) Proc (1) Pyro (4) RYamide(2) SIFamide (2)

Motor neurons (LPG): Orc (?)

Sensory neurons ACh AST 5-HT

Figure 8.3 Rhythm generating networks are controlled by numerous converging neuromodulators. (A) The respiratory network is controlled by numerous neuromodulators that act on a variety of intrinsic and synaptic membrane properties. (B) The stomatogastric ganglion receives numerous modulatory inputs from a variety of neuromodulatory and sensory neurons. A modified from Ramirez et al. (2012) with permission. B prepared by D. Bucher, used with permission.

Neural Networks for the Generation of Rhythmic Motor Behaviors

A

HN

HN

1, 2

1, 2

3, 4

3, 4

5

5 silent

6, 7

6, 7

Descending reticulospinal inputs

B

Axial muscles

EIN

CIN

CIN

EIN

MN

IIN

IIN

MN

Axial muscles

Right

Left

C preMNs V4 NA MNs XII

XII

preMNs 100 μm

NA

NA

R O pre-BötC neurons

IO XII nerve

500 μm

231

232

Neurobiology of Motor Control: Fundamental Concepts and New Directions

two sides is flexible and this flexibility occurs in the spinal cord (Frigon et al. 2013). The CPG for vertebrate hindlimb locomotion appears to be continuously distributed along the six segments in the lumbar spinal cord. Each segment alone can generate a rhythmic output, showing the distributed nature of the network; however, the rostral segments have greater rhythmic capabilities than the caudal (Kjaerulff and Kiehn 1996). How the segments are coordinated is not yet well understood. Unlike the left–right alternation of locomotion, in breathing both sides are typically activated synchronously (alternating activation of the left and right diaphragm would be biomechanically detrimental). Perhaps not surprisingly, the respiratory CPG is bilaterally organized, and excitatory neurons essential for breathing project bilaterally to synchronize left and right activity (Fig. 8.4C) (Koizumi et al. 2008). The Half-Center Concept versus the Unit Burst Generator Concept These considerations lead to a

conceptually important question: to what extent does the basic CPG depend on an alternating organization? This question has historical significance, as Brown proposed not only the existence of CPGs, but also that the network generating alternating rhythmic locomotor activity was based on reciprocal inhibition between two groups of neurons (Fig. 8.1C). This concept of a “half-center oscillator” as the basis of rhythmicity has influenced our ideas about rhythm generation for more than a century. However, in both fish and mammals, rhythmic locomotor activity in the spinal cord continues in the isolated hemicord (and even hemisegment), and also after pharmacological blockade of inhibitory pathways (Cohen and Harris-Warrick 1984; Cowley and Schmidt 1995). In the respiratory system, the preBötC primarily generates a one-phase inspiratory rhythm. Most neurons within this network are active in phase with inspiration (Carroll and Ramirez 2013), and the inspiratory rhythm continues after blockade of synaptic inhibition, indicating that the network can generate the rhythm in the absence of another phase (Shao and Feldman 1997). Moreover, there is increasing evidence that one phase of expiration, “active expiration”, is generated in a different area of the medulla, the RTN/pFRG region (Janczewski and Feldman 2006; Pagliardini et al. 2011). These data are a major challenge to the half center concept, in which rhythmicity arises from reciprocal synaptic inhibition between different groups of neurons. An alternate concept is that of a unit burst generator, particularly relevant to the control of the multiple segments of a limb (Grillner 1981; Stein et al. 1997). Separate unit burst generators could be coordinated to control the rhythmic alternations of flexions and extensions around each limb joint. The coordination between unit burst generators would be Figure 8.4 Bilateral organization of neural networks. (A) Synaptic interactions among interneurons (HN) that control leech heartbeat. Note the network’s bilateral organization and the reciprocal inhibitory connections between right and left HN (3,4) interneurons. (B) The lamprey swimming system contains glycinergic commissural interneurons (CIN) organized in a reciprocal manner on the left and right sides of the spinal cord. Other interneurons include glutamatergic neurons (EINs) and ipsilaterally-projecting inhibitory L-interneurons (IINs). Arrows represent excitatory synapses, circles inhibitory synapses. (C) Inspiratory neurons thought to be involved in respiratory rhythm generation within the preBötC project bilaterally. Other neurons in the same transverse plane are pre-motorneurons (preMNs) that project from the preBötC to hypoglossal motorneurons (XII MNs). Additional abbreviations: XII, hypoglossus motor nucleus; NA, nucleus ambiguus; IO, inferior olive; XII nerve, hypoglossal nerve; V4, fourth ventricle. A redrawn from Roffman et al. (2012), B from Mental et al. (2008). C taken from Koizumi et al. (2008) with permission. See also Koizumi et al. (2013).

Neural Networks for the Generation of Rhythmic Motor Behaviors

mediated by neurons that are not essential for the generation of the rhythm, but instead essential for determining the phases of components of the movement. Consistent with this idea, for some rhythmic behaviors, e.g., breathing, the neural networks that generate different aspects of the rhythm are located in different parts of the nervous system. Joint and limb coordination would also depend on sensory proprioceptive interneurons, which could regulate the timing and coordination of the unit burst generators based on environmental and behavioral conditions. Multiple Layers of Network Organization Recent experimental and modeling work in cat

and rodent spinal cord suggests an additional level of organization in rhythmogenic locomotor networks. Rybak and colleagues have proposed a two-layer model for the locomotor CPG (Rybak et al. 2006; Zhong et al. 2012). A distributed high level rhythm-generating network coordinates the basic timing of the rhythm along the cord. This provides rhythmic drive to a set of networks in the pattern-forming layer, which coordinate the phasing and intensity of flexion/extension, left–right movements, and joint coordination within the limb. This concept arose from studies in cats (Lafreniere-Roula and McCrea 2005) and mice (Zhong et al. 2012) looking at the consequences of spontaneous deletions of components of the fictive locomotor pattern, where, for example, all the flexor motorneurons of a limb would cease firing while the extensors fired tonically. In most (cat) or virtually all (mouse) cases, when the flexors resumed firing, they did so an integer number of locomotor cycles later, and resumed normal bursting within a single cycle. These “non-resetting” deletions demonstrate that some components of the locomotor CPG continue cycling rhythmically when there is no motor output. Recordings from motorneurons show that during a non-resetting deletion, they lose all synaptic drive. One set of pre-motor V2a interneurons also lose their drive. However, other V2a interneurons and commissural interneurons continue to receive rhythmic synaptic drive, showing explicitly that the rhythm-generating layer continues to operate normally in the absence of output from the pattern-forming layer (Zhong et al. 2012). Most models of mammalian locomotion have posited the rhythm generator layer to be a symmetrical half-center, with mutually inhibitory rhythmogenic flexor and extensor components, so that the rhythmic drive to the flexor and extensor pattern-forming networks alternates. However, the effects of locomotor deletions were markedly asymmetrical (Pearson and Duysens 1976; Duysens and Pearson 1980; Zhong et al. 2012). When flexor output fell silent, extensor output was continuously tonic until flexor activity resumed. However, when extensor output fell silent, flexor output continued its normal rhythmic bursting without any apparent change. This asymmetry is explained by an asymmetric rhythm-generating layer, where only the flexor component is endogenously rhythmic. The extensor component is tonic, and drives the extensor pattern-forming layer to fire whenever it is not inhibited by the flexor networks. Other work supporting this idea comes from studies of flexor and extensor phasing during locomotion at different speeds (Halbertsma 1983). At relatively high speeds, flexor (swing) and extensor (stance) phases are of roughly equivalent duration. However, as animals slow down, flexor duration changes very little while extensor duration progressively lengthens. Although sensory feedback plays in important role in regulating this asymmetry in vivo, a very similar pattern is seen in vitro during fictive locomotion in the absence of all sensory feedback (Talpalar et al. 2013; Shevstova et al. 2015). This

233

234

Neurobiology of Motor Control: Fundamental Concepts and New Directions

again suggests that the extensor component fires actively whenever it is not being inhibited by the flexor component, while the flexor component drives its own rhythm. These ideas are consistent with work by Duysens and Pearson (Pearson and Duysens 1976; Duysens and Pearson 1980; Duysens 2006) who generated a “flexor burst generator” model with dominant flexor rhythmicity to explain deletions during cat locomotion.

8.4 Identification of CPG Neurons and Synapses: The “Wiring Diagram” We now describe the detailed mechanisms by which rhythmic behaviors are organized and generated. A critical first step is defining the “wiring diagram” of the network, identifying the component neuron types and the pattern of their synaptic interactions. Defining “the” wiring diagram for specific behaviors has become an important mission in mammalian nervous system research, known as defining the “Connectome” (Lichtman et al. 2008). The search for connectomes began many years ago in invertebrate networks. In the smaller invertebrate systems, with relatively small numbers of neurons in the ganglion or region where the CPG is located, this can be accomplished by repeated intracellular recordings, which have yielded virtually complete lists of the component neurons for several rhythmic motor CPGs (STG, Selverston 1976; leech heartbeat, Lamb and Calabrese 2011; cardiac ganglion, Cooke 2002; crayfish swimmeret system, Mulloney and Smarandache-Wellmann 2012). For most of these, the neurons’ synaptic connections have also been determined, primarily by paired electrophysiological recordings, giving a complete wiring diagram of the CPG that drives each behavior. As we describe below, a wiring diagram is not sufficient to explain how the behavior is generated: it is also essential to understand the intrinsic electrophysiological properties of the neurons, which integrate synaptic input and affect the decision to fire action potentials. The pyloric network in the stomatogastric ganglion is one of the best understood neural networks (Selverston et al. 1976; Marder and Bucher 2007; Harris-Warrick and Johnson 2009). In lobsters, it contains 14 neurons divided among six major classes (Fig. 8.5A). These neurons are highly interconnected, but with only inhibitory and electrical synapses. Multiple half-center-like reciprocal inhibitions are present, as is electrical coupling to synchronize activity. The network generates a rhythmic motor pattern with bursts of action potentials of characteristic duration and phasing among the different neuron classes (Fig. 8.5B). The half-centers could drive rhythmogenesis, but the system is typically pacemaker-driven. The AB neuron is conditionally rhythmic, and in the presence of appropriate neuromodulatory input rhythmically fires burst of action potentials. It is electrically coupled to the two PD neurons; these three neurons together act as the network pacemaker. These neurons inhibit and silence all the other neurons. At the end of the pacemaker bursts, the follower neurons recover by post-inhibitory rebound, but burst at different times depending on their expression of subthreshold currents such as the transient potassium current, IA (which retards post-inhibitory rebound) and the hyperpolarization-activated inward current, Ih (which accelerates post-inhibitory rebound) (Harris-Warrick et al. 1995b). Follower neurons phasing is also determined by the inhibitory connections among the neurons. The entire process is reset by the next AB-PD burst.

Neural Networks for the Generation of Rhythmic Motor Behaviors

A PY -8-

LP

AB

PD -2-

VD

IC

B AB

PD

LP

PY

0.2 sec

IC VD

Figure 8.5 The pyloric network of the lobster stomatogastric ganglion. (A) Wiring diagram of the six classes of pyloric neurons. Cholinergic inhibitory synapses, filled circles; glutamatergic inhibitory synapses, open circles; non-rectifying electrical coupling, resistors; rectifying electrical coupling, diodes. There are no excitatory chemical synapses. (B) Rhythmic activity of the Anterior Burster (AB), Pyloric Dilator (PD), Lateral Pyloric (LP), Pyloric Constrictor (PY), Inferior Cardiac (IC), and Ventricular Dilator (VD) neurons. Vertical scales 5 mV for AB neuron, 10 mV for others. A redrawn from Harris-Warrick and Johnson (2010).

235

236

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Optogenetic Approaches (see also Chapter 4) Identifying wiring diagrams in vertebrates is

much more difficult due to their large number of neurons and unknown number of neuron classes. New technologies, such as automated electron microscopic sectioning of small pieces of mammalian brains (Helmstaedter et al. 2013) and brainbow mice, in which genetic methods are used to make individual neurons fluoresce, show promise (Livet et al. 2007). Genetic tools have been invaluable in attacking this problem (Goulding and Pfaff 2005). For example, a specific class of glutamatergic neurons in the preBötC respiratory network expresses the transcription factor Dbx1. These neurons can be selectively and optogenetically activated to explore their activity and functional connectivity in the network. This approach has shown that Dbx1 neurons are essential for generation of the mammalian respiratory rhythm (Wang et al. 2014b). A similar approach has been used in other networks (Fenno et al. 2011; Williams and Deisseroth 2013). These take advantage of the temporal patterns of selective expression of transcription factors and other genes that together determine neural identity in development. The roles of these genetically defined neuron classes in network function can then be studied using a combination of methods. For example, electrophysiological and optical recordings of fluorescently tagged neurons during behavior can tell whether the neurons are active in temporal relationship to the behavior and thus might (although not necessarily do) participate in its generation. Optogenetic molecules such as channelrhodopsins, halorhodopsin, and archaerhodopsin can be selectively expressed in genetically defined neuron classes to temporally excite and inhibit the targeted neurons (Madisen et al. 2012). While viral injections have distinct advantages, transgenic mouse lines with neuron-specific expression of Cre recombinase that can be combined with specific reporter mice provide unique opportunities to specifically and reliably manipulate identified neurons using optogenetic approaches. Cre-dependent mouse lines for the manipulation of most neuromodulatory neurons and specific classes of interneurons and motorneurons are now available, leading to a revolution in our understanding of mammalian neural circuits (Madisen et al. 2012). Genetic Lesioning (see also Chapter 4) Another approach used to explore vertebrate

networks is to study the consequences of eliminating specific classes of neurons. Gene knockout experiments of key transcription factors have been used to alter the development of specific neuron classes (Lanuza et al. 2004; Zhang et al. 2008), but the neurons usually re-specify to other classes, rendering interpretation of the data difficult. Alternatively, neurons that express a critical developmental gene can be engineered to express a toxin such as diphtheria toxin, killing the neurons and removing them from the network (Crone et al. 2008, 2009; Talpalar et al. 2013). Lesioning approaches can also be combined with optogenetic methods to activate or silence specific sets of neurons during the behavior. These methods allow identification of neuron classes in several mammalian CPG networks. Progress toward identifying network synaptic connectivity is being made with anatomical double-labeling studies, trans-synaptic pathway tracing with rabies virus and other constructs, and paired electrophysiological recordings. These methods have identified components of the hindlimb locomotor CPG in the mouse spinal cord and determined their function (Fig. 8.6). The isolated neonatal spinal cord is usually used; when nonspecifically excited (typically by addition of NMDA and serotonin, or by sensory stimulation), the CPG generates a rhythmic pattern with

Commissural Interneurons (CIN)

Right

Left

CINe1-F V0V

CINe1-F V0V

Excitatory rhythmic population Excitatory population Inhibitory population Excitatory synaptic inputs

Right-Left Coordination INe-F V2a

Rhythm Generation (RG)

Pattern Formation (PF)

In-RG-F V2b; V1 RG-E

In-RG-E V2b; V1

PF-E V2a type II, V0C, ?

to left extensor MNs

In-PF-F V2b, V1 In-PF-E V2b; V1

CINe-F V3

CINe-F V3

CINi-F V0D

CINi-F V0D

Inhibitory synaptic inputs

INi-F

INi-F

RG-F ( ) Hb9, dI6, shox2

RG-F ( ) Hb9, dI6, shox2

PF-F V2a type II, V0C, ?

PF-F V2a type II, V0C, ?

to left flexor MNs

to right flexor MNs

INe-F V2a

In-RG-F V2b; V1 RG-E

In-RG-E V2b; V1

In-PF-F V2b, V1

In-PF-E V2b; V1

PF-E V2a type II, V0C, ?

to right extensor MNs

Figure 8.6 Model of the mouse spinal locomotor CPG. The networks are duplicated symmetrically on the left and right sides of the cord, and consist of a rhythm generator kernel and a pattern-formation kernel coupled by commissural neurons. In the rhythm generator kernel, the flexor-related group (“RG-F”) shows intrinsic oscillatory ability while the extensor-related group (“RG-E) fires tonically. The “In-RG-F” and “In-RG-E” neurons provide inhibitory feedback onto the RG-F group. See text for details. Redrawn from Zhong et al. (2012).

238

Neurobiology of Motor Control: Fundamental Concepts and New Directions

alternating left–right activity and alternating hindlimb flexor–extensor activity, generating a slow simplified fictive walk. Based on transcription factor expression, four major classes of ventral spinal interneurons have been identified, called V0–V3, each of which has subgroups based on additional gene expression or differential activity during locomotion (Kiehn 2006; Gosgnach 2011; Harris-Warrick 2014a). V0 interneurons are predominantly commissural, helping coordinate alternating activity on the left and right sides of the cord; the more dorsally derived group (V0D ) is inhibitory and ensures alternation at slow speeds while the more ventrally derived group (V0V ) is excitatory and ensures alternation at high speeds by driving contralateral inhibitory interneurons. V1 neurons are ipsilaterally projecting inhibitory neurons that include the Renshaw cells, which provide recurrent feedback to motorneurons, and Ia inhibitory interneurons that play a role in crossed inhibition between flexors and extensors. V2a neurons are excitatory, ipsilaterally projecting neurons divided into two groups, one synapsing onto motorneurons and the other onto V0V commissural interneurons. V2b neurons are inhibitory; together with a subgroup of V1 neurons, they mediate inhibition between ipsilateral flexor and extensor-related interneurons to assure alternation of flexion and extension during locomotion. V3 interneurons are predominantly excitatory commissural interneurons which may mediate left–right synchrony such as occurs in hopping. Three other groups, Hb9, dI6, and a subset of shox2-expressing interneurons, are thought to play roles in rhythm generation, but the details of their roles are not yet clear. Similar approaches have been used to specifically activate and silence respiratory neurons. This has led to a better understanding of the key structures involved in respiratory rhythm generation, the mechanisms underlying sensory modulation, and the identification of important neuromodulatory inputs.

8.5 Cellular Properties That Shape Network Output: Building Blocks for Network Operation An important general principle is that CPG output depends on an intricate interplay between the intrinsic cellular properties of the component neurons and the pattern of synaptic interactions. We first discuss how intrinsic properties play a major role in how neurons integrate their synaptic inputs and make decisions about whether to spike. Getting (1988) published a groundbreaking review of neural network function in which he listed the “building blocks” used to make a functional motor network. Work on these building blocks has occupied many motor physiologists since then, and these concepts can be applied to virtually all motor systems. This property allows neurons to fire rhythmic bursts of action potentials in the absence of any synaptic input. Originally studied in invertebrates such as Aplysia and Crustacea (Adams and Benson 1985; Hartline and Graubard 1992), oscillatory neurons are also present in vertebrate motor networks (Llinas and Sugimori 1980; Dekin et al. 1985; Ramirez et al. 2004; Kolta et al. 2007; Brocard et al. 2010). In some motor networks (especially those that cycle continuously throughout the animal’s life), the oscillatory property is intrinsic and unconditional: it is always “on” so the neuron cycles rhythmically at all times. In other networks the property

Rhythmic Bursting of Pacemaker Activity (Fig. 8.7)

Neural Networks for the Generation of Rhythmic Motor Behaviors

A

C

B

D

Figure 8.7 Pacemaker neurons in rhythm generating networks. (A) Respiratory network. (B) Lamprey spinal cord. (C) PD neuron in the stomatogastric ganglion. (D) Rodent spinal cord. A from Viemari et al. (2013), B from Wang et al. (2014a), C from Weimann and Marder (1994), D from Brocard et al. (2013), all with permission.

is conditional, activated by modulatory or other activating inputs, which may play a role in initiating the rhythmic behavior. It is generated by a cyclic interaction of inward and outward currents, described in detail below. It allows neurons to potentially act as pacemakers for rhythmic motor behaviors. When embedded in a neural network, pacemaker activity can be entrained by synaptic inputs. In this situation the intrinsically generated bursts act as non-linear amplifiers for these driver synaptic inputs, as has been demonstrated in several motor systems including locust flight and the mammalian respiratory network (Fig. 8.8) (Ramirez and Pearson 1991; Ramirez and Richter 1996). Plateau Potentials, or Bistability This property enables a neuron to have two stable states,

one silent and the other tonically active. A brief depolarization can step the neuron up from the silent state to the plateau, which sustains tonic action potentials; a brief inhibition can terminate the plateau and return the neuron to the silent, hyperpolarized state (Hultborn et al. 2013). This mechanism is also frequently conditional, appearing only in the presence of specific modulatory inputs. It allows neurons to produce sustained activity with no synaptic input. Extensor postural motorneurons show marked bistability, allowing them to maintain a standing posture until signaled to do otherwise (Bouhadfane et al. 2013). Bistability can also determine the onset phase of neurons in the motor rhythm, with synaptic inhibition triggering activity offset: this is an example of intrinsic and synaptic mechanisms working together to shape network output. The plateau depends on activation of persistent inward currents that remain active throughout the plateau, thus essentially adding to the baseline resting currents to generate a new “rest potential” that is above threshold for spiking. Post-Inhibitory Rebound Many cells show a brief depolarizing rebound immediately

after the end of synaptic inhibition. This is often mediated by subthreshold inward

239

240

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B

75 ms

C

75 ms

25 ms

bursting bursting synaptic synaptic

Figure 8.8 Intrinsic bursting can amplify synaptic input in rhythmically active networks. (A) Mammalian respiration. (B) Mammalian neocortex. (C) Locust flight. Light gray indicates synaptic potentials, darker gray intrinsic bursting mechanisms. Modified from Koch et al. (2011) with permission.

currents such as Ih , the persistent sodium current INaP , the T-type calcium current, or even deinactivation of voltage-activated sodium currents (Harris-Warrick et al. 1995b). Post-inhibitory rebound converts inhibitory inputs into a bimodal inhibitory–excitatory response in the neuron. In some cases, such as the stomatogastric networks, it is adequate to trigger a plateau potential in follower neurons (Fig. 8.5). This helps shape the timing and intensity of neuron firing after synaptic inhibition. Post-inhibitory rebound has also been implicated in the generation of rhythmic activity in the mammalian nervous system (e.g., post-inspiratory neurons in the preBötC) (Ramirez et al. 1997). Spike Frequency Adaptation and Delayed Onset of Firing When tonically excited (by synaptic

input or step depolarization), neurons can fire in many patterns, ranging from a single action potential to tonic high-frequency firing. Most neurons show initial high-frequency firing whose rate declines with time, called spike-frequency adaptation. Spike-frequency adaptation is evoked by decreases in tonic inward currents or slow accumulation of outward currents such as IKCa (Hartline and Graubard 1992). Because this property progressively weakens synaptic output from the neuron, the rate of adaptation can determine when follower neurons resume firing in the motor pattern. Other neurons have a pronounced transient potassium current (IA ), causing them to show a marked pause before starting to fire, called delayed excitation. This can help shape the precise phasing of onset of firing of neurons in the motor pattern.

8.6 Combined Neural Mechanisms for Rhythmogenesis Rhythm generation has evolved as a collaboration between the synaptic organization in the CPG and the intrinsic firing properties of the CPG neurons. Arguments have often arisen about which of these mechanisms underlies rhythmogenesis. On the one hand,

Neural Networks for the Generation of Rhythmic Motor Behaviors

individual neurons or neuron classes can acquire intrinsic rhythmic bursting capability, thus allowing them to serve as pacemakers for the rhythm, acting as the clock that drives the rhythmic pattern. On the other, synaptic interactions can maintain rhythmically alternating neural firing, acting as a “network pacemaker”; in this model, no individual neuron is intrinsically rhythmic, and rhythmicity alternatively arises through synaptic inhibition, synaptic fatigue, mutual excitation, and electrical coupling. As so often occurs in such “binary” debates, the answer is that both sides are right. Intrinsic pacemaker and network oscillator mechanisms do not operate in isolation (Harris-Warrick 2010; Garcia et al. 2011; Ramirez et al. 2012; Carroll and Ramirez 2013). Synaptic and intrinsic mechanisms can be separated experimentally only by removing these properties pharmacologically or genetically. This typically results in rudimentary rhythms that do not reflect the complexity and plasticity of the complete network in which integration of synaptic, intrinsic, and modulatory properties can dynamically occur. The main challenge of understanding rhythm generation is not to distinguish between pacemaker and network mechanisms of rhythmogenesis, but instead to unravel how the different components of rhythm generation cooperate to generate reliable, stable rhythms over a variety of different conditions.

8.7 Ionic Currents Shaping CPG Network Neuron Intrinsic Firing Properties The intrinsic properties of CPG neurons are shaped by their patterns of expression of ion current genes. A number of different ion currents shape CPG network function (Harris-Warrick 2002, 2010; Ramirez et al. 2012). Many of these currents have slow kinetic properties that allow them to support rhythmic activity by maintaining the neuron in a depolarized or hyperpolarized state for long times. Many are active at subthreshold voltages, allowing them to help initiate activity after synaptic inhibition. Differences in the kinetic and voltage dependence of the different currents imbue network neurons with different overall characteristics that may be advantageous for their specific roles in a given neural network. 8.7.1 Role of Outward Currents in Regulating Pacemaker and Network Activity

Potassium currents play a major role in terminating bursts, regulating phasing, and shaping the intensity of firing of CPG neurons. There are many classes of K+ channels, whose different properties determine how they shape neural activity. Transient Potassium Current, IA The transient potassium current is activated at subthresh-

old voltages, but is mostly inactivated at rest. It thus requires hyperpolarization (e.g., synaptic inhibition) to remove inactivation in order to be activated after the hyperpolarization ends. The channel opens only transiently, and thus plays a prominent role in regulating rebound rate and the timing of spiking onset in rhythmic motor patterns. In the lobster pyloric network follower neuron rates of recovery after inhibition by the pacemaker neurons are determined in part by the amount of IA they express (Harris-Warrick et al. 1995a,b). The Kv4 (invertebrate shal) and Kv1 (Shaker) genes encode IA .

241

242

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Delayed Rectifier Potassium Current, IK(V) This family of channels (encoded primarily by the

Kv2/shab and Kv3/shaw gene families) generates voltage-gated potassium currents that play major roles in terminating the action potential. The level of expression of IK(V) thus regulates spike width (and hence neurotransmitter release) and can contribute to the post-spike afterhyperpolarization that regulates spike frequency. The role of this current has not been carefully analyzed in motor networks. Calcium-Activated Potassium Current, IK(Ca) These channels are activated by increases in

intracellular calcium. The SK family is voltage-independent and purely activated by intracellular calcium, while the BK (Slo) family is dually regulated by intracellular calcium and voltage. They undergo significant alternative splicing, generating many variants which differ in their voltage dependence and kinetics. This allows them to play many roles in motor networks. In mammals, they are often the major currents driving spike afterhyperpolarization (AHP); in neonatal rodent spinal cord commissural interneurons, the AHP is predominantly due to SK currents (Diaz-Rios et al. 2007). As such, they help shape neuron spike frequency, and are a major target of neuromodulation to increase spike frequency in motorneurons and interneurons (Diaz-Rios et al. 2007). At a very different time scale, they are often the major currents terminating bursting in intrinsically bursting neurons. During the burst, voltage-sensitive calcium channels cause a slow accumulation of [Ca2+ ]in , which finally activates enough IK(Ca) to outweigh the tonic inward currents sustaining the burst. IK(Ca) can be localized at presynaptic nerve terminals where it regulates pre-synaptic voltage and thus transmitter release (Gola and Selverston 1981). Sodium-Activated Potassium Current, IK(Na) The Slack (Slo2) and Slick (Slo2.1) genes are related to the Slo family of IK(Ca) channels, but are instead activated by increases in intracellular sodium (Kaczmarek 2013). They have several splice variants with different kinetics: rapidly activating IKNa repolarizes neurons quickly, allowing them to fire at high frequencies (Markham et al. 2013), while slower-activating IKNa causes spike frequency adaptation. Wallen et al. (2007) showed that a Slack-like IKNa regulates lamprey spinal neuron spike AHPs in a frequency-dependent manner, contributing less than 20% of the AHP to single spikes but up to 50% during a spike burst. IKNa channels also regulate synaptic transmission (Nanou et al. 2008), limiting EPSP amplitude and shortening EPSP duration. These actions will contribute to determining synaptic strength and the intensity of neuron firing in motor networks. Leak Potassium Channels, IK2P A large family of leak potassium channels in the KCNK fam-

ily help to set the resting potential and membrane resistance, playing major roles in setting neuron excitability (Enyedi and Czirjak 2010). As described below, in the respiratory preBötC, one subset of bursting neurons oscillates due to a finely balanced interaction between a persistent sodium current and leak currents (Butera et al. 1999). In rat neonatal motorneurons, bradykinin enhances bistability and persistent firing by a mixed mechanism of conductance decrease of a TWIK-type leak potassium current and a conductance increase of a TRP-type calcium-activated nonselective current (ICAN , see below; Bouhadfane et al. 2015). As major components of the cell’s input resistance, these channels also regulate the amplitude of synaptic potentials, thus regulating transmission between network neurons.

Neural Networks for the Generation of Rhythmic Motor Behaviors

Pump Currents Pumps regulate intracellular ion concentrations. Many pumps are elec-

trogenic, typically generating an outward current due to a greater export than import of cations. As it has become apparent that intracellular and extracellular ion concentrations can change markedly during motor network activity (see below), studying the roles of pump currents has assumed greater importance. Del Negro and Hayes (2008) have proposed a “Group Pacemaker Hypothesis” for rhythm generation in the preBötC consisting of three steps: (1) glutamatergic synaptic activity preceding a burst leads to an increase in intracellular calcium; (2) this activates ICAN , which underlies the high frequency burst; (3) accumulation of sodium during the burst activates the Na,K-ATPase, whose hyperpolarizing pump current helps terminate the burst. Thus, pump currents can be quantitatively large enough to shape motor network activity. 8.7.2 Role of Inward Currents in the Generation of Pacemaker and Network Activity

As with outward currents, differences in inward current properties shape motor network activity. Due to the advances in genetic tools and electrophysiological approaches, much is known about the inward currents in mammalian motor networks. For example, bursting properties in the respiratory system were described as early as 1985 (Dekin et al. 1985; Smith et al. 1991), and two types of preBötC pacemaker neurons can be differentiated on the basis of whether their bursting depends on the persistent sodium current (INaP ) or the calcium activated non-selective cation current (ICAN ) (Thoby-Brisson and Ramirez 2001). In this section we discuss the roles of these and other inward currents in CPG function. Calcium-Activated Nonselective Current, ICAN This current is mediated by ion channels

belonging to the transient receptor potential (TRP) family (Partridge 1994; Montell 2005). This family has seven subfamilies, and which contribute to bursting and rhythm generation in different systems remains a matter of ongoing research. The channels are activated by a rise in intracellular calcium and typically have very slow kinetics, allowing them to sustain bursting and plateau activity. Because of its calcium dependence, the ICAN bursting mechanism is ideal for integrating synaptic inputs and action potentials that lead to calcium influx. Thus, ICAN is an important boosting mechanism to amplify synaptic drive (Rubin et al. 2009). In the respiratory preBötC, metabotropic glutamate receptors appear to activate calcium waves that in turn activate ICAN . However, ICAN can also be activated by calcium released from internal stores. During severe hypoxia, respiratory neurons whose bursts depend on this current are inactivated. This shut-down is one of the mechanisms that contributes to the reconfiguration of the respiratory network as it transitions into a gasping state (Lieske et al. 2000; Tryba et al. 2006), in which the network relies primarily on neurons whose bursts are driven by the persistent sodium current, INaP . Both mechanisms are active under normoxic conditions. ICAN plays a critical role in many other neural networks. In the crustacean pyloric network, dopamine evokes bursting by causing release of calcium from intracellular stores, leading to activation of ICAN which sustains the burst (Kadiri et al. 2011). Persistent Sodium Current, INaP The persistent sodium current is a component of the

voltage-activated sodium currents generated by the NaV gene family. In respiratory

243

244

Neurobiology of Motor Control: Fundamental Concepts and New Directions

neurons, INaP is probably generated by the Nav1.1, Nav1.2, and Nav1.6 channels (Ptak et al. 2005). The voltage-dependent characteristics of INaP have been modeled and used to explain how these properties can initiate and terminate bursts in respiratory neurons (Butera et al. 1999). Its slow inactivation can explain burst duration and termination, while its slow recovery from inactivation can explain the timing of burst onsets. As described above, these mechanisms are essential for rhythm generation under hypoxic conditions. INaP also plays a role in generating the locomotor rhythm in the rodent spinal cord (Brocard et al. 2013), where changes in extracellular ionic concentrations are important in its activation (see below). T-Type Calcium Current, ICaT The low voltage activated T-type calcium current (encoded by

the Cav3 gene family) has also been implicated in rhythm generation in many regions of the brain. Its contribution to neural bursting is best described in thalamocortical projection neurons (Luthi and McCormick 1998). Under normal resting conditions the current is inactivated and the projection neurons are tonically active. Hyperpolarization removes channel inactivation and initiates rhythmic bursting. The timing of this ICaT dependent bursting is determined by Ih , which activates at the neuron’s most hyperpolarized state, and depolarizes the neuron into the activation range of ICaT . This mechanism allows inhibition to switch the neuron from tonic to burst firing and plays an important functional role in relaying sensory information to the neocortex (Luthi and McCormick 1998). ICaT has been implicated in the generation of spinal locomotor activity (Anderson et al. 2012) and the respiratory rhythm (Rybak et al. 2014). Computational and experimental studies suggest that ICaT is important for generating the rebound burst in post-inspiratory neurons following their hyperpolarization during inspiration. ICaT also has a significant window current which may contribute to neural activity even at rest. L-Type Calcium Current, ICaL The L-type calcium channels are encoded by the Cav1 subfam-

ily. Because of its slow activation and inactivation kinetics, ICaL is an ionic mechanism for the generation and maintenance of plateau-potentials. In turtle and adult rodent postural extensor motorneurons, this current seems to be important in the generation of bistability leading to persistent firing (Hounsgaard and Kiehn 1989). L-type calcium currents are ubiquitous in nervous systems, and contribute to pacemaker generation in substantia nigra neurons (Ramirez et al. 2004). These neurons have been implicated in Parkinson’s disease, which has devastating motor and cognitive consequences. Hyperpolarization-Activated Inward Current, Ih The HCN family of genes encodes channels

whose voltage dependence is unusual, being activated by hyperpolarization below the resting potential rather than depolarization (He et al. 2014). These channels have variable kinetics and voltage dependence, but all show relatively slow activation and deactivation. They can be activated during synaptic inhibition, generating a rebound current that in some cases can lead to post-inhibitory rebound spiking. Such currents can shape the phasing of neural activity in rhythmic networks. In follower cells of the crustacean pyloric network, the rate of post-inhibitory rebound is determined by the ratio of the opposing actions of IA and Ih , both of which are activated following synaptic inhibition by the network pacemaker kernel, with some neurons firing earlier than others because they have more Ih and/or less IA (Harris-Warrick et al. 1995a,b). As described above,

Neural Networks for the Generation of Rhythmic Motor Behaviors

Ih helps shape inhibition-evoked bursting in thalamocortical neurons (Luthi and McCormick 1998). In the respiratory network, Ih can slow the respiratory rhythm, presumably by slowing pacemaker activation properties (Thoby-Brisson et al. 2000). NMDA Currents NMDA receptors likely play important roles in generating rhythmic

motor activity. Because activation is both glutamate- and voltage-dependent, NMDA receptors do not simply act as excitatory glutamate receptors. They interact with intrinsic and synaptic properties, which depolarize the neurons enough to remove the channel’s Mg2+ block, leading to voltage-dependent opening of NMDA-receptor channels. The NMDA receptor current can act as a resonance amplifying current (Hutcheon and Yarom 2000; Martell et al. 2012). Although intrinsic NMDA-dependent bursting has not been established for mammalian motor CPGs, it is very likely that NMDA receptors play an important role in locomotion (Talpalar and Kiehn 2010), cortical network oscillation (Martell et al. 2010, 2012), and respiration (Foutz et al. 1988). 8.7.3 Interaction of Inward and Outward Currents in the Generation of Pacemaker Activity

While different ionic currents endow neurons with different intrinsic firing properties, a single current never determines a neuron’s activity: it is the balance of inward and outward currents that establishes a neuron’s activity pattern (Harris-Warrick 2010). This balance is not static, but continuously changes in response to alterations in the modulatory milieu and synaptic inputs (Garcia et al. 2011). On slower time scales, changes in ionic currents also occur during development and in response to the behavioral state of the organism. These short- and long-term changes determine the relative ratio between the different ionic conductances, which differs for different neurons and shapes their physiological role in network activity. 8.7.4 Homeostatic Plasticity and the Balance between Different Ion Channel Types

The stomatogastric ganglion was the first model system to establish the concept of homeostatic plasticity in rhythm generating networks. When isolated from its neuromodulatory inputs, the pyloric network loses its ability to generate a motor rhythm. However, after 3 days of inactivity, sodium and calcium current densities increase resulting in a return of rhythmicity (Thoby-Brisson and Simmers 1998). There is now ample evidence that various intracellular mechanisms continuously fine tune inward and outward current balance (Golowasch et al. 1999; Marder and Prinz 2003; Zhang et al. 2009). Such homeostatic mechanisms also occur in mammals, where regulation of synaptic strength has been studied in detail. When excitatory or inhibitory synaptic function is blocked, neurons respond by upregulating the corresponding receptors to become more sensitive to the blocked transmitter (Desai et al. 1999; Maffei and Turrigiano 2008). An important cellular mechanism that seems to underlie this homeostatic plasticity is regulation of internal calcium levels (Turrigiano 2008), which involves a large number of different second messenger pathways, kinases, growth factors (such as BDNF), and inflammatory molecules (TNFalpha) (Turrigiano 2012). Homeostatic regulation is critical for physiological and pathophysiological

245

246

Neurobiology of Motor Control: Fundamental Concepts and New Directions

neural network functions (Koch et al. 2011). In the respiratory network, inactivity leads to dramatic increases in phrenic motor output, a homeostatic mechanism that involves a variety of intracellular mechanisms including TNFalpha activation (Broytman et al. 2013). Intermittent hypoxia, seen in obstructive sleep apnea, Rett syndrome, and Familial dysautonomia, results in long-term frequency modulation in the preBötC (Blitz and Ramirez 2002), increased gain of sensory receptors, and increased amplitude of phrenic motor output. Similarly, after spinal cord injury, spinal motorneurons become hyperexcitable due to the loss of descending input, in part mediated by increases in persistent sodium and calcium currents. This homeostatic response is one cause of spasticity, which is very common after spinal cord injury (Bennett et al. 2001; Harvey et al. 2005; Li et al. 2007; Murray et al. 2010). 8.7.5 Rapid Changes in Extracellular Ion Concentrations during Rhythmic Network Function

It was previously thought that extracellular ion concentrations remain relatively constant during normal neural activity. Recent work challenges this view: natural changes in intracellular and extracellular ion concentrations may play critical roles in triggering rhythmic behavior. The rhythmogenic persistent sodium current, INaP , is enhanced, and its voltage activation curve shifted to more negative values, by reductions in extracellular calcium. In the mammalian chewing network, neurons in the dorsal principle sensory trigeminal nucleus initiate rhythmic bursting upon reductions in extracellular calcium, associated with an increase in INaP , suggesting that normal neural activity may arise from a drop in extracellular calcium (Kolta et al. 2007, 2010). Brocard et al. (2013) verified this in spinal locomotor network: ion-selective electrodes measured a rapid decline in extracellular calcium and elevation of extracellular potassium upon activation of fictive locomotion either pharmacologically or by stimulation of descending inputs. During this time, about half of recorded unidentified interneurons switched from tonic to burst firing, associated with increased INaP amplitude and left-shifted voltage activation curves. A mathematical model suggested that the reduction in calcium and increase in potassium enhanced INaP activation sufficiently to induce bursting in these neurons. Whether this is a general mechanism for initiating rhythmic activity remains to be seen.

8.8 Role of Network Synaptic Properties in Organizing Rhythmic Behaviors Synaptic Inhibition Synaptic properties were the first building blocks implicated in the

generation of rhythmic activity. The original mechanism for rhythmogenesis was reciprocal synaptic inhibition between two functionally antagonistic neuron groups, each being tonically excited by descending synaptic drive (Brown 1912) (Fig. 8.1). Reciprocal inhibition is indeed present in almost every neural network studied so far (Koch et al. 2011). Nonetheless, CPGs based solely on synaptic mechanisms may not exist: as described above, intrinsic membrane properties are critical for phase transitions in all well-studied CPGs. For example, Ih activated during the hyperpolarized inhibitory phase can lead

Neural Networks for the Generation of Rhythmic Motor Behaviors

to sufficient depolarization to trigger “escape” from synaptic inhibition, thus giving the inhibitory synapse a delayed excitatory function. Phase transitions can similarly result from a “release” mechanism, where spike frequency adaptation and accumulating synaptic fatigue weaken a given neuron’s inhibition of the opposing neurons, eventually allowing them to begin firing again and thus induce a phase switch. These escape and release mechanisms of rhythmic phase transition have been extensively modeled (Skinner et al. 1994) and tested in experiments injecting artificial currents into neurons using the dynamic clamp approach (Sharp et al. 1996). This work has demonstrated that such mechanisms could indeed generate alternating activity. However, the network oscillations are not robust, and depend on good matches between time constants and amplitudes of the synaptic and intrinsic membrane properties. With small shifts in these properties, for example, the same network can generate stable synchrony between the two reciprocally organized groups of neurons. Experimentally, almost all rhythm-generating networks continue to generate rhythmic activity in the absence of synaptic inhibition (Cohen and Harris-Warrick 1984; Cowley and Schmidt 1995; Janczewski et al. 2013), suggesting that mutual inhibition is not essential for rhythmogenesis. However, synaptic inhibition certainly plays critical roles in shaping network phasing. In the respiratory network, synaptic inhibition often occurs concurrently with depolarizing input. This concurrent inhibition and excitation seems to be important for constraining the amplitude and shape of neural activity in the network, and synaptic inhibition controls the gain of expiratory activity (Zuperku and McCrimmon 2002). In the pyloric network, synaptic inhibition is critical for generating the different phases of pyloric activity. When the pyloric network is active, all the follower neurons fire tonically unless inhibited and show pronounced post-inhibitory rebound. Thus, as described above, the timing of onset and offset of each follower neuron class is determined by the interaction of the pattern of synaptic inhibition and each neuron’s different intrinsic rate of rebound (Harris-Warrick and Johnson 2009). Synaptic Excitation Synaptic excitation plays a critical role in synchronizing neuron

activity. In the respiratory network, inspiratory neuron activity is synchronized via glutamatergic connections. In the mammalian respiratory and locomotor networks, blocking glutamatergic excitation leads to rapid cessation of rhythmic activity. Respiratory glutamatergic connections are mediated by AMPA receptors, but other ionotropic and metabotropic glutamate receptors also help generate network activity. Post-synaptically, glutamate leads to a transient calcium influx. This influx, along with activation of metabotropic glutamate receptors, facilitates ICAN activation, which sustains the inspiratory burst as described above (Del Negro and Hayes 2008). In the rodent locomotor network, pharmacological or optogenetic blockade of glutamatergic excitation completely stops rhythmic activity. In mouse and zebrafish, optogenetic activation of glutamatergic interneurons is sufficient to initiate rhythmic locomotor activity (Hagglund et al. 2010; Ljunggren et al. 2014). These data suggest that excitatory glutamatergic interneurons are at the heart of the oscillator kernel. Talpalar and Kiehn (2010) found that both NMDA and non-NMDA receptors play critical roles in mouse fictive locomotion, but at different speeds: non-NMDA mechanisms predominate at higher speeds while the NMDA mechanisms are most important at lower speeds, and in conveying the signal to the motorneurons to maintain stability.

247

248

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Electrical Synapses and Gap Junctions Electrical synapses are found in many motor

networks. These ion channels can function uni- and/or bi-directionally (rectifying or non-rectifying junctions, respectively), and their strength is subject to neuromodulation. In the pyloric network, the three pacemaker neurons are electrically coupled and normally fire synchronously. This electrical coupling can regulate cycle frequency (Kepler et al. 1990). Dopamine enhances the intrinsic burst frequency of the pacemaker AB neuron, but inhibits and decreases the input resistance of the two electrically coupled PD neurons. The hyperpolarized PD neurons exert an electrical drag on the AB neuron, resulting in an unexpected slowing of cycle frequency (Ayali and Harris-Warrick 1999). In mammalian nervous systems, gap junctions are strongly expressed in immature networks; in neonatal rat spinal cord, gap junctions between motorneurons can sustain synchronized rhythmic membrane potential oscillations even in the presence of TTX (Tresch and Kiehn 2000). Limited electrical coupling persists into adulthood between motorneurons and between several sets of identified locomotory CPG interneurons. The functional role of this coupling is not well understood. Complex Multi-Component Synapses In both invertebrates and vertebrates, many synap-

tic connections have multiple components. Stimulation of a single pre-synaptic neuron thus can evoke a complex post-synaptic response, often with components on different time scales. In the stomatogastric ganglion, electrical coupling is often associated with chemical synaptic inhibition, leading to short-term excitation followed by a slower inhibition (Harris-Warrick and Johnson 2009). These components can be independently modulated; at several synapses, dopamine weakens electrical coupling and activates a previously silent chemical inhibition, switching the synaptic interaction from excitatory to inhibitory (Johnson et al. 1993). Mixed electrical/chemical synapses are common in neonatal spinal cord, but the electrical component often weakens with maturation. At many if not most synapses, slow metabotropic and fast ionotropic receptors are intermingled, producing a mixed and prolonged response that outlasts the period of synaptic stimulation. In the Tritonia swim circuit, serotonergic inputs activate both rapid ionotropic and slow metabotropic receptors, leading to rapid synaptic excitation during the stimulus train and a long-lasting slow excitation which, coupled with enhancement of other synaptic inputs, maintains network activity for seconds to minutes (Katz and Frost 1995). More complicated multi-component modulatory effects that enhance, reduce, or reset the strength of other synapses are also present (Sakurai and Katz 2009). Interaction of Synaptic and Intrinsic Mechanisms for Rhythmic Pattern Generation Synaptic and

intrinsic neural mechanisms cooperatively regulate the activation and termination of rhythmic network activity. As described above, the pyloric motor pattern depends on both the intrinsic properties of the component neurons (e.g., conditional bursting, different rates of post-inhibitory rebound) and on the pattern of synaptic inhibition, which shapes phase differences among the follower neurons (Harris-Warrick and Johnson 2009). In the locust flight CPG, intrinsic neural bursting properties differentially amplify synaptic inputs derived from proprioceptive inputs relative to input from other CPG neurons. During flight, bursting amplifies proprioceptive input from a wing sensory tegula, which triggers wing elevation; input from CPG neurons remains

Neural Networks for the Generation of Rhythmic Motor Behaviors

subthreshold (Fig. 8.8C). This amplification results in the timing of the phase switch from wing depression to wing elevation being mediated by the proprioceptive input rather than central input (Orchard et al. 1993). Bursting also depends on octopamine, which is released only during flight; this guarantees that tegula sensory input does not initiate wing elevation in quiescent animals (Pfluger et al. 2004). Similarly, in cat locomotion, although the isolated CPG can generate the entire locomotor rhythm, swing phase initiation in vivo is triggered by synaptic input from extensor-related proprioceptors (Pearson 2008). Similar interactive processes between intrinsic, neuromodulatory, and synaptic mechanisms are also present in mammalian nervous systems. Intrinsically bursting neurons amplify synaptic input in the neocortex (Schwindt and Crill 1999) (Fig. 8.8B), hippocampus, spinal cord, and respiratory network (Fig. 8.8A) (Koch et al. 2011). This interaction between synaptic and intrinsic properties can facilitate rapid neuronal synchronization and thus trigger a network burst, as seen in the locust flight system. Since intrinsic firing properties are regulated by neuromodulators, the interaction between synaptic and intrinsic bursting properties can transform inactive, passive networks into rhythmically active ones. In locomotor systems the endogenous release of aminergic and peptidergic modulators can enable an inactive spinal network to respond to glutamatergic inputs dynamically and generate rhythmic stepping movements (Harris-Warrick 2014a).

8.9 Variable Output from Motor Networks As described above, the properties of neural network components are not fixed, but are instead highly variable so as to allow the behavior to adapt to the situation at hand. Every facet of neural network function can be changed as needed. Thus, even when the complete wiring diagram of a network is determined, it should be considered only a library of potential components with potential properties and connections which modulatory inputs, descending commands, and sensory feedback can reconfigure as needed. The mechanisms underlying network reconfiguration have been reviewed many times (Harris-Warrick 1988, 2011, 2014b; Doi and Ramirez 2008; Stein 2009; Marder 2012; Ramirez et al. 2012) and are therefore only briefly discussed below. Variable Neural Composition of CPG Networks Just because a neuron has synaptic interac-

tions that would allow it to participate in a CPG does not mean it will do so: CPG neural composition and synaptic wiring can vary continuously to generate flexible motor output. In the pyloric network with its six classes of neurons, variable subsets ranging from just two to all six can be active depending on network state. In part, this arises from variations in network modulatory inputs. These shape network activity by altering CPG neuron excitability and intrinsic properties. Peptides select which neurons are active by neuron-selective expression of their receptors, all of which activate a similar rectifying inward current. Monoamines have more complex mechanisms, acting on many different receptors, which modify the properties of many different ion channels to shape each neuron’s activity and thus network output (Harris-Warrick 2014b). Similarly, different motorneurons and interneurons are active at different speeds in the zebrafish swim CPG (Fetcho et al. 2008; McLean et al. 2008). The V2a interneurons are activated in a ventro-dorsal wave; more ventral neurons fire at low speeds but are

249

250

Neurobiology of Motor Control: Fundamental Concepts and New Directions

actively inhibited at higher speeds, and progressively more dorsal V2a interneurons are recruited as swim speed increases. A similar pattern of motorneuron recruitment is also seen. In part this reflects differences in neuron input resistance. The higher resistance ventral cells, with their longer membrane time constants over which to integrate synaptic input, are activated by the weaker descending drive during slow swimming. The lower resistance dorsal cells, with their shorter time constants, are active only during faster, stronger swimming (McLean et al. 2008; Wang and McLean 2014). Changes in CPG active synaptic pathways can also occur as a function of locomotor speed. In rodents a basic left–right alternating motor pattern is seen at almost all locomotor speeds; there is no gait change with acceleration, as occurs in cats and horses. This alternation is sustained by a dual commissural pathway, with direct inhibition of contralateral target neurons mediated by the inhibitory V0D interneurons, and indirect inhibition mediated by the excitatory V0V interneurons, which synapse onto contralateral inhibitory interneurons (Kiehn 2006; Quinlan and Kiehn 2007). The V0V interneurons are driven ipsilaterally by a subset of V2a interneurons (Crone et al. 2008). Interestingly, selective ablation of the V2a interneurons (Crone et al. 2009), or the V0V interneurons they drive (Talpalar et al. 2013), causes gait changes to occur as a function of speed: neither ablation substantially alters left–right alternation at slow locomotor speeds, but at intermediate speeds alternation is disturbed and at high speeds is replaced by a bounding gait characterized by left–right synchrony of forelimbs and of hindlimbs. A similar gait change is seen during drug or sensory-evoked fictive locomotion in isolated mutant spinal cords. This result suggests that the V2a/V0V pathway drives normal left–right alternation only at high speeds. Consistent with this, V2a interneurons are relatively quiet at low speeds and are increasingly recruited and fire more action potentials per cycle as speed increases (Zhong et al. 2011). In contrast, selective ablation of the direct inhibitory V0D pathway results in synchronized left–right activity at low speeds, but normal alternation at high speeds. Finally, ablation of all V0 interneurons results in synchronous bounding at all speeds (Talpalar et al. 2013). Shevtsova et al. (2015) have modeled this system as an interaction between a direct excitatory commissural pathway (as could be mediated by the V3 interneurons, and is uncovered by glycine/GABA inhibition) and the direct V0D and indirect V2a/V0V inhibitory pathways. Deletion of either inhibitory pathway duplicated the experimental results. Thus, although both pathways have been known for many years, it appears that they are active in largely non-overlapping speed domains, and thus that the composition of the network changes with locomotor speed. Variable Intrinsic Properties of Network Neurons and Their Synapses Behavioral flexibility often

results from changes in the firing properties of network neurons. Many behaviors are episodic and need to be turned on and off, or the speed changed, and this can be generated by changing neuron firing properties. As described above, in the pyloric network this can arise from modulatory inputs that alter the activity of ionic currents to activate rhythmic bursting in the conditional pacemaker neuron (Nusbaum and Blitz 2012; Harris-Warrick 2014b). Recruitment of neurons can also occur by modulatory excitation. In the vertebrate spinal cord, descending serotonergic activity increases when locomotion begins and when it accelerates. Serotonin enhances the excitability of some locomotor CPG interneurons and motorneurons by reducing IKCa or increasing slow inward currents, allowing them to fire more action potentials per burst (Zhong et al. 2005; Diaz-Rios et al. 2007; Abbinanti and Harris-Warrick 2012).

Neural Networks for the Generation of Rhythmic Motor Behaviors

Changes in network phasing underlie such behavioral changes as gait changes. This can arise from differential activation of subthreshold ionic currents that determine firing onset or offset in the cycle. Dopamine markedly phase advances follower neuron activity in the pyloric network, in part by reducing IA (Harris-Warrick et al. 1995a,b). In this system, multiple neuromodulators can evoke rhythmic bursting in the conditional pacemaker AB neuron, but they do so by different mechanisms. For example, dopamine acts primarily by the release of calcium from intracellular stores activating ICAN , while serotonin acts by enhancing INaP and decreasing potassium leak (Kadiri et al. 2011). In the mammalian preBötC, activating different 5-HT receptors, noradrenergic receptor subtypes, or the NK1 receptor by substance P, induces bursting in different pacemaker neuron types (Pena and Ramirez 2002; Viemari and Ramirez 2006; Doi and Ramirez 2008; Ramirez et al. 2012). Motor network synaptic strengths can be similarly gated up and down by modulatory and sensory inputs, resulting in changes in timing and phasing of the motor activity. Dopamine changes the strength of every synapse in the pyloric circuit, in many different ways (Johnson et al. 1995). Some synapses are strengthened, while others are weakened. Some synapses are silent at rest and activated by dopamine, while others are active at rest but silenced by dopamine. These modulations arise from a combination of pre- and post-synaptic actions. Dopamine can have opposing effects at these synapses, for example enhancing pre-synaptic transmitter release but suppressing post-synaptic response to the transmitter at the same synapse (Harris-Warrick and Johnson 2010). These opposing actions may function to limit the extent of synaptic modulation to a certain range in which the network is still functional, preventing over-modulation to a non-functional state. Changes during CPG Development During development, including in many species a period

after birth, motor networks mature slowly, changing their properties and their outputs. In the stomatogastric ganglion, early in development two normally separate networks are fused into a single conjoint network that separates only later (Fenelon et al. 2003). In the rodent spinal cord, the locomotor CPG begins to assemble about a week before birth, with spontaneous motorneuron bursts that are synchronized both across the cord and between flexors and extensors (Vinay et al. 2002). At this time, future inhibitory synapses are depolarizing and excitatory, due to low levels of expression of the KCC2 chloride/potassium transporter, which results in ECl being set well above rest potential. Near birth, KCC2 is upregulated, pumping Cl out of the neuron so that GABA and glycine synapses assume their normal hyperpolarizing form. Contemporaneous with this change, the CPG begins to generate alternating flexor–extensor and left–right activity patterns (Vinay and Jean-Xavier 2008; Viemari et al. 2011). This late switch in sign of GABA and glycine synapses presumably occurs to allow Hebbian synapse refinement at the earlier developmental stages. Interestingly, after spinal cord injury, KCC2 is again down-regulated, weakening inhibitory synapses and contributing to the motorneuron hyperactivity that results in spasticity (Boulenguez et al. 2010). In parallel with these developmental changes in synaptic function, the intrinsic properties of CPG neurons also change markedly in the first several weeks after birth (Gao and Ziskind-Conhaim 1998). Action potentials narrow and their afterhyperpolarization accelerates, allowing neurons to fire at higher frequencies. Spike threshold often hyperpolarizes. Several classes of interneurons develop bistability and the ability to fire

251

252

Neurobiology of Motor Control: Fundamental Concepts and New Directions

repetitively in the presence of neuromodulators, both not possible at birth (Abbinanti et al. 2012; Husch et al. 2015). These changes make the neurons more excitable and able to generate a motor pattern. Rat motorneurons show bistability at birth, but, due to the involvement of a thermoregulated TRP-type ICAN , only at elevated temperatures (Bouhadfane et al. 2013). Newborn pups typically do not express this capability because they are born with inefficient temperature regulation. Development of thermoregulation and an erect posture correlate well, suggesting that temperature-dependent regulation of motorneuron bistability is one barrier to neonatal locomotion. These data show that CPG properties undergo postnatal developmental changes, and may follow somewhat different rules in the adult than the neonate.

8.10 Conclusions Understanding how neural networks generate behavior, perception, and consciousness is a fundamental goal of neuroscience. Understanding of rhythmic motor networks is far in advance of sensory or cognitive networks for a number of reasons. The network output is known: a pattern of motorneuron activity and muscle contractions that can be directly measured. The behavior is cyclic, greatly simplifying identification of the neurons that participate in the network. Advances in genetic and optogenetic manipulation, calcium and voltage-dependent dye imaging, tract tracing, large array recording, and mathematical modeling coupled with traditional anatomical and electrophysiological approaches are accelerating progress in vertebrate networks. Motor networks have provided valuable general insights into network function in general and will continue to do so in the future. Despite these advances, our current insights also reveal that we will face major challenges in understanding neural network function. The stomatogastric ganglion has long served as a model for rhythm generation due its small neuron number, well-defined motor function, and known synaptic connectivity and intrinsic membrane properties. A major conclusion from this work is that rhythm generation does not result from simple clock-like pacemaker processes, and that the structure of a neural network provides only a first, limited insight into its operation. Neurons and networks are tuned on an individual animal basis and are extremely plastic. Networks can assume multiple states, and no single rhythm generating mechanism can explain all of a network’s outputs. We began with the expectation that a complete description of the ganglion’s connectome would allow us to understand its activity. This expectation was wrong: virtually identical motor outputs can be generated in multiple ways (Prinz et al. 2004). This is an important lesson that needs to reach investigators studying mammalian neural networks. All work to date indicates that mammalian motor networks are every bit as complex as invertebrate ones. As such, it is almost certain that the neuron conductances of mammalian neural networks in general are individually and homeostatically regulated, and that the networks rely on multiple rhythm generating mechanisms, contain large neural ensembles that are sparsely connected and dynamically regulated, and interact not only with a staggering number of neuromodulators but also with glial cells. Describing a network’s connectome is thus only the rudimentary first step towards understanding how a behavior or consciousness is produced. A connectome is

Neural Networks for the Generation of Rhythmic Motor Behaviors

only predictive when its multiple membrane and ionotropic and metabotropic synaptic conductances are added, when the strengths of these conductances are homeostatically fine-tuned, and when it is embedded in its network of neuromodulators. The ball and stick diagrams used to explain interactions between identified invertebrate neurons are inadequate to describe these highly complex networks, and new computational approaches will be required to describe the dynamic network interactions that underlie their rhythm generation. This is clearly a daunting task. But there are reasons for optimism. Many of these processes have been described in smaller invertebrate networks, and we have numerous examples that show that lessons learned in smaller networks can be applied to more complex ones. With advances in optogenetic targeting of identified neurons, we believe it will be possible to understand how CPG networks generate rhythmic behavior in all animal species.

Acknowledgements This work was supported by NIH grants NS17323 and NS081713 (RH-W), and by HL107084 and HL090554 (J-MR). We thank Dirk Bucher for preparing Fig. 8.3B

References Abbinanti MD, Harris-Warrick RM (2012) Serotonin modulates multiple calcium current subtypes in commissural interneurons of the neonatal mouse. J Neurophysiol 107:2212–2219. Abbinanti MD, Zhong G, Harris-Warrick RM (2012) Postnatal emergence of serotonininduced plateau potentials in commissural interneurons of the mouse spinal cord. J Neurophysiol 108:2191–2202. Adams WB, Benson JA (1985) The generation and modulation of endogenous rhythmicity in the Aplysia bursting pacemaker neurone R15. Prog Biophys Mol Biol 46:1–49. Anderson TM, Abbinanti MD, Peck JH, Gilmour M, et al. (2012) Low-threshold calcium currents contribute to locomotor-like activity in neonatal mice. J Neurophysiol 107:103–113. Ayali A, Harris-Warrick RM (1999) Monoamine control of the pacemaker kernel and cycle frequency in the lobster pyloric network. J Neurosci 19:6712–6722. Bennett DJ, Li Y, Siu M (2001) Plateau potentials in sacrocaudal motorneurons of chronic spinal rats, recorded in vitro. J Neurophysiol 86:1955–1971. Blitz DM, Ramirez JM (2002) Long-term modulation of respiratory network activity following anoxia in vitro. J Neurophysiol 87:2964–2971. Bouhadfane M, Kaszas A, Rozsa B, Harris-Warrick RM, et al. (2015) Sensitization of neonatal rat lumbar motoneurons by the inflammatory pain mediator bradykinin. eLife DOI: 10.7554/eLife.06195. Bouhadfane M, Tazerart S, Moqrich A, Vinay L, Brocard F (2013) Sodium-mediated plateau potentials in lumbar motorneurons of neonatal rats. J Neurosci 33:15626–15641. Boulenguez P, Liabeuf S, Bos R, Bras H, et al. (2010) Down-regulation of the potassium–chloride cotransporter KCC2 contributes to spasticity after spinal cord injury. Nat Med 16:302–307.

253

254

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Brocard F, Shevtsova NA, Bouhadfane M, Tazerart S, et al. (2013) Activity-dependent changes in extracellular Ca2+ and K+ reveal pacemakers in the spinal locomotor-related network. Neuron 77:1047–1054. Brocard F, Tazerart S, Vinay L (2010) Do pacemakers drive the central pattern generator for locomotion in mammals? Neuroscientist 16:139–155. Brown TG (1912) Rhythmic Movements: A Contribution to the Study of the Central Nervous System. Edinburgh, Thesis, University of Ediburgh. Broytman O, Baertsch NA, Baker-Herman TL (2013) Spinal TNF is necessary for inactivity-induced phrenic motor facilitation. J Physiol 591:5585–5598. Büschges A (2012) Lessons for circuit function from large insects: towards understanding the neural basis of motor flexibility. Curr Opin Neurobiol 22:602–608. Büschges A, El Manira A (1998) Sensory pathways and their modulation in the control of locomotion. Curr Opin Neurobiol 8:733–739. Butera R, Rinzel J, Smith JC (1999) Models of respiratory rhythm in the pre-Bötzinger complex. I. Bursting pacemaker neurons. J Neurophysiol 81:382–397. Carroll MS, Ramirez JM (2013) Cycle-by-cycle assembly of respiratory network activity is dynamic and stochastic. J Neurophysiol 109:296–305. Cohen AH, Harris-Warrick RM (1984) Strychnine eliminates alternating motor output during fictive locomotion in the lamprey. Brain Res 293:164–167. Combes D, Sillar KT, Simmers J (2012) A switch in aminergic modulation of locomotor CPG output during amphibian metamorphosis. Front Biosci (Schol Ed) 4:1364–1374. Cooke IM (2002) Reliable, responsive pacemaking and pattern generation with minimal cell numbers: the crustacean cardiac ganglion. Biol Bull 202:108–136. Cowley KC, Schmidt BJ (1995) Effects of inhibitory amino acid antagonists on reciprocal inhibitory interactions during rhythmic motor activity in the in vitro neonatal rat spinal cord. J Neurophysiol 74:1109–1117. Crone SA, Quinlan KA, Zagoraiou L, Droho S, et al. (2008) Genetic ablation of V2a ipsilateral interneurons disrupts left–right locomotor coordination in mammalian spinal cord. Neuron 60:70–83. Crone SA, Zhong G, Harris-Warrick R, Sharma K (2009) In mice lacking V2a interneurons, gait depends on speed of locomotion. J Neurosci 29:7098–7109. Dekin MS, Richerson GB, Getting PA (1985) Thyrotropin-releasing hormone induces rhythmic bursting in neurons of the nucleus tractus solitarius. Science 229:67–69. Del Negro CA, Hayes JA (2008) A “group pacemaker” mechanism for respiratory rhythm generation. J Physiol 586:2245–2246. Delcomyn F (1980) Neural basis of rhythmic behavior in animals. Science 210:492–498. Desai NS, Rutherford LC, Turrigiano GG (1999) Plasticity in the intrinsic excitability of cortical pyramidal neurons. Nat Neurosci 2:515–520. Diaz-Rios M, Dombeck DA, Webb WW, Harris-Warrick RM (2007) Serotonin modulates dendritic calcium influx in commissural interneurons in the mouse spinal locomotor network. J Neurophysiol 98:2157–2167. Doi A, Ramirez JM (2008) Neuromodulation and the orchestration of the respiratory rhythm. Respir Physiol Neurobiol 164:96–104. Duysens J (2006) How deletions in a model could help explain deletions in the laboratory. J Neurophysiol 95:562–563; author reply 563–565. Duysens J, Pearson KG (1980) Inhibition of flexor burst generation by loading ankle extensor muscles in walking cats. Brain Res 187:321–332.

Neural Networks for the Generation of Rhythmic Motor Behaviors

Enyedi P, Czirjak G (2010) Molecular background of leak K+ currents: two-pore domain potassium channels. Physiol Rev 90:559–605. Fenelon V, Le Feuvre Y, Bem T, Meyrand P (2003) Maturation of rhythmic neural network: role of central modulatory inputs. J Physiol Paris 97:59–68. Fenno L, Yizhar O, Deisseroth K (2011) The development and application of optogenetics. Annu Rev Neurosci 34:389–412. Fetcho JR, Higashijima S, McLean DL (2008) Zebrafish and motor control over the last decade. Brain Res Rev 57:86–93. Foutz AS, Champagnat J, Denavit-Saubie M (1988) Respiratory effects of the N-methyl-D-aspartate (NMDA) antagonist, MK-801, in intact and vagotomized chronic cats. Eur J Pharmacol 154:179–184. Frigon A, Hurteau MF, Thibaudier Y, Leblond H, et al. (2013) Split-belt walking alters the relationship between locomotor phases and cycle duration across speeds in intact and chronic spinalized adult cats. J Neurosci 33:8559–8566. Gabriel JP, Büschges A (2007) Control of stepping velocity in a single insect leg during walking. Philos T Roy Soc A 365:251–271. Gao BX, Ziskind-Conhaim L (1998) Development of ionic currents underlying changes in action potential waveforms in rat spinal motorneurons. J Neurophysiol 80:3047–3061. Garcia AJ, 3rd, Zanella S, Koch H, Doi A, Ramirez JM (2011) Chapter 3—networks within networks: the neural control of breathing. Prog Brain Res 188:31–50. Getting P (1988) Comparative analysis of invertebrate central pattern generators. In Neural Control of Rhythmic Movements. Cohen AH, Rosignol S, Grillner S (eds). New York, NY: Wiley. Gola M, Selverston A (1981) Ionic requirements for bursting activity in lobster stomatogastric neurons. J Comp Physiol 145:191–207. Golowasch J, Abbott LF, Marder E (1999) Activity-dependent regulation of potassium currents in an identified neuron of the stomatogastric ganglion of the crab Cancer borealis. J Neurosci 19:RC33 31–35. Gosgnach S (2011) The role of genetically-defined interneurons in generating the mammalian locomotor rhythm. Integr Comp Biol. 51:903–912. Goulding M, Pfaff SL (2005) Development of circuits that generate simple rhythmic behaviors in vertebrates. Curr Opin Neurobiol 15:14–20. Graham–Brown T (1911) The intrinsic factors in the act of progression in the mammals. Proc R Soc Lond B Biol Sci 84:308–319. Grillner S (1981) Control of locomotion in bipeds, tetrapods and fish. In Handbook of Physiology, Section 1: The Nervous System. Brooks VB, ed. Bethesda, American Physiological Society. pp. 1179–1236. Grillner S (2003) The motor infrastructure: from ion channels to neural networks. Nat Rev Neurosci 4:573–586. Hagglund M, Borgius L, Dougherty KJ, Kiehn O (2010) Activation of groups of excitatory neurons in the mammalian spinal cord or hindbrain evokes locomotion. Nat Neurosci 13:246–252. Halbertsma JM (1983) The stride cycle of the cat: the modelling of locomotion by computerized analysis of automatic recordings. Acta Physiol Scand Suppl 521:1–75. Harris-Warrick RM (1988) Chemical modulation of central pattern generators. In Neural Control of Rhythmic Movements. Cohen AH, Rossignol S, Grillner S (eds). New York, NY: Wiley.

255

256

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Harris-Warrick RM (2002) Voltage-sensitive ion channels in rhythmic motor systems. Curr Opin Neurobiol 12:646–651. Harris-Warrick RM (2010) General principles of rhythmogenesis in central pattern generator networks. Prog Brain Res 187:213–222. Harris-Warrick RM (2011) Neuromodulation and flexibility in Central Pattern Generator networks. Curr Opin Neurobiol 21:685–692. Harris-Warrick RM (2014a). Locomotor pattern generation in the rodent spinal cord. In Encyclopedia of Computational Neuroscience. Jung R, Jaeger D (eds). London, UK: Springer. Harris-Warrick RM (2014b) Modulation of crustacean networks for behavior. In Crustacean Nervous Systems and their Control of Behavior. Durby C, Thiel M (eds). Oxford, UK: Oxford University Press. Harris-Warrick RM, Coniglio LM, Barazangi N, Guckenheimer J, Gueron S (1995a) Dopamine modulation of transient potassium current evokes phase shifts in a central pattern generator network. J Neurosci 15:342–358. Harris-Warrick RM, Coniglio LM, Levini RM, Gueron S, Guckenheimer J (1995b) Dopamine modulation of two subthreshold currents produces phase shifts in activity of an identified motorneuron. J Neurophysiol 74:1404–1420. Harris-Warrick RM, Johnson BR (2009). Stomatogastric ganglion. In Encyclopedia of Neuroscience. Binder MD, Hirokawa N, Windhorst U (eds). London, UK: Springer. Harris-Warrick RM, Johnson BR (2010) Checks and balances in neuromodulation. Front Behav Neurosci 4:47. Hartline DK, Graubard K (1992) Cellular and synaptic properties in the crustacean stomatogastric nervous system. In Dynamic Biological Networks: the Stomatogastric Nervous System. Harris-Warrick RM, Marder E, Selverston AI, Moulins M (eds). Cambridge, MA: MIT Press. Harvey PJ, Li Y, Li X, Bennett DJ (2005) Persistent sodium currents and repetitive firing in motorneurons of the sacrocaudal spinal cord of adult rats. J Neurophysiol. 96:1141–1157. He C, Chen F, Li B, Hu Z (2014) Neurophysiology of HCN channels: from cellular functions to multiple regulations. Prog Neurobiol 112:1–23. Helmstaedter M, Briggman KL, Turaga SC, Jain V, et al. (2013) Connectomic reconstruction of the inner plexiform layer in the mouse retina. Nature 500:168–174. Hounsgaard J, Kiehn O (1989) Serotonin-induced bistability of turtle motorneurons caused by a nifedipine–sensitive calcium plateau potential. J Physiol (Lond) 414:265–282. Hultborn H, Zhang M, Meehan CF (2013) Control and role of plateau potential properties in the spinal cord. Curr Pharm Des 19:4357–4370. Husch A, Dietz SB, Hong DN, Harris-Warrick RM (2015) Adult spinal V2a interneurons show increased excitability and serotonin-dependent bistability. J Neurophysiol 113:1124–1134. Hutcheon B, Yarom Y (2000) Resonance, oscillation and the intrinsic frequency preferences of neurons. Trends Neurosci 23:216–222. Janczewski WA, Feldman JL (2006) Distinct rhythm generators for inspiration and expiration in the juvenile rat. J Physiol 570:407–420. Janczewski WA, Tashima A, Hsu P, Cui Y, Feldman JL (2013) Role of inhibition in respiratory pattern generation. J Neurosci 33:5454–5465. Johnson BR, Peck JH, Harris-Warrick RM (1993) Dopamine induces sign reversal at mixed chemical–electrical synapses. Brain Res 625:159–164.

Neural Networks for the Generation of Rhythmic Motor Behaviors

Johnson BR, Peck JH, Harris-Warrick RM (1995) Distributed amine modulation of graded chemical transmission in the pyloric network of the lobster stomatogastric ganglion. J Neurophysiol 74:437–452. Jones JG, Tansey EM, Stuart DG (2011) Thomas Graham Brown (1882–1965): behind the scenes at the Cardiff Institute of Physiology. J Hist Neurosci 20:188–209. Kaczmarek LK (2013) Slack, slick and sodium-activated potassium channels. ISRN Neurosci 2013:354262. Kadiri LR, Kwan AC, Webb WW, Harris-Warrick RM (2011) Dopamine-induced oscillations of the pyloric pacemaker neuron rely on release of calcium from intracellular stores. J Neurophysiol 106:1288–1298. Katz PS (1995) Intrinsic and extrinsic neuromodulation of motor circuits. Curr Opin Neurobiol 5:799–808. Katz PS, Frost WN (1995) Intrinsic neuromodulation in the Tritonia swim CPG: serotonin mediates both neuromodulation and neurotransmission by the dorsal swim interneurons. J Neurophysiol 74:2281–2294. Kepler TB, Marder E, Abbott LF (1990) The effect of electrical coupling on the frequency of model neural oscillators. Science 248:83–85. Kiehn O (2006) Locomotor circuits in the mammalian spinal cord. Annu Rev Neurosci 29:279–306. Kjaerulff O, Kiehn O (1996) Distribution of networks generating and coordinating locomotor activity in the neonatal rat spinal cord in vitro: A lesion study. J Neurosci 16:5777–5794. Koch H, Garcia AJ, 3rd, Ramirez JM (2011) Network reconfiguration and neural plasticity in rhythm–generating networks. Integr Comp Biol 51:856–868. Koizumi H, Koshiya N, Chia JX, Cao F, et al. (2013) Structural–functional properties of identified excitatory and inhibitory interneurons within pre-Bötzinger complex respiratory microcircuits. J Neurosci 33:2994–3009. Koizumi H, Wilson CG, Wong S, Yamanishi T, et al. (2008) Functional imaging, spatial reconstruction, and biophysical analysis of a respiratory motor circuit isolated in vitro. J Neurosci 28:2353–2365. Kolta A, Brocard F, Verdier D, Lund JP (2007) A review of burst generation by trigeminal main sensory neurons. Arch Oral Biol 52:325–328. Kolta A, Morquette P, Lavoie R, Arsenault I, Verdier D (2010) Modulation of rhythmogenic properties of trigeminal neurons contributing to the masticatory CPG. Prog Brain Res 187:137–148. Kristan WB, Jr., Calabrese RL, Friesen WO (2005) Neural control of leech behavior. Prog Neurobiol 76:279–327. Lafreniere-Roula M, McCrea DA (2005) Deletions of rhythmic motorneuron activity during fictive locomotion and scratch provide clues to the organization of the mammalian central pattern generator. J Neurophysiol 94:1120–1132. Lamb DG, Calabrese RL (2011) Neural circuits controlling behavior and autonomic functions in medicinal leeches. Neural Syst Circuits 1:13. Lanuza GM, Gosgnach S, Pierani A, Jessell TM, Goulding M (2004) Genetic identification of spinal interneurons that coordinate left–right locomotor activity necessary for walking movements. Neuron 42:375–386.

257

258

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Li X, Murray K, Harvey PJ, Ballou EW, Bennett DJ (2007) Serotonin facilitates a persistent calcium current in motorneurons of rats with and without chronic spinal cord injury. J Neurophysiol 97:1236–1246. Lichtman JW, Livet J, Sanes JR (2008) A technicolour approach to the connectome. Nat Rev Neurosci 9:417–422. Lieske SP, Ramirez JM (2006) Pattern-specific synaptic mechanisms in a multifunctional network. I. Effects of alterations in synapse strength. J Neurophysiol 95:1323–1333. Lieske SP, Thoby-Brisson M, Telgkamp P, Ramirez JM (2000) Reconfiguration of the neural network controlling multiple breathing patterns: eupnea, sighs and gasps. Nat Neurosci 3:600–607. Livet J, Weissman TA, Kang H, Draft RW, et al. (2007) Transgenic strategies for combinatorial expression of fluorescent proteins in the nervous system. Nature 450:56–62. Ljunggren EE, Haupt S, Ausborn J, Ampatzis K, El Manira A (2014) Optogenetic activation of excitatory premotor interneurons is sufficient to generate coordinated locomotor activity in larval zebrafish. J Neurosci 34:134–139. Llinas R, Sugimori M (1980) Electrophysiological properties of in vitro Purkinje cell somata in mammalian cerebellar slices. J Physiol 305:171–195. Luthi A, McCormick DA (1998) Periodicity of thalamic synchronized oscillations: the role of Ca2+ -mediated upregulation of Ih . Neuron 20:553–563. Madisen L, Mao T, Koch H, Zhuo JM, et al. (2012) A toolbox of Cre-dependent optogenetic transgenic mice for light-induced activation and silencing. Nat Neurosci 15:793–802. Maffei A, Turrigiano GG (2008) Multiple modes of network homeostasis in visual cortical layer 2/3. J Neurosci 28:4377–4384. Marder E (2012) Neuromodulation of neural circuits: back to the future. Neuron 76:1–11. Marder E, Bucher D (2007) Understanding circuit dynamics using the stomatogastric nervous system of lobsters and crabs. Annu Rev Physiol 69:291–316. Marder E, Calabrese RL (1996) Principles of rhythmic motor pattern generation. Physiol Rev 76:687–717. Marder E, Prinz AA (2003) Current compensation in neural homeostasis. Neuron 37:2–4. Markham MR, Kaczmarek LK, Zakon HH (2013) A sodium-activated potassium channel supports high-frequency firing and reduces energetic costs during rapid modulations of action potential amplitude. J Neurophysiol 109:1713–1723. Martell A, Dwyer J, Koch H, Zanella S, et al. (2010) N-methyl-D-aspartate-induced oscillatory properties in neocortical pyramidal neurons from patients with epilepsy. J Clin Neurophysiol 27:398–405. Martell AL, Ramirez JM, Lasky RE, Dwyer JE, et al. (2012) The role of voltage dependence of the NMDA receptor in cellular and network oscillation. Eur J Neurosci 36:2121–2136. McLean DL, Masino MA, Koh IY, Lindquist WB, Fetcho JR (2008) Continuous shifts in the active set of spinal interneurons during changes in locomotor speed. Nat Neurosci. 11:1419–1429. Mentel T, Cangiano L, Grillner S, Büschges A (2008) Neuronal substrates for state-dependent changes in coordination between motorneuron pools during fictive locomotion in the lamprey spinal cord. J Neurosci 28:868–879. Montell C (2005) The TRP superfamily of cation channels. Sci STKE 2005:re3. Moore JD, Deschenes M, Furuta T, Huber D, et al. (2013) Hierarchy of orofacial rhythms revealed through whisking and breathing. Nature 497:205–210.

Neural Networks for the Generation of Rhythmic Motor Behaviors

Mulloney B, Smarandache-Wellmann C (2012) Neurobiology of the crustacean swimmeret system. Prog Neurobiol 96:242–267. Murray KC, Nakae A, Stephens MJ, Rank M, et al. (2010) Recovery of motorneuron and locomotor function after spinal cord injury depends on constitutive activity in 5–HT2C receptors. Nat Med 16:694–700. Nanou E, Kyriakatos A, Bhattacharjee A, Kaczmarek LK, et al. (2008) Na+ -mediated coupling between AMPA receptors and KNa channels shapes synaptic transmission. Proc Natl Acad Sci USA 105:20941–20946. Nusbaum MP, Beenhakker MP (2002) A small-systems approach to motor pattern generation. Nature 417:343–350. Nusbaum MP, Blitz DM (2012) Neuropeptide modulation of microcircuits. Curr Opin Neurobiol 22:592–601. Orchard I, Ramirez J-M, Lange AB (1993) A multifunctional role for octopamine in locust flight. Ann Rev Entomol 38:227–249. Pagliardini S, Janczewski WA, Tan W, Dickson CT, et al. (2011) Active expiration induced by excitation of ventral medulla in adult anesthetized rats. J Neurosci 31:2895–2905. Partridge LD (1994) Calcium-activated non-selective channels in the nervous system. Brain Res Rev 19:319–325. Pearson K, Duysens J (1976) Function of segmental reflexes in the control of stepping in cockroaches and cats. In Neural Control of Locomotion. Herman KM, Grillner S, Stein PSG, Stuart G (eds). New York, NY: Plenum. Pearson KG (2008) Role of sensory feedback in the control of stance duration in walking cats. Brain Res Rev 57:222–227. Pena F, Ramirez JM (2002) Endogenous activation of serotonin-2A receptors is required for respiratory rhythm generation in vitro. J Neurosci 22:11055–11064. Pfluger HJ, Duch C, Heidel E (2004) Neuromodulatory octopaminergic neurons and their functions during insect motor behaviour. The Ernst Florey memory lecture. Acta Biol Hung 55:3–12. Prinz AA, Bucher D, Marder E (2004) Similar network activity from disparate circuit parameters. Nature Neurosci 7:1345–1352. Ptak K, Zummo GG, Alheid GF, Tkatch T, et al. (2005) Sodium currents in medullary neurons isolated from the pre-Bötzinger complex region. J Neurosci 25:5159–5170. Quinlan KA, Kiehn O (2007) Segmental, synaptic actions of commissural interneurons in the mouse spinal cord. J Neurosci 27:6521–6530. Ramirez JM, Doi A, Garcia 3rd AJ, Elsen FP, et al. (2012) The cellular building blocks of breathing. Comp Physiol 2:2683–2731. Ramirez JM, Pearson KG (1991) Octopaminergic modulation of interneurons in the flight system of the locust. J Neurophysiol 66:1522–1537. Ramirez JM, Pearson KG (1993) Alteration of bursting properties in interneurons during locust flight. J Neurophysiol 70:2148–2160. Ramirez JM, Richter DW (1996) The neural mechanisms of respiratory rhythm generation. Curr Opin Neurobiol 6:817–825. Ramirez JM, Telgkamp P, Elsen FP, Quellmalz UJ, Richter DW (1997) Respiratory rhythm generation in mammals: synaptic and membrane properties. Respir Physiol 110:71–85. Ramirez J-M, Tryba AK, Pena F (2004) Pacemaker neurons and neural networks: an integrative view. Curr Opin Neurobiol 14:665–674.

259

260

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Roffman RC, Norris BJ, Calabrese RL (2012) Animal-to-animal variability of connection strength in the leech heartbeat central pattern generator. J Neurophysiol 107:1681–1693. Rubin JE, Hayes JA, Mendenhall JL, Del Negro CA (2009) Calcium-activated nonspecific cation current and synaptic depression promote network-dependent burst oscillations. Proc Natl Acad Sci USA 106:2939–2944. Rybak IA, Molkov YI, Jasinski PE, Shevtsova NA, Smith JC (2014) Rhythmic bursting in the pre–Bötzinger complex: mechanisms and models. Prog Brain Res 209:1–23. Rybak IA, Shevtsova NA, Lafreniere-Roula M, McCrea DA (2006) Modelling spinal circuitry involved in locomotor pattern generation: insights from deletions during fictive locomotion. J Physiol 577:617–639. Sakurai A, Katz PS (2009) State-, timing-, and pattern-dependent neuromodulation of synaptic strength by a serotonergic interneuron. J Neurosci 29:268–279. Schwindt P, Crill W (1999) Mechanisms underlying burst and regular spiking evoked by dendritic depolarization in layer 5 cortical pyramidal neurons. J Neurophysiol 81:1341–1354. Selverston AI (1976) A model system for the study of rhythmic behavior. In Simpler Networks and Behavior. Fentress JC (ed). Sunderland, MA: Sinauer. Selverston AI, Russell DF, Miller JP, King DG (1976) The stomatogastric nervous system: structure and function of a small neural network. Prog Neurobiol 7:215–290. Shao XM, Feldman JL (1997) Respiratory rhythm generation and synaptic inhibition of expiratory neurons in pre-Botzinger complex: differential roles of glycinergic and GABAergic neural transmission. J Neurophysiol 77:1853–1860. Sharp AA, Skinner FK, Marder E (1996) Mechanisms of oscillation in dynamic clamp constructed two-cell half-center circuits. J Neurophysiol 76:867–883. Sherrington, C. S. (1906) The Integrative Activity of the Nervous System. New Haven, CT: Yale University Press. Shevtsova NA, Talpalar AE, Markin SN, Harris-Warrick RM, Kiehn O (2015) Organization of left-right coordination of neuronal activity in the mammalian spinal cord: insights from computational modelling. J Physiol 593:2403–2456. Skinner FK, Kopell N, Marder E (1994) Mechanisms for oscillation and frequency control in reciprocally inhibitory model neural networks. J Comput Neurosci 1:69–87. Smith JC, Ellenberger HH, Ballanyi K, Richter DW, Feldman JL (1991) Pre-Bötzinger complex: a brainstem region that may generate respiratory rhythm in mammals. Science 254:726–729. Stein PSG, Grillner S, Selverston AI, Stuart DG (1997) Neurons, Networks and Motor Behavior. Cambridge, MA: Bradford Books. Stein W (2009) Modulation of stomatogastric rhythms. J Comp Physiol A 195:989–1009. Talpalar AE, Bouvier J, Borgius L, Fortin G, et al. (2013) Dual-mode operation of neural networks involved in left–right alternation. Nature 500:85–88. Talpalar AE, Kiehn O (2010) Glutamatergic mechanisms for speed control and network operation in the rodent locomotor CPG. Front Neural Circuits 4:19. Thoby-Brisson M, Ramirez J-M (2001) Identification of two types of inspiratory pacemaker neurons in the isolated respiratory neural network of mice. J Neurophysiol 86:104–112. Thoby-Brisson M, Simmers J (1998) Neuromodulatory inputs maintain expression of a lobster motor pattern generating network in a modulation-dependent state: evidence from long–term decentralization in vitro. J Neurosci 18:2212–2225.

Neural Networks for the Generation of Rhythmic Motor Behaviors

Thoby-Brisson M, Telgkamp P, Ramirez JM (2000) The role of the hyperpolarization-activated current in modulating rhythmic activity in the isolated respiratory network of mice. J Neurosci 20:2994–3005. Tresch MC, Kiehn O (2000) Motor coordination without action potentials in the mammalian spinal cord. Nat Neurosci 3:593–599. Tryba AK, Pena F, Lieske SP, Viemari JC, et al. (2008) Differential modulation of neural network and pacemaker activity underlying eupnea and sigh-breathing activities. J Neurophysiol 99:2114–2125. Tryba AK, Pena F, Ramirez JM (2006) Gasping activity in vitro: a rhythm dependent on 5-HT2A receptors. J Neurosci 26:2623–2634. Turrigiano G (2012) Homeostatic synaptic plasticity: local and global mechanisms for stabilizing neural function. Cold Spring Harb Perspect Biol 4:a005736. Turrigiano GG (2008) The self-tuning neuron: synaptic scaling of excitatory synapses. Cell 135:422–435. Viemari JC, Bos R, Boulenguez P, Brocard C, et al. (2011) Chapter 1—importance of chloride homeostasis in the operation of rhythmic motor networks. Prog Brain Res 188:3–14. Viemari JC, Garcia AJ, 3rd, Doi A, Elsen G, Ramirez JM (2013) beta-noradrenergic receptor activation specifically modulates the generation of sighs in vivo and in vitro. Front Neural Circuits 7:179. Viemari JC, Ramirez JM (2006) Norepinephrine differentially modulates different types of respiratory pacemaker and nonpacemaker neurons. J Neurophysiol 95:2070–2082. Vinay L, Brocard F, Clarac F, Norreel JC, et al. (2002) Development of posture and locomotion: an interplay of endogenously generated activities and neurotrophic actions by descending pathways. Brain Res Brain Res Rev 40:118–129. Vinay L, Jean-Xavier C (2008) Plasticity of spinal cord locomotor networks and contribution of cation–chloride cotransporters. Brain Res Rev 57:103–110. Wallen P, Robertson B, Cangiano L, Low P, et al. (2007) Sodium-dependent potassium channels of a Slack-like subtype contribute to the slow afterhyperpolarization in lamprey spinal neurons. J Physiol 585:75–90. Wang D, Grillner S, Wallen P (2014a) Endogenous release of 5-HT modulates the plateau phase of NMDA-induced membrane potential oscillations in lamprey spinal neurons. J Neurophysiol 112:30–38. Wang WC, McLean DL (2014) Selective responses to tonic descending commands by temporal summation in a spinal motor pool. Neuron 83:708–721. Wang X, Hayes JA, Revill AL, Song H, et al. (2014b) Laser ablation of Dbx1 neurons in the pre-Bötzinger complex stops inspiratory rhythm and impairs output in neonatal mice. Elife 3:e03427. Weimann JM, Marder E (1994) Switching neurons are integral members of multiple oscillatory networks. Curr Biol 4:896–902. Williams SC, Deisseroth K (2013) Optogenetics. Proc Natl Acad Sci USA 110:16287. Wilson DM (1961) Central nervous control of flight in a locust. J Exp Biol 38:471–490. Zhang Y, Khorkova O, Rodriguez R, Golowasch J (2009) Activity and neuromodulatory input contribute to the recovery of rhythmic output after decentralization in a central pattern generator. J Neurophysiol 101:372–386. Zhang Y, Narayan S, Geiman E, Lanuza GM, et al. (2008) V3 spinal neurons establish a robust and balanced locomotor rhythm during walking. Neuron 60:84–96.

261

262

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Zhong G, Droho S, Crone SA, Dietz S, et al. (2010) Electrophysiological characterization of V2a interneurons and their locomotor–related activity in the neonatal mouse spinal cord. J Neurosci 30:170–182. Zhong G, Dias-Rios M, Zhang Y, Harris-Warrick RM (2005) Intrinsic and functional differences among commissural interneurons during fictive locomotion and serotonergic modulation in the neonatal mouse. J Neurosci 14:6509–6517. Zhong G, Sharma K, Harris-Warrick RM (2011) Frequency-dependent recruitment of V2a interneurons during fictive locomotion in the mouse spinal cord. Nature Comm. 2:274. Zhong G, Shevtsova NA, Rybak IA, Harris-Warrick RM (2012) Neural activity in the isolated mouse spinal cord during spontaneous deletions in fictive locomotion: insights into locomotor central pattern generator organization. J Physiol 590:4735–4759. Zuperku EJ, McCrimmon DR (2002) Gain modulation of respiratory neurons. Respir Physiol Neurobiol 131:121–133.

263

9 Sensory Feedback in the Control of Posture and Locomotion Donald H. Edwards 1 and Boris I. Prilutsky 2 1 2

Neuroscience Institute, Georgia State University, Atlanta, GA, USA School of Biological Sciences, Georgia Institute of Technology, Atlanta, GA, USA

9.1 Introduction All animals must find food, avoid predation, and reproduce. Living on land adds the additional challenges of gravity and functioning at the interface between a low-drag atmosphere and a textured substrate. Friction must be minimized to move quickly over this substrate, which can be accomplished by having a polarized, axial body structure raised on segmentally specialized, jointed legs. This set of adaptations was achieved by both vertebrates and invertebrate arthropods, including insects, crustaceans, and arachnids. Legged animals face the additional challenges of maintaining body posture against gravity and external perturbations both when stationary and when moving. The common body plan presents common control problems to both sets of animals: how to produce stable postures, how to produce coordinated and stable locomotor gaits, and how to shift efficiently between these states in a noisy, changing environment. These common control problems have led to common solutions. In both sets of animals descending motor commands excite similar motor circuits to produce similar leg and body movements. Feedback from similar sensors then provides information on the immediate consequences of the motor commands, which is particularly important in shaping their postural and locomotor behaviors. These common solutions have emerged despite large differences in the sizes, skeletal structures, and habitats of vertebrates and arthropods. In the last 50 years, human engineers have discovered the same strategies in their efforts to build devices that respond adaptively to the environment. Chapter Plan We begin with a brief history of feedback control and a review of classical

control theory. We then show how that theory informs our general understanding of the control of limb movement and posture. We then describe feedback control of posture and locomotion in arthropods and vertebrates, and interpret that in light of control theory. We end with a discussion of promising future research strategies.

Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

264

Neurobiology of Motor Control: Fundamental Concepts and New Directions

9.2 History and Background of Feedback Control Negative feedback first came into widespread use in fly-ball centrifugal governors that controlled the rotation speed of early steam engines (Maxwell 1868). As engine speed increased, the plane of the fly-ball rotation rose and reduced the flow of the steam that drove the engine. The valve openings and sensitivity of the system could be adjusted to enable it to rotate at a near constant speed independent of steam pressure. In the mid-19th century, Claude Bernard identified negative feedback as being responsible for maintaining the constancy of the internal body environment despite changes in outside conditions (Gross 1998). Early in the 20th century, Charles Sherrington showed how spinal reflexes used negative feedback to resist perturbations to maintain limb positions and body postures (Sherrington 1910). He showed that the reflex response consisted of several parallel pathways that excited ipsilateral muscles to resist the perturbation and inhibited their antagonists, while also exciting and inhibiting appropriate contralateral muscles to maintain body posture. Sherrington also showed how these reflexes could be chained in sequences to generate rhythmic locomotor movements (Sherrington 1913). An initial leg movement would trigger a reflex response in the opposing direction, and the resulting movement would trigger another reflex response in the original direction. While working in Sherrington’s laboratory, Brown showed that “central pattern generators”, as they later came to be known, were the primary drivers of rhythmic motor activity in cat spinal cord (Brown 1911; Stuart and Hultborn 2008) (see also Chapter 8). Recent work in many animals has established the importance of feedback-evoked reflex responses in helping centrally generated rhythmic motor patterns adjust to the physical demands of locomotion (Duysens et al. 2000). The development of radar during World War II made it possible to couple sensing the position and velocity of an enemy aircraft with directing and firing an anti-aircraft gun. Work on this problem led to the development of what is now called “classical” control theory, which showed how desired patterns of output could be evoked from any system in a noisy environment despite large differences between the power of the control signal and that of the output. In his book “Cybernetics: Command and Control in the Animal and the Machine”, Wiener (1961) showed that the same control problems are faced by animals.

9.3 Classical Control Theory Figure 9.1 shows how negative feedback enables a weak input signal to control the output of a much more powerful system in the presence of noise. I is the input to a system, e.g., the angle of a steering wheel, and O is the output, e.g., the angular direction of a large ship (these are Laplace variables that lend themselves to mathematical descriptions of negative feedback systems and provide a means of predicting how they behave). The power required to change the angle of the ship is much greater than the power needed to turn the steering wheel, and thus the steering wheel signal, I, must be amplified greatly. This is accomplished by a controller, C, and steering plant, P. Ship direction can be affected by external stimuli, or noise, N, such as a cross-wind perpendicular to the ship’s direction, which can be much larger than the input signal, I.

Sensory Feedback in the Control of Posture and Locomotion

Figure 9.1 Control systems without and with feedback. (A) Without feedback. I is input, O output, C the controller, P the plant, and N noise. Arrows pointing to a rectangle are its inputs. Arrow leaving a rectangle is its output, the product of the rectangle and its input. Output of a summing node (a circle with a “Σ”) is the sum of its inputs, according to the sign adjacent to the input arrow. (B) A proportional controller, e is error. (C) A PID controller, where s is the Laplace frequency variable, and CD , CP , and CI are the differential, proportional, and integral control gains, respectively.

A

N + I

P

C

Σ

+

B

O

N Σ

I +

+

e

Σ

P

C

O

+ – CDS

C I

Σ

+

e

N +

CP

Σ

Σ

P

O

+

– CI/S

Without feedback, the output is given by O = N + ICP,

(9.1)

so that the ship’s direction depends directly on the noise N and the input amplified by the controller and the plant, ICP (Fig. 9.1A). If the noise is comparable in amplitude to ICP, the output O will be significantly distorted. However, if the output O (the ship’s angular direction) is “fed back” to the input and subtracted from it (Fig. 9.1B), the effect of noise is greatly reduced. The difference between input and output is the error, e, in the output of the system, e = (I − O).

(9.2)

If the error is amplified by the controller, C, and by the plant, P, the product is the output, O: O = eCP = (I − O)CP.

(9.3)

Because the output is modified by noise, N, the complete expression becomes O = N + (I − O)CP,

(9.4)

which can be re-arranged to O = (N + ICP)∕(1 + CP).

(9.5)

If CP ≫ N and ≫1, then O ∼ I. This means that despite perturbations produced by noise, system output will equal system input. This is very useful in many systems, such as that described here; feedback allows the ship to be steered in the desired direction despite powerful perturbations.

265

266

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Delay The negative feedback system in equation 9.5 assumes that the feedback is effec-

tively instantaneous relative to the rate of change of any system variable, including input, output, and noise. When this is not true, interesting (and occasionally catastrophic) things can happen. Assume, for instance, that it takes some time, 𝜏, for the input to activate the plant and generate an output, but that the feedback from that output is immediately subtracted from the input. If the input varies with a period twice the time needed to produce the output, then an output peak will produce a peak in feedback that is subtracted from the input at a time when the input is in a trough. As a result, the delay has transformed the negative feedback into positive feedback, where the error signal grows with each cycle of the feedback loop. The frequency where the system blows up is f∞ = 1∕2𝜏. Longer delays thus cause the system to blow up at lower input frequencies. The best way to expand the range of input frequencies over which the system operates without blowing up is to minimize the delay; for many systems the delay can be reduced enough so that f∞ is outside the frequency range of normal inputs. However, sometimes there is a limit to how fast the output can be produced in response to an input. In animals, for instance, the delay in responding to a motor command results from the time needed for excitation–contraction coupling, for muscle contraction, and to overcome limb and load inertia. These times can be substantial. In cats the shortest delay of the ankle muscle monosynaptic stretch reflex is about 12 ms, in humans about 30 ms. Unaided, this system will blow up if the frequencies of the input to the motorneurons are near f∞ , 42 Hz for cats and 17 Hz for humans. However, there is also an instantaneous negative feedback to postural perturbations caused by intrinsic muscle properties—so-called short-range muscle stiffness—in which muscle stretch increases muscle force. This response is not mediated by sensory afferents. This instantaneous negative feedback can be made to work for both flexing and extending perturbations by co-contracting opposing muscles. PID Controllers The feedback systems described above use a signal proportional to the response error, e, to drive the plant:

(9.6)

Plant input = Ce = C(I − O).

For systems functioning in rapidly changing environments, this mechanism may be inadequate to prevent the error from becoming disastrously large. It may also leave a residual difference between input and output after steady-state is achieved. The residual error may be both large enough to impair system effectiveness and too small for the feedback to reduce further. These problems can be addressed by adding an additional pair of controllers, one that responds to the rate of change of the error and one that responds to its time integral, to create a proportional-integral-derivative (PID) controller (Fig. 9.1C). In this scheme, the plant, P, is driven by three controllers, CP , CI , and CD , that respond to the amount of error (CP ), the time-integral of the error (CI ), and the rate of change of the error (CD ). The response of the controller becomes ( C(t) =

(

t

e(t)CP +

∫0

e(t)CI dt +

) ) de CD . dt

(9.7)

Sensory Feedback in the Control of Posture and Locomotion

The Laplace transform yields an algebraic formula that can be used to revise equation 9.7: ) ]/[ ( ) ] [ ( CI CI 1 + CP + (9.8) O = N + I CP + + sCD P + sCD P , s s where s is the frequency variable in the Laplace domain. For low values of s (low frequencies), PCI ∕s is large compared to both P(CP + sCD ) and N, and thus O ∼ I. For high values of s (high frequencies) PsCD is large relative to P(CP + CI ∕s) and to N, and so again O ∼ I. In the middle range of frequencies, CP P is large compared to P(sCD + CI ∕s) and to N and O ∼ I as in equation 9.5.

9.4 Nervous System Implementation in the Control of Posture and Limb Movements Feedback systems that conform to the formal description given above are implemented by the nervous system at several levels to help control movement, beginning with control of limb movement around a single joint. To illustrate this, we used the neuromechanical simulator AnimatLab (Cofer et al. 2010a; see www.animatlab.com) to create a single jointed limb model (Fig. 9.2A), in which flexor (black) and extensor (gray) muscles stretch across opposing sides of a hinge (elbow or knee) joint to flex and extend a distal limb segment. The flexor and extensor muscles are Hill type models consisting of a spring in series with a parallel combination of a spring, a dashpot, and an actuator that transduces membrane depolarization into contraction force (Fig. 9.2B). The muscles are excited by integrate-and-fire motorneurons (MN). Without feedback (Fig. 9.2C), tonic current stimulation of the α Flexor MN sufficient to make it fire at 50 Hz flexed the resting limb against the forces of gravity and the stretching extensor muscle. The flexor EPSPs produced an oscillating flexor muscle membrane potential and a flexor muscle tension that varied around a mean of just less than 2 N. These oscillations are greater than would be observed in real limbs because the muscle models each have the same single time-constant and are each excited by single motorneurons. Despite these oscillations, the motorneuron activity produced a steady 30∘ flexion of the limb. Response to Limb Perturbation without Feedback Limb posture was disturbed by an upward

1N force applied for 100 ms to the distal limb segment at 1 s (arrow, Fig. 9.2C). The perturbation evoked a large, very slowly decaying oscillation of the distal limb segment. This oscillation, and that described in the next paragraph, resulted from the spring-like nature of the flexor and extensor muscles and the pendular arrangement of the joint and distal limb. Response to Step Excitation of the 𝛼 Flexor MN Without Feedback A step depolarization was then

applied to increase α Flexor MN firing rate to 100 Hz (time = 8 s, Fig. 9.2C). Flexor potential and tension increased, causing the distal limb segment to flex quickly. An oscillation was again evoked that only very slowly declined to the new limb flexion angle (full decay not shown).

267

268

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

C

B

Sensory Feedback in the Control of Posture and Locomotion

Figure 9.2 A neuromechanical single joint limb model. (A) The limb consisted of a fixed Proximal Segment connected by a hinge joint to a Distal Segment. Flexor (black) and Extensor (gray) muscles span the anterior and posterior sides, respectively, of the joint. (B) Top: Hill muscle model, with a parallel spring (spring constant, Kp ), serial spring (spring constant, Kse ), dashpot (dashpot constant, B), and actuator, A. T is tension developed across the model. Bottom: Integrate and fire models of single α Flexor and α Extensor MNs drive the Flexor and Extensor Muscles. (C) Model activity. Black and gray traces show flexor and extensor activity, respectively. 1st trace: α MN membrane potential; 2nd : α MN firing rate; 3rd : muscle membrane potential; 4th : muscle tension; 5th : joint angle. Stimuli: A 30∘ limb flexion was maintained at the outset by a constant +12 nA injected in the α Flexor MN. At 1 s (arrow) the joint was perturbed by a 1 N, 0.1 s upward force applied to the distal limb segment. At 8 s, an additional +20 nA current step was injected into the α Flexor MN to increase limb flexion. Note oscillations with perturbation and changes in MN firing rate.

Sensors for Feedback: Muscle Spindles and Muscle Receptor Organs Vertebrate muscle spindles

and arthropod muscle receptor organs consist of small groups of muscle fibers that lie parallel to the muscle they help control. Both include terminals of motorneurons that innervate them, and stretch-sensitive endings of afferents that respond to stretch or active contraction of the sense organ. In the spindle, stretch or contraction of a nuclear bag fiber (one of two types of spindle muscle fibers) excites Ia afferents. In cat, Ia afferents respond phasically to the onset of spindle stretch, increase their firing rate linearly with increasing stretch, and drop to a lower rate when stretch is maintained (Fig. 9.3A, left) (Prochazka 1996). The phasic and linear responses provide the derivative and proportional feedback used in a PD controller (Fig. 9.1). Stimulation of one of the γ motorneurons that excite the spindle evokes a phasic Ia response at stimulation onset and a sustained response with maintained γ MN stimulation (Fig. 9.3A, right). Spindle stretch during γ MN stimulation evokes enhanced Ia responses. To illustrate how spindles help control joint movement, we placed flexor and extensor muscle spindles in parallel with the flexor and extensor muscles in the limb model (see Fig. 9.4). Each spindle had the same origin and insertion attachments to the limb as its corresponding muscle. The spindle model is a Hill model that, like the muscle model, represents spindle elastic, viscous, and force-generating properties (Fig. 9.2B). Because the spindle contains many fewer muscle fibers than the main muscle, the spring constants of the spindle model were reduced to 10%, and the dashpot constant was reduced to 75%, of the corresponding muscle model values. The model spindle transduces spindle tension and the serial spring’s rate of stretch to generator currents that are applied to an integrate and fire model of an Ia afferent. Parameter values of the spindle and Ia afferent were set to make afferent responses to imposed stretch very similar to real responses of cat Ia afferents both with and without Ia firing (compare Fig. 9.3A and B). Adding Negative Feedback from Muscle Spindles to Control Movement Flexor and extensor feed-

back loops were enabled by having the flexor and extensor Ia afferents synaptically excite the α Flexor and Extensor MNs (Fig. 9.4A). Flexor and Extensor Spindles were excited separately from Flexor and Extensor Muscles by the γ Flexor and Extensor MNs. In addition, cross-inhibitor interneurons prevented excitation of antagonist MNs. These are negative feedback loops because the muscle contraction each MN produces reduces tension in the associated spindle, and so reduces the excitation of the Ia afferent that excites the MN.

269

270

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

Real Spindle

γ MN Stim

Ia rate 100/s

1s Spindle length

B

Model Spindle

γ MN Stim

150 Ia rate 100 (Hz) 50 0

Spindle length

1 cm

1s

Figure 9.3 Ia responses to stretch of real and model spindles. (A) Ia firing rate responses (top) to stretch (bottom) of a mammalian spindle before (left) and during (right) γ MN stimulation. Modified from Prochazka (1996). (B) Model Ia firing rate responses (top) to stretch (bottom) of a model spindle before (left) and during (right) γ MN stimulation.

Resistance Reflex Responses to External Perturbation A constant stimulus was applied to the

α Flexor MN so as to induce an initial joint angle of about 33∘ . The same upward, 1 N force was applied at 1 s to cause limb flexion. This upward movement stretched the extensor spindle and phasically excited the extensor Ia afferent, which in turn excited the α Extensor MN and inhibited the α Flexor MN. These responses dramatically reduced the initial flexion induced by the perturbation (compare to Fig. 9.2C). As the limb fell back, a similar extension reflex response prevented oscillation and brought the limb position to equilibrium within 200 ms. This postural control system functions like a PD controller. The afferent response is the error signal, and is proportional to both the amount of spindle tension and its rate of change (Fig. 9.3B, left).

Feedback Control of Voluntary Movement When the 𝛼 and γ Flexor MNs were simultane-

ously depolarized by injected current (time = 3 s, Fig. 9.4B), the α Flexor MN responded with a step increase in firing that was immediately augmented by a burst of Flexor Ia

Sensory Feedback in the Control of Posture and Locomotion

271

B 0 α Mns –25 (mV) –50

A Flexor PADI

200 α Mns rate(Hz) 100

Ia

α Flexor MN

0 –20 Muscle –40 (mV) –60

γ Flexor MN Flexor X-inhib

γ Extensor

Flexor muscle & spindle

4 Tension (N)

0

Extensor X-inhib

75 Angle 50 (Deg) 25 0

MN

α Extensor MN

Extensor PADI

2

400 Iarate (Hz) 200 Ia

0 0 Extensor muscle & spindle

1

2 1N

3

4

5

Time (s)

Figure 9.4 Neuromechanical single joint limb model with feedback control. (A) Model in Fig. 9.2 with γ Flexor and γ Extensor MNs, Flexor and Extensor PAD Interneurons (PADI), Flexor and Extensor Ia afferents, Flexor and Extensor X-inhib (cross-inhibitory) interneurons, and Flexor and Extensor Muscle Spindles (Fig. 9.3B). (B) Flexor α and γ MNs were each initially excited by constant injected current (4 nA and 8 nA, respectively) to maintain limb flexion at 30∘ against gravity and extensor tension. At 1 s (arrow), a 1 N, 0.1 s vertical force was applied to the distal segment. At 3 s, the current stimuli to the α and γ Flexor MNs were increased by 15 nA and 20 nA, respectively. The Extensor PADI was excited by 1 nA from 5.145 s to 5.33 s. 1st trace: α MN membrane potential; 2nd : α MN firing rate; 3rd : muscle membrane potential; 4th : muscle tension; 5th : joint angle; 6th : Ia afferent firing rate. Flexor element responses are in black; extensor element responses are in gray. Note the lack of oscillation of the joint angle after perturbation and the smooth trajectory to a new limb position following the change in MN firing rate.

activity. The Flexor Ia burst was evoked by the sudden increase in Flexor Spindle tension produced by the step activation of the Flexor γ MN. The spindle contracted isometrically because the limb had yet to move, and as a result spindle tension increased suddenly. The α Flexor MN burst provided a transient increase in flexor tension that helped initial joint flexion. Joint flexion stretched the Extensor Spindle and excited the Extensor Ia afferent. The Extensor Ia excited the α Extensor MN, which slowed the rate of flexion. The timing and amplitudes of afferent and motorneuron responses were such that limb flexion drove smoothly to the desired equilibrium angle, thus again removing the oscillation seen without feedback. After the initial flexion, the crossed inhibitory IN (Fig. 9.4A) prevented the Extensor MN from responding to the Extensor Ia even though it was continuously excited after the limb reached its new position. The α Flexor MN

6

272

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Limb movement + Xα

αMN

+

+

Muscle tension

+

Joint torque

–

Joint angle

+

Muscle length

+

Descending limb flexion command + + Xγ

γMN

+

Spindle tension

d/dt +

Ia (error)

Figure 9.5 Flow chart of α/γ feedback loops. Positive and negative signs indicate whether the influence is in the same or opposite direction. Note that the sign reversal, where increased joint torque decreases joint angle, participates in two negative feedback loops, one mediated by increased muscle stiffness, the other by the spindle Ia afferent.

was thus unopposed after the transient responses of the α Extensor MN, and steady state joint flexion was set solely by flexor muscle contraction and limb weight. Although the Ia afferents provide feedback in the limb position control system, how that system maps onto the negative feedback system shown in Fig. 9.1 is not straightforward. It may initially seem that the Ia response is a negative feedback signal and the error is calculated in the central nervous system. However, the error is actually calculated peripherally from the tensions in both the working muscle and the muscle spindle. This is apparent if we examine the control system for one muscle type (flexor or extensor) (Fig. 9.5). Motor control is shared by both the α and γ MNs, where Xα , the alpha command, provides a baseline for the control system to operate on, and X𝛾 , the gamma command, is the control command. Increased α MN activity increases muscle tension and joint angle torque, which reduces muscle and muscle spindle lengths. This sign inversion makes the feedback negative. The reduced muscle length immediately reduces muscle tension, and so creates the instantaneous negative feedback loop described earlier that depends on intrinsic, short range muscle stiffness. In this loop muscle tension is the error signal, subtracting the effects of muscle shortening from the effects of α MN excitation. Spindle tension similarly results from the difference between the effects of γ MN activity and of muscle length, which varies with limb position. In this loop, spindle tension is the error signal and produces the generator current for the Ia afferent. The afferent responds to both spindle tension and its rate of increase, and its response adds to the excitation of the α Flexor MN. PAD Inhibition of the Antagonist Ia Response Negative feedback (resistance) reflexes stabilize

limb position and so can interfere with voluntary movements. To block unwanted feedback at the time of the expected reflex response, movement commands therefore also excite neurons that inhibit spindle afferents. In the spinal cord of vertebrates and the ventral nerve cord of crustaceans, pre-synaptic inhibition in the form of primary afferent depolarization (PAD) prevents afferents from reflexively exciting antagonist MNs that would oppose the voluntary movement (Clarac 2008). As in the vertebrate myotactic

Sensory Feedback in the Control of Posture and Locomotion

reflex explained above, PAD is invoked in crustaceans during walking to prevent resistance reflexes from interfering with walking movements. Indeed, the resistance reflexes are transformed into assistance reflexes, in which the agonist MN is excited to enhance the movement that triggered the reflex (Cattaert and Le Ray 2001). This sort of reflex reversal also occurs in vertebrates (Ekeberg and Pearson 2005). In the limb flexion model, the Extensor PAD Inhibitory interneuron (PADI) inhibited the Extensor Ia afferent (Fig. 9.4A). The Extensor PADI was depolarized with a brief (185 ms) current pulse at the onset of Flexor MN stimulation. The resulting inhibition shortened and reduced Extensor Ia response and allowed the limb to flex nearly to its target position before extensor tension and Flexor MN inhibition slowed it and brought it to rest (Fig. 9.4B). The flexion response was truly tri-phasic: an initial, strong Flexor MN excitation, a subsequent dip in Flexor MN response and a burst in the Extensor MN, and finally a sustained Flexor MN discharge. Role in Postural Control Animal posture consists of a set of joint angles the limbs adopt.

For each joint, any particular angle can be achieved with a range of opposing muscle excitations, each of which changes joint stiffness. Increased tension in the opposing muscles increases the gain of the instantaneous feedback described above, and thus makes the limb more resistant to perturbation. Varying γ MN excitation to muscle spindles changes spindle tension and thus the gain of the Ia feedback loop, and hence also changes resistance to perturbation. Role in Movement Commands Feedback allows α/γ co-activation step commands to pro-

duce step-like limb flexions (Fig. 9.4B). Additional simulations demonstrate that flexion movement scales nearly linearly with step command and will follow a sinusoidal command, diminishing in amplitude with command frequency, like a low-pass filter. Thus, despite the complexities of muscle contraction and perturbation by external forces, feedback, including γ modulation of the feedback and PAD inhibition of unwanted feedback, enables relatively simple motor commands to produce proportional limb movements around a joint. Load Compensation and Gain If the gain of the loop is sufficiently large, feedback can

enable a limb to produce the same movement under increased loads. The loads can be represented by the noise term, N in (5), so minimizing their effect requires that the product CP, the open loop gain, is large compared to N. If it is not, then the error will increase and limb flexion will fall short of the target by an amount proportional to the load and inversely proportional to the gain. Better compensation occurs when gain is increased, but that risks making the system unstable. Positive Feedback Negative feedback becomes positive when the feedback is added to,

rather than subtracted from, the input, so that “error” growth accelerates. The resulting explosive growth of output can produce useful power surges and rapid state transitions, but, to be useful, these rapid changes must be controlled. Positive feedback is often controlled by nesting it within a larger negative feedback system that limits total change. For example, the rapid membrane depolarization of an action potential is produced by the positive feedback between depolarization and sodium channel opening. The membrane depolarization is limited by the sodium equilibrium potential (ENa ), because inward

273

274

Neurobiology of Motor Control: Fundamental Concepts and New Directions

sodium current is proportional to the difference between membrane potential and ENa (see Chapter 2). In locomotion, positive feedback assistance reflexes promote rapid transitions between stance and swing that are stabilized by negative feedback. At the end of stance, a leg perturbation or an attempt to lift it may trigger an assistance reflex that excites leg flexor motor neurons. The assistance reflex may increase as the leg rises and be extinguished as the leg reaches the limits of elevation (Chung et al. 2015).

9.5 Organization and Function in Arthropods Arthropods engage in complex and sophisticated movements. A cockroach walking at high speed over a rough terrain, crayfish wrestling for social dominance, and a praying mantis carefully stalking its prey before launching a directed attack, all require a high degree of coordination between brain centers, thoracic motor centers, and multiple streams of sensory feedback continuously arriving from the periphery. Posture and locomotion are controlled in arthropods using sensory feedback from a wide variety of sensors and musculoskeletal structures and the control principles and organizational patterns described above. Arthropod bodies, limbs, and nervous systems have been widely reviewed (Burrows 1996; Derby and Thiel 2014; Niven et al. 2008; Strausfeld 2012; Watling and Thiel 2013). We therefore here provide only a brief summary to acquaint readers unfamiliar with the general organization. 9.5.1 Locomotory System Gross Anatomy

Like a suit of armor, the arthropod exoskeleton provides a protective hard cuticular surface for the limb or body segment with a flexible soft cuticular covering at each joint. Muscles originate from interior points of the exoskeleton and project across a joint to attach on the cuticle or an apodeme, an internal tendon-like structure. Insect and crustacean legs are segmented tubular structures that emerge from the thorax, itself a fused multi-segmental structure. The legs contain five (insects) or seven (decapod crustaceans) serially arranged segments, ending with single or paired (pincer) claws. The leg segments are linked together by bicondylar hinge joints that allow the claw to move through a wide range of the space around the body. The joints have wide angles of movement and are generally oriented orthogonal to each other: the proximal, thoracic-coxa (TC) joint moves the entire leg forward and back; the adjacent joint (coxobasal (CB) in Crustacea, coxa-trochanter (CT) in insects) levates and depresses the leg; the next major joint (carpus-propodus (CP) in Crustacea, femur-tibia (FT) in insects) flexes and extends the distal portion of the leg. 9.5.2 Proprioceptors and Exteroceptors

The legs contain multiple proprioceptors and exteroceptors that provide sensory feedback to the nervous system (Mill 1976). Like muscle spindles, muscle receptor organs (MROs) are clusters of muscle fibers with embedded stretch receptors that give rise to length-sensitive sensory afferents (DiCaprio and Clarac 1983). Like spindles, they receive efferent innervation that enables them to play a similar role in feedback control (Head and Bush 1991). Chordotonal organs (CO) are elastic strands with embedded sensory afferents but no efferent supply. The afferents respond to CO stretch

Sensory Feedback in the Control of Posture and Locomotion

and release as the joint opens and closes. Tendon receptors, like Golgi tendon organs, respond to apodeme force (Hartman 1985). Cuticular sensors, campaniform sensilla (CS) in insects (Zill et al. 2010; Zill and Moran 1981), and cuticular strain detectors (CSD) (Klarner and Barth 1986) and funnel canal organs (FCO) in crustaceans (Libersat et al. 1987a), respond to forces produced in the leg cuticle by muscle contraction or leg torque. Hair-like receptors respond to contact with the ground, with another body segment at a joint, or with an external object or force on the leg. 9.5.3 Arthropod Nervous Systems

Arthropod nervous systems are ladder-like structures segmentally organized (see also Fig. 10.2) into ganglia connected by paired intersegmental connectives, which are bundles of axons. Each bilaterally symmetric thoracic ganglion controls a pair of legs and contains about 2,000 neurons, 1,900 of which are interneurons (Wiersma 1961). Sensory afferents project to their segmental hemi-ganglion and motorneurons project to their segmental muscles. While the distal portion of the leg is supplied by hundreds of sensory afferents from cuticular hairs (Marchand et al. 1997), several other sense organs contain small numbers of sensory afferents, some of which are individually identifiable or as one of an identifiable small group (Cattaert 2014). Even fewer motorneurons innervate the legs; each walking leg of a lobster or crayfish is innervated by approximately 95 motorneurons, nearly all of which are individually identifiable (Chrachri and Clarac 1989). Unlike vertebrates, arthropods possess inhibitory motorneurons that prevent excitation of a muscle or speed muscle relaxation (Govind and Atwood 1982). One inhibitory motorneuron, the common inhibitor, innervates all the muscles of a leg. 9.5.4 Postures and Movement Commands (see also Chapter 7.3)

Postural and locomotor movements are initiated by descending commands from anterior parts of the nervous system. In Crustacea, a variety of static postures and locomotor movements can be released by stimulating discrete “command fibers” in the circumesophageal connectives that link the brain and ventral nerve cord (Bowerman and Larimer 1974a,b; Evoy and Ayers 1982). Neurons in the central complex of the insect brain are important for decisions related to walking speed and movement around barriers (Bender et al. 2010; Ritzmann et al. 2012) and neurons in protocerebral areas near the central body of the crayfish brain help control initiation of walking (Kagaya and Takahata 2011). Coordinating neurons that project through crustacean thoracic ganglia help coordinate the half-center oscillators for leg depression/elevation and promotion/remotion that underlie forward and backward walking (Cattaert and Le Ray 2001). These networks presumably translate unpatterned descending “commands” into patterns of coordinated leg movement, although the details are unknown. Nonetheless, the thoracic motor patterns bear a remarkable resemblance to analogous patterns in vertebrate spinal cord (Clarac and Pearlstein 2007; Duysens et al. 2000). 9.5.5 Sensory Feedback in the Maintenance of Posture

Little is known of how descending commands for a given static posture are distributed among the elements of the thoracic nervous system. Nonetheless, it is clear that static postures are maintained through negative feedback resistance reflexes in which nearly

275

276

Neurobiology of Motor Control: Fundamental Concepts and New Directions

all the proprioceptors and exteroceptors of the leg participate. This process is particularly well understood in crayfish. Two TC joint sense organs mediate resistance reflexes that maintain leg anterior/posterior position. Leg movement forward excites afferents from the TC chordotonal organ (TCCO), which in turn excite leg remotor MNs and inhibit leg promotor MNs (Skorupski and Sillar 1986). Backward leg movement excites afferents of the TC muscle receptor organ (TCMRO), which then excite promotor MNs and inhibit remotor MNs. At the CB joint, an elastic strand receptor, the coxobasal chordotonal organ (CBCO), mediates resistance reflexes that maintain leg position against elevation and depression. Twenty release-sensitive CBCO afferents respond to leg elevation by exciting depressor MNs and inhibiting levator MNs. Leg depression excites 20 stretch-sensitive CBCO afferents with opposite effects (El Manira et al. 1991a). As in insects and vertebrates, the excitatory resistance reflexes are monosynaptic and the inhibitory reflexes are di- or polysynaptic. For both the stretch- and release-sensitive CBCO afferents, some respond to CBCO stretch or length, some to rate of stretch, and some to both. Like the vertebrate Ia resistance reflex, the CBCO resistance reflex acts as a PID controller to help maintain posture. Cuticular stress detectors (CSDs), located near the CBCO, respond to cuticle stress evoked by depression of the leg against the substrate and excite the levator MNs in a negative feedback reflex (Leibrock et al. 1996). Insects have similar posture-maintaining resistance reflexes. Campaniform sensilla (CS) are functionally similar to crustacean CSD receptors. In cockroach tibia, proximal CS excite extensor MNs and inhibit flexor MNs and distal CS have reverse effects (Zill et al. 1981). In stick insect, an array of CS responds to the amplitude and rate of loading or muscle forces in directions out of the plane of leg movement. They excite leg retractor and protractor MNs in response to cuticular stress applied to the leg in a negative feedback reflex to reduce the stress (Schmitz 1993; Zill et al. 2012). Many resistance reflexes span more than one joint, so that the leg movements are directed in more than a single plane. For example, stretch-sensitive CBCO afferents excited by leg depression excite levator MNs (CB joint) to resist the depression. They also excite leg remotor MNs (TC joint) that move the leg backward, and leg extensor (mero-carpopodite joint) and leg bender (CP joint) MNs to extend and pronate the leg. Raising the leg excites the antagonist MNs at all these joints (Clarac et al. 1978). Wider divergence of sensory effects is also seen. Polysynaptic reflex responses in locusts are produced by afferents that excite non-spiking interneurons with wide effects in the thoracic ganglion (Burrows et al. 1988) and, through projecting intersegmental interneurons, on adjacent legs as well (Laurent and Burrows 1989). Reflexes evoked by single proprioceptors or exteroceptors are thus part of a complex, but poorly understood, postural control system that works to stabilize posture through many parallel negative feedback pathways. These pathways diverge from each sense organ and converge from several sense organs onto sets of MNs that stabilize movements around multiple joints. 9.5.6 Sensory Feedback in Movement and Walking

The limb simulation above (Fig. 9.4) shows that the negative feedback control typical of resistance reflexes can also be used to control dynamic movement patterns by changes in the efferent control of the stretch receptor. Among arthropod leg receptors, only muscle receptor organs (MROs) provide for efferent control, with motor neurons projecting

Sensory Feedback in the Control of Posture and Locomotion

to the receptor muscle (RM) of the MRO either alone or to the homonymous working muscle as well (Head and Bush 1991). As in the spindle model (Fig. 9.4), co-activation of the RM and working muscle can excite afferent feedback to efferents of the working muscle to provide additional muscle activation at the beginning of each contraction phase, with afferent discharge decreasing as the RM shortens with limb movement. Nervous systems exist in “quiescent” states where rhythmic motor patterns are inactive and resistance reflex responses are the rule, and “active” states characterized by at least some rhythmic motor activity and active resistance and assistance reflex responses (Chung et al. 2015; Le Ray and Cattaert 1997). Rhythmically activated resistance reflexes can produce rhythmic motor patterns in quiescent nervous systems, but these are readily seen to be artifactual. For example, when recording from an isolated crayfish nervous system with an intact CBCO, imposed sinusoidal CBCO stretch and release induces alternating levator/depressor and remotor/promotor activity suggestive of backward walking (El Manira et al. 1991b). However, the stretches and releases that would occur during actual backward walking are opposite to the imposed CBCO movements. Thus, if these resistance reflexes occurred in a walking animal, they would oppose the movement. This does not happen in active preparations or during walking, indicating that such unwanted resistance reflexes are prevented from occurring when the nervous system is in the active state. Unwanted reflexes are prevented in both arthropods and vertebrates by pre-synaptic inhibition of the synapses between the afferents and the motor neurons that mediate the reflex (Büschges and Wolf 1999; Cattaert et al. 1992). Pre-synaptic inhibition targets the afferent-to-interneuron connection that mediates the reflex, and often spares the outputs of the same afferents to other targets and the responses of the motor neurons to other inputs. Pre-synaptic inhibition often takes the form of the PAD explained earlier (Eccles et al. 1962; Kennedy et al. 1974). In locust, PAD occurs in central terminals of proprioceptor (FCO) afferents produced by di-synaptic inputs arising from other afferents (Burrows and Laurent 1993). In crayfish, PAD occurs in the central terminals of CBCO afferents and of dactyl sensory afferents (Cattaert et al. 1992; Marchand et al. 1997), where it acts by shunting or inactivating spikes at or before they reach the pre-synaptic terminal (Cattaert and El Manira 1999; Edwards 1990). Strong pre-synaptic inputs can also sufficiently depolarize the pre-synaptic terminal to evoke spikes that travel antidromically to the distal spike generating mechanism of the afferent. Antidromic spikes can inhibit the spike generating mechanism for hundreds of milliseconds, thereby preventing afferent participation in all reflex responses (Cattaert et al. 1999; Gossard et al. 1999). Not all resistance reflexes are unwanted during walking. During the early part of stance, when the leg is firmly planted and bearing a load, a resistance reflex to upward perturbation of the leg would be helpful. Depolarization of crayfish leg depressor MNs, which occurs in stance, increases their resistance reflex responses to CBCO afferents, and so may help resist upward leg perturbations during early stance (Le Bon-Jego et al. 2006). Reflex Reversal As mentioned above, in active preparations, imposed sinusoidal CBCO

movement entrains a rhythm with reversed reflex responses: CBCO stretch, which occurs during leg depression, evokes a depressor MN burst, and CBCO release, which occurs during leg elevation, evokes a levator MN burst (Le Ray and Cattaert 1997).

277

278

Neurobiology of Motor Control: Fundamental Concepts and New Directions

CSDs also reverse their reflexes in active preparations and entrain levator and depressor MN activity (Leibrock et al. 1996). Reflex reversal also occurs at other joints in the same leg upon transition from a quiescent to an active state. For example, protraction and retraction of TCMRO/TCCO stretch receptors evoke resistance reflexes in the quiescent state and assistance reflexes in the active state (DiCaprio and Clarac 1981; Sillar et al. 1986; Skorupski et al. 1992). Similar reflex reversals occur in insects. In stick insect, the femoral chordotonal organ (fCO) responds to flexion and extension of the flexor–tibia joint. fCO-mediated resistance reflexes become assistance reflexes when nervous system state shifts from quiescent to active (Bässler 1976; Büschges 2012; Zill et al. 2012). In cockroach, leg remotion during walking produces strains in the leg cuticle that excite the trochanteral campaniform sensilla; they then excite leg depressor MNs that produce leg remotion. This positive feedback, assistance reflex response is enhanced when an added load increases resistance to leg remotion during walking (Pearson 1972). Reflex reversal depends on context as well as nervous system state (Hellekes et al. 2012). Reflex reversal in stick insects is enhanced during forward, but not backward, walking, and by stepping of the rostral ipsilateral leg, but not contralateral legs. During curve walking, fCO promotes leg flexion of the inside middle leg but not the outside leg. Mechanisms of Reflex Reversal Network pathways for assistance reflexes are less well

understood than resistance reflex pathways. Two mechanisms for assistance reflexes have been proposed: persistent changes in the relative strength of parallel excitatory and inhibitory pathways from sensory neurons to motor neurons, and a switch in the reflex pathway to the opposite side of the half-center CPG (Büschges et al. 2011; Cattaert and Le Ray 2001). Details of these mechanisms have been obtained for the leg depressor assistance reflex in crayfish (Clarac and Cattaert 1996; Le Ray and Cattaert 1997). In quiescent preparations, 8 of the 12 leg depressor MNs were excited (resistance reflex response) by CBCO release (leg lift) and one depressor MN was excited (assistance reflex response) by CBCO stretch (leg depression). In active preparations, CBCO release excited few or no depressor MNs, whereas CBCO stretch excited strong assistance reflex responses from several depressor MNs. These assistance responses were mediated by an assistance reflex interneuron (ARIN) that forms a di-synaptic pathway between stretch-sensitive CBCO afferents and depressor MNs (Le Ray and Cattaert 1997). ARIN was inhibited in the quiescent state, but was responsive in the active state, in which post-synaptic responses were amplified by a voltage-sensitive inward current. Phase-sensitive modulation of this network may come from CPG inputs that depolarize ARIN at the swing-to-stance transition (Cattaert and Le Ray 2001). Similar inputs may gate pre-synaptic inhibition of primary afferents to inhibit resistance reflexes when they would inappropriately oppose CPG driven leg movements. The role of assistance reflexes during walking became apparent in experiments in which the sensory feedback loop was opened or closed at will. This was possible with a “hybrid preparation” in which an in vitro crayfish thoracic nervous system and CBCO were attached to a computational neuromechanical model of the crayfish body and 5th leg (Fig. 9.6A) (Chung et al. 2015). Recorded activity of levator and depressor motor nerves from the 5th thoracic ganglion excited model levator and depressor leg muscles, and their contractions drove the model leg up and down (Fig. 9.6B). The live CBCO was

Sensory Feedback in the Control of Posture and Locomotion

A

Crayfish leg and body model as

le Re ch ret

St

CO CB

Thorax Depressors

e

Crayfish in vitro nerve cord

Interface CBCO

OXO in Probe drive

CBCO n Lev n

Levators

T4

Lev

Release

T5 Openclose loop

CBCO Stretch Elevate Depress

T3

Coxa

Leg

Dep n

Dep CB joint Basipodite

B CB angle

OXO out

Closed loop

A1

Open loop 100°

CBCO length

1 mm

CBCO n

Lev n

Dep n

60 s

Figure 9.6 Closed loop responses in a hybrid biological/neuromechanical preparation of crayfish leg elevation/depression. (A) Experimental set-up. Recorded activity from the depressor (Dep n) and levator (Lev n) nerves (right) drives corresponding muscles of a neuromechanical model of the leg and body (left). Leg up and down movements release and stretch a model CBCO. Model CBCO movements cause a speaker-driven probe to move the real CBCO identically. CBCO afferent projections to the nervous system complete the feedback loop. Switches in the interface between the preparation and model (middle) permit opening the feedback loop. Oxotremorine, a muscarinic agonist, was applied to the preparation to make it active. (B) Activity of model leg (CB) angle (top), CBCO length, and CBCO n, Lev n, and Dep n activity in trials with feedback loop closed, open, and when simulation was not running (rest of time). Adapted from Chung et al. (2015).

279

280

Neurobiology of Motor Control: Fundamental Concepts and New Directions

released and stretched according to the model leg movements, and the resulting CBCO afferent activity projected back to the thoracic ganglion to complete the feedback loop. When the nerve cord was in a quiescent state, brief imposed lifts of the model leg evoked chained resistance reflexes, exciting depressor MNs during leg lift and levator MNs during leg fall (Chung et al. 2015). Exposure to oxotremorine, a muscarinic agonist, caused the cord to gradually enter the active state. Leg lifts made early in this state evoked levator MN assistance reflexes, a levator MN burst immediately followed by a depressor MN burst. Shortly thereafter, these levator/depressor burst pairs occurred spontaneously at low frequency (∼1/25 s) when the feedback loop was open and became nearly three times faster when the feedback loop was intact. During these closed loop periods, spontaneous levator/depressor burst pairs produced rhythmic up and down leg movements every 10 s that ended only when the feedback loop was opened at the end of the model simulation. The assistance reflexes evoked by upward leg movement in closed loop conditions appear to be responsible for the increased frequency of motor burst pairs and up/down leg movements. The levator/depressor burst pairs that occurred in open loop were largely unchanged in closed loop, but the interval between the depressor burst of one burst pair and the levator burst of the next shortened significantly. Levator activity increased gradually following each depressor burst, and in closed loop the levator MNs evoked small upward leg movements that were then amplified by the assistance reflex. As occurred with imposed leg lifts, the assisted upward movements triggered the levator/depressor burst pair well before it would have occurred without the intact feedback loop. The early triggering of the burst pair in each cycle reset the levator/depressor CPG each cycle and accounted for the increased frequency of the levator/depressor rhythm. These conclusions were supported by simulations using a model of the hybrid preparation (Fig. 9.7A) (Bacque-Cazenave et al. 2015). In this model, a model neural network was developed from descriptions of the networks that control levator and depressor MNs during quiescent and active states (Cattaert and Le Ray 2001). The model network was substituted for the in vitro nervous system, linked to the same leg and body model used in the hybrid experiments, and used to simulate the leg lift experiments in the quiescent and active states and the open and closed loop configurations in the active state. The model reproduced all of the results of the hybrid experiments (Chung et al. 2015), including increased MN bursting frequency in closed loop conditions (Fig. 9.7B) (Bacque-Cazenave et al. 2015). This showed that the circuit configurations of the quiescent and active states (Cattaert and Le Ray 2001) could account for the reflex reversal and the effects of opening and closing the feedback loop on levator/depressor bursting behavior. Moreover, it showed that the ARINs and their assistance reflexes played an essential role in speeding the bursting rhythm in the closed loop, active state preparation. These data suggest that the TCMRO/TCCO feedback loops that govern TC joint forward/backward movements likely have similar accelerating effects on motor rhythm frequency. Open loop experiments have shown that assistance reflexes occur on both stretch and release when the TCMRO and TCCO are stimulated together as they are in the intact leg, and that these reflexes can entrain the promotor/remotor CPG (Sillar et al. 1986; Skorupski et al. 1992). Moreover, the subpopulation of promotor and remotor MNs excited in assistance reflexes has lower thresholds for excitation

Sensory Feedback in the Control of Posture and Locomotion

A

B

Figure 9.7 (A) The neuromechanical model of the hybrid preparation. A neural network containing integrate and fire model neurons replaces the in vitro preparation. (B) Model activity under open loop and closed loop conditions when the network was activated by stimulating the OXO neuron to simulate oxotremorine effects. Excitatory synapses: forks; inhibitory synapses: filled circles. Adapted from Bacque-Cazenave et al. (2015).

281

282

Neurobiology of Motor Control: Fundamental Concepts and New Directions

by the promotor/remotor CPG. It appears likely that, when leg elevation assistance reflexes mediated by the CBCO trigger the levator/depressor burst pair, leg promotion assistance reflexes mediated by the TCMRO/TCCO will trigger a near-simultaneous promotor/remotor burst pair. Feedback in Locomotor Phase Transitions Stance/swing and swing/stance transitions are

also promoted in decapod crustaceans by feedback pathways from other leg sensory organs. Stimulation of funnel canal organs (FCO) in the dactyl of the crab leg facilitates stance/swing transitions during walking by exciting levator and promotor MNs and inhibiting depressor and remotor MNs (Libersat et al. 1987a,b). In response to FCO stimulation in crayfish, levator excitation was accompanied by depressor inhibition in the stimulated leg to facilitate its swing phase and by depressor excitation in adjacent legs to promote their stance phase (Clarac et al. 1991). In rock lobster, FCO nerve stimulation cut stance short and triggered swing, whereas stimulation at the end of swing delayed the onset of the next stance (Muller and Clarac 1990). CS play similar roles in insects by responding to the changes in force in the leg cuticle that accompany swing-stance transitions. In cockroach, distal CS receptors in the middle leg were strongly excited towards the end of stance as force on the leg decreased when the hind leg entered stance and took up the ipsilateral load (Zill et al. 2009). A similar pattern of CS activation occurs in locust legs, in which the decline in longitudinal tensile forces on the dorsal tibia prior to swing excites CS receptors (Newland and Emptage 1996). These help excite flexor motorneurons and inhibit extensor motorneurons, promoting the stance/swing transition. In the stick insect, CS input enhances retractor motor neuron activity during forward walking (Büschges 2012). The same CS input inhibits retractor motor neuron activity during backward walking, providing an example of context-dependent reflex reversal. Insect joint stretch receptors also facilitate stance/swing transitions. The position response of the stick insect femoral chordotonal organ (fCO1 ) promotes stance/swing transition (Bässler 1993). Synergy between CS and fCO responses further promotes these transitions. Femoral CS stimulation alone inhibits leg extensor motorneurons at the end of stance more reliably but more slowly than fCO stimulation alone. Simultaneous stimulation of both receptor organs inhibits extensors with the reliability of CS stimulation and the speed of fCO stimulation (Akay and Büschges 2006). Swing/stance transitions similarly depend on feedback, primarily ground contact (Duysens et al. 2000). Motor neuron discharge during stance also depends on stance velocity, which varies throughout stance phase (Gabriel and Büschges 2007).

9.6 Organization and Function in Vertebrates 9.6.1 Sensory Feedback in the Maintenance of Posture

As in arthropods, postural control in vertebrates involves maintaining a specific body configuration, a set of joint angles, that keeps the body above the feet during standing, locomotion, and other movements, and resists sudden perturbations of posture. Proprioception, along with tactile, visual, and vestibular information, are involved in postural control to varying degrees depending on the number of legs on the ground, animal body

Sensory Feedback in the Control of Posture and Locomotion

design, postural task, and perturbation characteristics. Quiet standing in quadrupeds is statically stable, meaning that the vertical projection of body center of mass is well within the area of support and small displacements of this projection therefore do not threaten body balance (Gray 1968; Horak and Macpherson 2011). In bipeds (birds, kangaroos, humans), quiet standing, and especially locomotion, are inherently mechanically unstable; even when the vertical projection of the center of mass is within the support area, without appropriate postural responses, small body sways accelerate the center of mass toward the border of support and lead to loss of balance (Hof et al. 2005; Winter 1995). Neural control mechanisms in human standing and postural reactions to perturbation are often studied by modeling the body as a single or double inverted pendulum with rotational axes at the ankle or ankle and hip joints. Body equilibrium in these models is controlled by joint moments generated by a feedback PID controller (Fig. 9.1) (Li et al. 2012; Masani et al. 2006; Peterka 2002). Postural control involves integration of multiple sources of sensory feedback with feed forward motor commands at multiple levels in the central nervous system (Horak and Macpherson 2011; Nashner et al. 1989). The specific contribution each input makes to spinal motorneuron pools is difficult to establish experimentally in vivo. Therefore, various experimental paradigms, reduced animal preparations, and neuromechanical modeling are necessary to fully understand the mechanisms of sensory control of posture. Muscle Length Feedback Unexpected shifts of the vertical projection of body center of

mass toward the border of the support area lead to re-configuration of body limbs and corresponding changes in the length of muscles affected by the limb joint motion, i.e., some muscles stretch while their anatomical antagonists shorten. Greater lengthening and shortening occur in muscles with longer moment arms (An et al. 1984; Landsmeer 1961). Muscles with greater stretch have a greater mechanical advantage in generating movements opposing joint angle and limb configuration changes. Muscle stretch produces force responses due to intrinsic muscle properties and stretch-evoked muscle reflexes. Intrinsic musculoskeletal properties contribute substantially to postural control. The position restoring response to externally imposed limb displacements, the apparent stiffness of a multi-joint limb, depends in part on limb configuration. This property derives directly from geometric analysis of the transformation between small displacements of the limb endpoint and the corresponding changes in limb joint angles (Hogan 1985; Mussa-Ivaldi et al. 1985). Human arm and cat hindlimb forces generated in response to sudden small horizontal shifts of limb endpoint are restrained to directions passing close to the limb proximal joint (Macpherson 1988a; Mussa-Ivaldi et al. 1985). This response appears to be a consequence of limb musculoskeletal anatomy irrespective of muscle activation (Bunderson et al. 2010). Intrinsic muscle force-length and force-velocity properties (muscle short-range stiffness) instantaneously resist muscle stretch before any stretch reflexes are evoked (Joyce et al. 1969; Malamud et al. 1996; Rack and Westbury 1969, 1974). Short-range stiffness is explained by elastic deformation of engaged cross-bridges and thus operates only over a relatively short muscle fiber elongation range, beyond which the bridges disengage and the rate of muscle force development decreases or changes sign (Flitney and Hirst 1978). Because short-range stiffness depends on the number of engaged cross-bridges,

283

284

Neurobiology of Motor Control: Fundamental Concepts and New Directions

its magnitude increases with muscle activation. Stretch reflexes, in particular the fastest monosynaptic stretch reflex, increase the length range over which muscle intrinsically resists imposed stretch (Huyghues-Despointes et al. 2003; Nichols and Houk 1976). Short-range stiffness is the first line of defense against postural perturbations, operating in the time period before muscle length feedback reaches the central nervous system and engages spinal stretch reflexes. Stretch reflexes play the major role in generating postural responses to modest perturbations, ones not causing dramatic body displacements (Fig. 9.4). Muscle responses to small horizontal displacement of the support surface are directionally tuned (Henry et al. 1998; Macpherson 1988b)—muscle response is greater in directions in which a muscle is more stretched by the perturbation and provides more effective resistance to limb configuration change (Bunderson et al. 2010; Honeycutt and Nichols 2014). This directional tuning of muscle postural responses is at least partially mediated by muscle length sensitive afferents: activity of group Ia and II spindle afferents from medial gastrocnemius and biceps femoris muscles of anesthetized or decerebrate cats is tuned in the directions of postural responses of these muscles and opposite to the directions of support surface displacement (Honeycutt et al. 2012). This directional tuning arises from the directional dependence of limb muscle moment arms, which controls the amount the muscle is stretched by perturbation. This dependence results in perturbations activating muscles with the largest mechanical advantage for producing moments that correct the perturbation. The magnitude of muscle stretch and spindle afferent responses to postural perturbations depends on the size and rate of the perturbations (Honeycutt et al. 2012). In cats (Eng and Hoffer 1997) and humans (Day et al. 2013; Loram et al. 2009), during quiet standing and relatively small perturbations of the ankle joint, muscle–tendon unit length changes in ankle extensors do not correlate well with muscle responses. This poor correlation is due to the long and compliant muscle tendon, which absorbs substantial length changes in the muscle-tendon unit complex. During standing, length changes in muscle fascicles are more closely related to muscle activity changes (Day et al. 2013; Eng and Hoffer 1997), which is expected given that the muscle spindles are embedded in parallel with muscle fascicles. Other sensory signals are also involved in muscle directionally tuned postural responses. Selective destruction of large-diameter sensory axons (primarily Ia and Ib fibers) in cat hindlimb delays muscle responses to horizontal support displacements two to three times but does not affect directional tuning of muscle activity (Stapley et al. 2002). Removal of hindlimb cutaneous feedback in decerebrate cats reduces the magnitude, but not the directional tuning, of muscle responses to support perturbations (Honeycutt and Nichols 2010). Load-Tactile Feedback Postural perturbations change the pressure on the skin of the limb

end-segment in contact with the support and loading on affected muscles. Skin pressure and deformation are detected by skin mechanoreceptors (Abraira and Ginty 2013) whereas changes in muscle forces are registered by Golgi tendon organs (Jami 1992). Even a very light fingertip touch of a stationary object profoundly reduces postural sway in standing humans (Jeka and Lackner 1994), possibly by providing information about the direction of body sway. The ability of skin mechanoreceptors in contact with the support surface to sense the direction of surface shift has been implicated in shaping

Sensory Feedback in the Control of Posture and Locomotion

muscle and joint moment directional responses to postural perturbations (Jacobs and Macpherson 1996; Meyer et al. 2004). Detection of perturbing horizontal forces by tactile sensitive afferents in the foot or paw provides information about the direction of the center of mass acceleration, as the ground reaction forces reflect the resultant external forces, and thus acceleration, at the body center of mass. This may provide an explanation for the fact that the body center of mass acceleration and other center of mass kinematic variables are predictive of muscle long-latency postural responses in humans (Safavynia and Ting 2013). Acute denervation of the paw pad in decerebrate cats, alternatively, does not affect the directional tuning and latency of muscle postural responses to horizontal shifts of the support surface and reduces only the background activity and magnitude of muscle responses. In these experiments, the animal’s head was fixed, and hence visual and vestibular inputs were unchanged by the support shifts. Human postural control possibly differs from that of cats because a standing human, accurately represented as an inverted pendulum with the rotational axis at the ankle joint, is inherently unstable (Winter 1995). Ankle joint angle, fascicle length (and thus muscle spindle length) of ankle extensors, and center of mass position changes, do not correlate with ankle extensor activity. Rather, the human postural control system responds to changes in center of mass velocity with a delay of several hundred milliseconds (Loram et al. 2005), thus exhibiting integral feedback control (Fig. 9.1). Center of mass velocity during postural sway could be potentially derived from horizontal forces sensed by cutaneous afferents on the plantar surface of the foot, as discussed above. 9.6.2 Sensory Feedback and its Integration with Motor Commands in Movement

During movements, muscle length and force-dependent reflexes are modulated, and sensory feedback pathways reorganized, to meet a wide range of behavioral requirements. This reflex modulation is required, as in invertebrates, because posture stabilizing reflexes would be detrimental for movement production. For example, during quiet standing in the cat, extensor muscles are active to support the body against gravity and antagonistic flexor muscles are relaxed (Torres-Oviedo et al. 2006). If standing posture is perturbed by a horizontal shift of the support, muscles stretched by the perturbation, irrespective of their extensor or flexor function, respond with strong corrective activity (Torres-Oviedo et al. 2006). This response is automatic and mediated in part by length-dependent sensory pathways in the spinal cord (Macpherson and Fung 1999). If these reflexes were operational when the cat made a step by flexing and then extending the limb, extensor muscles being stretched during flexion, and flexor muscles stretched during extension, would activate stretch reflexes opposing the intended movement. The nervous system resolves this problem (von Holst and Mittelstadt 1950) by changing spinal reflex properties and reconfiguring interneuronal pathways. Taskand phase-dependent reorganizations of reflexes during movements occur at the levels of synaptic transmission of afferent signals, spinal locomotor pattern generating networks, and supraspinal centers. One important mechanism disabling stretch reflex in antagonistic muscles during movements is reciprocal inhibition via Ia inhibitory interneurons—agonist motorneuron activity coincides with activity of Ia inhibitory interneurons that inhibit antagonist motorneurons (Feldman and Orlovsky 1975) (X-Inhib interneurons in Fig. 9.4A).

285

286

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Regulation of Spinal Stretch Reflexes The gain of stretch reflex from agonists is substan-

tially reduced during movements compared to the rest condition or postural tasks. For example, the Hoffman (H-) reflex, which is evoked by electrical stimulation of Ia afferents instead of by spindle stretch (Misiaszek 2003), is lower during human running than walking, and lower during walking than standing (Stein and Capaday 1988). In decerebrate cats, the gain of the stretch reflex is similarly reduced during walking compared to rest (Bennett et al. 1996). Compared to rest, the amplitude of motorneuron monosynaptic excitatory post-synaptic potentials (EPSPs) evoked by group I afferent electrical stimulation in the decerebrate cat is significantly depressed during brainstem-evoked fictive locomotion (Gosgnach et al. 2000). This depression arises from tonic suppression by spinal GABAergic interneurons of neurotransmitter release at group Ia afferent synapses onto motorneurons and interneurons (Fig. 9.4). Genetic elimination of spinal GABAergic interneurons in mice, and hence removal of pre-synaptic inhibition of length-dependent afferent feedback, results in severe forearm oscillations during voluntary reaching movements but does not impair locomotion (Fink et al. 2014). These results, and simulations of limb movements with high gain length feedback (Fig. 6 in Fink et al. 2014), suggest that reduced gains of length feedback during voluntary reaching ensures smooth performance. Other mechanisms of reflex gain regulation also occur during movements. One of these is fusimotor control of muscle length feedback (Hulliger et al. 1989; Taylor et al. 2006), which adjusts static and dynamic responses of length sensitive muscle spindle afferents to specific motor task requirements (Fig. 9.4) (Hunt 1990; Matthews 1981). Tendon stretches during walking in decerebrate cats (Stein et al. 2000) and computer simulations (Yakovenko et al. 2004) show that length dependent feedback from ankle extensor muscles contribute about 30% to their locomotor activity. Similar estimates have been obtained in humans by imposing quick stretches in ankle extensor muscles during walking (Mazzaro et al. 2005). On the other hand, permanent removal of stretch reflexes from selected ankle (Gregor et al. 2014; Pantall et al. 2016) and knee extensors (Mehta et al. 2014) by self-reinnervation of these muscles (Alvarez et al. 2010; Bullinger et al. 2011; Cope et al. 1994; Cope and Clark 1993; Lyle et al. 2016) does not decrease muscle activity during level, upslope, and downslope walking (Fig. 9.8) and causes (minor) kinematic deficits in only downslope walking (Abelew et al. 2000; Maas et al. 2007). Thus, effects of length dependent feedback on activity of ankle and knee extensors during locomotion are unclear and require further study. In another rhythmic task, cat paw shake response, removing knee extensor length-dependent feedback changes vastus medialis activity (Mehta et al. 2014), but removing ankle extensor length-dependent feedback does not alter ankle extensor activity (Mehta and Prilutsky 2014). This dependence of vastii activity and independence of ankle extensor activity on length dependent feedback during paw shake is consistent with data obtained by constraining knee motion during paw shake and hindlimb deafferentation in spinal cats (Smith and Zernicke 1987). Thus, in some rhythmic tasks, length dependent feedback from knee extensors is critical for correct movement execution. Stance-Swing and Swing-Stance Transitions Length-sensitive hip muscle spindle group Ia

and II afferents help regulate locomotor phase transitions. Manipulating hip joint angle

Sensory Feedback in the Control of Posture and Locomotion

B EMG

C

No SO-LG length feedback SO

0.5 mV

LG

EMG

0.5 mV

SO

0.5 mV

EMG

EMG

Intact SO-LG length feedback

LG

0.5 mV

A

D SO

34 MG

20

SO

40

Fascicle length (mm)

Fascicle length (mm)

40

34 MG

20

14

14 0.5 s

0.5 s Intact VL length feedback 3 weeks after VL nerve cut & repair No VL length feedback

E

F 0.8 Cat CL

Cat ES

EMG (mV)

EMG (mV)

0.2

0.1

0

0.4

0 0

20

40

60

80

Normalized cycle time(%)

100

0

20

40

60

80

100

Normalized cycle time(%)

Figure 9.8 Locomotor EMG activity and fascicle length in ankle and knee extensor muscles with length feedback removed by muscle self-reinnervation. Downward vertical arrows indicate stance onset, upward vertical arrows swing onset. (A, C) Low-pass filtered EMG activity and fascicle length of intact soleus (SO), lateral gastrocnemius (LG), and medial gastrocnemius (MG) muscles of a cat during level walking. (B, D) Low-pass filtered EMG activity and fascicle length of self-reinnervated SO and LG and intact MG muscles in the same animal as in A and C. Adapted from Pantall et al. (2016). (E, F) Low-pass filtered EMG activity of vastus lateralis (VL, mean) of two cats during 27∘ upslope walking in intact condition (black filled lines), 3 weeks after nerve cut and repair (gray filled lines), and after muscle self-reinnervation (12–14 weeks after nerve repair; unfilled gray lines).

287

288

Neurobiology of Motor Control: Fundamental Concepts and New Directions

entrains locomotor rhythms in spinal or decerebrate cat (Andersson and Grillner 1983; Hiebert et al. 1996; Kriellaars et al. 1994). Resisting or assisting hip flexion during walking increases and decreases flexor activity, respectively, in decerebrate or intact cats (Lam and Pearson 2001), whereas vibrating or stretching hip flexors or stimulating their nerves at intensities that activate group I and II afferents during the extensor (stance) phase of real or fictive locomotion shortens ongoing extensor activity and promotes the onset of flexor activity (Hiebert et al. 1996; Lam and Pearson 2002; Perreault et al. 1995). Thus, length-sensitive group I and II spindle afferents from hip flexor muscles regulate extensor (stance)-flexor (swing) phase transitions by providing excitatory input to the flexor CPG half-center (Fig. 9.9A, B). The role of length feedback from hip and ankle flexors in extensor–flexor phase transitions in cat locomotion has also been investigated using a neuromechanical model of spinal locomotion (Markin et al. 2016). In the model the central pattern generator for each hindlimb is controlled by hindlimb afferent signals from group I and II muscle and paw pad cutaneous afferents (Fig. 9.9A, B) and reproduces the muscle activity and mechanics of the walking cat. Removing length feedback from hip and ankle flexors causes the model to collapse after three strides (Fig. 9.9C3). This shows that length feedback from flexors regulating CPG stance-swing transitions is required for proper control of stance phase duration. There is also a strong correlation between hip flexion angle (McVea et al. 2005) or muscle-tendon unit length of hip extensors (Gregor et al. 2006) and the onset of extensor activity during level, downslope, and upslope cat walking, indicating the importance of length sensitive afferents from hip extensors in regulating flexor (swing)–extensor (stance) phase transitions during locomotion via excitatory input to the CPG half-extensor center (Fig. 9.9B). Hip muscle length-dependent feedback is necessary, but not sufficient, for locomotor phase resetting. This is illustrated by comparing hip joint angles during walking on a split belt treadmill with the two belts operating at different speeds. The slower moving leg has a shorter stride length and smaller joint range of motion than the hip of the faster leg (Malone and Bastian 2010). Stance-swing phase transitions can thus occur at different lengths of hip flexors and extensors. Cat Ankle Extensor Tension The other sensory cue for stance–swing transition is

load-dependent signals from group I ankle extensor afferents. Electrical stimulation of ankle extensor nerves at intensities activating group I spindle and Golgi tendon afferents during flexor activity in fictive or real locomotion resets locomotion to extensor activity, whereas the same stimulations during the extensor phase increase extensor activity and duration throughout the hindlimb (Gossard et al. 1994; Guertin et al. 1995). The locomotor rhythm resetting and changes in activity across hindlimb extensors suggest that the responses are mediated by the extensor half-center of the CPG (Fig. 9.9B). Thus, unloading of muscle force-sensitive group Ib Golgi tendon organ afferents and the reduction of input from length-sensitive group Ia spindle afferents due to muscle shortening at stance end diminish the excitatory input to the extensor CPG half-center and contribute to initiation of the extensor–flexor transition (Duysens and Pearson 1980; McCrea and Rybak 2008). Feedback signals from group I ankle extensor afferents may also provide inhibitory input to the flexor CPG half-center that modulates its activity and phase transitions (Pearson 2008).

Sensory Feedback in the Control of Posture and Locomotion

A

m hyth

rator

gene

B

R Pattern formation (controls motor synergies) Extensor

Flexor

ns

ro

u ne

or

ot

M

Afferent signals

Efferent signals

C

Figure 9.9 Neuromechanical simulation of spinal locomotion with selective removal of muscle proprioceptive feedback. (A) Model schematic. A 10 degree-of-freedom musculoskeletal system of two hindlimbs (only one shown) with 18 Hill-type muscles controlled by a two-level locomotor CPG consisting of a rhythm generator and a pattern formation network (Rybak et al. 2006a,b). CPG activity is regulated by sensory feedback from group I and group II muscle and paw pad cutaneous afferents. (B) Afferent projections onto the rhythm generator. f-RGF interneurons are excited by Ia and II spindle afferents from hip flexor iliopsoas (IP) and ankle flexor tibialis anterior (TA) muscles and these afferents excite the flexor half-center of the rhythm generator (RG-F). i-RGE interneurons are excited by Ia and II afferents of hip extensor biceps femoris anterior (BFA), Ib afferents of ankle extensor soleus (SO) and gastrocnemius (GA), and paw pad cutaneous afferents (cut. tibial) and these afferents excite the extensor half-center of the RG (RG-E). MNs are motor neurons. (C) Walking kinematics before and after removal of selected sensory feedback. Black stick figures are last 4 cycles of 40 s of walking with intact feedback before turning off specific feedback types; gray figures show effects of feedback removal. 1) Removal of length dependent Ia feedback from SO and GA. 2) Removal of force dependent Ib feedback from SO and GA. 3) Removal of length dependent Ia and II feedback from IP and TA. Adapted from Markin et al. (2016).

289

290

Neurobiology of Motor Control: Fundamental Concepts and New Directions

It is difficult to separate the contribution of length-dependent and force-dependent afferent feedback to activity in fictive or real locomotion because activation thresholds of group Ia and Ib afferents to muscle nerve electrical stimulations are similar and evoking the stretch reflex by stretching the muscle also increases muscle force due to muscle force-length-velocity properties (Joyce et al. 1969; Rack and Westbury 1969). Modeling can help identify the relative contributions of length and force feedback (e.g., Bacque-Cazenave et al. 2015; Ekeberg and Pearson 2005; Markin et al. 2010). In such neuromechanical modeling the CPGs of the neural control system and the motion-dependent afferent signals of the musculoskeletal system are implemented together (Fig. 9.9). Parameters of the two parts of the model are typically tuned separately to ensure that the neural model reproduces the activity in muscle nerves during fictive locomotion (i.e., without motion dependent feedback) (Rybak et al. 2006a,b), and the musculoskeletal model reproduces real locomotion using recorded muscle activity as input (Prilutsky et al. 2016). The neuromechanical model of cat hindlimb locomotion developed in Markin et al. (2016) reproduces muscle activity and mechanics of level walking and thus allows simulation of the effects of removal of length-dependent feedback from ankle extensors (Fig. 9.9C1) and hip and ankle flexors (Fig. 9.9C3). These simulations suggest that length feedback from ankle extensors during walking is not critical because its removal does not cause visible changes in walking mechanics. This conclusion is supported by the small changes in soleus and gastrocnemius EMG locomotor activity induced by removal of their stretch-reflexes in vivo (Fig. 9.8A-D). Furthermore, during early stance in level walking, when the muscle-tendon length of medial gastrocnemius muscle is increasing (Gregor et al. 2006; Maas et al. 2007), the muscle’s fascicles do not lengthen and hence the spindles embedded in parallel with the fascicles should not either (see also Griffiths 1991; Hoffer et al. 1989; Maas et al. 2009). Soleus fascicles, alternatively, lengthen in early stance (Fig. 9.8C, D), which should cause spindle stretching. Length-sensitive group Ia and II afferents from the triceps surae do increase their activity during early stance in freely walking cats (Loeb and Duysens 1979; Prochazka et al. 1977), however it is not known whether the recordings were made from soleus or gastrocnemius. Removing group Ib afferent input from soleus and gastrocnemius muscles in the Markin et al. model sharply increases stance duration and causes loss of walking stability within 3 strides (Fig. 9.9C2). Muscle load-dependent sensory signals from ankle extensors thus appear important for maintaining stable walking (see also Ekeberg and Pearson 2005). Load-dependent signals from cat paw pad cutaneous afferents also appear important for promoting extensor activity during the extensor (stance) phase and for controlling phase transitions by resetting the locomotor rhythm (McCrea 2001; Rybak et al. 2006b). Afferent Activity during Cat Locomotion The data presented above were obtained from

experimental or computational manipulations of sensory feedback and analysis of the resulting changes in motorneuron or muscle activity. Additional insight into the role of sensory feedback in controlling locomotion can be gained from analysis of afferent activity in freely moving cats. These recordings are typically made from dorsal root ganglia using implanted wire electrodes (Loeb et al. 1977; Prochazka et al. 1977) or an electrode array (Weber et al. 2007). Figure 9.10 shows the activity of muscle

Sensory Feedback in the Control of Posture and Locomotion

A

Figure 9.10 Ensemble activity of cat hindlimb muscle and paw cutaneous afferents during the walk cycle. (A) Each panel is mean in vivo recorded muscle EMGs, muscle-tendon unit lengths, and firing rates of group Ia, II, and Ib afferents from nine cat hindlimb muscles. The number of afferents contributing to each trace is shown in parentheses to the left of the traces. Question marks indicate that afferent activity was not recorded. (B1, B2) Activity of individual hair-cell receptors, recorded from L7 and S1 dorsal root ganglia, respectively, with receptive fields on the paw shown by arrows at right. (B3) Activity of a cat paw stretch-sensitive mechanoreceptor. Cell receptive field shown at right. Horizontal bars under traces indicate stance phase of walking. Adapted from Prochazka and Gorassini (1998), B1 and B2 from Loeb et al. (1977), B3 from Loeb (1981), all with permission.

291

292

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Figure 9.10 (Continued)

B Hair cell receptors 1

2 Cutaneous mechanoreceptors 3

length-sensitive spindle group Ia and II afferents and muscle force-sensitive Golgi tendon organ group Ib afferents from nine cat hindlimb muscles and the activity of three skin pressure mechanoreceptors from the paw (Loeb et al. 1977; Prochazka and Gorassini 1998). Activity of Ia and II spindle afferents from ankle flexors—tibialis anterior, extensor digitorum longus, peroneus longus—reaches peak values at stance end. This is consistent with their length changes during walking and the proposed role of this sensory feedback in triggering the stance-swing transition (Hiebert et al. 1996). Activity of group II spindle afferents from a hip flexor sartorius is also close to maximum at stance end when the muscle-tendon unit length approaches its peak. This feedback again appears to help reset locomotor activity from extension to flexion (Hiebert et al. 1996; Kriellaars et al. 1994). Activity of length sensitive spindle group Ia and II afferents from hip extensor posterior hamstrings increases monotonically during swing and reaches maximum values before stance onset (Fig. 9.10A). This activity is consistent with length sensitive afferents from hip extensors triggering the swing-to-stance transition (McVea et al. 2005) and the earlier onset of extensor muscle activity during downslope walking compared to level or upslope walking (Gregor et al. 2006). Group Ib afferents from ankle extensors—triceps surae, plantaris, and flexor digitorum longus (Fig. 9.10A)—and paw plantar cutaneous mechanoreceptors (Fig. 9.10B) are active during stance and thus could enhance extensor activity during stance and contribute to flexor phase onset at stance end when their activity decreases (McCrea and Rybak 2008; Pearson 2008). Some of the somatosensory activity recorded during free movements and shown in Fig. 9.10 can be substantially depressed by phasic and tonic pre-synaptic inhibition. Specifically, synaptic transmission from group Ia afferents to motorneurons and interneurons is reduced during locomotion compared to rest (Duenas and Rudomin 1988; Gosgnach et al. 2000). The above account shows that very similar problems confront the use of feedback to control posture and locomotion in arthropods and vertebrates, and evolution has responded with very similar mechanisms. Both groups employ efferent-controlled muscle receptors (MROs and muscle spindles) as key parts of negative feedback PID controllers to stabilize posture. Both use multiple other sensors to provide needed information for feedback control. Both use positive feedback embedded within negative feedback systems to speed locomotory phase transitions. The positive feedback is produced by reversing resistance reflexes to become assistance reflexes. Vertebrates and arthropods accomplish this by pre-synaptically inhibiting afferent terminals onto

Sensory Feedback in the Control of Posture and Locomotion

motor antagonists with PAD, and by disinhibiting di- and polysynaptic pathways to motor agonists. Phase transitions within a cycle are accomplished through sensory feedback, often mediated by assistance reflexes, that resets the rhythm. It is likely that more complex patterns of control, which may include feed-forward models, prediction of reafference and efference copy (Franklin and Wolpert 2011), are realized by similar mechanisms in arthropods and vertebrates.

9.7 Conclusions The descriptions of the role of sensory feedback in posture and locomotion given above illustrate how complex these systems are and how difficult it is to obtain a comprehensive understanding of how they work in any one animal. This difficulty also stems from the experimental necessity of focusing on individual elements and how they contribute to the function of small parts of the system or the entire system itself, as seen through the animal’s behavior. The role of individual sense organs, or of particular network elements, in locomotion and posture is difficult to determine because of the many simultaneous interactions that occur among the parts of the system in any one state, and because of the transformations in those interactions caused by changes in system state. These difficulties beg the question of how to understand feedback in the context of the entire system as the animal adopts particular postures, shifts from posture to walking, and changes gait or direction. Since the work of Hodgkin and Huxley (1952), computational models have been used to describe dynamic relationships among the elements of complex neural systems. Such approaches are particularly helpful for understanding sensorimotor integration and complex motor systems and movements (Bacque-Cazenave et al. 2015; Chiel et al. 2009; Cofer et al. 2010b; Ekeberg et al. 2004; Ekeberg and Grillner 1999; Markin et al. 2016; Pearson et al. 2006; Shadmehr and Wise 2005). Most models are built to account for specific, limited sets of phenomena seen in experiments or displayed by the animal under well-defined conditions. They capture the properties of specific network elements, sensory receptors, or muscle and limb biomechanics, and ask whether and over what range of parameter values their interaction can account for emergent properties of the system. When such models fail they are useful because they demonstrate gaps in our understanding. Interrogating the model can then suggest how those gaps might be filled through new experiments or by changes in the model. When they succeed they are useful because they provide insights into how adaptive properties emerge from the system and can be used to simulate future experiments that provide new tests of the model. We believe, however, that models attempting to reconstruct the entire system governing an animal’s postural and locomotor behavior are also needed. Such models would include the musculoskeletal structure of the animal, its relevant sense organs, sensory afferents, local networks, and descending inputs. It must also include relevant portions of the physical environment: ground and water surfaces, gravity, surface friction, hydrodynamics. These models would not replace models focused on understanding particular aspects of a system, such as specific network configurations (Fig. 9.7) or biophysical details (De et al. 2005). They would also not replace top-down models that capture overall postural or locomotor behavior (Ekeberg 1993, 2004). Rather, they would enable

293

294

Neurobiology of Motor Control: Fundamental Concepts and New Directions

relevant aspects of these models to be incorporated into an overarching model that would reveal how their dynamic interactions contribute to behavior. Building an entire system model is substantially easier in arthropods, with 1,000 neurons in each segmental ganglion and 95 leg motor neurons (Cattaert 2014), than in vertebrates, where the numbers are many thousands time larger. However, even for an arthropod this is not a task that a single research team can do. It is rather the task of a community of researchers, each of whom contributes to the parts of the system she or he knows best. Each addition creates opportunities for examining previously unexamined interactions. For example, adding visual systems to a model of motor control creates an opportunity to investigate visually guided motor control. This presupposes a common modeling environment in which community members have access to the relevant set of published models and can build on or revise them. It also assumes that the model becomes a knowledge base, in which each model part—neuron, synapse, muscle, sense organ—is linked to the source publications of the part. These annotations identify the experimental contexts from which model parameter values were obtained, and so identify the contexts in which simulations are valid. Such a model would provide a statement of what is currently understood about an animal, a statement that would be continually revised in light of new experimental knowledge, computational capabilities, and tests of model validity.

Acknowledgements Support was provided by NSF Grant IOS1120291 to DHE and NIH grants NS-048844, EB-012855, HD032571, and support from the Center for Human Movement Studies, Georgia Tech, to BIP.

Endnote 1 The lower case “f” in “fCO” distinguishes the femoral chordotonal organ of insects from

the “FCO”, the funnel canal organ of Crustacea, with a capital “F”.

References Abelew TA, Miller MD, Cope TC, Nichols TR (2000) Local loss of proprioception results in disruption of interjoint coordination during locomotion in the cat. J Neurophysiol 84:2709–2714. Abraira VE, Ginty DD (2013) The sensory neurons of touch. Neuron 79:618–639. Akay T, Büschges A (2006) Load signals assist the generation of movement dependent reflex reversal in the femur–tibia joint of stick insects. J Neurophysiol 96:3532–3537. Alvarez FJ, Bullinger KL, Titus HE, Nardelli P, Cope TC (2010) Permanent reorganization of Ia afferent synapses on motoneurons after peripheral nerve injuries. Ann NY Acad Sci 1198:231–241. An KN, Takahashi K, Harrigan TP, Chao EY (1984) Determination of muscle orientations and moment arms. J Biomech Eng 106:280–282.

Sensory Feedback in the Control of Posture and Locomotion

Andersson O, Grillner S (1983) Peripheral control of the cat’s step cycle. II. Entrainment of the central pattern generators for locomotion by sinusoidal hip movements during “fictive locomotion.”. Acta Physiol Scand 118:229–239. Bacque-Cazenave J, Chung BP, Cofer D, Cattaert D, Edwards DH (2015) The effect of sensory feedback on crayfish posture and locomotion: II. Neuromechanical simulation of closing the loop J Neurophysiol 113:1772–1783. Bässler U (1976) Reversal of a reflex to a single motoneuron in the stick insect Carausius morosus. Biol Cyber 24:47–49. Bässler U (1993) The femur–tibia control system of stick insects—a model for the study of the neural basis of joint control. Brain Res Rev 18:207–226. Bender JA, Pollack AJ, Ritzmann RE (2010) Neural activity in the central complex of the insect brain is linked to locomotor changes. Curr Biol 20:921–926. Bennett DJ, De Serres SJ, Stein RB (1996) Gain of the triceps surae stretch reflex in decerebrate and spinal cats during postural and locomotor activities. J Physiol 496:837–850. Bowerman RF, Larimer JL (1974a) Command fibres in the circumoesophageal connectives of crayfish. I. Tonic fibres. J Exp Biol 60:95–11. Bowerman RF, Larimer JL (1974b) Command fibres in the circumoesophageal connectives of crayfish. II. Phasic fibres. J Exp Biol 60:119–134. Brown TG (1911) The intrinsic factors in the act of progression in the mammal. Proc R Soc Lond 84:308–319. Bullinger KL, Nardelli P, Pinter MJ, Alvarez FJ, Cope TC (2011) Permanent central synaptic disconnection of proprioceptors after nerve injury and regeneration. II. Loss of functional connectivity with motoneurons. J Neurophysiol 106:2471–2485. Bunderson NE, McKay JL, Ting LH, Burkholder TJ (2010) Directional constraint of endpoint force emerges from hindlimb anatomy. J Exp Biol 213:2131–2141. Burrows M (1996) The Neurobiology of an Insect Brain. Oxford, UK: Oxford University Press. Burrows M, Laurent G (1993) Synaptic potentials in the central terminals of locust proprioceptive afferents generated by other afferents from the same sense organ. J Neurosci 13:808–819. Burrows M, Laurent GJ, Field LH (1988) Proprioceptive inputs to nonspiking local interneurons contribute to local reflexes of a locust hindleg. J Neurosci 8:3085–3093. Büschges A (2012) Lessons for circuit function from large insects: towards understanding the neural basis of motor flexibility. Curr Opin Neurobiol 22:602–608. Büschges A, Scholz H, El Manera A (2011) New moves in motor control. Curr Biol 21:R513–R524. Büschges A, Wolf H (1999) Phase-dependent pre-synaptic modulation of mechanosensory signals in the locust flight system. J Neurophysiol 81:959–962. Cattaert D (2014) Neural control of posture and walking in crustaceans. In Crustacean Nervous Systems and Their Control of Behavior, Derby CD, Thiel M (eds). Oxford, UK: Oxford University Press. Cattaert D, El Manira A (1999) Shunting versus inactivation: analysis of pre-synaptic inhibitory mechanisms in primary afferents of the crayfish. J Neurosci 19:6079–6089. Cattaert D, El Manira A, Bevengut M (1999) Pre-synaptic inhibition and antidromic discharges in crayfish primary afferents. J Physiol Paris 93:349–358.

295

296

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Cattaert D, El Manira A, Clarac F (1992) Direct evidence for pre-synaptic inhibitory mechanisms in crayfish sensory afferents. J Neurophysiol 67:610–624. Cattaert D, Le Ray D (2001) Adaptive motor control in crayfish. Prog Neurobiol 63:199–240. Chiel HJ, Ting LH, Ekeberg O, Hartmann MJ (2009) The brain in its body: motor control and sensing in a biomechanical context. J Neurosci 29:12807–12814. Chrachri A, Clarac F (1989) Synaptic connections between motor neurons and interneurons in the fourth thoracic ganglion of the crayfish, Procambarus clarkii. J Neurophysiol 62:1237–1250. Chung B, Bacque-Cazenave J, Cattaert D, Edwards DH (2015) The effect of sensory feedback on crayfish posture and locomotion: I. Experimental analysis of closing the loop J Neurophysiol 113:1763–1771. Clarac F (2008) Some historical reflections on the neural control of locomotion. Brain Res Rev 57:13–21. Clarac F, Cattaert D (1996) Invertebrate pre-synaptic inhibition and motor control. Exp Brain Res 112:163–180. Clarac F, Chrachri A, Cattaert D (1991) Interneuronal control of a walking leg reflex in an in vitro crayfish preparation. In Locomotor Neural Mechanism in Arthropods and Vertebrates, Armstrong DM, Bush BMH (eds). Manchester, UK: Manchester University Press. Clarac F, Pearlstein E (2007) Invertebrate preparations and their contribution to neurobiology in the second half of the 20th century. Brain Res Rev 54:113–161. Clarac F, Vedel JP, Bush BM (1978) Intersegmental reflex coordination by a single joint receptor organ (CB) in rock lobster walking legs. J Exp Biol 73:29–46. Cofer D, Cymbalyuk G, Reid J, Zhu Y, et al. (2010a) AnimatLab: A 3D graphics environment for neuromechanical simulations. J Neurosci Methods 187:280–288. Cofer DW, Cymbalyuk G, Heitler WJ, Edwards DH (2010b) Neuromechanical simulation of the locust jump. J Exp Biol 213:1060–1068. Cope TC, Bonasera SJ, Nichols TR (1994) Reinnervated muscles fail to produce stretch reflexes. J Neurophysiol 71:817–820. Cope TC, Clark BD (1993) Motor-unit recruitment in self-reinnervated muscle. J Neurophysiol 70:1787–1796. Day JT, Lichtwark GA, Cresswell AG (2013) Tibialis anterior muscle fascicle dynamics adequately represent postural sway during standing balance. J Appl Physiol 115:1742–1750. De SE, Ekeberg O, Kotaleski JH, Achard P, Lansner A (2005) Biophysically detailed modelling of microcircuits and beyond. Trends Neurosci 28:562–569. Derby CD, Thiel M editors (2010) Crustacean Nervous Systems and Control of Behavior. Oxford, UK: Oxford University Press. DiCaprio RA, Clarac F (1981) Reversal of a walking leg reflex elicited by a muscle receptor. J Exp Biol 90:197–203. DiCaprio RA, Clarac F (1983) Reversal of an intersegmental reflex elicited by a muscle receptor organ. J Exp Biol 103:303–306. Duenas SH, Rudomin P (1988) Excitability changes of ankle extensor group Ia and Ib fibers during fictive locomotion in the cat. Exp Brain Res 70:15–25. Duysens J, Clarac F, Cruse H (2000) Load-regulating mechanisms in gait and posture: comparative aspects. Physiol Rev 80:83–133.

Sensory Feedback in the Control of Posture and Locomotion

Duysens J, Pearson KG (1980) Inhibition of flexor burst generation by loading ankle extensor muscles in walking cats. Brain Res 187:321–332. Eccles JC, Magni F, Willis WD (1962) Depolarization of central terminals of group I afferent fibres from muscle. J Physiol 160:62–93. Edwards DH (1990) Mechanisms of depolarizing inhibition at the crayfish giant motor synpase. I. Electrophysiology. J Neurophysiol 64:532–540. Ekeberg O (1993) A combined neuronal and mechanical model of fish swimming. Biol Cybern 69:373–374. Ekeberg O, Blumel M, Büschges A (2004) Dynamic simulation of insect walking. Arthropod Struct Dev 33:287–300. Ekeberg O, Grillner S (1999) Simulations of neuromuscular control in lamprey swimming. Philos Trans R Soc Lond B 354:895–902. Ekeberg O, Pearson K (2005) Computer simulation of stepping in the hind legs of the cat: an examination of mechanisms regulating the stance-to-swing transition. J Neurophysiol 94:4256–4268. El Manira A, Cattaert D, Clarac F (1991a) Monosynaptic connections mediate resistance reflex in crayfish (Procambarus clarkii) walking legs. J Comp Physiol A 168:1432–1351. El Manira A, DiCaprio RA, Cattaert D, Clarac F (1991b) Monosynaptic interjoint reflexes and their central modulation during fictive locomotion in crayfish. Eur J Neurosci 3:1219–1231. Eng JJ, Hoffer JA (1997) Regional variability of stretch reflex amplitude in the cat medial gastrocnemius muscle during a postural task. J Neurophysiol 78:1150–1154. Evoy WH, Ayers J (1982) Locomotion and control of limb movements. In Neural Integration and Behavior, Sandeman DC, Atwood HL (eds). New York, NY: Academic Press. Feldman AG, Orlovsky GN (1975) Activity of interneurons mediating reciprocal 1a inhibition during locomotion. Brain Res 84:181–194. Fink AJ, Croce KR, Huang ZJ, Abbott LF, et al. (2014) Pre-synaptic inhibition of spinal sensory feedback ensures smooth movement. Nature 509:43–48. Flitney FW, Hirst DG (1978) Cross-bridge detachment and sarcomere ‘give’ during stretch of active frog’s muscle. J Physiol 276:449–465. Franklin DW, Wolpert DM (2011) Computational mechanisms of sensorimotor control. Neuron 72:425–442. Gabriel JP, Büschges A (2007) Control of stepping velocity in a single insect leg during walking. Phil Trans R Soc A 365:251–271. Gosgnach S, Quevedo J, Fedirchuk B, McCrea DA (2000) Depression of group Ia monosynaptic EPSPs in cat hindlimb motoneurones during fictive locomotion. J Physiol 526:639–652. Gossard JP, Bouyer L, Rossignol S (1999) The effects of antidromic discharges on orthodromic firing of primary afferents in the cat. Brain Res 825:132–145. Gossard JP, Brownstone RM, Barajon I, Hultborn H (1994) Transmission in a locomotor-related group Ib pathway from hindlimb extensor muscles in the cat. Exp Brain Res 98:213–228. Govind CK, Atwood HL (1982) Organization of neuromuscular systems. In The Biology Of Crustacea. Volume 3: Neurobiology, Structure and Function. Bliss DE, Sandeman DC, Atwood HL (eds). Waltham, MA: Academic Press. Gray J (1968) Animal locomotion. London, UK: William Clowes & Sons.

297

298

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Gregor RJ, Bulgakova MA, Maas H, Oliver A, Prilutsky BI (2014) Locomotor activity of feline ankle extensors and kinematics during level and slope walking after removal of stretch reflex from soleus and lateral gastrocnemius by self–reinnervation. Program No. 827.01. Soc Neurosci Abstr. Online. Gregor RJ, Smith DW, Prilutsky BI (2006) Mechanics of slope walking in the cat: quantification of muscle load, length change, and ankle extensor EMG patterns. J Neurophysiol 95:1397–1409. Griffiths RI (1991) Shortening of muscle fibres during stretch of the active cat medial gastrocnemius muscle: the role of tendon compliance. J Physiol 436:219–236. Gross CG (1998) Claude Bernard and the Constancy of the Internal Environment. Neuroscientist 4:380–385. Guertin P, Angel MJ, Perreault MC, McCrea DA (1995) Ankle extensor group I afferents excite extensors throughout the hindlimb during fictive locomotion in the cat. J Physiol 487:197–209. Hartman HB (1985) Tension receptors on the closer muscle apodeme in the walking legs of the blue crab, Callinectes sapidus. J Comp Physiol 157:355–362. Head SI, Bush BMH (1991) Reflex actions of one proprioceptor on the motorneurones of a muscle receptor and their central modulation in the shore crab. J Physiol 437:49–62. Hellekes K, Blincow E, Hoffmann J, Büschges A (2012) Control of reflex reversal in stick insect walking: effects of intersegmental signals, changes in direction, and optomotor–induced turning. J Neurophysiol 107:239–249. Henry SM, Fung J, Horak FB (1998) EMG responses to maintain stance during multidirectional surface translations. J Neurophysiol 80:1939–1950. Hiebert GW, Whelan PJ, Prochazka A, Pearson KG (1996) Contribution of hind limb flexor muscle afferents to the timing of phase transitions in the cat step cycle. J Neurophysiol 75:1126–1137. Hodgkin AL, Huxley AF (1952) A quantitative description of membrane current and its application to conduction and excitation in nerve. J Physiol (London) 117:500–544. Hof AL, Gazendam MG, Sinke WE (2005) The condition for dynamic stability. J Biomech 38:1–8. Hoffer JA, Caputi AA, Pose IE, Griffiths RI (1989) Roles of muscle activity and load on the relationship between muscle spindle length and whole muscle length in the freely walking cat. Prog Brain Res 80:75–85. Hogan N (1985) The mechanics of multi-joint posture and movement control. Biol Cybern 52:315–331. Honeycutt CF, Nardelli P, Cope TC, Nichols TR (2012) Muscle spindle responses to horizontal support surface perturbation in the anesthetized cat: insights into the role of autogenic feedback in whole body postural control. J Neurophysiol 108:1253–1261. Honeycutt CF, Nichols TR (2010) Disruption of cutaneous feedback alters magnitude but not direction of muscle responses to postural perturbations in the decerebrate cat. Exp Brain Res 203:765–771. Honeycutt CF, Nichols TR (2014) The mechanical actions of muscles predict the direction of muscle activation during postural perturbations in the cat hindlimb. J Neurophysiol 111:900–907.

Sensory Feedback in the Control of Posture and Locomotion

Horak FB, Macpherson JM (2011) Postural orientation and equilibrium. Comp Physiol, Supplement 29: Handbook of Physiology, Exercise: Regulation and Integration of Multiple Systems:255–292. First published in print. doi: 101002/cphycp120107. Hulliger M, Durmuller N, Prochazka A, Trend P (1989) Flexible fusimotor control of muscle spindle feedback during a variety of natural movements. Prog Brain Res 80:87–101. Hunt CC (1990) Mammalian muscle spindle: peripheral mechanisms. Physiol Rev 70:643–663. Huyghues-Despointes CM, Cope TC, Nichols TR (2003) Intrinsic properties and reflex compensation in reinnervated triceps surae muscles of the cat: effect of activation level. J Neurophysiol 90:1537–1546. Jacobs R, Macpherson JM (1996) Two functional muscle groupings during postural equilibrium tasks in standing cats. J Neurophysiol 76:2402–2411. Jami L (1992) Golgi tendon organs in mammalian skeletal muscle: functional properties and central actions. Physiol Rev 72:623–666. Jeka JJ, Lackner JR (1994) Fingertip contact influences human postural control. Exp Brain Res 100:495–502. Joyce GC, Rack PM, Westbury DR (1969) The mechanical properties of cat soleus muscle during controlled lengthening and shortening movements. J Physiol 204:461–474. Kagaya K, Takahata M (2011) Sequential synaptic excitation and inhibition shape readiness discharge for voluntary behavior. Science 332:365–368. Kennedy D, Calabrese RL, Wine JJ (1974) Pre-synaptic inhibition: primary afferent depolarization in crayfish neurons. Science 186:451–454. Klarner D, Barth FG (1986) The cuticular stress detector (CSD2) of the crayfish. I. Physiological properties. J Exp Biol 122:149–159. Kriellaars DJ, Brownstone RM, Noga BR, Jordan LM (1994) Mechanical entrainment of fictive locomotion in the decerebrate cat. J Neurophysiol 71:2074–2086. Lam T, Pearson KG (2001) Proprioceptive modulation of hip flexor activity during the swing phase of locomotion in decerebrate cats. J Neurophysiol 86:1321–1332. Lam T, Pearson KG (2002) Sartorius muscle afferents influence the amplitude and timing of flexor activity in walking decerebrate cats. Exp Brain Res 147:175–185. Landsmeer JM (1961) Studies in the anatomy of articulation. I. The equilibrium of the “intercalated” bone. Acta Morphol Neerl Scand 3:287–303. Laurent G, Burrows M (1989) Intersegmental interneurons can control the gain of reflexes in adjacent segments of the locust by their action on nonspiking local interneurons. J Neurosci 9:3030–3039. Le Bon-Jego M, Masante-Roca I, Cattaert D (2006) State-dependent regulation of sensory–motor transmission: role of muscarinic receptors in sensory–motor integration in the crayfish walking system. Eur J Neurosci 23:1283–1300. Le Ray D, Cattaert D (1997) Neural mechanisms of reflex reversal in coxo-basipodite depressor motor neurons of the crayfish. J Neurophysiol 77:1963–1978. Leibrock CS, Marchand AR, Barnes WJ, Clarac F (1996) Synaptic connections of the cuticular stress detectors in crayfish: mono- and polysynaptic reflexes and the entrainment of fictive locomotion in an in vitro preparation. J Comp Physiol A 178:711–725. Li Y, Levine WS, Loeb GE (2012) A two-joint human posture control model with realistic neural delays. IEEE T Neural Sys Reh 20:738–748.

299

300

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Libersat F, Clarac F, Zill S (1987a) Force-sensitive mechanoreceptors of the dactyl of the crab: single-unit responses during walking and evaluation of function. J Neurophysiol 57:1618–1637. Libersat F, Zill S, Clarac F (1987b) Single-unit responses and reflex effects of force-sensitive mechanoreceptors of the dactyl of the crab. J Neurophysiol 57:1601–1617. Loeb GE (1981) Somatosensory unit input to the spinal cord during normal walking. Can J Physiol Pharmacol 59:627–635. Loeb GE, Bak MJ, Duysens J (1977) Long-term unit recording from somatosensory neurons in the spinal ganglia of the freely walking cat. Science 197:1192–1194. Loeb GE, Duysens J (1979) Activity patterns in individual hindlimb primary and secondary muscle spindle afferents during normal movements in unrestrained cats. J Neurophysiol 42:420–440. Loram ID, Maganaris CN, Lakie M (2005) Human postural sway results from frequent, ballistic bias impulses by soleus and gastrocnemius. J Physiol 564:295–311. Loram ID, Maganaris CN, Lakie M (2009) Paradoxical muscle movement during postural control. Med Sci Sports Exerc 41:198–204. Lyle MA, Prilutsky BI, Gregor RJ, Abelew TA, Nichols TR (2016) Self-reinnervated muscles lose autogenic length feedback, but intermuscular feedback can recover functional connectivity. J Neurophysiol 116:1055–1067. Maas H, Gregor RJ, Hodson-Tole EF, Farrell BJ, Prilutsky BI (2009) Distinct muscle fascicle length changes in feline medial gastrocnemius and soleus muscles during slope walking. J Appl Physiol 106:1169–1180. Maas H, Prilutsky BI, Nichols TR, Gregor RJ (2007) The effects of self–reinnervation of cat medial and lateral gastrocnemius muscles on hindlimb kinematics in slope walking. Exp Brain Res 181:377–393. Macpherson JM (1988a) Strategies that simplify the control of quadrupedal stance. I. Forces at the ground. J Neurophysiol 60:204–217. Macpherson JM (1988b) Strategies that simplify the control of quadrupedal stance. II. Electromyographic activity. J Neurophysiol 60:218–231. Macpherson JM, Fung J (1999) Weight support and balance during perturbed stance in the chronic spinal cat. J Neurophysiol 82:3066–3081. Malamud JG, Godt RE, Nichols TR (1996) Relationship between short-range stiffness and yielding in type-identified, chemically skinned muscle fibers from the cat triceps surae muscles. J Neurophysiol 76:2280–2289. Malone LA, Bastian AJ (2010) Thinking about walking: effects of conscious correction versus distraction on locomotor adaptation. J Neurophysiol 103:1954–1962. Marchand AR, Barnes WJ, Cattaert D (1997) Primary afferent depolarizations of sensory origin within contact-sensitive mechanoreceptive afferents of a crayfish leg. J Neurophysiol 77:3340–3354. Markin SN, Klishko AN, Shevtsova NA, Lemay MA, et al. (2010) Afferent control of locomotor CPG: insights from a simple neuromechanical model. Ann NY Acad Sci 1198:21–34. Markin SN, Klishko AN, Shevtsova NA, Lemay MA, et al. (2016) A neuromechanical model of spinal control of locomotion. In Neuromechanical Modeling of Posture and Locomotion. Prilutsky BI, Edwards DH Jr, (eds). New York, NY: Springer.

Sensory Feedback in the Control of Posture and Locomotion

Masani K, Vette AH, Popovic MR (2006) Controlling balance during quiet standing: proportional and derivative controller generates preceding motor command to body sway position observed in experiments. Gait Posture 23:164–17. Matthews PB (1981) Evolving views on the internal operation and functional role of the muscle spindle. J Physiol 320:1–30. Maxwell JC (1868) On governors. Proc R Soc Lond 16:270–283. Mazzaro N, Grey MJ, Sinkjaer T (2005) Contribution of afferent feedback to the soleus muscle activity during human locomotion. J Neurophysiol 93:167–177. McCrea DA (2001) Spinal circuitry of sensorimotor control of locomotion. J Physiol 533:41–50. McCrea DA, Rybak IA (2008) Organization of mammalian locomotor rhythm and pattern generation. Brain Res Rev 57:134–146. McVea DA, Donelan JM, Tachibana A, Pearson KG (2005) A role for hip position in initiating the swing-to-stance transition in walking cats. J Neurophysiol 94:3497–3508. Mehta R, Kajtaz E, Gregor RJ, Prilutsky BI (2014) Effects of stretch-reflex removal by self-reinnervation of one-joint vastii and two-joint rectus femoris on hindlimb muscle activity and mechanics during walking in the cat. Program No. 734.06. Soc Neurosci Abstr. Online. Mehta R, Prilutsky BI (2014) Task-dependent inhibition of slow-twitch soleus and excitation of fast–twitch gastrocnemius do not require high movement speed and velocity-dependent sensory feedback. Front Physiol 5:410. Meyer PF, Oddsson LI, De Luca CJ (2004) Reduced plantar sensitivity alters postural responses to lateral perturbations of balance. Exp Brain Res 157:526–536. Mill PJ (1976) Structure and function of proprioceptors in the invertebrates. London, UK: Chapman and Hall. Misiaszek JE (2003) The H-reflex as a tool in neurophysiology: its limitations and uses in understanding nervous system function. Muscle Nerve 28:144–160. Muller U, Clarac F (1990) Dactyl sensory influences on rock lobster locomotion. II. Role in interleg coordination. J Exp Biol 148:113–128. Mussa-Ivaldi FA, Hogan N, Bizzi E (1985) Neural, mechanical, and geometric factors subserving arm posture in humans. J Neurosci 5:2732–2743. Nashner LM, Shupert CL, Horak FB, Black FO (1989) Organization of posture controls: an analysis of sensory and mechanical constraints. Prog Brain Res 80:411–418. Newland PL, Emptage NJ (1996) The central connections and actions during walking of tibial campaniform sensilla in the locust. J Comp Physiol 178:749–762. Nichols TR, Houk JC (1976) Improvement in linearity and regulation of stiffness that results from actions of stretch reflex. J Neurophysiol 39:119–142. Niven JE, Graham CM, Burrows M (2008) Diversity and evolution of the insect ventral nerve cord. Annu Rev Entomol 53:253–271. Pantall A, Gregor RJ, Mehta R, Prilutsky BI (2014) Effects of 60-min sciatic nerve stimulation immediately after cut and repair of feline soleus and lateral gastrocnemius nerves on locomotor EMG activity of ankle muscles. Program No. 539.19. Soc Neurosci Abstr. Online. Pantall A, Hodson-Tole EF, Gregor RJ, Prilutsky BI (2016) Increased intensity and reduced frequency of EMG signals from feline self-reinnervated ankle extensors during walking do not normalize excessive lengthening. J Neurophysiol 115:2406–2420.

301

302

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Pearson K, Ekeberg O, Büschges A (2006) Assessing sensory function in locomotor systems using neuro-mechanical simulations. Trends Neurosci 29:625–631. Pearson KG (1972) Central programming and reflex control of walking in the cockroach. J Exp Biol 56:173–193. Pearson KG (2008) Role of sensory feedback in the control of stance duration in walking cats. Brain Res Rev 57:222–227. Perreault MC, Angel MJ, Guertin P, McCrea DA (1995) Effects of stimulation of hindlimb flexor group II afferents during fictive locomotion in the cat. J Physiol 487:211–220. Peterka RJ (2002) Sensorimotor integration in human postural control. J Neurophysiol 88:1097–1118. Prilutsky BI, Klishko AN, Weber DJ, Lemay MA (2016) Computing motion dependent afferent activity during cat locomotion using a forward dynamics musculoskeletal model. In Neuromechanical Modeling of Posture and Locomotion. Prilutsky BI, Edwards DH Jr, (eds). New York, NY: Springer. Prochazka A (1996) Proprioceptive feedback and movement regulation. In Handbook of Physiology Section 12: Exercise and Regulation of Multiple Systems. Rowell LB, Shepherd JT (eds). Bethesda, MD: American Physiological Society. Prochazka A, Gorassini M (1998) Ensemble firing of muscle afferents recorded during normal locomotion in cats. J Physiol 507:293–304. Prochazka A, Westerman RA, Ziccone SP (1977) Ia afferent activity during a variety of voluntary movements in the cat. J Physiol 268:423–448. Rack PM, Westbury DR (1969) The effects of length and stimulus rate on tension in the isometric cat soleus muscle. J Physiol 204:443–460. Rack PM, Westbury DR (1974) The short range stiffness of active mammalian muscle and its effect on mechanical properties. J Physiol 240:331–350. Ritzmann RE, Harley CM, Daltorio KA, Tietz BR, et al. (2012) Deciding which way to go: how do insects alter movements to negotiate barriers? Front Neurosci 6:97. Rybak IA, Shevtsova NA, Lafreniere-Roula M, McCrea DA (2006a) Modelling spinal circuitry involved in locomotor pattern generation: insights from deletions during fictive locomotion. J Physiol 577:617–639. Rybak IA, Stecina K, Shevtsova NA, McCrea DA (2006b) Modelling spinal circuitry involved in locomotor pattern generation: insights from the effects of afferent stimulation. J Physiol 577:641–658. Safavynia SA, Ting LH (2013) Long-latency muscle activity reflects continuous, delayed sensorimotor feedback of task-level and not joint-level error. J Neurophysiol 110:1278–1290. Schmitz J (1993) Load-complensating reactions in the proximal leg joints of stick instects during standing and walking. J Exp Biol 183:15–33. Shadmehr R, Wise SP (2005) The Computational Neurobiology of Reaching and Pointing. Cambridge, MA: MIT Press. Sherrington CS (1910) Flexion–reflex of the limb, crossed extension–reflex, and reflex stepping and standing. J Physiol 40:28–121. Sherrington CS (1913) Further observations on the production of reflex stepping by combination of reflex excitation with reflex inhibition. J Physiol 47:196–214. Sillar KT, Skorupski P, Elson RC, Bush BMH (1986) Two identified afferent neurones entrain a central locomotor rhythm generator. Nature 323:440–443.

Sensory Feedback in the Control of Posture and Locomotion

Skorupski P, Rawat BM, Bush BM (1992) Heterogeneity and central modulation of feedback reflexes in crayfish motor pool. J Neurophysiol 67:648–663. Skorupski P, Sillar KT (1986) Phase-dependent reversal of reflexes mediated by the thoracocoxal muscle receptor organ in the crayfish, Pacifastacus leniusculus. J Neurophysiol 55:689–695. Smith JL, Zernicke RF (1987) Predictions for neural control based on limb dynamics. Trends Neurosci 10:123–128. Stapley PJ, Ting LH, Hulliger M, Macpherson JM (2002) Automatic postural responses are delayed by pyridoxine–induced somatosensory loss. J Neurosci 22:5803–5807. Stein RB, Capaday C (1988) The modulation of human reflexes during functional motor tasks. Trends Neurosci 11:328–332. Stein RB, Misiaszek JE, Pearson KG (2000) Functional role of muscle reflexes for force generation in the decerebrate walking cat. J Physiol 525:781–791. Strausfeld NJ (2012) Arthropod Brains: Evoluton, Functional Elegance and Historical Significance. Cambridge, MA: Belknap Press. Stuart DG, Hultborn H (2008) Thomas Graham Brown (1882–1965), Anders Lundberg (1920–), and the neural control of stepping. Brain Res Rev 59:74–95. Taylor A, Durbaba R, Ellaway PH, Rawlinson S (2006) Static and dynamic gamma-motor output to ankle flexor muscles during locomotion in the decerebrate cat. J Physiol 571:711–723. Torres-Oviedo G, Macpherson JM, Ting LH (2006) Muscle synergy organization is robust across a variety of postural perturbations. J Neurophysiol 96:1530–1546. von Holst E, Mittelstaedt H (1950) Daz Reafferezprincip. Wechselwirkungen zwischen Zentralnerven-system und Peripherie. Naturwissenschaften 37:467–476. Translation: The reafference principle. In The Behavioral Physiology of Animals and Man. The Collected Papers of Erich von Holst (Translated by Martin R) (1973) Coral Gables, FL: University of Miami Press. Watling L, Thiel M (eds). (2013) Functional Morphology and Diversity. Oxford, UK: Oxford University Press. Weber DJ, Stein RB, Everaert DG, Prochazka A (2007) Limb-state feedback from ensembles of simultaneously recorded dorsal root ganglion neurons. J Neural Eng 4:S168–180. Wiener N (1961) Cybernetics: or Control and Communication in the Animal and the Machine. Cambridge, MA: MIT Press. Wiersma CA (1961) Reflexes and the central nervous system. In The Physiology of Crustacea. II Sense Organs, Integration, and Behavior. Waterman TH (ed). New York, NY: Academic Press. Winter DA (1995) Human balance and posture control during standing and walking. Gait Posture 3:193–214. Yakovenko S, Gritsenko V, Prochazka A (2004) Contribution of stretch reflexes to locomotor control: a modeling study. Biol Cybern 90:146–155. Zill SN, Keller BR, Chaudhry S, Duke ER, et al. (2010) Detecting substrate engagement: responses of tarsal campaniform sensilla in cockroaches. J Comp Physiol A 196:407–420. Zill SN, Keller BR, Duke ER (2009) Sensory signals of unloading in one leg follow stance onset in another leg: transfer of load and emergent coordination in cockroach walking. J Neurophysiol 101:2297–2304.

303

304

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Zill SN, Moran DT (1981) The exoskeleton and insect proprioception. I. Responses of tibial campaniform sensilla to external and muscle–generated forces in the American cockroach, Periplaneta americana. J Exp Biol 91:1–24. Zill SN, Moran DT, Varela FG (1981) The exoskeleton and insect proprioception: II. Reflex effects of tibial campaniform sensilla in the American cockroach, Periplaneta americana. J Exp Biol 94:43–55. Zill SN, Schmitz J, Chaudhry S, Büschges A (2012) Force encoding in stick insect legs delineates a reference frame for motor control. J Neurophysiol 108:1453–1472.

305

10 Coordination of Rhythmic Movements Jean-Patrick Le Gal 1 , Réjean Dubuc 1,2 , and Carmen Smarandache-Wellmann 3 1 Groupe de Recherche sur le Système Nerveux Central, Département de neurosciences, Université de Montréal, Montréal, Québec, Canada 2 Groupe de recherche en activité physique adaptée, Département des sciences de l’activité physique, Université du Québec à Montréal, Montréal, Québec, Canada 3 Biozentrum Köln, Institut für Zoologie, Universität zu Köln, Köln, Germany

10.1 Introduction A fundamental goal of neuroscience is to explain, on the cellular level, how nervous systems generate behavior. Progress toward this goal often comes from studying nervous systems with particularly advantageous cellular organizations or anatomies, and that continue to produce behaviorally relevant activity in reduced preparations or even in isolation. Cellular explanations gained in such “model” systems often lead to insights that are widely applicable across phyla (Marder and Calabrese 1996). One such insight is that in both invertebrates and vertebrates the basic rhythmicity and pattern of activity of most rhythmic behaviors is generated by central pattern generators (CPGs), neural networks that can generate rhythmic, patterned activity in the absence of rhythmic central input or sensory feedback (although both typically play major roles in initiation and on-going control of CPG activity) (see Chapters 8, 9). A good example is locomotion, for which we now have fairly well-developed understanding on the cellular level. Generating behavior often requires coordinating the activities of multiple CPGs. The cellular basis of inter-CPG coordination is less understood than our present fairly deep understanding of CPG structure and function. Synchronization of oscillators in perfect phase or antiphase have been studied into detail experimentally and theoretically. Perfect in-phase synchrony can be established by electric coupling of neurons through gap junctions (Beierlein et al. 2003; MacMillan and Deller 1989), with the ability of electrical coupling to synchronize two oscillators depending on coupling strength and the activity the oscillators produce when disconnected (Mancilla et al. 2007). Reciprocal inhibition can produce antiphase or (somewhat counterintuitively) in phase (Wang and Rinzel 1992) synchrony; which occurs depends on the synaptic and cellular properties of the network. Synchronizing oscillators at phase lags other than 0.0 (in-phase) or 0.5 (antiphase) is more difficult (Winifree 2001). Locomotory behaviors are particularly good systems in which to study how such “not simple” patterns of coordination occur, as the movements that comprise these behaviors often occur at phase lags other than 0 or 0.5 (e.g., Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

306

Neurobiology of Motor Control: Fundamental Concepts and New Directions

lamprey, leech, and crayfish nearest neighbor oscillators synchronize with phase lags of 0.01, 0.05, and 0.25, respectively) (Grillner 1974; Mulloney and Smarandache 2010; Pearce and Friesen 1984; Wallén and Williams 1984). In legged systems, the legs must similarly be activated with particular phase lags (in bipeds, 0.0 or 0.5, but in animals with more legs again sometimes other phase relationships). Moreover, terrestrial animals alter locomotion speed by preferentially decreasing the stance phase of leg movements (Grillner 1981). Consequently, the duty cycles (movement duration divided by cycle period, the equivalent of phase for durations) of swing and stance change with locomotion period. Despite this, phase relationships among limb components (ankle, knee, hip), and between limbs, that maintain functional motor output, must continue to be produced. In segmental limbless animals each segment’s muscle contractions are generated by segmentally repeated CPGs located one per segment or hemisegment; similarly, in limbed animals the movements of each limb (and indeed, perhaps the movements of each limb joint) are activated by their own CPG (Büschges 2005; Grillner 1981). Coordination of these modules depends on a mix of sensory feedback and central coordinating influences (Marder et al. 2005): sensory activity due to movements generated by one CPG feeding back onto other CPGs; central connections between CPG elements themselves, both to nearest neighbors and more widely; specialized coordinating neurons not themselves CPG neurons but driven by CPG activity; modulation of central drive to the CPGs by input from any of the above levels, with the resulting changes in central drive resulting in the desired changes in CPG activity (Fig. 10.1). Coordination is studied in many preparations, each with specific advantages. The invertebrate section of this chapter is organized by preparation. The vertebrate section, alternatively, is organized by CPG function (locomotion, respiration). Despite these different approaches, both sections support the most fundamental conclusion of this chapter—that coordination is typically mediated by multiple mechanisms (central connections, sensory feedback loops, varying amounts of central drive) (Fig. 10.1).

10.2 Overview of Invertebrate CPGs We now present coordination mechanisms used in several particularly well-studied invertebrate preparations (Fig. 10.2). In some of these preparations only a few, or

Figure 10.1 Different coordinating mechanisms. (A) Classical synaptic connections between networks. (B) Sensory feedback. An anterior to posterior chain of five CPGs, each represented as two interacting cells or cell groups, A and B. Each CPG induces the muscles in its segment to contract. These contractions generate sensory feedback to their own CPGs and, importantly, to anterior or posterior ones. The CPGs produce coordinated rhythmic activity because of this sensory feedback from neighboring segments. (C) Central connections coordinating a CPG chain. Chain conventions as in B. i. Each CPG sends coordinating information to only neighboring CPGs. This can be in only the ascending direction, only the descending direction, or in both directions. ii. Short and long range connections up and down the chain. (D) Changes in central drive coordinating a CPG chain. Chain conventions as in B. i. Central drive ascending from tail ganglion. ii. Central drive descending from head ganglion. In vertebrates, Di and Dii would represent central drive arising in different brain regions. Brain schematic modified from Menzel (1983) with permission.

Coordination of Rhythmic Movements

B

A A

sensory drive

B

C

D

E

F

1 A

B 2

A

B 3

A

neuronal network

Ci

A

B

C

D

E

F

4 A

B 5

A

Cii

1 A

B

B

1

B

A

B

2

2 A

B

A

B

3

3 A

B

A

B

4

4 A

B

A

B

5

5 A

B

A

Di

B

Dii 1 A

B

central drive

2 A

B

1

3 A

A

B

A

A

B B

B

3

5 A

B

2

4

A

B

4 A

B

5 A

B

even one, of the above mechanisms are used, or at least are of primary importance. It is possible, indeed likely, that this specialization is present because preparations in which particularly strong coordination is present, and is maintained in reduced or in vitro preparations, were chosen for study. These choices may thus have biased work towards evolutionarily extreme systems (analogous to using bats to study echolocation). Such a bias would not diminish the generality of the mechanisms described in these systems—that something is exaggerated in some species does not mean it is not present,

307

308

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B L

C

R R1

L1

R2

L2

R1 R2 R3 R4

L3

R3

Figure 10.2 Schematic drawings of annelids (A), long-tailed crustaceans (B), and insects (C). Segmental ventral nerve cord shown in light grey. All possess a cephalic and a tail ganglion. Ganglia are connected by connectives. Each ganglion innervates the segment and appendages in its immediate proximity. (A) Annelids (e.g., leech) have no appendages, movement is generated by body wall undulations. (B) In long-tailed Crustacea, appendages (swimmerets) attach to four (or five, not shown) of the six abdominal segments. (C) Most insects possess three walking legs. Appendages are labeled with respect to which side they attach (R, right; L, left) and enumerated from anterior to posterior.

to a lesser degree, across species. However, it does suggest that, in many, perhaps most, systems a mixture of mechanisms may be used. 10.2.1 Stomatogastric Nervous System: Feeding Circuits in Decapod Crustacea

The stomachs of decapod Crustacea (lobsters, crabs) not only store and enzymatically digest food. They also possess teeth in the gastric mill that grind (chew) food, and apparatus in the pylorus that divide the chewed food into streams for additional chewing, for absorption, or for excretion. The nervous system driving these multiple motor patterns is the stomatogastric nervous system (STNS). The STNS contains neural networks that generate four semi-independent rhythmic activities: the esophageal, cardiac sac, pyloric, and gastric mill rhythms (Johnson and Hooper 1992). Food is stored in the cardiac sac, and cardiac sac network rhythmic activity helps move food through multiple cycles of digestion and chewing. The gastric mill network helps move food from the cardiac sac to the gastric mill, and controls the movement of the three teeth of the gastric mill that grind the food. The pyloric network generates the filtering movements of the pylorus. These neural networks can function as separate but interacting networks (Bartos et al. 1999; Bucher et al. 2006; Mulloney 1977; Nadim et al. 1998; Thuma and Hooper 2002, 2003), neurons can switch with which network they are active (Hooper and Moulins 1989; Katz and Harris-Warrick 1991; Weimann et al. 1991), and the networks can be fused to create larger networks that produce new motor patterns involving multiple stomach regions (Dickinson et al. 1990; Meyrand et al. 1991). The networks are modulated by multiple descending projection neurons (Blitz et al. 1999; Norris et al. 1996; Nusbaum and Marder 1989) and are influenced by sensory feedback (Beenhakker

Coordination of Rhythmic Movements

et al. 2007; Katz et al. 1989). The networks of the most commonly used species (crabs, Cancer borealis, C. pagurus; lobster, Homarus gammarus, H. americanus, Panulirus interruptus, Palinurus vulgaris), although homologous, have substantial differences in neural make-up, synaptic connectivity, and coordination mechanisms. This large variety in such a closely related group of species supports the major theme of this chapter, that multiple mechanisms coordinate neural network activity, and which play the most important roles shows great diversity across species. Morphology of Stomatogastric Nervous System Crustacea, like all arthropods, possess a

decentralized nervous system. Because of the smaller number of neurons present in these anatomically separated ganglia, describing neural network neuronal make-up and synaptic connectivity in them is easier than in vertebrate central nervous system. The STNS also has the advantage of many invertebrate preparations that network neuron make-up and synaptic connectivity are the same across individuals. The STNS is an extension of the ventral nerve cord and is located on top of the foregut. It contains four interconnected but distinct ganglia: two Commissural (CoGs, each approximately 500 neurons), the Esophogeal (OG, 14 neurons; so abbreviated because first publication in a British journal, hence oesophageal), and the Stomatogastric (STG, 25–30 neurons). In addition to containing some CPG neurons, neurons in the CoGs and OG send modulatory projections downstream to the STG. STNS nerves are mixed nerves containing motor axons driving foregut movement and afferent sensory axons. Pyloric, Gastric Mill, and Cardiac Sac Rhythms The synaptic connections, active properties of

their neurons (e.g., inherent rhythmicity, plateau potentials, post-inhibitory rebound), and the ionic currents underlying these properties, of the pyloric and gastric mill networks are known in detail (Marder and Bucher 2001, 2007; Nusbaum and Beenhakker 2002) (see also Chapter 8). The neurons of both networks are all located in the STG. Almost all synapses in the networks are inhibitory chemical (the gastric mill network has one excitatory chemical synapse) or electrical, a pattern observed in many other invertebrate networks. A noteworthy aspect of these networks is that pattern generation is not performed by a set of pre-motor interneurons which then drive motor neurons that play little, or no, role in generating network activity. Instead, almost all gastric mill and pyloric neurons are motor neurons that both (via their active properties and synaptic interconnections in the STG) help generate their network’s respective rhythms and (through their peripheral axons) drive the muscles that generate each pattern’s movements. This “double duty” is likely one reason these networks can produce such complicated activity patterns with such small numbers of neurons. The activity of all these neurons in both networks can be uniquely identified in recordings of the system’s nerves. The activity of all the neurons of the two networks can therefore be easily and completely assessed from simple extracellular recordings. The pyloric rhythm is spontaneously active and cycles rapidly (cycle period 0.5–3 s) (Fig. 10.3A, PY and PD traces; 10.3B, pdn trace; 10.3C left panel, PY and PD traces) (Marder and Bucher 2007). The network has six neuron types, some of which fire together, producing a tri-phasic rhythm. Network rhythmicity results from the presence of both an endogenous oscillator neuron and multiple embedded half-center

309

310

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

PY

10 mV

PD

10 mV 1s

aln

B pdn

LPG

5 mV 2s

C MG

GM

PY

5s PD

D

CS

Dorsal PD muscle

20 s

60 μm

10 mV

Coordination of Rhythmic Movements

oscillators. The activity of the network’s neurons is modulated by input from modulatory neurons located in the OG and CoGs. The net effect of this input is typically excitatory; without it pyloric cycle period increases and many pyloric neurons cease firing. The gastric mill rhythm is much slower (cycle period 5–15 s) than the pyloric (Fig. 10.3A, aln trace; Fig. 10.3B right panel, LPG trace; Fig. 10.3C left panel, MG and GM traces). It produces a two phase rhythm that drives alternating protraction and retraction of the medial tooth and closing and opening of the lateral teeth. It has no endogenous oscillator neurons, with its rhythmicity instead arising solely from embedded half-center oscillators. Rhythmic activity obligatorily depends on input from modulatory projection neurons from the CoGs or application of appropriate transmitters (Blitz et al. 1999; Nusbaum et al. 1992). The cardiac sac rhythm is the slowest, with periods of 20 s to several minutes (Fig. 10.3D, CS “trace”). The neurons (IV, CD1, CD2) forming the network are distributed among the different ganglia and project to them all. This network generates only active dilation of the cardiac sac. During the long cardiac sac interburst intervals, a gastric mill network neuron induces cardiac sac constrictions. The movements of the cardiac sac thus appear to be strong, infrequent dilations, which would transfer food anteriorly from the gastric mill to the cardiac sac, separated by gastric-timed contractions that transfer the food back into the gastric mill, in time with gastric mill tooth movements, for further chewing. Central Coordination of the Rhythms Extensive synaptic connectivity among the four STNS

networks, and among STNS neurons and descending inputs to the networks, exists. The networks also receive sensory feedback. As such, the basis for multiple coordinating mechanisms exists. The simplest form of coordination involves classical (non-modulatory) chemical synapses and electrical coupling (mechanism in Fig. 10.1A). This type is best exemplified by the connections between the gastric mill and pyloric networks. In lobster these connections are relatively weak, and result in (1) pyloric-timed variations in spike frequency during gastric mill neuron bursts (Bucher et al. 2006; Mulloney 1977) (Fig. 10.3B right panel, LPG trace; Fig. 10.3C left panel, MG and GM traces); and (2) various of the pyloric neurons firing different numbers of spikes per burst, and the pyloric network cycling slightly faster or slower, in different phases of the gastric mill

Figure 10.3 Stomatogastric system coordination. (A) Small changes in pyloric neuron activity (changes in PY neuron spike number per burst, changes in PD neuron spike frequency in its spike burst, changes in pyloric cycle period) occur in different phases of the gastric mill (aln) cycle period. (B) The LPG neuron fires with the pyloric network (pdn trace) when the gastric mill is not cycling (left) but with the gastric mill when it is active (right trace). Note again changes in pyloric activity as a function of gastric mill phase. (C) After stimulation of a sensory input pathway to the STNS, the gastric mill (MG, GM) and pyloric (PY, PD) networks fuse into a single network that produces a new, unified output. (D) The dorsal PD muscle, which is innervated by only the PD neuron, nonetheless contracts in phase with the cardiac sac neural network (black rectangles represent cardiac sac network bursts). The PD neuron continues to fire with the pyloric network (around 1 s cycle periods and 0.25 s burst durations, see inset) throughout the entirety of the panel. Inset shows a time expansion of muscle trace and associated PD neuron activity. Small contractions and relaxations are muscle response to each PD neuron spike burst. A from Thuma and Hooper (2002); B from Weimann et al. (1991); C from Faumont et al. (2005); D from Morris et al. (2000); all with permission.

311

312

Neurobiology of Motor Control: Fundamental Concepts and New Directions

cycle (Bucher et al. 2006; Thuma and Hooper 2002) (Fig. 10.3A). These changes can be relatively small on the neuronal level, but even small changes in motor neuron firing can cause very large changes in muscle activity (see Fig. 10.3D and below). In these coordinations the identity and neural make-up of the two neural networks are always maintained. These connections can also cause n:m coupling between the rhythms, with 4 to 11 pyloric cycles occurring in each gastric mill cycle (the wide variation in coupling ratios because of wide variation of gastric mill cycle periods) (Bucher et al. 2006). The connections between the two networks are much stronger in crab. As a result, many crab neurons (pyloric and gastric) fire with the pyloric pattern when it alone is active. Alternatively, when the gastric mill is also active, these neurons fire either exclusively with it (Fig. 10.3B), or produce a melded activity in which they fire with the pyloric during one gastric mill phase but are silent during the other (Weimann et al. 1991). The extent to which these gastro-pyloric neurons fire with which rhythm differs from preparation to preparation, and across individuals great variation in which neurons are coordinated with which rhythm is thus observed. That the gastro-pyloric neurons are regardless truly members of both networks is demonstrated by their being able, regardless of which network they are active with, to reset both rhythms (Weimann and Marder 1994). A similar example of neurons switching between networks is observed in lobster in response to stimulation of a sensory afferent to the STNS, which induces one neuron to switch from the pyloric to the cardiac sac network, firing long bursts with the CD1 and CD2 neurons and being silent, instead of firing with the pyloric network, during the long cardiac sac interburst intervals. This switch is due to the input silencing active properties (post-inhibitory rebound) in the switching neuron, which results in it no longer being able to generate bursts in response to pyloric-timed inhibition (Hooper and Moulins 1989). In crab another coordination mechanism, in which one CPG gates descending input to another (a specialized form of Fig. 10.1Dii) has been observed. In this case a descending projection neuron from the CoGs, MCN1, activates both the pyloric and gastric mill rhythms. MCN1’s terminals in the STG are inhibited by a gastric mill protractor neuron. The pyloric network thus does not receive MCN1 input during the gastric mill protraction phase and, as a result, cycles faster during gastric mill retraction than protraction (Bartos and Nusbaum 1997). The interaction of MCN1 input and an effective excitation from the pyloric to the gastric mill network also results, as in lobster, in crab gastric mill cycle periods being an integer number of pyloric cycle periods (Nadim et al. 1998). In all the above examples, the identity of the different neural networks was maintained, i.e., although a neuron might in some cases cycle with the pyloric network and sometimes with the gastric mill, that there was a pyloric and a gastric network to switch between was always clear. Sensory input or neuromodulator application can induce a qualitatively different type of coordination in which this concept of identity disappears. In these cases two or more networks fuse to create a new, larger functional network that generates a new, unified rhythm unlike those produced by any of the smaller networks (Dickinson et al. 1990; Kwiatkowski et al. 2013; Meyrand et al. 1991) (Fig. 10.3C). This is similar to the mechanism in Fig. 10.1A, except the boundaries surrounding the two networks disappear. A natural question is whether network fusion should be considered an example of coordination, since on the network level there no longer remain two entities whose

Coordination of Rhythmic Movements

activity is being coordinated. However, it is important to remember that coordination is typically defined behaviorally. Thus, considering the muscles of the crustacean stomach, when networks with clearly separate identities are being coordinated, one might observe that a muscle being driven by the pyloric network contracts five times for every contraction of a muscle being driven by the gastric mill network. When the neural networks have fused and produce a single output, the two muscles might instead contract one-for-one. From a behavioral point of view, in each case the activities of the two muscles are still coordinated, just in a different fashion. As such, what we know from the neural work to be actually the result of a network fusion appears, on the behavioral level, simply as a different coordination pattern. Sensory Feedback Induced Coordination Two STNS sensory pathways have been described

in which movement generated by one neural network alters the activity of another, and thus could serve to coordinate the activity of the two. The first pathway consists of the bilaterally symmetric Gastropyloric Receptor cells (GPR1 and GPR2), which innervate two gastric muscles. They are rhythmically activated during middle tooth retraction when the gastric mill motor pattern is active, and otherwise fire tonically at a low frequency. They use as serotonin and most likely acetylcholine (Katz et al. 1989; Katz and Harris-Warrick 1989) to modulate the activity of their target neurons. One effect of this modulation is to enhance gastric mill network activity by direct connections to gastric mill neurons. The GPR neurons serve a coordinating role by altering the activity of multiple modulatory projection neurons, including MCN1, CPN2, MCN5, and MCN7 (Blitz et al. 2004). These projection neurons alter pyloric as well as gastric mill network activity. This coordination is thus a combination of the mechanisms in Fig. 10.1B and Dii, in that sensory feedback acts to coordinate the pyloric and gastric mill networks by altering the activity of descending pathways that project to both networks. The second is the Anterior Gastric Receptor Cell (AGR), first identified in crayfish (Larimer and Kennedy 1966) and later in lobster and crabs (Combes et al. 1995, 1997, 1999; Simmers and Moulins 1988a,b; Smarandache and Stein 2007). AGR is a single, bipolar STNS neuron whose cell body is immediately posterior to the STG. It sends one axon to a tendon-like structure close to gastric mill protractor muscles and the other anteriorly through the STG to both CoGs (Elson et al. 1994; Simmers and Moulins 1988a; Smarandache and Stein 2007). In the absence of gastric mill movements, AGR fires tonically at 1–3 Hz, sufficient to alter both gastric mill and pyloric network activity (Daur et al. 2009). Without gastric mill movement these effects cannot, of course, be coordinating, but they do indicate the capacity for coordination exists. This possibility was confirmed in closed-loop preparations, which showed that gastric mill retractor movements activate AGR, and this feedback enhances gastric mill network activity (Smarandache et al. 2008). AGR activation also coordinates pyloric and gastric mill network activity by increasing pyloric network period and decreasing the activity of several pyloric neurons (Smarandache and Stein 2007). All these effects are mediated by AGR directly altering the activity of modulatory projection neurons, in particular MCN1, MCN5, and CPN2 (Hedrich et al. 2009), and thus belong to the same mechanism class as GPR coordination, in which sensory feedback alters a descending projection neuron pathway (Fig. 10.1B, Dii). Two other examples of sensory input altering network coordination are ones in which, although the sensory input is not activated by either network, the input nonetheless

313

314

Neurobiology of Motor Control: Fundamental Concepts and New Directions

changes their interaction, and hence their coordination. One of these are the Inferior Ventricular (IV) neurons, which project from the brain to the STNS (when the cardiac sac is active, the IV neurons fire rhythmically with the CD1 and CD2 neurons), and are most likely activated by food-induced stimulation of chemosensory neurons. Rhythmic activation of these neurons always activates a gastric mill rhythm and induces pronounced changes in pyloric activity, and thus the coordination between the two networks. This change in coordination state is achieved by direct effects of IV neuron synapses onto pyloric and gastric mill neurons in the STG and indirect effects mediated by IV neuron input altering projection neuron activity (Hedrich et al. 2009, 2011; Hedrich and Stein 2008; Sigvardt and Mulloney 1982a,b). The other sensory input pathway, the ventral cardiac neuron system, is most likely activated by stretch of the cardiac sac by food and again alters both rhythms and their coordination by activation of projection neurons (Beenhakker et al. 2004, 2007; Blitz et al. 2004). Taken together, these data show that the coordination of STNS motor networks, on the neuronal level, is mediated by multiple mechanisms: direct interconnections between the neurons of different networks (Fig. 10.1A), switching of neurons between networks, or fusion of networks, by descending input (a combination of Fig. 10.1A and Dii, where in Fig. 10.1A the network boundaries disappear); and sensory alteration of descending input (a combination of Fig. 10.1B and Dii) where the sensory input can be activated either by movements induced by one of the networks, or by stimuli independent of both networks. Coordination via Low-Pass Filtering at the Muscle Level The coordination discussed here stems

from the, often quite small, changes of pyloric neuron activity that occur in phase with gastric mill and cardiac sac network activity due to direct synaptic connections between the networks. These changes consist of the pyloric neurons firing different number of action potentials per burst, or cycling at different cycle periods, during different phases of the gastric mill or cardiac sac cycle periods (Fig. 10.3A). One might naturally expect these changes in neuron activity to result simply in pyloric-timed contractions with different amplitudes and occurring at the same cycle period as the pyloric neuron cycle period. However, pyloric muscles contract and relax very slowly, in most cases too slowly to relax fully between pyloric neuron spike bursts (Morris and Hooper 1998; Morris et al. 2000). They consequently slow (low pass) filter the pyloric neuron input they receive. Given that gastric mill and cardiac sac cycle periods are so much longer than pyloric cycle periods, the result is that many pyloric-innervated muscles contract primarily or almost exclusively (Fig. 10.3D) in time with the gastric mill and cardiac sac networks, not the pyloric network (Thuma et al. 2003). As in the case with network fusion, the question arises of whether this slow filtering should truly be called coordination. In considering this question, it is again useful to think of a naïve observer of stomach muscle contractions. Such an observer would see that the pyloric muscle in Fig. 10.3D contracted one-for-one with the cardiac sac muscles. Two set of muscles contracting one-for-one are as coordinated as coordinated can be. It is only with knowledge of the neural networks innervating these muscles, and hence that the cardiac sac muscles are innervated by neurons firing 5–10 s long bursts every minute, and the pyloric muscle by neurons firing 0.25 s long bursts every 1 s, that one realizes the neurons driving the muscles belong to different networks. As such, from

Coordination of Rhythmic Movements

a behavioral point of view, the muscle slow filtering of the rapid pyloric input is unarguably coordinating pyloric muscle contractions to the cardiac sac movement pattern. This type of coordination is not expected to occur in vertebrate skeletal muscle because it has much faster dynamics than stomatogastric muscles. Alternatively, many invertebrate muscles are as slow as stomatogastric muscles, and thus it could occur in many invertebrate systems. 10.2.2 Leech Locomotion

Medicinal leeches are widely used to study motor pattern coordination (reviewed in Brodfuehrer et al. 1995; Hill et al. 2003; Kristan Jr et al. 2005). We focus here on coordination in swimming and crawling. Leeches swim by undulating their body when water levels are high (Kristan Jr et al. 1974) and crawl with body wall elongation and contraction when water levels are low (Stern-Tomlinson et al. 1986) (see also Chapter 7.3). Central Nervous System Morphology Leeches are annelids (Fig. 10.2A) and as such have seg-

mented bodies and a ventral nerve cord. Each of the 21 body segments are innervated by its own bilaterally symmetric ganglia. Only the head and tail ganglia, which are involved in decision tasks, show segment specificity: the head is innervated by the cephalic ganglion, formed from six fused ganglia, and the last segment, the tail, is innervated by a terminal ganglion formed from seven fused ganglia. Swimming Leech swimming starts with a flattening of the body followed by dorso-

ventral (s-shaped) undulations (Kristan Jr et al. 1974). The antagonistic dorsal and ventral (longitudinal) muscles are thus active in antiphase. Leeches swim only forward; swimming therefore consists of an anterior to posterior metachronal wave (Fig. 10.4A). Cycle period is 0.35 to 2 s. One body undulation is always one period long, independent of swimming speed. In intact animals the phase shift from one body segment to the next is nearly 20∘ (Pearce and Friesen 1984). The swimming and crawling CPGs (premotor interneurons and motorneurons) generating each segment’s rhythmic activity are located in that segment’s ganglion. Obtaining robust fictive swimming CPG activity in single ganglia removed from the animal requires stimulation of sensory afferents or modulatory input (Weeks 1981). Robust swimming with an isolated nerve cord requires a chain of at least six ganglia (Hill et al. 2003). CPG neuron make up and synaptic connectivity are known (Kristan Jr et al. 2005);

A

Anterior to posterior metachronal wave

B

Posterior to anterior metachronal wave

R1 R2 R3 R4

Figure 10.4 Idealized examples of metachronal waves, as observed in leech and lamprey forward swimming (A) and in crayfish swimmeret movement and lamprey backward swimming (B).

315

316

Neurobiology of Motor Control: Fundamental Concepts and New Directions

all 21 mid-body ganglia have identical networks. The CPG network consist of three pairs of bilaterally symmetric interneurons, two pairs of bilaterally symmetric inhibitory motor neurons, and one unpaired interneuron. The network is dominated by electrical coupling and inhibitory chemical synapses, having only one excitatory chemical synapse (see Fig. 6.9). The CPG produces a three phase neural output, which is transformed into two-phase muscle activity as a result of complicated connections to the motor neurons, which are not part of the CPG. The network contains no endogenous oscillator neurons and only two classical half-center oscillator connections. However, the network’s extensive synaptic interconnectivity contains many chains of synaptic connections appropriate for reciprocal cyclic inhibition, in which activity reverberates along a chain of sequentially active neurons (Brodfuehrer et al. 1995; Friesen et al. 1978; Mullins et al. 2011). In vitro the phase shift between adjacent ganglia (segments) is 10∘ . Since in this situation there is no sensory feedback, this coordination must be due to central connections. These connections are made by a subset of the CPG interneurons themselves, all of which project either anteriorly or posteriorly 6 to 10 ganglia (mechanism shown in Fig. 10.1Cii). (Brodfuehrer et al. 1995). The connectivity is asymmetric in that some inhibitory interneurons extend anteriorly and others posteriorly, and the network’s single excitatory interneuron projects only posteriorly. This asymmetric coupling is enough to achieve the observed anterior to posterior phase lags (Cang and Friesen 2002; Skinner and Mulloney 1998). The long distances the CPG neurons project results in each ganglion receiving input from CPGs in multiple anterior and posterior ganglia. Receiving input from multiple CPGs is critical to maintaining the 10∘ phase shifts observed in vitro. Removing input from distant ganglia increases the phase lag; when only two segments are connected, the phase lag between them is 40∘ (Kristan Jr et al. 2005; Mullins et al. 2011). In intact animals the intersegmental phase lag is 20∘ , not the 10∘ in isolated nerve cords. This difference is presumably due to the lack of sensory feedback. Thus, although the inter-CPG coordinating network can produce qualitatively correct coordinated activity (Kristan Jr and Calabrese 1976; Pearce and Friesen 1984), fully reproducing in vivo activity also requires sensory feedback (mechanism in Fig. 10.1B). Analogous to the issue raised in the stomatogastric discussion about whether fused networks should be considered an example of coordination, the inability of individual ganglia to spontaneously produce rhythmic output, and the need to have six coupled ganglia for robust rhythmic activity, raises the question of whether each ganglion’s network should be considered an oscillator, and thus the complete system a set of coupled oscillators. It is probably best to state that, although individual networks have a rudimentary oscillatory capability, full expression of this ability requires an ensemble of multiple networks. Crawling In crawling the same muscles are activated as in swimming. The same pre-

motor interneurons could thus be involved in both behaviors. Unfortunately, no data exist about the crawling CPG. There are, however, data about movement initiation, the decision to crawl or swim, and most importantly for this chapter, mechanisms of coordination. Crawling can be elicited in isolated nervous system and individual ganglia when either specific sensory cells are activated or dopamine is applied (Puhl and Mesce 2008). Unlike swimming, in dopamine individual ganglia produce robust crawling. The

Coordination of Rhythmic Movements

crawling network in each individual ganglion extends to at least its two nearest neighbor ganglia, and these local connections play an important role in coordinating activity along the chain (Puhl and Mesce 2010). However, activity in bilaterally symmetric neurons located in the head ganglion that project posteriorly through all segments (the R3b1 neurons), that are (in the absence of dopamine) necessary and sufficient to produce crawling, are also required for correct coordination (Puhl et al. 2012; Puhl and Mesce 2010). Thus in this system both local connections (at least the mechanism in Fig. 10.1Ci, and perhaps 10.2Cii) and descending input (Fig. 10.1Dii) are involved in coordination. 10.2.3 Crayfish Swimmeret System Morphology of the Central Nervous System Crustacea, like annelids, have a segmented (one

ganglion per segment) ventral nerve cord, with the neural networks responsible for generating the movements of each segment’s appendages typically located in that segment. In long tailed crustaceans (shrimp, crayfish, lobster, Fig. 10.2B), the abdomen consists of six segments. Its limbs, the swimmerets, are bilaterally present on the first five segments in shrimp, but on only the second to the fifth segments in lobster and crayfish (Mulloney and Smarandache-Wellmann 2012). The neurons responsible for the movements of each swimmeret are located in that segment’s hemiganglion (Mulloney and Smarandache 2010). Swimming The swimmerets are used for forward swimming; their activity pattern is

conserved, stereotypic, and cannot be easily altered. Each swimmeret performs a very simple movement pattern: the power stroke (PS) creates forward thrust and the return stroke (RS) returns the limb to the start position close to the abdomen. These limbs are very well coordinated across segments during swimming. Activity starts in the most posterior segment with a PS and a metachronal wave travels, in this system, from posterior to anterior (Fig. 10.4B). The phase lag between segments is approximately 25%, and right and left sides are active in phase. This coordinated activity is also present in isolated preparations, and in contrast to leech swimming intersegmental phase lags are identical in vivo and in vitro. This pattern (ipsilateral synchrony, 25% intersegmental phase lag) is maintained across a very wide (10-fold) frequency range (Mulloney and Smarandache 2010). Neural Network For this system the CPG and coordinating network are known on the

cellular level (Fig. 10.5). The coordinating neural network alone is all that is required for behaviorally relevant coordination. The CPG in each hemiganglion consists of five nonspiking interneurons, three Inhibitors of PS (IPS) and two Inhibitors of RS (IRS), which form a half center oscillator and induce the alternating activity of the PS and RS motor neurons (Smarandache-Wellmann et al. 2013). Coordination of the ipsilateral segmental CPGs is accomplished by a neural network consisting of two spiking Coordinating Neurons and one nonspiking Commissural Interneuron in each hemiganglion. The coordinating neurons encode information about each oscillator’s motor output and project to the neighboring neural networks. The Ascending Coordinating Neuron (ASCE ) encodes PS activity and projects to all anterior segments, the Descending Coordinating (DSC) encodes RS activity and projects to all posterior segments (the long-range mechanism shown in Fig. 10.1.Cii).

317

318

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A2 C1

IPS ASCE

IRS DSC

PS

RS

A3 C1

IPS

IRS

ASCE

DSC PS

RS

Figure 10.5 Synaptic connectivity of intersegmental circuit that coordinates swimmeret CPGs in abdominal ganglia A2, A3, A4, and A5. The neural network in each hemiganglion contains the same type and number of neurons. The CPG is a half center oscillator formed by the pool of nonspiking IPS and IRS interneurons. They drive rhythmic activity in the motor neurons (Power Stroke, PS; Return Stroke, RS) via a direct inhibitory synapse. The coordinating neurons, ASCE and DSC, carry information from the CPG they originate in to all neighboring ganglia. ComInt1 (abbreviated C1 in figure) receives the coordinating information as a gradient of synaptic strength: ASCE input from the nearest neighbor segment is stronger than DSC input from the nearest neighbor and input from nearer ganglion is stronger than input from more distant ganglia. ComInt1 alters CPG activity via an electrical synapse to only one IRS neuron. The grey scale in leaving each hemiganglion is consistent in each figure. Arrows indicate the direction of impulse traffic in coordinating axons and motor axons.

A4 C1

IPS

IRS

ASCE

DSC PS

RS

A5 C1

IPS ASCE

PS

IRS

RS

inhibitory synapse excitatory synapse (size corresponds to synaptic strength) electrical synapse

ASCE and DSC receive the same input from the CPG neurons as do the motor neurons and have very similar firing patterns; they therefore can be described as efference copy neurons (Smarandache-Wellmann and Grätsch 2014). ASCE and DSC make excitatory synapses onto a third neuron, the nonspiking Commissural Interneuron 1 (ComInt1; in figure abbreviated as C1), not in their own but in all other ganglia. This coordinating information forms a gradient of synaptic

Coordination of Rhythmic Movements

strength onto ComInt1: input from the neighboring ASCE is always stronger than from the neighboring DSC, and input from coordinating neurons originating in more distant segments induces progressively weaker EPSPs (Smarandache-Wellmann et al. 2009). Each ComInt1 alters the activity of the CPG in its hemiganglion through a low-pass electrical synapse onto one of the IRS-CPG neurons (Smarandache-Wellmann et al. 2014). This weak asymmetric coupling is enough to synchronize the neural oscillators with the observed phase lag of 20-25% (Zhang et al. 2014). Sensory Feedback That phase lags do not differ in intact and isolated nerve cords imme-

diately shows that sensory feedback is not required for coordination. In fact, the swimmeret system in general seems to receive very little sensory regulation. Its only known sensory input is from a nonspiking stretch receptor. This pathway can alter the activity of the hemisegment in which it is located (Mulloney et al. 2014). However, the coordinated rhythm can be entrained only if the sensory pathways in at least three consecutive limbs are stimulated simultaneously (Deller and MacMillan 1989). 10.2.4 Insect Locomotion

In terrestrial locomotion the legs must both support the body weight and deal with uneven terrain. Presumably for these reasons, sensory feedback plays a much more important role in the neural systems generating terrestrial locomotion than in the preparations we have discussed thus far. Similar to the stomatogastric and leech systems, these neural systems must also produce multiple motor patterns to support walking at different speeds, forward and backward walking, and turning. The neural networks must therefore be highly tunable and react quickly to sensory information about the environment. A tripod gait, in which three legs are on the ground and the other three in swing, is used during fast hexapodal walking (Fig. 10.6A) in cockroach, stick insects, and Drosophila. In slow walking a tetrapod gait is used (Fig. 10.6B) in which only two, diagonally opposed, legs are simultaneously in swing while the others are on the ground (Cruse et al. 2009; Delcomyn 1971; Grabowska et al. 2012; Wosnitza et al. 2013).

A

B R1 R2 R3 L1 L2 L3 Tripod coordination

Tetrapod coordination

Figure 10.6 Idealized walking coordination patterns in hexapodal animals. (A) Insects use a tripod coordination pattern in fast walking: at a given time, a group of three legs (one from one side, two from the contralateral side) are on the ground (rectangles) and the other legs are in swing. (B) Most insects use a tetrapod coordination pattern during slow walks. In this case four legs are always on the ground and two legs on diagonally opposite sides of the body are in swing.

319

320

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Morphology of the Central Nervous System of Insects As arthropods, insects have segmented

bodies (Fig. 10.2C). These segments are organized into three larger structures, the head, thorax, and abdomen. The nerve cord is ventral, with a ganglion in each segment or fused or semi-fused ganglia serving several segments; for instance, the head with its cephalic and esophageal ganglia. The three pairs of walking legs attach to the thorax, and are innervated by the three thoracic ganglia: the first pair by the prothoracic ganglia, the middle pair by the mesothoracic, and the posterior pair by the metathoracic. The abdomen is segmented, but has no appendages. The brain and tail ganglia are involved in decision tasks. Stick Insect Walking As in all locomotor systems, stick insect walking activity is driven by segmental CPGs. In stick insects as well as in other animals with segmented legs, the antagonistic muscles of each joint are driven by their own CPGs (Büschges 2005). Sometimes the term unit bursts generators (UBG), defined as a CPG that activates a group of antagonistic motor neurons (Grillner 1981), is used. Stick insect legs have three main joints, the thoraco-coxal (TC), which moves the leg back (retraction) and forward (protraction); the coxa-trochanter (CTr), responsible for levation and depression; and the femur-tibia joint (FTi), responsible for flexion and depression. Each limb is thus controlled by three independent CPGs or UBGs (Bässler and Büschges 1998; Büschges 1995), although as yet these have not been described on the cellular level. The conclusion that the CPGs are independently active was reached by recording motor neuron activity in isolated systems. In these experiments no correlated activity was observed in motor neurons driven by different CPGs/UBGs, suggesting that this coordination resulted from sensory feedback from the limbs. Büschges (2005) reviews in detail how load and movement sensory information (Akay 2002; Bucher et al. 2003; Hellekes et al. 2012; Hess and Büschges 1999) control interjoint coordination of CPGs/UBGs in stick insect (Fig 10.1B mechanism). For example, sensory information for femur position (FTi joint) is reported to the nervous system by the femoral chordotonal organ. When this input indicates that the femur is extended (FTi CPG is active), the CPG for the CTr joint is activated to start leg depression. The opposite effect occurs when the input indicates the joint is flexed. Sensory information is required not only for interjoint coordination, but also interleg coordination (the tripod or tetrapod gaits described above). No coordinating neural network has been identified that could account for these coordinations, and it appears that sensory information plays the primary role in them. Sensory input from front leg stepping (prothoracic ganglion) can entrain middle segment (mesothoracic ganglion) rhythmic activity, and increases posterior segment (metathoracic ganglion) activity (Borgmann et al. 2009, 2012; Ludwar et al. 2005). That sensory information is sufficient for leg coordination has also been shown in modelling studies where the only signals for interjoint and intersegmental coordination are correctly timed sensory input (Borgmann et al. 2012; Daun-Gruhn 2011; Toth and Daun-Gruhn 2011; Toth et al. 2015). Walking in Other Insects Other small insects, e.g., Drosophila (Wosnitza et al. 2013), walk

with a variety of coordination patterns. These gaits change fluently and rapidly (i.e., over only a few cycle periods) from one to the next. Because of their small size, the locomotory CPGs in these animals have not been described on the cellular level. Presumably

Coordination of Rhythmic Movements

the same organization exists as in other hexapodal insects, with one CPG for each joint and sensory feedback playing a critical role in coordinating joint and leg movements (Berendes et al. 2016). With respect to more global aspects of locomotion, Bidaye et al. (2014) used genetic assays to identify two command neurons in the brain (MDN), and two in the tail ganglion (MAN), that change walking direction and therefore interleg coordination (see also Chapter 4). The MDN neurons make descending projections and trigger backward walking, even when animals do not need to do so. The MAN neurons make ascending projections and elicit persistent backward movement, accompanied by inhibition of forward walk inducing pathways. The morphology of the MAN and MDN neurons resembles the morphology of neurons found in stick insects and cockroach. There are thus likely homologous projection neurons in all insects that control global aspects (e.g., forward vs. backward) of walking (mechanisms in Fig. 10.1Di and ii). 10.2.5 Multiple Mechanisms Mediate Coordination in Invertebrate Systems

In the above examples are some cases in which single mechanisms dominate the coordination (e.g., central connections in swimmerets, sensory feedback in stick insect walking). As noted above, it is possible that these extremes are due to the choice of experimental preparation and the more common case is co-existence of multiple, cooperating coordination mechanisms. Relevant to this issue is work in leech swimming investigating the necessity of central coordinating connections. Isolated leech nerve cords can produce coordinated swimming movement without sensory feedback (Hocker et al. 2000; Kristan Jr and Calabrese 1976), but the phase lag is much shorter than that observed in vivo. Sensory input thus clearly plays a critical coordinating role. This observation raised the question whether sensory information alone could maintain behaviorally correct intersegmental phase lags. Yu et al. (1999) answered this question by showing that leech maintain coordinated swimming behavior even when the nerve cord is transected. As such, leech have two coordination mechanisms, one central (Fig. 10.1.Cii) and one based on sensory feedback (Fig. 10.1.B), that work together in vivo to maintain swimming-appropriate phase lags along the body.

10.3 Overview of Vertebrate CPGs As in invertebrates, vertebrate rhythmic movements are generated by CPGs, and CPGs involved in different motor functions are located in different regions of the central nervous system. In vertebrates, the locomotor CPGs are located in the spinal cord and those for respiration, mastication, swallowing, and airway defensive behaviors (cough, sneeze) are located in the brainstem (see also Fig. 7.2.1). CPG neural organization differs for different motor patterns. Although subject to modification, the motor patterns of respiration and mastication are relatively invariant. The CPGs underlying these activities are correspondingly simpler than those generating more plastic motor outputs. For instance, the respiratory CPG is composed of a recurrent excitatory network, sufficient to produce an adequate respiratory motor pattern from birth until death. Other behaviors require more flexible rhythmic movement generators to produce highly adaptable motor patterns. For example, locomotion is a

321

322

Neurobiology of Motor Control: Fundamental Concepts and New Directions

highly flexible behavior: individuals can walk fast, slow, forward, backward, and sideways. To generate multiple locomotor patterns, the underlying neural networks presumably require a complex organization. For locomotion in mammals, Grillner (1981) proposed that locomotor CPGs are composed of a series of “unit burst generators”, each of which controls the synergistic muscles of a single joint, with the multiple interactions possible between all these units CPGs providing the requisite flexibility. These basic units may have been conserved through evolution and the different modes of locomotion in different species (swimming, flying, walking) achieved simply by changing the interactions between these units. In absence of such a unit organization, each different locomotor pattern would require evolution of a new dedicated neural network. The CPGs involved in the rhythmic activities must interact with one another to produce motor behaviors adapted to the biomechanical and physiological constraints of the organism and its changing needs. For instance, swallowing is associated with a pause in respiration function to prevent food entering the airways. Another example is the significant increase in respiratory rhythm observed during exercise (locomotion) to meet the need for additional gas exchange caused by the increased metabolic demand. Interactions can also lead to a coupling between rhythms, e.g., coupling of the respiratory and locomotor motor patterns (Fig. 10.7) (Bramble and Carrier 1983; Reilly et al. 2009; Reilly and White 2009). We describe here the current knowledge about the neural mechanisms underlying CPG coordination in vertebrates. 10.3.1 General Characteristic of Vertebrate CPGs 10.3.1.1 Locomotor CPGs Intrinsic Function of Neural Networks In vertebrates, the rhythmic timing and coordination

of muscle contractions during locomotion are generated by CPGs located in the spinal cord. At rest, the CPGs are inactive. To initiate movement, a locomotor command is generated by supraspinal locomotor control centers that activate the spinal locomotor Figure 10.7 Coordination of respiratory and locomotor patterns in humans. In both panels upper trace is locomotory rhythm, bottom trace is respiration. (A) Four steps per breath. (B) One step per breath. Modified, with permission, from Bramble and Carrier (1983).

Coordination of Rhythmic Movements

networks (see Chapter 7.2). The locomotor CPGs consist of neurons playing two crucial functions: rhythm generation, constituting the clock, and pattern generation, producing left–right and flexor–extensor alternation. Anatomical Localization and Functional Organization In animals that locomote axially, e.g.,

lampreys, the locomotor CPGs are distributed along the entire spinal cord, whereas in limbed species, they are located in areas of the spinal cord controlling each limb. In mammalian quadrupeds, the CPGs controlling the forelimbs are located in the cervical enlargement and those controlling the hindlimbs are located in the lower thoracic and higher lumbar spinal cord. In all species, locomotor-associated neurons are located in the ventral part of the spinal cord in laminae VII, VIII, and X. The timing of the locomotor rhythm relies on non-commissural glutamatergic neurons. In lamprey and tadpole, these excitatory neurons are distributed segmentally along the cord and seem to function as burst-generating units that provide excitatory synaptic drive to motorneurons and inhibitory interneurons involved in left–right and intersegmental CPG coordination. Indeed, blocking synaptic transmission in the hemicord of lampreys revealed that the commissural inhibitory projection and ipsilateral inhibitory interneurons are not necessary for rhythm generation (Cangiano and Grillner 2003, 2005). Similar results were obtained in rodents and cats (Bonnot et al. 2002; Bracci et al. 1996; Kato 1987), suggesting that excitatory networks are also crucial for locomotor rhythm generation in mammals. In addition to generating the rhythm, locomotor CPGs have to generate the appropriate muscle activation pattern. Based on the unit burst generator theory (Grillner 1981), one such generator controls the activity of synergist muscles acting on each limb joint. Efficient locomotion requires that the activities of these generators be coordinated to produce correctly timed left–right and flexor–extensor alternation in limbed animals, and intersegmental coordination in animals that locomote axially. 10.3.1.2 Respiratory CPGs

Respiration is an automatic rhythmic behaviour generated by CPGs located in the brainstem. Due to its vital role for regulating oxygen uptake and expelling carbon dioxide, respiratory activity must continue throughout life. Depending on the species, respiration consists of one to three phases. In lampreys, respiratory activity consists of synchronous contractions of branchial muscles to expel water from the gills sacs, whereas frogs exhibit buccal ventilation coordinated with lung ventilation. In mammals, ventilation consists of inspiration, post-inspiration and late-expiration. The pre-Bötzinger complex (pre-BötC) is a small neural structure localized in the medulla, ventral to the nucleus ambiguous (Smith et al. 1991), that is both necessary and sufficient to generate inspiration. Rhythmic activity in pre-BötC neurons is believed to be generated through pacemaker properties acting in conjunction with excitatory synaptic interconnections within the network. The pre-BötC is interconnected with a more rostral region, the parafacial respiratory group (pFRG)/retrotrapezoid nucleus, believed to be the oscillator for active expiration. Together, the pre-BötC and pFRG form the core of the neural circuitry that generates respiration. However, because the metabolic needs of animals are not constant, the activity of respiratory CPGs must be finely controlled to match breathing frequency to metabolic need.

323

324

Neurobiology of Motor Control: Fundamental Concepts and New Directions

10.3.1.3 Feeding CPGs

As originally postulated by Meltzer (1899), swallowing is centrally programmed by a neuronal network. The neural substrate of the swallowing reflex consists of two groups of brainstem interneurons: the dorsal swallowing group (DSG) in the nucleus tractus solitarii, including some of the adjacent reticular formation, and the ventral swallowing group (VSG), located in the ventrolateral medulla. The DSG contains neurons that trigger, shape, and time the swallowing motor pattern whereas the VSG contains neurons that distribute the swallowing drive to the motor neurons to produce the motor activity (reviewed in Jean 2001; Lang 2009). Mastication consists of rhythmic jaw openings and closings and coordinated movements of cheeks, lips, and tongue. The masticatory pattern is generated by a CPG located in the brainstem between the rostral poles of the trigeminal and the facial motor nuclei. The interneurons involved in pattern generation are located in the peritrigeminal area surrounding the Vth motor nucleus, the nucleus reticularis parvocellularis, the rostral subdivision of the Vth spinal nucleus, and the dorsal part of the Vth principal nucleus (NVpr); the rhythmicity is probably produced by neurons with intrinsic burst properties located in the dorsal part of the NVpr (Brocard et al. 2006; Kolta et al. 2007; Morquette et al. 2015). 10.3.2 CPG Interactions within One Motor Function

As noted above, producing even a single motor locomotory pattern requires coordinating the activities of multiple CPGs. We first review the present state of knowledge on these interactions in limbless and limbed animals. 10.3.2.1 Unit Generators in Limbless Swimming Vertebrates

In lamprey and tadpole, swimming consists of an undulatory wave that propagates along the body and all body segments contribute to the propulsive movement. The lamprey spinal cord has been examined extensively and the neurons of the locomotor CPG have been defined. The core of the spinal networks generating swimming is excitatory interneurons (EINs) with axons projecting ipsilaterally (Buchanan and Grillner 1987). The EINs excite motor neurons and project to other rostral and caudal EINs on the same side (Fig. 10.8) (see also Fig. 6.9). Experiments performed on the isolated spinal cord show that bursting activity can be generated in each segment or hemi-segment. Left–right alternation at each segment is also required for efficient swimming. It is produced by glycinergic inhibitory interneurons that project contralaterally and caudally (CCINs). EINs activate the CCINs, which in return inhibit the contralateral CCINs. This inhibition coordinates and stabilizes the EIN networks on both sides, leading to an efficient left–right alternation. Sensory feedback from excitatory and inhibitory stretch receptors (SR-Es and SR-Is) also plays a role in left–right coordination. SR-Es excite ipsilateral MNs and CCINs and SR-Is inhibit contralateral CCINs and SR-Es (Di Prisco et al. 1990; Grillner et al. 1982, 1984). During swimming, contraction of right-side myotomes stretches the left side of the body, exciting both left-side SR-Es and left-side SR-Is. The excited SR-Es induce bursting on the left side of the cord (Wallén 1997) while the SR-Is help terminate the ongoing bursting activity on the right side, thus helping trigger the transition from right side to left side contraction. Left side contraction, in turn, excites right side SR-Es and

Coordination of Rhythmic Movements

Figure 10.8 Schematic representation of the neuronal network responsible for generating lamprey swimming. On each side, excitatory interneurons (EINs) project to the motorneurons (MN) and two populations of inhibitory interneurons (LINs, CCINs). The LINs inhibit the CCINs on the same side, whereas the axons of the CCINs project across the midline and inhibit the contralateral CCINs. Stretch-receptor neurons excite neurons on the same side (SR-Es) or inhibit neurons on the contralateral side (SR-Is). Newly drawn, but based on data from Grillner et al. (1995).

SR-Is, whose activity then helps end the ongoing left side contraction and trigger the next right side contraction, and the process repeats. The lamprey spinal cord consists of 100 segments that must be coordinated along the rostrocaudal axis. In forward swimming, the rostral segments are activated first and the other segments follow from head to tail (as in Fig. 10.4A) with a positive intersegmental phase lag of 1% of cycle duration (Grillner 1974; Wallén and Williams 1984). The lamprey can also produce backward swimming with a caudorostral traveling wave from the tail to the head (as in Fig. 10.4B), in this case with an intersegmental phase lag of –1% of the cycle duration (Islam et al. 2006). The neural circuitry responsible for the rostrocaudal coordination is not identified yet. It was proposed that neurons with ascending axons play a significant role (Cohen et al. 1992; Guan et al. 2001), but their cellular properties are still unknown. The excitability gradient between segments was also found to be crucial (Matsushima and Grillner 1992; reviewed in Grillner and Wallén 2002). When the excitability level is high in the rostral segments near the head, the wave of contraction propagates from the head to the tail to propel the animal forward. When caudal segments are more excited, the wave of contraction is reversed and the animal swims backward. Interestingly, the same conclusions were drawn from computational models created to describe swimming in salamanders (Bicanski et al. 2013; Ijspeert et al. 2007; Knüsel et al. 2013; Ryczko et al. 2015). Coordination of swim CPG activity in lamprey thus consists of multiple mechanisms, including multi-segment influences up and down the cord (Fig. 10.1Cii), crossed connections between specific neurons in the right and left CPGs in single segments (Fig. 10.1A), and sensory induced feedback (Fig. 10.1B). 10.3.2.2 Unit Generators in Mammalian Limbs (see also Chapter 8 and Fig. 8.6)

According to the unit burst generator theory (Grillner 1981), separate burst generators rhythmically activate synergist muscles at each limb joint. Data obtained with optogenetic tools (see Chapter 4) have demonstrated that locomotor activity can indeed

325

326

Neurobiology of Motor Control: Fundamental Concepts and New Directions

be induced unilaterally or independently in flexor or extensor networks by specific activation of glutamatergic neurons (Hagglund et al. 2013). These results support the hypothesis that locomotor networks are composed of distributed and interconnected unit burst generators that are selectively recruited to produce specific locomotor patterns. Although the neurons responsible for the rhythm generation itself remain unknown, recent genetic and optogenetic work studying transcription factor expression has revealed multiple subclasses of spinal interneurons responsible for interconnecting the unit burst generators. Left–Right Alternation As in lamprey, left–right alternation is mediated by commissural

interneurons (CINs) whose axons cross the midline. In limbed vertebrates, however, this coordination consists of two inhibitory pathways. The first pathway is composed of inhibitory CINs and the second of excitatory CINs that activate inhibitory neurons. Ablating commissural V0 interneurons (V0 INs) in mice alters left–right coordination during drug-induced locomotion (Lanuza et al. 2004) and, when all V0 INs are deleted, hopping remains the only gait generated (Bellardita and Kiehn 2015). The V0 INs are thus crucial for left–right alternation. However, V0 INs are not a homogeneous population, with some being excitatory (V0V ) and others inhibitory (V0D ). Selective ablation of the inhibitory V0D INs substantially alters left–right alternation at low locomotor frequencies, but does not affect left–right alternation at high frequencies (Talpalar et al. 2013). Selective ablation of excitatory V0v INs, alternatively, results in left–right alternation being maintained at low locomotor frequencies but in a hopping gait at high frequencies. These results suggest that the left–right coordination in quadrupeds uses different mechanisms at different locomotor speeds. In addition to the CINs, ipsilateral excitatory INs are also involved in left–right alternation. Genetic ablation of ipsilateral glutamatergic V2a INs changes bilateral alternation (Crone et al. 2008), especially at high locomotor speeds (Crone et al. 2009). How these ipsilaterally active neurons affect left–right coordination is not completely understood. The V2a INs are known to directly excite the V0 CINs. The effects of the V2a INs in left–right alternation presumably result from an indirect inhibitory pathway. Flexor–Extensor Alternation Locomotion in limbed animals also requires precise coor-

dination of flexor and extensor muscle activity. Lesion experiments, show that commissural projections are not involved in flexor–extensor alternation (Kjaerulff and Kiehn 1996). Flexor and extensor activities are generated by individual but interconnected neural networks. During locomotion, flexor and extensor motor neurons receive inhibitory and excitatory inputs, with the inhibitory input playing a particularly important role in flexor/extensor alternation. Renshaw cells (Renshaw 1946) and reciprocal Ia interneurons are two populations of inhibitory interneurons that project to motor neurons. Renshaw cells are activated by motor neuron axon collaterals and recurrently inhibit the motor neurons that activate them. Renshaw cells fire rhythmically in phase with motor neuron activity, most strongly at its end (Pratt and Jordan 1987). Renshaw cells may thus help terminate motor neuron discharge during locomotion. Ia interneurons also regulate flexor–extensor alternation. They are activated by Ia afferents from muscle spindles and fire rhythmically during locomotion and participate in reciprocal inhibition of motor neurons of antagonist muscles (Jankowska and Roberts 1972; see

Coordination of Rhythmic Movements

also Chapter 8). Both Renshaw cells and Ia interneurons are thus good candidates for producing flexor–extensor alternation. Anatomical studies in postnatal rodents show that Renshaw cells and Ia INs derive from the inhibitory and ipsilateral V1 IN population. The involvement of Renshaw cells and Ia interneurons in flexor-extensor alternation was therefore further characterized by reversibly inactivating or completely suppressing V1 IN activity in mice. Although both treatments increase cycle period, they do not change the phase relationship of flexor–extensor motor neuron activity, suggesting that populations of ipsilateral-projecting inhibitory interneurons other than the V1 INs are involved in flexor–extensor alternation. Complete abrogation of the V1 and V2 INs does suppress flexor–extensor alternation, suggesting that V1–V2-derived interneurons play a crucial role in flexor–extensor coordination in limbed vertebrates. 10.3.3 CPGs Interactions for Different Motor Functions 10.3.3.1 Coordination of Respiration and Swallowing

Swallowing requires coordinated activity of muscles in the mouth, pharynx, larynx, and esophagus. It is divided in three phases: oral, pharyngeal, and esophageal. The oral phase is characterized by the preparation of the bolus during mastication. The pharyngeal phase is defined by the sequential contraction of the tongue and pharyngeal muscles that transfers the bolus to the esophagus. The third phase is the peristalsis of the esophagus muscles that propels the bolus to the stomach. Because the pharynx constitutes a common pathway for both air flow and food, it is crucial that swallowing and respiration be well-coordinated to allow passage of food through the pharynx without tracheal aspiration (Doty and Bosma 1956; Gestreau et al. 2000; Shiba et al. 1999), and the respiratory rhythm is indeed modulated during swallowing (Sumi 1975). Deglutition (swallowing) in humans causes an abrupt decrease in airflow leading to apnea, followed by a period of expiration (Paydarfar et al. 1995). In cats, swallowing similarly results in an inhibition of inspiration associated with a prolongation of the late expiratory phase (Dick et al. 1993). These specific resettings of the respiratory rhythm demonstrate that during swallowing the normal respiratory pattern is reconfigured to produce what is referred to as a “swallowing-respiratory pattern”. How this coordination is achieved is not yet completely understood on the cellular level. One hypothesis posits that this coordination arises from an interaction between the dorsal swallow group (DSG), located in the nucleus tractus solitarii, and the ventral swallow group (VSG), located in the ventral lateral medulla. Swallowing is initiated by activating inputs from the superior laryngeal nerve (SNL) and supramedullary brain centers to the DSG. The swallowing-generated command is relayed via the VSG, which sends efferent projections to motor neurons controlling muscles involved in deglutition and respiration. Some DSG neurons are also part of the dorsal respiratory group. The effect of swallowing on respiration could thus result from regulation and reconfiguration of these respiratory elements by the DSG swallowing CPG. Although inspiration stops during a swallowing, a small contraction of the diaphragm is still present. The contraction is reduced compared to that during normal respiration, and phrenic premotor neurons of the dorsal respiratory group show decreased peak discharge frequency during swallowing (Gestreau et al. 1996). Although the mechanism underlying these changes is unknown, electrophysiological recordings of brainstem

327

328

Neurobiology of Motor Control: Fundamental Concepts and New Directions

neurons show substantial changes in neuron activity in regions involved in respiratory rhythm or pattern generation. In particular, swallowing induced by superior laryngeal nerve stimulation is associated with inhibition of inspiratory-modulated neurons of the pre-Bötzinger region, a structure crucially involved in inspiratory rhythm generation. The Bötzinger complex could also be involved in the inspiratory inhibition present during swallowing. In addition to its role in respiratory rhythm genesis, the Bötzinger complex also plays a role in terminating inspiration and in delaying inspiration onset by prolongation of expiration. Neurons in the Bötzinger complex display an altered pattern of activity during SNL superior laryngeal nerve stimulation. Expiratory decrementing (E-DEC) neurons are activated during swallowing and expiratory augmenting (E-AUG) neurons are activated following each swallow (Saito et al. 2003). Because E-DEC neurons may contribute to the inhibition of phrenic premotor neurons, it has been suggested that the prolonged expiration following swallowing could result from the activation of E-DEC and E-AUG neurons. By modulating the activation of the E-DEC and E-AUG neurons, the swallowing CPG could also reset the respiratory rhythm, ensuring that swallowing is followed by expiration, not inspiration. Multi-neuronal recordings in the brainstem reveal that neurons quiescent during respiration are recruited during swallowing. Cross-correlation of these recordings shows that these neuron populations have excitatory or inhibitory connections with inspiratory and expiratory neurons. Recruitment of these specific neurons also excites E-DEC neurons, known to inhibit inspiration and prolong expiration. These data suggest that recruitment of specific non-respiratory modulated neurons during swallowing plays a central role in generating the swallowing-respiratory pattern. In summary, coordination between swallowing and respiration is based on interactions between the two CPGs (Fig. 10.9). It appears that the swallowing-respiratory pattern results from (1) activation of the swallowing CPG to produce appropriately-timed contractions of muscles involved in the pharyngeal phase; (2) reconfiguration of the respiratory CPG to stop inspiration and produce the respiratory-swallowing pattern; and (3) modulation of the swallowing and respiratory CPGs by additional neural structures in the brainstem, such as the parabrachial and Kölliker-Fuse nuclei. As such, coordination of swallowing and respiration appears to involve a mixture of direct interconnections between the neurons of the two CPGs (Fig. 10.1A) and alterations in the controlling drive to the CPGs (Fig. 10.1Di, Dii). 10.3.3.2 Coordination of Locomotion and Respiration

Respiration is a vital motor behavior whose intensity must be modulated to satisfy metabolic demands. Changes in respiratory output are considerable during exercise as respiratory frequency increases to meet the larger O2 needs of the body. Locomotor and respiratory CPGs therefore need to be coordinated to produce motor output appropriate to the physiological needs and biomechanical constraints of the animal. During high-speed locomotion, this coordination can lead to a coupling between the two rhythms (Fig. 10.7). According to current theories, locomotor-respiratory coordination can result from either central or peripheral mechanisms, which we now describe.

Coordination of Rhythmic Movements

Figure 10.9 Schematic representation of the interactions between the respiratory and swallowing CPGs. Respiration and swallowing depend on two overlapping CPGs. Inputs from the superior laryngeal nerve and central structures activate the swallowing CPG. Some respiratory neurons are involved in swallowing, resulting in a partial overlap (middle portion of figure) between the two CPGs.

Feedback Mechanisms Chemoreception Hypothesis Exercise elevates O2 demand and increases CO2 production.

The influence of these two compounds on human respiration has been long known: increasing the CO2 /O2 ratio in inspired air increases respiratory frequency (Haldane and Priestley 1905). The possibility that central chemoreceptors are involved in respiratory increases during exercise has been investigated in ponies by measuring the changes in brainstem cerebrospinal fluid acidity and gas concentrations during exercise (Bisgard et al. 1978). Surprisingly, 9 minutes of exercise only slightly increased pH and decreased CO2 partial pressure, suggesting that central chemoreceptors unlikely play a crucial role. The role of peripheral chemoreceptors was investigated by silencing them, which reduced the exercise-induced increase in respiratory frequency by 15–20%. Bilateral carotid body resection in humans produces conflicting results: Honda et al. (1985) observed a reduced respiratory response during exercise whereas Lugliani et al. (1971) saw no changes in respiratory response. Taken together, these data suggest that central and peripheral chemoreceptors play at most a moderate role in increasing respiratory drive during locomotion. To the extent they do, they would combine the mechanisms shown in Fig. 10.1B and Di, Dii in that they would function by locomotion-triggered sensory input altering central drive to the respiratory network. Peripheral Nervous Feedback Hypothesis An alternative hypothesis is that sensory feedback

induced by muscle contractions modulates respiratory activity. Inducing muscle contraction by stimulation of the spinal ventral roots modulates respiratory activity in anesthetized dogs (Comroe and Schmidt 1943). Passive limb movements increase respiratory frequency in humans (Bell and Duffin 2004) and passive wing flapping in Canada geese entrains the respiratory rhythm (Funk et al. 1992). These responses are probably mediated by activation of joint, muscle, and/or tendon proprioceptive afferents. Repeated stimulation of skin or muscle afferents, known to be rhythmically active during locomotion, also entrains the respiratory rhythm (Howard et al. 1969; Iscoe and Polosa 1976), and single stimulations applied during the expiratory phase can

329

330

Neurobiology of Motor Control: Fundamental Concepts and New Directions

reset the rhythm. Cutting the dorsal roots abolishes the responses, further supporting a role for sensory afferents. Rhythmic activation of limb somatic afferents by low-threshold electrical stimulation can reset or entrain respiration in the in vitro isolated central nervous system of newborn rats (Giraudin et al. 2008; Morin and Viala 2002) or reduced in situ preparations of juvenile rats (Potts et al. 2005). These ascending influences from the limb afferents are conveyed to a brainstem pontine relay located in the parabrachial/Kölliker-Fuse (PB/KF) complex (Giraudin et al. 2012; Potts et al. 2005) (Fig. 10.10). It thus appears that feedback from movement-activated sensory afferents (Fig. 10.1B mechanism) plays a substantial role in coordinating locomotor and respiratory activity. Feedforward Mechanisms

In addition to feedback mechanisms, feedforward connections also play a critical role in modulating respiration during locomotion. Krogh and Lindhard (1913) showed that, during exercise in humans, changes in breathing frequency occur less than a second

Figure 10.10 Schematic representation of the neural mechanisms involved in locomotor and respiratory CPG coordination. Both feedforward and feedback mechanisms are present. The feedforward mechanism relies on central connections from supraspinal structures involved in the control of locomotion (MLR) or spinal locomotor CPGs onto the brainstem respiratory CPG (left and middle pathways). The feedback mechanism results from the activation of chemoreceptors or proprioceptors activated during locomotion (right pathways).

Coordination of Rhythmic Movements

after locomotion initiation. Because blood from the contracting muscles cannot reach the respiratory networks in the medulla within this time, the authors postulated that the quick-onset changes they observed in respiratory rate could not result from changes in CO2 or O2 blood concentrations. They therefore hypothesized that a purely neurogenic mechanism modulated respiratory frequency during exercise. This hypothesis was supported many years later by experiments that electrically or chemically stimulated supraspinal regions known to be involved in the control of locomotion. This work showed that electrical stimulation of the lateral hypothalamic area elicited a rhythmic activity similar to locomotion (fictive locomotion) in paralysed cats (Eldridge et al. 1981, 1985), and that this activity was associated with an increase in respiratory frequency similar to that observed during active locomotion in intact animals. The authors concluded that central neural networks, probably supraspinal, actively modulated respiratory activity during locomotion and postulated that a feedforward mechanism was involved. Locomotion triggered by stimulation of the dorsolateral funiculi of the cervical spinal cord also increases respiratory activity (Romaniuk et al. 1994). This work, however, does not exclude that the modulation originates from the spinal locomotor CPGs. This issue was also addressed by work in lampreys showing that respiratory frequency increases before the initiation of spontaneous locomotion (Gariepy et al. 2012; Gravel et al. 2007). These data suggest that the increase of respiratory frequency does not require activation of spinal locomotor CPGs, but is programmed centrally. Data from the in vitro lamprey brainstem-spinal cord preparation revealed the central mechanism involved. Stimulating the mesencephalic locomotor region (MLR) in lampreys significantly increases respiratory drive and elicits synaptic responses in respiratory generator neurons (Gariepy et al. 2012). Removing the spinal cord does not prevent these effects. A direct glutamatergic projection from the MLR to the respiratory CPGs has been described (Fig. 10.10) and proposed be a key central player in coordinating respiration with locomotion. Connections between spinal locomotor CPGs and respiratory networks also exist. Work in rabbits showed that spinal locomotor CPGs must be maintained intact for MLR stimulation to increase respiratory frequency (Corio et al. 1993). Consistent with this, the lumbar locomotor CPG modulates respiratory activity through an ascending pathway acting via a substance P-releasing projection within the brainstem (Le Gal et al. 2014). In addition, pre-I neurons in the pFRG substantially depolarize during spinal locomotor CPG activation. Spinal locomotor CPG projections to the respiratory network thus also likely play an important role in coordinating respiration and locomotion. These data describe some of the connections responsible for central, feedforward neural connections coordinating locomotor and respiratory activity. With respect to coordination mechanism type, present data are too little for certain assignment, but likely involve both direct connections between the two CPGs and modifications of drive to the respiratory CPG (Fig. 10.10; Fig. 10.1A and Di, Dii).

10.4 Conclusion In both invertebrates and vertebrates, coordination arises through multiple mechanisms: direct interconnections between the relevant neural networks, sensory input

331

332

Neurobiology of Motor Control: Fundamental Concepts and New Directions

triggered by one network’s movements that feeds back onto the other network, chains of networks connected by either short or long range connections, differential alterations in drive from higher centers to the networks, global changes in animal internal state that differentially affect the networks. In systems with slow muscles, behavioral coordination can also occur by muscle slow filtering extracting modulation of neural activity imposed by one network on another. A common occurrence in the invertebrate examples, and a constant theme in the vertebrate examples, are multiple mechanisms working in concert to produce coordination. Having multiple mechanisms would allow fine control of the degree of coordination and close matching of behavior to need. It would therefore not be surprising if concerted action via multiple mechanisms is not the most common condition.

References Akay T (2002) The role of sensory signals for interjoint coordination in stick insect legs (Carausius morosus and Cuniculina impigra). PhD thesis, Universität zu Köln, Köln, Germany. Bartos M, Manor Y, Nadim F, Marder E, Nusbaum MP (1999) Coordination of fast and slow rhythmic neuronal circuits. J Neurosci 19:6650–6660. Bartos M, Nusbaum MP (1997) Intercircuit control of motor pattern modulation by presynaptic inhibition. J Neurosci 17:2247–2256. Bässler U, Büschges A (1998) Pattern generation for stick insect walking movements—multisensory control of a locomotor program. Brain Res Rev 27:65–88. Beenhakker MP, Blitz DM, Nusbaum MP (2004) Long-lasting activation of rhythmic neuronal activity by a novel mechanosensory system in the crustacean stomatogastric nervous system. J Neurophysiol 91:78–91. Beenhakker MP, Kirby MS, Nusbaum MP (2007) Mechanosensory gating of proprioceptor input to modulatory projection neurons. J Neurosci 27:14308–14316. Beierlein M, Gibson JR, Connors BW (2003) Two dynamically distinct inhibitory networks in layer 4 of the neocortex. J Neurophysiol 90:2987–3000. Bell HJ, Duffin J (2004) Respiratory response to passive limb movement is suppressed by a cognitive task. J Appl Physiol 97:2112–2120. Bellardita C, Kiehn O (2015) Phenotypic characterization of speed-associated gait changes in mice reveals modular organization of locomotor networks. Curr Biol 25:1426–1436. Berendes V, Zill SN, Büschges A, Bockemühl T (2016) Speed-dependent interplay between local pattern-generating activity and sensory signals during walking in Drosophila. J Exp Biol 219:3781–3793. Bicanski A, Ryczko D, Cabelguen J, Ijspeert A (2013) From lamprey to salamander: an exploratory modeling study on the architecture of the spinal locomotor networks in the salamander. Biol Cybern 107:565–587. Bidaye SS, Machacek C, Wu Y, Dickson BJ (2014) Neuronal control of Drosophila walking direction. Science 344:97–101. Bisgard GE, Forster HV, Byrnes B, Stanek K, et al. (1978) Cerebrospinal fluid acid–base balance during muscular exercise. J Appl Physiol 45:94–101. Blitz DM, Beenhakker MP, Nusbaum MP (2004) Different sensory systems share projection neurons but elicit distinct motor patterns. J Neurosci 24:11381–11390.

Coordination of Rhythmic Movements

Blitz DM, Christie AE, Coleman MJ, Norris BJ, et al. (1999) Different proctolin neurons elicit distinct motor patterns from a multifunctional neuronal network. J Neurosci 19:5449–5463. Bonnot A, Whelan PJ, Mentis GZ, O’Donovan MJ (2002) Locomotor-like activity generated by the neonatal mouse spinal cord. Brain Res Brain Res Rev 40:141–151. Borgmann A, Hooper SL, Büschges A (2009) Sensory feedback induced by front-leg stepping entrains the activity of central pattern generators in caudal segments of the stick insect walking system. J Neurosci 29:2972–2983. Borgmann A, Toth TI, Gruhn M, Daun–Gruhn S, Buschges A (2012) Dominance of local sensory signals over inter-segmental effects in a motor system: experiments. Biol Cybern 105:399–411. Bracci E, Ballerini L, Nistri A (1996) Spontaneous rhythmic bursts induced by pharmacological block of inhibition in lumbar motorneurons of the neonatal rat spinal cord. J Neurophysiol 75:640–647. Bramble DM, Carrier DR (1983) Running and breathing in mammals. Science 219:251–256. Brocard F, Verdier D, Arsenault I, Lund JP, Kolta A (2006) Emergence of intrinsic bursting in trigeminal sensory neurons parallels the acquisition of mastication in weanling rats. J Neurophysiol 96:2410–2424. Brodfuehrer PD, Debski EA, O’Gara BA, Friesen WO (1995) Neuronal control of leech swimming. J Neurobiol 27:403–418. Buchanan JT, Grillner S (1987) Newly identified glutamate interneurons and their role in locomotion in the lamprey spinal cord. Science 236:312–314. Bucher D, Akay T, DiCaprio RA, Büschges A (2003) Interjoint coordination in the stick insect leg-control system: the role of positional signaling. J Neurophysiol 89:1245–1255. Bucher D, Taylor AL, Marder E (2006) Central pattern generating neurons simultaneously express fast and slow rhythmic activities in the stomatogastric ganglion. J Neurophysiol 95:3617–3632. Büschges A (1995) Role of local nonspiking interneurons in the generation of rhythmic motor activity in the stick insect. J Neurobiol 27:488–512. Büschges A (2005) Sensory control and organization of neural networks mediating coordination of multisegmental organs for locomotion. J Neurophysiol 93:1127–1135. Cang JJ, Friesen WO (2002) Model for intersegmental coordination of leech swimming: central and sensory mechanisms. J Neurophysiol 87:2760–2769. Cangiano L, Grillner S (2003) Fast and slow locomotor burst generation in the hemispinal cord of the lamprey. J Neurophysiol 89:2931–2942. Cangiano L, Grillner S (2005) Mechanisms of rhythm generation in a spinal locomotor network deprived of crossed connections: the lamprey hemicord. J Neurosci 25:923–935. Cohen A, Ermentrout G, Kiemel T, Kopell N, et al. (1992) Modelling of intersegmental coordination in the lamprey central pattern generator for locomotion. Trends Neurosci 15:434–438. Combes D, Meyrand P, Simmers J (1999) Dynamic restructuring of a rhythmic motor program by a single mechanoreceptor neuron in lobster. J Neurosci 19:3620–3628. Combes D, Simmers J, Moulins M (1995) Structural and functional characterization of a muscle tendon proprioceptor in lobster. J Comp Neurol 363:221–234. Combes D, Simmers J, Moulins M (1997) Conditional dendritic oscillators in a lobster mechanoreceptor neurone. J Physiol (Lond) 499:161–177.

333

334

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Comroe JH, Schmidt CF (1943) Reflexes from the limbs as a factor in the hyperpnea of muscular exercise. Am J Physiol 138:536–547. Corio M, Palisses R, Viala D (1993) Origin of the central entrainment of respiration by locomotion facilitated by MK 801 in the decerebrate rabbit. Exp Brain Res 95:84–90. Crone SA, Quinlan KA, Zagoraiou L, Droho S, et al. (2008) Genetic ablation of V2a ipsilateral interneurons disrupts left–right locomotor coordination in mammalian spinal cord. Neuron 60:70–83. Crone SA, Zhong G, Harris-Warrick R, Sharma K (2009) In mice lacking V2a interneurons, gait depends on speed of locomotion. J Neurosci 29:7098–7109. Cruse H, Dürr V, Schilling M, Schmitz J (2009) Principles of insect locomotion. In Spatial Temporal Patterns for Action-Oriented Perception in Roving Robots. Arena P, Patanè L (eds). Berlin, Germany: Springer. Daun-Gruhn S (2011) A mathematical modeling study of inter–segmental coordination during stick insect walking. J Comput Neurosci 30:255–278. Daur N, Nadim F, Stein W (2009) Regulation of motor patterns by the central spike-initiation zone of a sensory neuron. Eur J Neurosci 30:808–822. Delcomyn F (1971) The locomotion of the cockroach Periplaneta americana. J Exp Biol 54: 443–452. Deller SRT, MacMillan DL (1989) Entrainment of the swimmeret rhythm of the crayfish to controlled movements of some of the appendages. J Exp Biol 144:257–278. Di Prisco GV, Wallén P, Grillner S (1990) Synaptic effects of intraspinal stretch receptor neurons mediating movement-related feedback during locomotion. Brain Res 530:161–166. Dick TE, Oku Y, Romaniuk JR, Cherniack NS (1993) Interaction between central pattern generators for breathing and swallowing in the cat. J Physiol (Lond) 465:715–730. Dickinson PS, Mecsas C, Marder E (1990) Neuropeptide fusion of two motor-pattern generator circuits. Nature 344:155–158. Doty RW, Bosma JF (1956) An electromyographic analysis of reflex deglutition. J Neurophysiol 19:44–60. Eldridge FL, Millhorn DE, Kiley JP, Waldrop TG (1985) Stimulation by central command of locomotion, respiration and circulation during exercise. Respir Physiol 59:313–337. Eldridge FL, Millhorn DE, Waldrop TG (1981) Exercise hyperpnea and locomotion: parallel activation from the hypothalamus. Science 211:844–846. Elson RC, Panchin YV, Arshavsky YI, Selverston AI (1994) Multiple effects of an identified proprioceptor upon gastric pattern generation in spiny lobsters. J Comp Physiol A 174:317–329. Faumont S, Combes D, Meyrand P, Simmers J (2005) Reconfiguration of multiple motor networks by short- and long-term actions of an identified modulatory neuron. Eur J Neurosci 22:2489–2502. Friesen WO, Poon M, Stent GS (1978) Neural control of swimming in the medicinal leech. IV. Identification of a network of oscillatory interneurones. J Exp Biol 75:25–43. Funk GD, Milsom WK, Steeves JD (1992) Coordination of wingbeat and respiration in the Canada goose. I. Passive wing flapping. J Appl Physiol 73:1014–1024. Gariepy JF, Missaghi K, Chevallier S, Chartre S, et al. (2012) Specific neural substrate linking respiration to locomotion. Proc Natl Acad Sci USA 109:84–92. Gestreau C, Grelot L, Bianchi AL (2000) Activity of respiratory laryngeal motorneurons during fictive coughing and swallowing. Exp Brain Res 130:27–34.

Coordination of Rhythmic Movements

Gestreau C, Milano S, Bianchi AL, Grelot L (1996) Activity of dorsal respiratory group inspiratory neurons during laryngeal-induced fictive coughing and swallowing in decerebrate cats. Exp Brain Res 108:247–256. Giraudin A, Cabirol-Pol MJ, Simmers J, Morin D (2008) Intercostal and abdominal respiratory motorneurons in the neonatal rat spinal cord: spatiotemporal organization and responses to limb afferent stimulation. J Neurophysiol 99:2626–2640. Giraudin A, Le Bon-Jego M, Cabirol MJ, Simmers J, Morin D (2012) Spinal and pontine relay pathways mediating respiratory rhythm entrainment by limb proprioceptive inputs in the neonatal rat. J Neurosci 32:11841–11853. Grabowska M, Godlewska E, Schmidt J, Daun-Gruhn S (2012) Quadrupedal gaits in hexapod animals—inter-leg coordination in free-walking adult stick insects. J Exp Biol 215:4255–4266. Gravel J, Brocard F, Gariepy JF, Lund JP, Dubuc R (2007) Modulation of respiratory activity by locomotion in lampreys. Neuroscience 144:1120–1132. Grillner S (1974) On the generation of locomotion in the spinal dogfish. Exp Brain Res 20:459–470. Grillner S (1981) Control of locomotion in bipeds, tetrapods and fish. In Handbook of Physiology. The Nervous System. Motor Control. Brooks V (ed). Bethesda, MD: Am Physiol Soc. Grillner S, Deliagina T, El Manira A, Hill RH, et al. (1995) Neural networks that co-ordinate locomotion and body orientation in lamprey. Trends Neurosci 18:270–279. Grillner S, McClellan A, Sigvardt K (1982) Mechanosensitive neurons in the spinal cord of the lamprey. Brain Res 235:169–173. Grillner S, Wallén P (2002) Cellular bases of a vertebrate locomotor system—steering, intersegmental and segmental co-ordination and sensory control. Brain Res Rev 40: 92–106. Grillner S, Williams T, Lagerback PA (1984) The edge cell, a possible intraspinal mechanoreceptor. Science 223:500–503. Guan L, Kiemel T, Cohen A (2001) Impact of movement and movement-related feedback on the lamprey central pattern generator for locomotion. J Exp Biol 4:2361–2370. Hagglund M, Dougherty KJ, Borgius L, Itohara S, et al. (2013) Optogenetic dissection reveals multiple rhythmogenic modules underlying locomotion. Proc Natl Acad Sci USA 110:11589–11594. Haldane JS, Priestley JG (1905) The regulation of the lung-ventilation. J Physiol (Lond) 32:225–266. Hedrich UB, Diehl F, Stein W (2011) Gastric and pyloric motor pattern control by a modulatory projection neuron in the intact crab Cancer pagurus. J Neurophysiol 105:1671–1680. Hedrich UB, Smarandache CR, Stein W (2009) Differential activation of projection neurons by two sensory pathways contributes to motor pattern selection. J Neurophysiol 102: 2866–2879. Hedrich UBS, Stein W (2008) Characterization of a descending pathway: activation and effects on motor patterns in the brachyuran crustacean stomatogastric nervous system. J Exp Biol 211:2624–2637. Hellekes K, Blincow E, Hoffmann J, Büschges A (2012) Control of reflex reversal in stick insect walking: effects of intersegmental signals, changes in direction, and optomotor–induced turning. J Neurophysiol 107:239–249.

335

336

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Hess D, Büschges A (1999) Role of proprioceptive signals from an insect femur–tibia joint in patterning motorneuronal activity of an adjacent leg joint. J Neurophysiol 81: 1856–1865. Hill AAV, Masino MA, Calabrese RL (2003) Intersegmental coordination of rhythmic motor patterns. J Neurophysiol 90:531–538. Hocker CG, Yu XT, Friesen WO (2000) Functionally heterogeneous segmental oscillators generate swimming in the medicinal leech. J Comp Physiol A 186:871–883. Honda Y (1985) Role of carotid chemoreceptors in control of breathing at rest and in exercise: studies on human subjects with bilateral carotid body resection. Jap J Physiol 35:535–544. Hooper SL, Moulins M (1989) Switching of a neuron from one network to another by sensory-induced changes in membrane properties. Science 244:1587–1589. Howard P, Bromberger-Barnea B, Fitzgerald RS, Bane HN (1969) Ventilatory responses to peripheral nerve stimulation at different times in the respiratory cycle. Respir Physiol 7:389–398. Ijspeert A, Crespi A, Ryczko D, Cabelguen J (2007) From swimming to walking with a salamander robot driven by a spinal cord model. Science 315:1416–1420. Iscoe S, Polosa C (1976) Synchronization of respiratory frequency by somatic afferent stimulation. J Appl Physiol 40:138–148. Islam SS, Zelenin PV, Orlovsky GN, Grillner S, Deliagina TG (2006) Pattern of motor coordination underlying backward swimming in the lamprey. J Neurophysiol 96: 451–460. Jankowska E, Roberts WJ (1972) An electrophysiological demonstration of the axonal projections of single spinal interneurones in the cat. J Physiol (Lond) 222:597–622. Jean A (2001) Brain stem control of swallowing: neural network and cellular mechanisms. Physiol Rev 81:929–969. Johnson BR, Hooper SL (1992) Overview of the stomatogastric nervous system. In Dynamic Biological Networks: The Stomatogastric Nervous System. Harris-Warrick RM, Marder E, Selverston AI, Moulins M (eds). Cambridge, MA: MIT Press. Kato M (1987) Motorneuronal activity of cat lumbar spinal cord following separation from descending or contralateral impulses. Cent Nerv Syst Trauma 4:239–248. Katz PS, Eigg MH, Harris-Warrick RM (1989) Serotonergic/cholinergic muscle receptor cells in the crab stomatogastric nervous system. I. Identification and characterization of the gastropyloric receptor cells. J Neurophysiol 62:558–570. Katz PS, Harris-Warrick RM (1989) Serotonergic/cholinergic muscle receptor cells in the crab stomatogastric nervous system. II. Rapid nicotinic and prolonged modulatory effects on neurons in the stomatogastric ganglion. J Neurophysiol 62:571–581. Katz PS, Harris-Warrick RM (1991) Recruitment of crab gastric mill neurons into the pyloric motor pattern by mechanosensory afferent stimulation. J Neurophysiol 65:1442–1451. Kjaerulff O, Kiehn O (1996) Distribution of networks generating and coordinating locomotor activity in the neonatal rat spinal cord in vitro: a lesion study. J Neurosci 16:5777–5794. Knüsel J, Bicanski A, Ryczko D, Cabelguen J, Ijspeert A (2013) A salamander’s flexible spinal network for locomotion, modeled at two levels of abstraction. Integr Comp Biol 53:269–282.

Coordination of Rhythmic Movements

Kolta A, Brocard F, Verdier D, Lund JP (2007) A review of burst generation by trigeminal main sensory neurons. Arch Oral Biol 52:325–328. Kristan Jr WB, Calabrese RL (1976) Rhythmic swimming activity in neurones of the isolated nerve cord of the leech. J Exp Biol 65:643–668. Kristan Jr WB, Calabrese RL, Friesen WO (2005) Neuronal control of leech behavior. Prog Neurobiol 76:279–327. Kristan Jr WB, Stent GS, Ort CA (1974) Neuronal basis of swimming in the medicinal leech. I. Dynamics of the swimming rhythm. J Comp Physiol A 94:97–119. Krogh A, Lindhard J (1913) The regulation of respiration and circulation during the initial stages of muscular work. J Physiol (Lond) 47:112–136. Kwiatkowski MA, Gabranski ER, Huber KE, Chapline MC, et al. (2013) Coordination of distinct but interacting rhythmic motor programs by a modulatory projection neuron using different co–transmitters in different ganglia. J Exp Biol 216:1827–1836. Lang IM (2009) Brain stem control of the phases of swallowing. Dysphagia 24:333–348. Lanuza GM, Gosgnach S, Pierani A, Jessell TM, Goulding M (2004) Genetic identification of spinal interneurons that coordinate left–right locomotor activity necessary for walking movements. Neuron 42:375–386. Larimer JL, Kennedy D (1966) Visceral afferent signals in the crayfish stomatogastric ganglion. J Exp Biol 44:345–354. Le Gal JP, Juvin L, Cardoit L, Thoby-Brisson M, Morin D (2014) Remote control of respiratory neural network by spinal locomotor generators. PLoS One 9:e89670. Ludwar BC, Göritz ML, Schmidt J (2005) Intersegmental coordination of walking movements in stick insects. J Neurophysiol 93:1255–1265. Lugliani R, Whipp BJ, Seard C, Wasserman K (1971) Effect of bilateral carotid-body resection on ventilatory control at rest and during exercise in man. New Engl J Med 285:1105–1111. MacMillan DL, Deller SRT (1989) Sensory systems in the swimmerets of the crayfish Cherax destructor and their effectiveness in entraining the swimmeret rhythm. J Exp Biol 144:279–301. Mancilla JG, Lewis TJ, Pinto DJ, Rinzel J, Connors BW (2007) Synchronization of electrically coupled pairs of inhibitory interneurons in neocortex. J Neurosci 27:2058–2073. Marder E, Bucher D (2001) Central pattern generators and the control of rhythmic movements. Curr Biol 11:R986–R996. Marder E, Bucher D (2007) Understanding circuit dynamics using the stomatogastric nervous system of lobsters and crabs. Annu Rev Physiol 69:291–316. Marder E, Bucher D, Schluz DJ, Taylor AL (2005) Invertebrate central pattern generation moves along. Curr Biol 15:R685–R699. Marder E, Calabrese RL (1996) Principles of rhythmic motor pattern generation. Physiol Rev 76:687–717. Matsushima T, Grillner S (1992) Neural mechanisms of intersegmental coordination in lamprey: local excitability changes modify the phase coupling along the spinal cord. J Neurophysiol 67:373–388. Meltzer SJ (1899) On the causes of the orderly progress of the peristaltic movements in the oesophagus. Am J Physiol 2:266–272. Menzel R (1983) Neurobiology of learning and memory: the honeybee as a model system. Naturwissenschaften 70:504–511.

337

338

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Meyrand P, Simmers J, Moulins M (1991) Construction of a pattern-generating circuit with neurons of different networks. Nature 351:60–63. Morin D, Viala D (2002) Coordinations of locomotor and respiratory rhythms in vitro are critically dependent on hindlimb sensory inputs. J Neurosci 22:4756–4765. Morquette B, Morquette P, Agostinone J, Feinstein E, et al. (2015) REDD2-mediated inhibition of mTOR promotes dendrite retraction induced by axonal injury. Cell Death Differ 22:612–625. Morris LG, Hooper SL (1998) Muscle response to changing neuronal input in the lobster (Panulirus interruptus) stomatogastric system: slow muscle properties can transform rhythmic input into tonic output. J Neurosci 18:3433–3442. Morris LG, Thuma JB, Hooper SL (2000) Muscles express motor patterns of non-innervating neural networks by filtering broad-band input. Nat Neurosci 3:245–250. Mullins OJ, Hackett JT, Buchanan JT, Friesen WO (2011) Neural control of swimming behavior: comparison of vertebrate and invertebrate model systems. Prog Neurobiol 93:244–269. Mulloney B (1977) Organization of the stomatogastric ganglion of the spiny lobster. V. Coordination of the gastric and pyloric systems. J Comp Physiol A 122:227–240. Mulloney B, Smarandache-Wellmann C (2012) Neurobiology of the crustacean swimmeret system. Prog Neurobiol 96:242–267. Mulloney B, Smarandache-Wellmann C, Weller C, Hall WM, DiCaprio RA (2014) Proprioceptive feedback modulates coordinating information in a system of segmentally distributed microcircuits. J Neurophysiol 112:2799–2809. Mulloney B, Smarandache C (2010) Fifty years of CPGs: two neuroethological papers that shaped the course of neuroscience. Front Behav Neurosci 4:45. Nadim F, Manor Y, Nusbaum MP, Marder E (1998) Frequency regulation of a slow rhythm by a fast periodic input. J Neurosci 18:5053–5067. Norris BJ, Coleman MJ, Nusbaum MP (1996) Pyloric motor pattern modification by a newly identified projection neuron in the crab stomatogastric nervous system. J Neurophysiol 75:97–108. Nusbaum MP, Beenhakker MP (2002) A small-systems approach to motor pattern generation. Nature 417:343–350. Nusbaum MP, Marder E (1989) A modulatory proctolin-containing neuron (MPN). I. Identification and characterization. J Neurosci 9:1591–1599. Nusbaum MP, Weimann JM, Golowasch J, Marder E (1992) Presynaptic control of modulatory fibers by their neural network targets. J Neurosci 12:2706–2714. Paydarfar D, Gilbert RJ, Poppel CS, Nassab PF (1995) Respiratory phase resetting and airflow changes induced by swallowing in humans. J Physiol 483(Pt 1):273–288. Pearce RA, Friesen WO (1984) Intersegmental coordination of leech swimming: comparison of in situ and isolated nerve cord activity with body wall movement. Brain Res 299:363–366. Potts JT, Rybak IA, Paton JF (2005) Respiratory rhythm entrainment by somatic afferent stimulation. J Neurosci 25:1965–1978. Pratt CA, Jordan LM (1987) Ia inhibitory interneurons and Renshaw cells as contributors to the spinal mechanisms of fictive locomotion. J Neurophysiol 57:56–71. Puhl JG, Masino MA, Mesce KA (2012) Necessary, sufficient and permissive: a single locomotor command neuron important for intersegmental coordination. J Neurosci 32:17646–17657.

Coordination of Rhythmic Movements

Puhl JG, Mesce KA (2008) Dopamine activates the motor pattern for crawling in the medicinal leech. J Neurosci 28:4192–4200. Puhl JG, Mesce KA (2010) Keeping it together: mechanisms of intersegmental coordination for a flexible locomotor behavior. J Neurosci 30:2372–2383. Reilly SM, McElroy EJ, White TD (2009) Abdominal muscle function in ventilation and locomotion in new world opossums and basal eutherians: breathing and running with and without epipubic bones. J Morphol 270:1014–1028. Reilly SM, White TD (2009) Breathing with your belly: abdominal exhalation, loco-ventilatory integration and size constraints on locomotion in small mammals. Zoology 112:161–168. Renshaw B (1946) Central effects of centripetal impulses in axons of spinal ventral roots. J Neurophysiol 9:191–204. Romaniuk JR, Kasicki S, Kazennikov OV, Selionov VA (1994) Respiratory responses to stimulation of spinal or medullary locomotor structures in decerebrate cats. Acta Neurobiol Exp 54:11–17. Ryczko D, Knüsel J, Crespi A, Lamarque S, et al. (2015) Flexibility of the axial central pattern generator network for locomotion in the salamander. J Neurophysiol 113:1921–1940. Saito Y, Ezure K, Tanaka I, Osawa M (2003) Activity of neurons in ventrolateral respiratory groups during swallowing in decerebrate rats. Brain Dev 25:338–345. Shiba K, Satoh I, Kobayashi N, Hayashi F (1999) Multifunctional laryngeal motorneurons: an intracellular study in the cat. J Neurosci 19:2717–2727. Sigvardt KA, Mulloney B (1982a) Properties of synapses made by IVN command–interneurones in the stomatogastric ganglion of the spiny lobster Panulirus interruptus. J Exp Biol 97:153–168. Sigvardt KA, Mulloney B (1982b) Sensory alteration of motor patterns in the stomatogastric nervous system of the spiny lobster Panulirus interruptus. J Exp Biol 97:137–152. Simmers J, Moulins M (1988a) A disynaptic sensorimotor pathway in the lobster stomatogastric system. J Neurophysiol 59:740–756. Simmers J, Moulins M (1988b) Nonlinear interneuronal properties underlie integrative flexibility in a lobster disynaptic sensorimotor pathway. J Neurophysiol 59:757–777. Skinner FK, Mulloney B (1998) Intersegmental coordination in invertebrates and vertebrates. Curr Opin Neurobiol 8:725–732. Smarandache CR, Daur N, Hedrich UB, Stein W (2008) Regulation of motor pattern frequency by reversals in proprioceptive feedback. Eur J Neurosci 28:460–474. Smarandache CR, Stein W (2007) Sensory-induced modification of two motor patterns in the crab, Cancer pagurus. J Exp Biol 210:2910–2922. Smarandache-Wellmann C, Grätsch S (2014) Mechanisms of coordination in distributed neural circuits: encoding coordinating information. J Neurosci 34:5627–5639. Smarandache-Wellmann C, Hall WM, Mulloney B (2009) Coordination of rhythmic motor activity by gradients of synaptic strength in a neural circuit that couples modular neural oscillators. J Neurosci 29:9351–9360. Smarandache-Wellmann C, Weller C, Mulloney B (2014) Mechanisms of coordination in distributed neural circuits: decoding and integration of coordinating information. J Neurosci 34:793–803. Smarandache-Wellmann C, Weller C, Wright Jr, TM, Mulloney B (2013) Five types of nonspiking interneurons in local pattern-generating circuits of the crayfish swimmeret system. J Neurophysiol 110:344–357.

339

340

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Smith JC, Ellenberger HH, Ballanyi K, Richter DW, Feldman JL (1991) Pre-Bötzinger complex: a brainstem region that may generate respiratory rhythm in mammals. Science 254:726–729. Stern–Tomlinson W, Nusbaum MP, Perez LE, Kristan Jr WB (1986) A kinematic study of crawling behavior in the leech, Hirudo medicinalis. J Comp Physiol A 158:593–603. Sumi T (1975) Coordination of neural organization of respiration and deglutition: its change with postnatal maturation. In Development of Upper Respiratory Anatomy and Function. Bosma JF, Showacre J (eds). Bethesda, MD: Department of Health, Education, and Welfare. Talpalar AE, Bouvier J, Borgius L, Fortin G, et al. (2013) Dual-mode operation of neural networks involved in left–right alternation. Nature 500:85–88. Thuma JB, Hooper SL (2002) Quantification of gastric mill network effects on a movement related parameter of pyloric network output in the lobster. J Neurophysiol 87:2372–2384. Thuma JB, Hooper SL (2003) Quantification of cardiac sac network effects on a movement–related parameter of pyloric network output in the lobster. J Neurophysiol 89:745–753. Thuma JB, Morris LG, Weaver AL, Hooper SL (2003) Lobster (Panulirus interruptus) pyloric muscles express the motor patterns of three neural networks, only one of which innervates the muscles. J Neurosci 23:8911–8920. Toth TI, Daun-Gruhn S (2011) A putative neural network controlling the activity of the leg motorneurons of the stick insect. Neuroreport 22:943–946. Toth TI, Grabowska M, Rosjat N, Hellekes K, et al. (2015) Investigating inter–segmental connections between thoracic ganglia in the stick insect by means of experimental and simulated phase response curves. Biol Cybern 109:349–362. Wallén P (1997) Spinal networks and sensory feedback in the control of undulatory swimming in lamprey. In Neurons, Networks, and Motor Behavior. Stein PSG, Grillner S, Selverston AI, Stuart DG (eds). Cambridge, MA: MIT Press. Wallén P, Williams TL (1984) Fictive locomotion in the lamprey spinal cord in vitro compared with swimming in the intact and spinal animal. J Physiol (Lond) 347:225–239. Wang X-J, Rinzel J (1992) Alternating and synchronous rhythms in reciprocally inhibitory model neurons. Neural Comput 4:87–97. Weeks JC (1981) Neuronal basis of leech swimming—separation of swim initiation, pattern generation, and intersegmental coordination by selective lesions. J Neurophysiol 45:698–723. Weimann JM, Marder E (1994) Switching neurons are integral members of multiple oscillatory networks. Curr Biol 4:896–902. Weimann JM, Meyrand P, Marder E (1991) Neurons that form multiple pattern generators: identification and multiple activity patterns of gastric/pyloric neurons in the crab stomatogastric system. J Neurophysiol 65:111–122. Winifree AT (2001) The Geometry of Biological Time. Heidelberg:Springer Verlag. Wosnitza A, Bockemuhl T, Dubbert M, Scholz H, Buschges A (2013) Inter-leg coordination in the control of walking speed in Drosophila. J Exp Biol 216:480–491. Yu XT, Nguyen B, Friesen WO (1999) Sensory feedback can coordinate the swimming activity of the leech. J Neurosci 19:4634–4643. Zhang C, Guy RD, Mulloney B, Zhang Q, Lewis TJ (2014) Neural mechanism of optimal limb coordination in crustacean swimming. Proc Natl Acad Sci USA 111:13840–13845.

341

11 Prehensile Movements Till Bockemühl Biozentrum Köln, Institut für Zoologie, Universität zu Köln, Köln, Germany

11.1 Introduction: Prehension as Goal-Directed Behavior In the context of behavior and motor control the term prehension describes the process of a limb or extremity reaching out to a certain target, often a movable or manipulable object, and grasping or touching that target with a specialized gripping structure situated at the end of the limb. Extremities like the elephant’s trunk, an octopus tentacle, or the raptorial forelegs of praying mantises are examples. The most prominent instance of a prehensile limb is the primate arm with its highly evolved pentadactyle hand. As a true general purpose extremity it contributes to or is capable of such diverse behaviors as locomotion, feeding, grooming, digging, climbing, communication, and tool use. The primate arm is homologous to the forelimb of vertebrates in general and mammals in particular. In most vertebrates and many mammals the forelimb is mainly or exclusively used for locomotion. In primates, alternatively, the hand has evolved in such a way that prehension has become a dominant function of the forelimb. This phylogenetic development has progressed farthest in humans, in which the upper extremity is used for locomotion only during a relatively short crawling period during infancy. After the development of stable bipedalism, although swinging of the arms plays some role in optimizing the energy cost of bipedal locomotion (Collins et al. 2009), the human upper limb can be regarded as an almost exclusively manipulatory extremity. Primates, and especially humans, can perform a very large range of arm movements. The same manipulatory motor system that skillfully folds an origami sculpture or assembles the tiny parts of a mechanical watch can, in the very next moment, wield a heavy sledge hammer or deal a knock-out blow to an opponent in a boxing match. Humans have a large and differentiated repertoire of grasping movements that are highly specialized for various situations and objects (Cutkosky 1989; Napier 1956). The opposable thumb gives the hand an especially high level of versatility. The importance of the hand in the primate motor system is also illustrated by measuring the area of primary motor cortex (M1) devoted to the control of various body parts. Based on this measure, Penfield and Boldrey (1937) developed a depiction of the human body as it is represented in the motor areas of the brain, the so-called motor homunculus (an analogous depiction, the sensory homunculus, exists for the somatosensory cortex). In this depiction, the size of a body part corresponds to the area of cortical tissue that, Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

342

Neurobiology of Motor Control: Fundamental Concepts and New Directions

when electrically stimulated, evokes movement in the body part. Penfield and Boldrey discovered that the hands are vastly overrepresented in M1. In most descriptions of the motor homunculus the hands cover more than half of the total motor cortical area. The hand has also become progressively more dexterous during primate phylogeny. In phylogenetically older primate species direct monosynaptic projections from cortex to upper limb and hand motor neurons are much weaker than in, for instance, macaques or humans (Lemon and Griffiths 2005; Nakajima et al. 2000). In the older species, hand function is mainly conveyed via oligosynaptic pathways involving propriospinal neurons. In contrast, in humans these functions are almost exclusively carried by direct monosynaptic connections between motor cortical neurons and alpha motor neurons via the lateral corticospinal tract. Control of hand motor function in higher primates is thus much more direct than in other primates. In combination with the greatly enlarged cortical area responsible for the hand, this forms the basis for a highly dexterous prehensile extremity and hand. Prehension is a type of goal-directed behavior. Prehension is goal-directed on both a high and a low level. The high-level aspect refers to the question what we want to achieve, what our aim is, when we, for instance, grasp a tool. The high-level goal of picking up a hammer might be to drive a nail into a wall; analogously, the goal of reaching for a glass of water might be to quench one’s thirst. It is these high-level intentions that we typically refer to as “goals” in everyday language. The low level of goal-directedness in prehension refers to the problem of how exactly we have to reach and grasp; essentially, it refers to the motor control of prehension and the low-level details of movement. This problem includes how the muscles controlling the arm and hand are activated, the speed with which the hand moves, the temporal development of posture during reaching, the grip configuration of the hand, and the amount of pressure each finger exerts. All these low-level aspects can also be seen as goals. This chapter mainly focuses on these low-level aspects of prehensile movements, i.e., the various levels of motor control. Of course, both levels are intricately intertwined. Reaching and Grasping Two aspects of prehension are generally distinguished from each

other: reaching and grasping. Reaching refers to positioning the end-effector of the extremity at a certain target location within the working range of the extremity. This location is typically coincident with the current or, in the case of interceptive action such as catching, future position of an object. Pointing movements, i.e., movements directed at targets in extra-personal space (beyond arm’s length), are also reaching movements, though they lack a specific target object in peri-personal space (within arm’s length). In that sense, reaching is mainly a function of the arm consisting of shoulder girdle, upper arm, and forearm. Grasping refers to the subsequent manipulation of the targeted object or surface. This includes the concrete process of taking hold of an object, but also pushing it, operating a tool, or sampling surface structures by repeatedly touching the object. Grasping is therefore a function of the hand and especially the fingers and the thumb. Reaching and grasping are closely related and often happen in parallel or direct succession (Yang and Feldman 2010). Reaching movements are typically characterized as consisting of three distinct discrete states. The first is an initial static state at which the extremity is at rest. The second is the transitory dynamic state during which the actual reaching movement takes place and the hand is transported to its target location. The third is the static goal state achieved

Prehensile Movements

when the movement ends. This segmentation distinguishes reaching movements from continuous or cyclic movements such as walking, swimming, or flying, which are characterized by their repetitiveness and lack of clearly defined initial or end states (see also Chapter 8). There are also intermediate behaviors that combine discrete and continuous aspects of movement. Examples are scratching or grooming behaviors in which an extremity is first brought to a certain body region with a discrete reaching movement and is then moved cyclically across the region. Here I focus mainly on discrete movements. Because of its clearly defined initial and end states, reaching is ideally suited for investigating the redundancy problem in motor control.

11.2 The Redundancy Problem in Motor Control Most animal species have a remarkable behavioral repertoire. They can modulate these behavioral patterns in flexible, fast, and context-dependent manners and can also combine them to create more elaborate behaviors or sequences of behavior. The biomechanical complexity of the motor apparatus is an important foundation for this diversity. Complex movements are only possible if the motor apparatus has many degrees of freedom (DOFs) and if the DOFs can be employed with little restrictions and in many combinations. At first glance, elaborate bodies with many limbs would seem evolutionarily advantageous. Consider a highly complex locomotion system with many multi-segment legs. Such an organism can adjust its movement patterns to many different substrates. It can choose from multiple gaits, each of which might be ideally suited for a specific context. Individual legs can be temporarily repurposed for other uses like grooming without compromising locomotion. Loss of an extremity might be compensated for by changes in the movement patterns of other legs. In sum, the more complex the body, presumably the greater the number of possible behavioral capabilities, and thus the greater ability to choose optimal behaviors. As such, complex organisms would seem to be better able to adapt to the environment than ones with simpler bodies. Although these considerations are presumably correct, complexity necessarily also entails certain costs. One cost of an intricate musculoskeletal system is increased control complexity and hence computational demands on the nervous system. Increases in biomechanical complexity similarly require that the neural processes providing control become more powerful as well. A fundamental issue that inevitably arises from increased biomechanical complexity is the redundancy problem in motor control. Nikolai Bernstein (1967) first identified it as an important issue and it is therefore also called Bernstein’s problem. Since then it has become a central focus of motor control research and has spawned a vast body of literature. In the context of behavioral control the term redundancy refers to the fact that a complex motor system typically has, in principle, several, many, or even an infinite number of ways to perform a given task. Selecting among these many possibilities and coordinating all the DOFs involved is non-trivial and poses a fundamental challenge for the neural system. The redundancy problem in motor control is closely related to the problem of action selection (reviewed in Prescott et al. 2007)). The redundancy problem’s conceptual focus lies primarily on how to perform a certain act (the level of actual motor control). The action selection problem, alternatively, relates more strongly to the question of what to

343

344

Neurobiology of Motor Control: Fundamental Concepts and New Directions

x1 F

x2 x3

y1 y2

Figure 11.1 General and simplified input/output relationship for one stage within the neuro-motor control chain. Inputs x1 , x2 , and x3 represent the state of one level in the chain. This state is transformed by the function F into the state of the next level indicated by y1 and y2 .

do next (the level of cognition and decision making). However, this distinction may be only semantic: both problems deal with the question of how an organism chooses among several or even infinitely many possible actions. In that sense, the redundancy problem can be regarded as an instance of the action selection problem with a more specific focus on motor control. Redundancy emerges on several levels in the motor control chain linking neural activity and behavior (Saltzman 1979). Consider the human motor system consisting of the skeletal muscle system controlled by the central nervous system (CNS). Here, approximately 5 million motor axons descending from higher brain centers project onto roughly 150,000 alpha motor neurons in the spinal cord. These, in turn, innervate 600 to 700 skeletal muscles that actuate 100 to 150 mechanical degrees of freedom, i.e., joints (Neilson 1993). In addition, we might consider a further level in this chain and define a small number of end-effectors capable of targeted movements (hands, feet, eyes, head) whose positions, orientations, velocities, or the forces they exert are the final control variables. A central aspect of this chain is that each stage has more input variables than output variables. The decrease in variable number at each stage can be formalized as shown as in Fig. 11.1. The variables x1 , x2 , and x3 represent the state of a certain level in the neuro-motor control chain. The variables y1 and y2 represent the state of the subsequent level in the control chain. The box denoted F describes all processes that transform x1 , x2 , and x3 into y1 and y2 . As a first approximation, this block diagram can be represented as a system of equations (Eqns. 11.1 and 11.2). For simplicity, linear equations are used here. However, most transformations in the neuro-motor chain are actually non-linear, a fact that complicates matters further. y1 = f11 x1 + f12 x2 + f13 x3

(11.1)

y2 = f21 x1 + f22 x2 + f23 x3

(11.2)

In matrix notation these equations can be formulated as y⃗ =

) ⎛x ⎞ ( f11 f12 f13 ⎜ 1 ⎟ − x = F→ x f21 f22 f23 ⎜ 2 ⎟ ⎝x3 ⎠

(11.3)

Given the state of y in such a system of equations, the state of x can be determined from −y → − (11.4) x = F −1→ However, this involves finding the inverse F −1 of matrix F, an operation defined only for square matrices. As shown in Eqn. 11.3, F is not square but rectangular. As such, no unique solution for Eqn. 11.4 exists.

Prehensile Movements

A

B

y dimension in arm plane

p

j3

s3

β s2

s1

j2

j1 α

x dimension in arm plane

y dimension in arm plane

γ γ′ j3

p′ s3

β′ s2 j1

s1 α′

j2

x dimension in arm plane

Figure 11.2 (A) Schematic of a planar three-DOF arm with three segments (s1 , s2 , s3 ) representing an upper arm, lower arm, and hand. The hand segment s3 also has a stylized thumb. Joints (j1 , j2 , j3 ) indicated by filled circles. The arm’s posture is uniquely defined by three joint angles (𝛼, 𝛽, 𝛾). The position of the endpoint of the arm, the fingertip, is the two-dimensional point p (open circle). (B) A different state of the system characterized by another set of joint angles (𝛼’, 𝛽’, 𝛾’) giving rise to a different endpoint position (p’).

Put another way, this is a redundant system of equations, one with more unknowns than equations. Determining the state of one motor control stage based on the state of a subsequent one is therefore an ill-posed problem for which it is impossible to arrive at a unique solution. Instead, one can find an infinite number of equally valid solutions, among which there is no basis to prefer one over the others. To ground this theoretical analysis in a specific problem in motor control, consider the reaching problem in a simplified motor system: a planar arm with three segments and three hinge joints (Fig. 11.2). As the name implies, movement of this arm is restricted to a plane; each joint axis is perpendicular to the plane of motion and has only one DOF. Following an anatomical naming convention, the basal joint (j1 ) can be regarded as a simplified shoulder joint, the second (j2 ) as an elbow joint, and the most distal (j3 ) as a wrist joint. In the same fashion, the most proximal segment (s1 ) can be thought of as an upper arm, the second (s2 ) as a lower arm, and the most distal (s3 ) as a hand. With the help of this simple system we can examine the general aspects of forward and inverse problems with regard to motor control. Forward Problems in Motor Control The term forward problem (also called direct problem)

refers to the task of predicting what the output of a system or a model is, given a certain system state or set of system parameters. For the planar arm in Fig. 11.2 the system’s state is the orientation of each segment, as described by the arm’s joint angles. The system output is the end-effector position. Since the segment lengths are constant the posture of the arm, i.e., its spatial configuration, is uniquely defined by the three joint angles. Each set of three joint angles is a valid posture and has an associated end-effector position p in the plane of motion. Within this plane, the end-effector position of this kinematic chain can be unambiguously described by a 2D representation, for instance Cartesian

345

346

Neurobiology of Motor Control: Fundamental Concepts and New Directions

x and y coordinates. Given a set of segment lengths and joint angles one can readily calculate the position of the end-effector. For each set of joint angles there is exactly one solution. This is equivalent to transforming or mapping a point in a three-dimensional space (here, joint angle space) to a point in a two-dimensional space (here, the workspace of the arm). In abstract form, this relationship is captured by Fig. 11.2. Such a problem is also called a well-posed problem, because one unique solution exists. Because of the relative simplicity of the example chosen here, it is important to note that just because a system is well-posed does not mean, particularly for biological systems, that it is easy to solve. Motor neuron activity, for example, results in postures and movements by activating muscles that apply force to joints that, typically, are parts of complicated, multi-segmented appendages. Muscles respond non-linearly to neuron input and have non-linear inherent properties (e.g., fatigue, passive muscle force generation), and torques generated at one joint can cause torques at others (interaction torques) (see also Chapter 12). For most movements in water and some in air, frictional forces generated by the medium must be considered, as must also, for large limbs moving in air (Hooper 2012; Hooper et al. 2009), gravity, the effect of which varies as a function of limb posture relative to the gravity field. Thus, even for well-posed problems (that is, ones in which the input variables are specified), calculating the output can be daunting. Inverse Problems in Motor Control The above discussion showed that each set of input vari-

ables gives rise to exactly one output. What about the reverse question: is each output associated with only one set of input variables? This question of what state a system must be in (what its set of parameters must be) to generate a desired output is called the inverse problem. For the planar arm, this question becomes what set(s) of joint angles result in the end-effector of the arm being positioned at any given point p? Figure 11.3A–C shows three sets of 𝛼, 𝛽, and 𝛾 joint angles that do so, and considering the figure makes it clear that infinitely many arm configurations would do so. The inverse problem described here is therefore ill-posed. Note that inverse problems do not necessarily have to be ill-posed; in particular, in systems with equal numbers of input and output parameters, the inverse problem is generally well-posed. However, as noted above, in motor systems this equality is typically not present. The redundancy problem thus raises the question of how motor systems select which of the many valid and basically equivalent solutions to use. Before turning to this issue, however, it is important to describe other levels in motor control at which redundancy exists.

11.3 Redundancy Occurs on Multiple Levels of the Motor System Redundancy exists at multiple levels in motor systems (Saltzman 1979). In the context of reaching, three levels are frequently considered: the trajectory level as the highest level, the postural or joint angle level as an intermediate level, and the muscle level as the lowest. It is often assumed that these problems have to be consecutively solved by the CNS before a movement can be executed. However, as we will see later, in some approaches higher level redundancies are addressed implicitly by solving the redundancy problem on lower levels.

Prehensile Movements

B γ′

γ

y dimension in arm plane

y dimension in arm plane

A

p

β

p β′

α′

α x dimension in arm plane

x dimension in arm plane

y dimension in arm plane

C

γ ′′ p β′′

α′′ x dimension in arm plane

Figure 11.3 Multiple sets of joint angles give the same end-effector position. (A–C) Three solutions to the problem. Each posture has different joint angles but still reaches p.

Trajectory, Postural, and Muscle Redundancy The first and highest level of redundancy is

choosing an end-effector trajectory: in order to execute a successful reaching movement, i.e., move the hand to a desired target, a curve in space connecting the current hand position to the target position, a trajectory, must be chosen. Furthermore, each point along this curve has an associated speed. Because only the first and the last point are fixed, in principle there are infinitely many ways of connecting two points in space. The trajectory can therefore assume any shape; another example of redundancy in motor control. Although it can be argued that some trajectories, e.g., a straight line, “should” be better, these implicitly assume a cost function the motor system has evolved to minimize (for a straight line choice, distance the hand travels). Present understanding of the constraints driving the evolution of motor control systems is still too limited to predict with

347

348

Neurobiology of Motor Control: Fundamental Concepts and New Directions

any likelihood of success what these cost functions are. Indeed, the intellectual process works in the opposite direction; discovering on what basis motor control systems choose the trajectories they do is most likely to illuminate what priorities drive motor system evolution. As such, at the moment it is unclear why motor systems choose one trajectory over another. Understanding on what basis nervous systems do so is a non-trivial problem, and the subject of extensive research (see below). The second, intermediate level of redundancy is at the joint angle level. For any specific hand trajectory the CNS must specify a series of postures, i.e., sets of joint angles, to use to move the hand along the chosen trajectory. This is called the inverse kinematics problem. Figure 11.3 shows that infinitely many postures can produce any given hand position. Selecting among these infinite possibilities is another example of redundancy. The third and lowest level of redundancy arises in specifying the forces and muscle contractions necessary in order to move the arm. In biomechanics these can be described via joint torques. For very slow movements, in which no or negligible inter-segmental interaction forces occur, the torques required to produce any desired series of postures can be calculated unambiguously (for more rapid movements, interaction torques make this calculation more difficult). Even in this case, however, redundancy arises because the CNS cannot produce joint torques directly, but must do so by activating muscles that in turn act on the joints. Because joints typically have both agonist and antagonist muscles and, in addition, many joints have more than one agonist muscle, there are again infinitely many muscle activation patterns that can produce a desired torque about a joint. Selecting among these patterns is again a problem of redundancy. Relevance for Motor Control As explained above, when described in the inverse direction (how many inputs can result in a given output), redundancy exists on multiple levels of the motor control system. However, in the forward direction, no such ambiguity exists: a given input state uniquely specifies a single output state. Why therefore is the inverse problem relevant to motor control? The relevance of the redundancy problem becomes apparent when we examine the direction in which information has to flow before a movement can be carried out. Desired behavior is typically formulated in high level goals that assume rather unspecific forms: “grasp this cup of tea”, “pick that berry”, “point at the moon”. When a person decides to reach for an object (the high level goal), the brain has to generate a neural activity pattern in motor cortex that will move the hand to the object. Information thus has to flow from the object’s position in space to a high level pattern of neural activity that, when fed through multiple, redundancy-containing, intermediate stages, results in the appropriate action on the causal level, i.e., implementation at the lowest stage (hand trajectory in Fig. 11.4). One way to envisage this is as a series of transformations that traverse the neuro-motor chain in the inverse direction. Information has to flow from the level of target position, here the location of an object, to the level of trajectory, then to the level of posture, further on to the level of muscle activation until finally arriving at the level of neural activity (dashed lines, Fig. 11.4). The entirety of this process is commonly called sensorimotor transformation because it requires the transformation of sensory information into motor activity (reviewed in Kalaska and Crammond 1992; Pouget and Snyder 2000). At each step of this information transfer, the CNS has to solve one of the ill-posed

Prehensile Movements

CNS

Muscle activation

Posture

Hand trajectory

Target

Figure 11.4 Flow of information in the neuro-motor control chain. Right arrows indicate forward problems. Left arrows indicate inverse and potentially ill-posed problems. Neural activity that moves a limb to a certain target ultimately depends on target position (solid left arrow). Each stage between target level and neural level is an inverse and ill-posed problem (dashed arrows).

problems described above in which, for example, a multitude of different hand trajectories, arm postures, and muscle activation patterns can accomplish the same task.

11.4 Overcoming the Redundancy Problem Motor systems effortlessly produce seamless, swift, and accurate movements, and clearly thus successfully solve the redundancy and other problems outlined above. Consider, for example, an expert table-tennis player attempting not only to counter an extremely fast topspin attack by the opponent, but also to attack in turn. Immediately after (or even before) the opponent has hit the ball, the player has to extrapolate where the ball will probably move, taking into account not only the ball’s direction, but also its speed and potential spin. The player then has to develop neural commands that not only move the racket to the right location at the right time, but also control racket velocity and orientation at that location to a high degree of accuracy. These commands have to travel to her or his arm muscles which, in turn, have to contract. The time for all this to occur is often less than 300 ms and the temporal accuracy required for a successful return is in the single-digit millisecond range (Bootsma and van Wieringen 1990). To achieve all this requires both that neural and biomechanical reaction times be very fast and spatial and temporal accuracy be very high. Typical reaction times of expert table-tennis players are around 100 to 120 ms (Bootsma and van Wieringen 1990), so the actual time available for the movement is barely 200 ms. Remarkably, for well-trained players this time is sufficient to execute the movement reliably and repeatedly. This example illustrates the impressive capabilities of the human sensorimotor system. The CNS clearly contains algorithms that efficiently overcome the theoretical limitations and constraints imposed by the redundancy problem. What, in principle, might these be? A formally valid, but trivial and very inefficient, strategy for solving a given redundancy problem would be to choose randomly among the infinity of available solutions. This strategy would entail highly erratic behavior, producing unpredictable and inefficient movements. Actual behavior in everyday life and in experiments with humans and animals, alternatively, is the opposite of random. It is repeatable and often highly stereotypical in many respects, such as duration, speed, kinematics, and accuracy. These observations suggest that movements are generated according to rules; purely random selection seems highly unlikely.

349

350

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B x1 x2 x3

F

y1

x1

y2

x2

y3

x3

y1 F

y2

Figure 11.5 Theoretical considerations for solving the redundancy problem. Making matrix F (see Fig. 11.1) invertible requires either increasing the number of outputs (A) or decreasing the number of inputs (B).

In principle two systematic ways of solving the motor redundancy problem exist. Both are tightly linked to Eqns. 11.3 and 11.4. As explained above, the core of any redundancy problem is the non-invertibility of matrix F (Fig. 11.1) that maps the state of one stage of the motor system to the next. Overcoming the redundancy problem can therefore be thought of as making matrix F invertible, i.e., making it square. The size of the matrix is determined by the number of input and output variables. Therefore, F can be made square either by increasing the number of outputs (Fig. 11.5A) or reducing the number of inputs (Fig. 11.5B). Equation 11.5 illustrates making F into a 3-by-3 square matrix by increasing the number of output variables. Equation 11.6 illustrates making F into a 2-by-2 square matrix by decreasing the number of input variables. In both cases the modified matrix is square, i.e., invertible, and the formerly ill-posed problem is transformed into a well-posed problem. ⎛y1 ⎞ ⎛f11 f12 f13 ⎞ ⎛x1 ⎞ − ⎜y2 ⎟ = ⎜f21 f22 f23 ⎟ ⎜x2 ⎟ = F → x ⎟⎜ ⎟ ⎜ ⎟ ⎜ y f f f x ⎝ 3 ⎠ ⎝ 31 32 33 ⎠ ⎝ 3 ⎠ ( ) ( ) ⎛x1 ⎞ y1 f f f − x = 11 12  13 ⎜x2 ⎟ = F → y2 f21 f22 f 23 ⎜ ⎟ x ⎝3 ⎠

(11.5)

(11.6)

What do these abstract considerations mean for motor control? Reducing the number of inputs is equivalent to reducing the number of available degrees of freedom or, more precisely, reducing the number of controlling variables. Increasing the number of outputs is equivalent to increasing the number of conditions that have to be met in order to solve the task, i.e., increasing the number of controlled variables. In other words, either the number of DOFs in the motor system has to be decreased to match that of the task, or the number of task requirements has to be increased so that task complexity matches that of the motor system. Which mechanism does the CNS use? Both mechanisms have an impressive array of supporting evidence. Current research thus suggests that both are realized in the nervous system. A common approach to identify which mechanism is used at which level in the motor system is to find invariant movement features. 11.4.1 Invariant Movement Features

Biological movements exhibit invariant features. These features are called invariant because they are present across a set of otherwise variable and changing movements. Ideally, they are found across many classes of movements and across individuals. The

Prehensile Movements

term is often applied rather broadly and can refer to any feature that is constant across movements. It is implicitly assumed that invariance of a movement feature indicates it is of general importance and might therefore represent a variable that the CNS actively controls. It is therefore often hypothesized that movements, or crucial aspects of them, are planned on the level on which these invariances occur. Invariant features found on the hand trajectory level, for instance, would suggest that hand kinematics are planned first and everything else, e.g., muscle activation or joint angles, is calculated subsequently in order to realize the desired hand kinematics. To illustrate the concept of invariance, consider the human eye. It has three rotational DOFs; it can rotate about a vertical axis (adduction, abduction), a horizontal axis (elevation, depression), and a torsional axis (excyclotorsion, incyclotorsion). The torsional angle is coincident with gaze direction and thus varies with gaze direction; however, for any gaze direction realized by the vertical and horizontal axes there are, in principle, infinitely many torsional angles. Therefore, the eye is a redundant system and the question arises how the torsional angle is determined. Two early studies showed that the eye’s torsional angle follows certain invariant rules. Donders (1848) found that when the eye is rotated to a certain gaze direction it always adopts the same torsional orientation, independent of the gaze direction before the movement. von Helmholtz (1867) extended this finding and qualitatively described how torsional angle depends on gaze direction. He attributed the first formulation of this law to Johann Benedict Listing, who never published his findings. These two invariant features are referred to as Donders’ law and Listing’s law and were the first rules formulated for motor control (reviewed in Wong 2004). Later work has identified a large number of motor invariants. One of the first studies systematically investigating regularities in the control of human arm movement was carried out by Fitts (1954). He found a close relation between duration, amplitude, and accuracy in simple human pointing movements. Reformulated as a trade-off between movement duration and accuracy, this relation became known as Fitts’ law. It predicts that it takes longer to point to targets that are smaller or farther away and, conversely, that faster movements are less accurate. Besides unraveling basic mechanisms in motor control, Fitts’ law plays an important role in, for example, usability considerations for software graphical user interfaces in which the user has to interact via a computer mouse. Later work found that in planar point-to-point reaching tasks humans tend to produce straight hand trajectories with symmetric, bell-shaped velocity profiles (Abend 1982; Flash and Hogan 1985; Morasso 1981). Soechting and Lacquaniti (1981) showed that hand trajectories were invariant with regard to movement speed. Lacquaniti et al. (1983) formulated a law relating kinematics and shape in drawing movements. This empirical law, known as the two-thirds power law, states that the instantaneous tangential velocity and curvature of the hand trajectory are closely related. Later work, however, argued that this law is an epiphenomenon of kinematically coupled joint oscillators (Sternad and Schaal 1999). Furthermore, it is violated in more complex three-dimensional movements and thus may be unique to planar or small-scale movement (Schaal and Sternad 2001). The importance of the two-thirds power law as an invariant feature, and thus its suggestion that movement planning occurs at the trajectory level, may thus be much less than previously thought.

351

352

Neurobiology of Motor Control: Fundamental Concepts and New Directions

With the notion of invariance explained, I now turn to mechanisms to solve the redundancy problem. I first discuss studies about increasing the number of task conditions and then focus on the idea of limiting the number of DOFs. 11.4.2 Increasing the Number of Task Conditions

In the example of the planar arm in Fig. 11.3a the output variables are the x and y coordinates of the arm’s end-effector. At first glance, it seems counterproductive to increase complexity (i.e., to add a further output variable) with the aim of making the redundancy problem manageable. Furthermore, it is not immediately clear what it means to increase the number of controlled variables. Since the arm can only move in the XY plane it follows that extending the output variables by adding a Z component, for instance, is not helpful. One approach is to consider other, more abstract variables that the CNS might control. Variables discussed in the literature are typically global in nature and would have to be calculated before the actual movement. They are commonly formulated as global cost functions (often also called loss functions). In order to produce a movement a particular cost function is minimized by an optimization procedure (Engelbrecht 2001), with movements with the lowest cost being selected for execution. Early approaches were also often influenced by theoretical considerations that mainly took physical variables into account (Nelson 1983). I now discuss several approaches found in the literature describing how the redundancy problem might be solved by introducing additional task requirements or constraints on the trajectory, postural, and muscular levels of control. Increasing the Number of Task Conditions on the Trajectory Level The minimum-jerk principle is

arguably the best-known minimum approach in motor control (Flash and Hogan 1985; Hogan 1984). It can be applied to resolve redundancy on the level of hand trajectories and is based on the kinematic quantity jerk. Jerk is defined as the third derivative of position or, equivalently, the first derivative of acceleration, i.e., the change in acceleration. When applied to motions of the hand, each hand trajectory has an associated series of jerk values over time. The minimum-jerk model assumes that, for a given movement duration, the CNS chooses hand trajectories that minimize the time integral of the square of jerk magnitude. When first proposed, the minimum-jerk model quite successfully captured several known invariant features of point-to-point reaching movements, such as straight hand paths and hand velocity profiles that are single-peaked, bell-shaped, and symmetric. Since then, it has been extended to account for perturbations, movement variability, and target shifts by blending and superimposing several minimum-jerk trajectories (Flash and Henis 1991; Henis and Flash 1995). A further modification showed the applicability of the minimum-jerk principle to cyclical movements during a figure tracing task (Viviani and Flash 1995). However, later work showed that features like straightness or symmetry of velocity profiles are not present in all movements, suggesting that the minimum-jerk principle may apply to only a subset of arm movements. For example, hand trajectories become curved near the boundaries of the workspace (Osu et al. 1997; Suzuki et al. 1997; Uno et al. 1989). Hand trajectories are thus not invariant with regard to workspace location

Prehensile Movements

but seem to depend on posture. Furthermore, in very slow or very fast movements hand velocity profiles are skewed and therefore the minimum jerk principle also does not hold for these classes of movements (Suzuki et al. 1997; Wiegner and Wierzbicka 1992). Another test of the minimum-jerk principle stems from its implication that motor control of the upper limb in primates is organized around the kinematics of the hand and especially its trajectories. Initial neurophysiological data supported the idea that hand kinematics are the central aspect around which motor planning for reaching revolves. Georgopoulos et al. (1982) found that during reaching in primates the activity of a large population of M1 neurons correlated with hand movement direction. Individual neurons in M1 were maximally active when the hand moved in a particular direction, gradually and continuously becoming less active as the hand moved in directions further and further from the preferred one, a behavior termed tuning (Fig. 11.6A). In a landmark paper, Georgopoulos et al. (1986) suggested that movement direction was represented by this differentially active population of tuned M1 neurons and introduced the concept of the population vector (Fig. 11.6B). In this view, each neuron in a population encodes a preferred movement direction represented by a vector pointing in that direction. A population mean vector is then calculated by multiplying each neuron’s preferred direction by the neuron’s relative activity and summing all the individual neuron vectors. Georgopoulos and coworkers showed that this population mean vector very closely corresponds to the hand movement direction observed during the neural recordings. This work did not, however, conclusively show that MI activity correlated exclusively with hand kinematics and was not also tuned to other movement related aspects, e.g., the activity of individual muscles or joint angle positions. Since then, theoretical and empirical arguments have been made that challenge the claim that M1 exclusively encodes hand kinematics. Theoretical work argues that the apparent correlation between hand movement direction and M1 activity might be an epiphenomenon due to the fact that many movement parameters (e.g., joint angle, hand position, muscle contraction and forces) are by necessity at least partially correlated (Mussa-Ivaldi 1988; Todorov 2000). These studies argue that one will therefore inevitably find correlations between a multitude of movement parameters and neural activity, without being easily able to determine which are causally relevant and which are mere correlations. The general applicability of the population vector with regard to hand movements was called further into question by Scott and colleagues in an elegant experiment that dissociated higher level task requirements from the motor commands and movements required to fulfill the task (Scott and Kalaska 1997; Scott et al. 1997). They trained monkeys to carry out reaching movements with a constant hand path but two different arm postures, a natural posture and one in which the shoulder was abducted. During these movements neuron activity in M1, the dorsal premotor cortex, and parietal area 5 was recorded. These areas are believed to contribute to various stages of planning during goal-directed reaching movements. The tuning characteristics of most recorded neurons changed markedly with the change in posture, something not be expected if the neurons represented hand kinematics. A follow-up study more closely investigating the distribution of preferred

353

354

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

−500

−500

0

500

1000

0

500

90°

1000

−500

0

500

1000

0° −500

−500

0

500

0

−500

1000

500

−500

1000

−500

0

500 1000 (ms)

B

90° 0°

0

0

500

500

1000

1000

Prehensile Movements

directions in motor cortex neurons and the corresponding directions and lengths of the population vector found that not all directions were encoded equally across the recorded cells (Scott et al. 2001). Instead, the distribution of preferred directions was clearly bimodal; one peak was cells that preferred directions away from and to the left of the body, the other one directions towards and to the right of the body. To a certain extent this non-uniform distribution could be explained by peak joint velocity, which moderately correlated with cell number. However, a better predictor was peak joint power, a measure that combines joint velocity and torque. These later findings thus support the idea that the role M1 plays in the generation of movements cannot readily explained based on hand kinematics. Instead, joint kinematics and dynamics seem to play a crucial role that must be taken into account. Increasing the Number of Task Conditions on the Postural Level With regard to posture at the

end of the movement, participants tend to adopt postures they subjectively identify as most comfortable (Cruse et al. 1990). Based on this finding, a maximum comfort hypothesis can be formulated in which each joint has an associated comfort function. Perceived comfort depends on both kinematic and dynamic factors: it is maximal near the center of the joint’s working range and minimal at the limits of the joint’s working range. Comfort is also influenced by the force needed to maintain a certain joint angle; the higher the force the lower the comfort. Based on these two measures, postures would be chosen so as to maximize comfort (or minimize discomfort, using the terminology of optimization) while positioning the end-effector at the desired location. In further studies along these lines, Rosenbaum and Jorgensen (1992) coined the term end-state comfort. This term refers to the observation that subjects choose movement or movement sequences with transiently awkward postures during the movement execution but maximally comfortable postures at movement end over ones without awkward intermediate postures but with uncomfortable final postures. These data indicate that the CNS plans ahead and takes future states, actions, and goals into account when issuing new motor commands. Increasing the Number of Task Conditions on the Muscle Level Harris and Wolpert (1998) made

an important contribution to the field with the formulation of the minimum-variance theory. This theory, which is explicitly formulated for eye and hand movements, assumes that all neural motor control signals are subject to noise degradation causing the end-effector to deviate from its ideal trajectory. Furthermore, the noise level is assumed to depend on the amplitude of the control signal, i.e., the force with which the controlled muscles contract. Based on these assumptions, Harris and Wolpert propose that in the presence of such signal-dependent noise the CNS selects a control signal that minimizes the variance of the final hand or eye position (Fig. 11.7). This hypothesis is an interesting hybrid approach that elegantly unifies kinematics and kinetics of Figure 11.6 Population vector coding of movement direction. (A) Individual M1 neurons are broadly tuned to fire in certain directions relative to the body center (for the neuron shown, to about 160∘ ). In each panel, “0” is the time of movement beginning. (B) Summing the activity-weighted direction vectors across the neuron population gives mean vectors (filled arrows) that match observed movement direction (open arrows). A modified from Georgopoulos et al. (1982); B from Georgopoulos et al. (1983), both with permission.

355

356

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Position distribution

Movement A

Movement B

Figure 11.7 Expected final position distribution (ellipsoid) is larger for more complicated (less smooth) trajectories (Movement B) than for less complicated (smoother) trajectories (Movement A). This difference is predicted to occur if the motor command signal contains noise that scales with motor command magnitude. Non-smooth movements require larger motor commands, and thus have greater total error. Modified with permission from Wolpert and Ghahramani (2000).

movement control in one approach. Movement kinematics are taken into account by formulating the objective function of the optimization in terms of endpoint error, a purely kinematic measure; however, calculating this error and minimizing it explicitly requires considering neural motor commands, muscle parameters, and a dynamic environment (also see Wolpert and Ghahramani 2000). The minimum-variance model predicts several key features of eye and hand movements, such as Fitts’ Law (Fitts 1954) and the two-thirds power law (Lacquaniti et al. 1983). For movements with small amplitudes it produces bell-shaped velocity profiles and straight trajectory paths (Abend 1982; Morasso 1981), for large amplitude movements velocity profiles are skewed as found in saccadic eye movements (Collewijn et al. 1988). Conclusion for Additional Task Conditions We can see a clear development in the field from

early and rather descriptive approaches to later considerations that explicitly take into account behavioral and task-relevant, i.e., ecological, aspects of the task (that consider accuracy, for instance). Many of the approaches revolve around the minimization of one or several parameters that serve as additional constraints and which thus increase the number of actively controlled variables. An important aspect of many of these models is the global nature of the additional task requirement. This means that the complete movement has to be taken into account when the minimization process takes place. As a consequence, the complete movement has to be planned in advance. This is a very strong assumption which cannot easily be brought into agreement with two empirical facts:

Prehensile Movements

(a) Goal-directed movements, such as reaching movements, can be corrected swiftly and smoothly on-the-fly. This can be demonstrated by shifting the original target during an ongoing movement. It is difficult to see how such a goal shift can be integrated into a controller that minimizes global movement features. One possibility would be that a target shift triggers an immediate re-calculation of the hand trajectory. However, this hypothesis raises the issue of what exactly constitutes a target shift. Do small fluctuations in perceived target location due to sensory noise qualify? (b) The models described so far are based on very simple experimental paradigms. For reaching movements the situation changes considerably when, for instance, obstacles are introduced that have to be avoided during reaching. In order to reconcile results from such obstacle studies with classic minimization approaches it is usually necessary to introduce additional via-points, virtual points through which the hand trajectory has to pass. Although highly trained and stereotypical motor behavior like writing can be described quite well by such an approach (Meulenbroek et al. 1996), via-points have the disadvantage that they introduce the need for an explicit calculation of the position of these points. Most of these approaches are also either incompatible with each other or their defining features are merely emergent properties of more general principles. Furthermore, minimum principles tend to be descriptive rather than explanatory. Their ecological relevance is mostly either neglected or justified in retrospect. It is, for instance, unclear what the ultimate reason for minimum-jerk trajectories is, other than the fact that the model predicts certain aspects of the data. An exception might be the minimum-variance theory proposed by Harris and Wolpert (1998). Although its key variable, motor noise, is found on a very low level of the neuro-motor chain, it is at the same time explicitly linked to accuracy, an ecologically relevant and high level aspect of movement. 11.4.3 Reducing the Number of DOFs

The second way of solving the redundancy problem is to reduce the number of independent DOFs (Fig. 11.5B). For example, a trivial manner to do this at the muscle/joint to endpoint matrix would be to reduce the number of muscles activated, or hold fixed the position of a sufficient number of joints, so that the dimensionality of the input (the x’s representing the muscle and joints in Fig. 11.5B) equaled that of the output (the y’s defining endpoint position in Fig. 11.5B). This however seems rather arbitrary and also contradicts the general observation that in most movements each DOF capable of contributing in principle typically does so to a certain extent in fact. How else could the number of DOFs be reduced? An important idea in this context can also be traced back to Bernstein (1967). Not only did he identify redundancy as a problem in motor control, he also proposed a potential solution: motor synergies. The term is often used interchangeably with similar terms such as muscle synergies, motor primitives, and movement primitives. (See also Chapter 12.) Although these terms differ in their exact meaning, each is centered around the concept that the CNS solves the redundancy problem by combining or linking the control of DOFs on various levels (reviewed in Flash and Hochner 2005; Lee 1984; Ting and McKay 2007; Tresch and Jarc 2009).

357

358

Neurobiology of Motor Control: Fundamental Concepts and New Directions

If each DOF is independently controlled, then each requires a separate control variable. From a control-theoretical point of view, this would mean that controlling an arm like the one in Fig. 11.2 at the joint angle level would require three control parameters. However, DOFs need not be independent. Two or more DOFs might share the same control parameter, in which case their behavior would be linked or coupled. This linkage would reduce the number of control variables and hence the number of DOFs that need to be controlled. The linkage would thus shift control of the system from a redundant space, i.e., from joint angle space, to a non-redundant space, motor synergy space. Within that space, the redundancy problem never arises, because there are no redundant DOFs and the number of controllable parameters corresponds to the task. In this view, the functional groups that constitute motor synergies are the basic units. More complex movements are composed by combining and superposing these basic building blocks. As above, the concept of a modular organization can be applied at several levels of movement generation: trajectory, joint angles, and muscle activation. Reducing the Number of DOFs on the Trajectory Level In an early study, Morasso and

Mussa-Ivaldi (1982) considered various classes of hand movements ranging from simple point-to-point movements to handwriting. Based on the invariant features of hand trajectories that were then known, they proposed a model that used parametric curves, e.g., splines or Beziér curves, as fundamental building blocks for the generation of trajectories. In this concept, one of these elemental curves is called a stroke. Because of the properties of the used parametric curves, smooth transitions from one stroke to another can be easily realized. Simple point-to-point movements consist of single strokes. More complex movements, such as handwriting, consist of several sequentially executed strokes. Sanger (2000) studied planar hand movements in humans. The participants in his study repeatedly copied simple template trajectories presented to them on a computer screen. The template trajectory for a given trial was generated from trajectories produced by the same participant during the two preceding trials. Sanger found that movements produced during this iterated practice progressively became smoother and more accurate. Using principal components analysis (PCA) he further showed that the movements mapped to a small number of principal components. He interpreted these data as arising from the combined effects of a modular movement controller and the properties of the musculoskeletal system. In a similar study, Polyakov et al. (2009) investigated hand movements in monkeys during a scribbling task. The monkeys were trained to move a cursor on a computer screen and were rewarded when they hit randomly appearing targets with the cursor. Similar to the findings of Sanger (2000), hand trajectories generated in this manner became more stereotypical with practice, and well-practised movements could be represented as sequences of a small set of simple parabolic trajectories. Polyakov et al. proposed that these parabolic trajectories are the basic motor primitives of hand trajectories. In analogy to written language, in which complex expressions consist of simpler and more basic elements, they termed longer movements “words” and the basic parabolic trajectories that these words consist of “letters”. Reducing the Number of DOFs on the Postural Level Thomas et al. (2005) studied whole body

reaching movements in humans based on six DOFs. These movements involved not

Prehensile Movements

only the arm but also necessitated movement in the trunk and legs. Using PCA to extract motor synergies, they found that about 90% of the variance in joint angle kinematics during these movements could be explained by a single component. They interpreted these data as strong evidence not only for a modular organization of reaching movements but also for integration of the two rather distinct processes of reaching and postural control. Using a similar experimental paradigm, Berret et al. (2009) examined the exact relationship between reaching and postural control. To this end, they compared joint angle time courses during unconstrained whole-body reaching movements with similar movements in which they imposed certain spatial or equilibrium constraints on hand trajectories or posture. While corroborating the basic findings of Thomas et al. (2005) for unconstrained movements, they found that constraining kinematic or equilibrium aspects of these movements altered the motor synergies used. Introducing a kinematic constraint led to the emergence of an additional motor synergy for the hand while leaving the part of the controller responsible for posture control unchanged. These results show that complex reaching movements can not only be controlled on the joint angle level by motor synergies, but that motor synergies can arise flexibly when tasks require more elaborate control. These findings thus illustrate the need for higher level control structures that select and combine motor synergies in a behavior-dependent manner. Bockemühl et al. (2010) extended these findings towards a more general description when they studied fast, intuitive, and unconstrained catching movements of human shoulder girdle and arm in a large portion of the natural 3D working range and decomposed these movements using PCA. The results suggest that the ten joint angles studied in these experiments are strongly kinematically coupled; a large part of the kinematic variance in these movements can be explained by three principal components which can be readily interpreted as motor synergies. Interestingly, three is the minimum number of synergies necessary to place the end-effector of an extremity at an arbitrary point in 3D space. The authors also showed that the contribution of specific synergies varies systematically and smoothly with the final target position of the catching movement, immediately suggesting a mapping between peri-personal Cartesian space and intrinsic joint angle space. Reducing the Number of DOFs on the Muscle Level Landmark studies on hind-limb wiping

movements in frogs revealed that the spinal cord is modularly organized on a functional level (Bizzi et al. 1991; Saltiel et al. 1998; reviewed in Bizzi et al. 1995, 2000, 2002) (Fig. 11.8). This work suggests that the spinal cord contains a set of neural modules that control the force vector producing movement of the distal end of the leg. When a module is active, it produces a force vector in the leg. When the starting position of the leg is varied, the vector the module produces changes. Mapping the set of vectors produced from all leg start positions defines the module’s force field. Strikingly, each module’s force field has a single convergence point at which the hind-limb comes to rest. That is, the module induces leg movements that vary in such a way that, from any starting position, the leg distal end moves toward a single point. Furthermore, simultaneous activation of two modules results in a force field which is the linear sum of the two activated force fields. This organization immediately implies that the core function of a putative controller for target-oriented hind-limb movements involves the combination of these force fields rather than control on the joint or muscle level.

359

360

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B

C

D

1 cm

0.5 N

Figure 11.8 Frog spinal movement modules. (A, B) The force fields induced by stimulation of two different modules; in both panels arrows show the force produced when the module is stimulated and the starting position of the distal end of the leg is located at the arrow’s position. (C) Force field predicted by summation of vector fields in A and B. (D) Forces actually produced when two modules are simultaneously stimulated. Modified with permission from Bizzi et al. (2000).

Graziano and colleagues (2005, 2002) obtained comparable results in macaque primary motor cortex. These authors activated various sites of the arm and hand areas of the motor cortex (M1 and premotor areas) using electrical microstimulation on a behaviorally relevant time scale of approximately 500 milliseconds. Previous experimental paradigms had used impulses 10 to 50 milliseconds in duration that induced only transient muscle twitches. The prolonged stimulations elicited complex and behaviorally meaningful changes in arm and hand posture. Typically, these movements had bell-shaped hand velocity profiles whose amplitudes varied systematically with movement amplitude (Morasso 1981). Stimulation of a specific site in the motor cortex always resulted in the same final arm posture, regardless of the posture prior to stimulation. Although these findings are based on motor cortical areas and not spinal circuits, they are remarkably similar to the force fields found in frog spinal cord (Bizzi et al. 1995). Even on the level of M1, long thought to be a mere “switchboard” between premotor areas and the spinal cord, high level aspects of movements, such as whole-limb position and posture, seem to be represented. Conclusion for Motor Synergies Approaches There is ample evidence for the existence of

modular and synergistic control on several levels of the neuro-motor control chain

Prehensile Movements

in targeted limb movements. Nonetheless, important questions with regard to the modular organization of the CNS remain. One question pertains to the adaptive value of motor synergies: why would evolution favor complex body geometries in the first place and then limit their usefulness by restraining the potential behavioral repertoire through motor synergies? Work by Berret et al. (2009) provides a potential answer. These authors show that whole-body reaching movements are indeed organized by motor synergies. They also show, however, that new tasks with additional movement constraints activate previously inactive synergies. This might be a general concept used by the CNS: motor synergies restrain the behavioral repertoire in a given context, but do so differently in each context. In this idea, for every general class of motor behavior there exists a set of motor synergies optimally suited to the task. There might be specialized sets of motor primitives for catching, reaching, writing, and other distinct behaviors. Selection among these sets of motor primitives might be implemented by high-level processes, e.g., in prefrontal cortex and premotor areas. In this concept, the CNS would be a hierarchically organized selection machine that solves each redundancy problem by distributing the decision process onto several levels, on each of which the redundancy problem never arises due to the use of synergistic mechanisms like motor synergies. Although motor synergies and primitives potentially simplify motor control, a controller is necessary nonetheless. That is, although the number of inputs in Fig. 11.5b has been reduced to make the matrix invertible, a controller must still exist that sets the values of the inputs. Most studies investigating motor synergies and primitives merely confirm the existence of modules but remain silent or vague with regard to the nature of this controller. A modular controller must combine two potentially opposing features: on the one hand, it must be complex enough to compensate for the reduced capabilities of the controlled motor synergies. On the other hand, it must be simple enough to not add new redundancies. There is evidence that, when muscle synergies are organized to take advantage of the natural dynamic properties of the limb, such controllers can be realized (Berniker et al. 2009).

References Abend W (1982) Human arm trajectory formation. Brain 105:331–348. Berniker M, Jarc A, Bizzi E, Tresch MC (2009) Simplified and effective motor control based on muscle synergies to exploit musculoskeletal dynamics. Proc Natl Acad Sci USA 106:7601–7606. Bernstein N (1967) The co-ordination and regulation of movements. Oxford, UK: Pergamon Press Ltd. Berret B, Bonnetblanc F, Papaxanthis C, Pozzo T (2009) Modular control of pointing beyond arm’s length. J Neurosci 29:191–205. Bizzi E, D’Avella A, Saltiel P, Tresch M (2002) Modular organization of spinal motor systems. Neuroscientist 8:437–442. Bizzi E, Giszter SF, Loeb E, Mussa-Ivaldi FA, Saltiel P (1995) Modular organization of motor behavior in the frog’s spinal cord. Trends Neurosci 18:442–446. Bizzi E, Mussa-Ivaldi F, Giszter S (1991) Computations underlying the execution of movement: A biological perspective. Science 253:287–291.

361

362

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Bizzi E, Tresch MC, Saltiel P, d’Avella A (2000) New perspectives on spinal motor systems. Nat Rev Neurosci 1:101–108. Bockemühl T, Troje NF, Dürr V (2010) Inter-joint coupling and joint angle synergies of human catching movements. Hum Movement Sci 29:73–93. Bootsma RJ, van Wieringen PCW (1990) Timing an attacking forehand drive in table tennis. J Exp Psychol Human 16:21–29. Collewijn H, Erkelens CJ, Steinman RM (1988) Binocular co-ordination of human horizontal saccadic eye movements. J Physiol 404:157–182. Collins SH, Adamczyk PG, Kuo AD (2009) Dynamic arm swinging in human walking. P Roy Soc B: Biol Sci 276:3679–3688. Cruse H, Wischmeyer E, Brüwer M, Brockfeld P, Dress A (1990) On the cost functions for the control of the human arm movement. Biol Cybern 62:519–528. Cutkosky MR (1989) On grasp choice, grasp models, and the design of hands for manufacturing tasks. IEEE T Robotic Autom 5:269–279. Donders FC (1848) Beitrag zur Lehre von den Bewegungen des menschlichen Auges. Holländische Beiträge zu den anatomischen und physiologischen Wissenschaften 1:104–145. Engelbrecht SE (2001) Minimum principles in motor control. J Math Psychol 45:497–542. Fitts PM (1954) The information capacity of the human motor system in controlling the amplitude of movement. J Exp Psychol Gen 47:381–391. Flash T, Henis E (1991) Arm trajectory modifications during reaching towards visual targets. J Cognit Neurosci 3:220–230. Flash T, Hochner B (2005) Motor primitives in vertebrates and invertebrates. Curr Opin Neurobiol 15:660–666. Flash T, Hogan N (1985) The coordination of arm movements: an experimentally confirmed mathematical model. J Neurosci 5:1688–1703. Georgopoulos A, Schwartz A, Kettner R (1986) Neuronal population coding of movement direction. Science 233:1416–1419. Georgopoulos AP, Caminiti R, Kalaska JF, Massey JT (1983) Spatial coding of movement: a hypothesis concerning the coding of movement direction by motor cortical populations. Exp Brain Res Suppl 7:327–336. Georgopoulos AP, Kalaska JF, Caminiti R, Massey JT (1982) On the relations between the direction of two-dimensional arm movements and cell discharge in primate motor cortex. J Neurosci 2:1527–1537. Graziano MS, Aflalo TN, Cooke DF (2005) Arm movements evoked by electrical stimulation in the motor cortex of monkeys. J Neurophysiol 94:4209–4223. Graziano MS, Taylor CS, Moore T (2002) Complex movements evoked by microstimulation of precentral cortex. Neuron 34:841–851. Harris CM, Wolpert DM (1998) Signal-dependent noise determines motor planning. Nature 394:780–784. Henis E, Flash T (1995) Mechanisms underlying the generation of averaged modified trajectories. Biol Cybern 72:407–419. Hogan N (1984) An organizing principle for a class of voluntary movements. J Neurosci 4:2745–2754. Hooper SL (2012) Body size and the neural control of movement. Curr Biol 22:R318–R322.

Prehensile Movements

Hooper SL, Guschlbauer C, Blümel M, Rosenbaum P, et al. (2009) Neural control of unloaded leg posture and of leg swing in stick insect, cockroach, and mouse differs from that in larger animals. J Neurosci 29:4109–4119. Kalaska JF, Crammond DJ (1992) Cerebral cortical mechanisms of reaching movements. Science 255:1517–1523. Lacquaniti F, Terzuolo C, Viviani P (1983) The law relating the kinematic and figural aspects of drawing movements. Acta Physiol 54:115–130. Lee WA (1984) Neuromotor synergies as a basis for coordinated intentional action. J Motor Behav 16:135–170. Lemon RN, Griffiths J (2005) Comparing the function of the corticospinal system in different species: organizational differences for motor specialization? Muscle Nerve 32:261–279. Meulenbroek RG, Rosenbaum DA, Thomassen AJ, Loukopoulos LD, Vaughan J (1996) Adaptation of a reaching model to handwriting: how different effectors can produce the same written output, and other results. Psychol Res 59:64–74. Morasso P (1981) Spatial control of arm movements. Exp Brain Res 42:223–227. Morasso P, Mussa Ivaldi FA (1982) Trajectory formation and handwriting: a computational model. Biol Cybern 45:131–142. Mussa–Ivaldi FA (1988) Do neurons in the motor cortex encode movement direction? An alternative hypothesis. Neurosci Lett 91:106–111. Nakajima K, Maier MA, Kirkwood PA, Lemon RN (2000) Striking differences in transmission of corticospinal excitation to upper limb motoneurons in two primate species. J Neurophysiol 84:698–709. Napier JR (1956) The prehensile movements of the human hand. J Bone Jt Surg (Br) 38–B:902–913. Neilson PD (1993) The problem of redundancy in movement control: the adaptive model theory approach. Psychol Res 55:99–106. Nelson WL (1983) Physical principles for economies of skilled movements. Biol Cybern 46:135–147. Osu R, Uno Y, Koike Y, Kawato M (1997) Possible explanations for trajectory curvature in multijoint arm movements. J Exp Psychol Human 23:890–913. Penfield W, Boldrey E (1937) Somatic motor and sensory representation in the cerebral cortex of man as studied by electrical stimulation. Brain 60:389–443. Polyakov F, Drori R, Ben-Shaul Y, Abeles M, Flash T (2009) A compact representation of drawing movements with sequences of parabolic primitives. PLoS Comp Biol 5:e1000427. Pouget A, Snyder LH (2000) Computational approaches to sensorimotor transformations. Nat Neurosci 3 Suppl:1192–1198. Prescott TJ, Bryson JJ, Seth AK (2007) Introduction. Modelling natural action selection. Philos T Roy Soc B 362:1521–1529. Rosenbaum DA, Jorgensen MJ (1992) Planning macroscopic aspects of manual control. Hum Movement Sci 11:61–69. Saltiel P, Tresch MC, Bizzi E (1998) Spinal cord modular organization and rhythm generation: an NMDA iontophoretic study in the frog. J Neurophysiol 80:2323–2339. Saltzman E (1979) Levels of sensorimotor representation. J Math Psychol 20:91–163. Sanger TD (2000) Human arm movements described by a low-dimensional superposition of principal components. J Neurosci 20:1066–1072.

363

364

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Schaal S, Sternad D (2001) Origins and violations of the 2/3 power law in rhythmic three-dimensional arm movements. Exp Brain Res 136:60–72. Scott SH, Gribble PL, Graham KM, Cabel DW (2001) Dissociation between hand motion and population vectors from neural activity in motor cortex. Nature 413:161–165. Scott SH, Kalaska JF (1997) Reaching movements with similar hand paths but different arm orientations. I. Activity of individual cells in motor cortex. J Neurophysiol 77:826–852. Scott SH, Sergio LE, Kalaska JF (1997) Reaching movements with similar hand paths but different arm orientations. II. Activity of individual cells in dorsal premotor cortex and parietal area 5. J Neurophysiol 78:2413–2426. Soechting JF, Lacquaniti F (1981) Invariant characteristics of a pointing movement in man. J Neurosci 1:710–720. Sternad D, Schaal S (1999) Segmentation of endpoint trajectories does not imply segmented control. Exp Brain Res 124:118–136. Suzuki M, Yamazaki Y, Mizuno N, Matsunami K (1997) Trajectory formation of the center-of-mass of the arm during reaching movements. Neuroscience 76:597–610. Thomas JS, Corcos DM, Hasan Z (2005) Kinematic and kinetic constraints on arm, trunk, and leg segments in target–reaching movements. J Neurophysiol 93:352–364. Ting LH, McKay JL (2007) Neuromechanics of muscle synergies for posture and movement. Curr Opin Neurobiol 17:622–628. Todorov E (2000) Direct cortical control of muscle activation in voluntary arm movements: a model. Nat Neurosci 3:391–398. Tresch MC, Jarc A (2009) The case for and against muscle synergies. Curr Opin Neurobiol 19:601–607. Uno Y, Kawato M, Suzuki R (1989) Formation and control of optimal trajectory in human multijoint arm movement. Minimum torque-change model. Biol Cybern 61:89–101. Viviani P, Flash T (1995) Minimum-jerk, two-thirds power law, and isochrony: converging approaches to movement planning. J Exp Psychol Human 21:32–53. von Helmholtz H (1867) Handbuch der Physiologischen Optik, vol 3. Leipzig: Leopold Voss. Wiegner AW, Wierzbicka MM (1992) Kinematic models and human elbow flexion movements: quantitative analysis. Exp Brain Res 88:665–673. Wolpert D, Ghahramani Z (2000) Computational principles of movement neuroscience. Nat Neurosci 3:1212–1217. Wong AM (2004) Listing’s law: clinical significance and implications for neural control. Surv Ophthalmol 49:563–575. Yang F, Feldman AG (2010) Reach-to-grasp movement as a minimization process. Exp Brain Res 201:75–92.

365

12 Muscle, Biomechanics, and Implications for Neural Control Lena H. Ting 1,2 and Hillel J. Chiel 3 1

Department of Biomedical Engineering, Emory University and Georgia Institute of Technology, Atlanta, GA, USA Department of Rehabilitation Medicine, Division of Physical Therapy, Emory University, Atlanta, GA, USA 3 Departments of Biology, Neurosciences, and Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA 2

12.1 Introduction Understanding neural and biomechanical interactions is fundamental to motor control. Nervous systems arose in the earliest motile animals, allowing them to move purposefully through the environment. For example, some tunicates have a complex nervous system in their juvenile stages that allows them to swim as tadpoles, before affixing to a rock and becoming sessile filter feeders with greatly reduced and simplified nervous systems as adults (Zaniolo et al. 2002; Meinertzhagen et al. 2004). Across species and at multiple scales, the biomechanics of muscular structure, body form, and environmental forces influence how animals move (Vogel 1988). Neural mechanisms must act through the complex biomechanics of the body and the environment to generate motor behaviors (Chiel and Beer 1997; Chiel et al. 2009). Of equal importance, neuromechanical interactions define whether and how neural signals can influence motor function, and can reveal seemingly paradoxical relationships between neural signals and body movements. Because there are no one-to-one relationships between neural signals and biomechanical variables, variations in neuromotor signals must be carefully interpreted with respect to the biomechanical properties of the organism and its environment, which can vary with movement context. As a result, biomechanical properties determine whether neural signals simply excite the natural dynamics of the system, in which case variability has little effect on motor output, or whether neural signals must be precisely controlled to achieve the correct motor function. Our ability to “read” the neural motor code is thus intimately entwined with decoding the physical dynamics of the system receiving the neural signals. In this chapter, we focus on the key principles of biomechanics and motor control that underlie motor behaviors. How do neural and biomechanical systems interact to produce functional sensorimotor behaviors? While muscles and neural circuits need to be studied in isolation to understand many of their properties, recent studies suggest strongly that it is only in a

Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

366

Neurobiology of Motor Control: Fundamental Concepts and New Directions

functional and physiological context that their role in producing movement can be fully understood. There is astonishing variety across animals of the types of muscle, body plans, and environments that determine the neuromechanical properties and interactions that underlie movement. We use examples from invertebrates and vertebrates to demonstrate how understanding the mechanics of motor behaviors helps make sense of this great diversity. In the first section of this chapter, we describe how the transformation of neural activity into individual muscle forces depends on both intrinsic neuromuscular properties and muscle function. We then examine how the effects of activating a muscle in producing a behavior depend on body structure, interaction with environmental forces, and behavioral context. The third section discusses how multi-functional biomechanical interactions influence neural strategies for movement and our interpretation of neuromotor patterns. We argue throughout that, in both vertebrates and invertebrates, muscle multi-functionality versus specialization drives the relative complexity of neural versus muscular adaptations in generating motor behavior.

12.2 Behavioral Context Determines How Motorneuron Activity Is Transformed into Muscle Force and Power In this section, we review the neuromuscular transform from motorneuron activity to muscle force, the first step in the transformation of neural signals to biomechanical outputs. Classically, motorneurons have been considered the “final common pathway”, acting as a relay that transmits information from descending commands to muscle. More recently, motorneurons from many species, including humans, have been shown to perform complex processing and modulation of descending commands that can greatly alter the pattern of muscle activation. Moreover, although motorneuron action potentials are reliably transformed into neurotransmitter release at the neuromuscular junction, the resulting force of the muscle is not uniquely determined by this transformation, but also depends upon the state of the muscle at the time of activation as well as its recent and long-term history of activity (Hooper and Weaver 2000). Classical analysis of isolated muscle separates forces into independent components dependent on muscle length, velocity, and activation level. Studies of muscle properties have therefore often been done on isolated muscle under conditions in which muscle force, velocity, or length were held constant (Hill 1938, 1953; Gordon et al. 1966; Rack and Westbury 1969). This work has defined multiple well-known muscle properties (Enoka and Pearson 2013) that have been used in a wide variety of phenomenological muscle models (Zajac 1989). Muscle Force Summation over Time Varies with Activation Frequency Muscles act as low pass

filters of motorneuronal activity (for an example of an extreme functional consequence, see Fig. 10.3d). Muscle activity can therefore outlast neural excitation, and muscles may fail to respond to low levels of neuronal activity. This property is often referred to as the “force-frequency” or “muscle twitch” characteristic, describing the rate of muscle force activation and deactivation. In response to repeated stimuli, force summation depends on the difference between motorneuron interspike interval and muscle twitch duration; no summation occurs for interspike intervals longer than twitch duration, and

Muscle, Biomechanics, and Implications for Neural Control

summation becomes increasingly greater as interspike interval becomes increasingly less than twitch duration. A consequence of this interplay is that activation amplitude can depend on either spike number or spike frequency (Morris and Hooper 1997; Hooper et al. 2007). Muscle Length Alters Muscle Force Production The number of myosin heads than can engage

the actin filament increases as thick and thin filament overlap increases. For any given activation level, muscle force therefore varies as a function of muscle length. This property is referred to as the “force-length” or “length-tension” characteristic of a muscle. Thus, when muscle is held at a fixed length (i.e., undergoes an isometric contraction), if the fixed length is other than the optimal length, the muscle develops less than its maximum possible tension. Muscle Velocity Alters Muscle Force Production The ability to form cross-bridges between

actin and myosin also depends on the speed with which the thick and thin filaments move relative to each other. This property is referred to as the “force-velocity” characteristic of a muscle. Generally, forces decrease as the rate of muscle shortening increases, and increase as the rate of muscle lengthening increases. Even When Muscles Receive no Neuronal Activation, They Resist Lengthening Stretched un-

activated muscles generate force, primarily due to stretching of sarcomere-spanning muscle giant proteins (for references see Hooper and Thuma 2005). This property is referred to as “passive muscle force” and increases with muscle length. Passive force is independent of acto-myosin interactions (Thuma and Hooper 2010), and is therefore typically considered to act in parallel with the active muscle force arising from cross-bridge interactions. Muscle models often treat force-frequency, force-length, and force-velocity properties as independent functions that can be simply multiplied by one another, and to which the passive properties are added. They also typically assume that whole muscle force generation properties are the same as the properties of individual muscle fibers or sarcomeres. However, as studies have been extended to analyses within behaving animals, and to physiological activation patterns of muscles and motorneurons, more complexity has emerged, which must be understood to properly predict muscle function. 12.2.1 The Neuromuscular Transform Is History-Dependent

The transformation from motor neural action potentials to muscle force can have varying time-histories. A classical view of the neuromuscular junction (Enoka and Pearson 2013) starts with the activation of the motorneuron innervating the muscle, which faithfully responds to its synaptic inputs and then releases transmitter at its pre-synaptic terminal, i.e., the neuromuscular junction. Transmitter binding by post-synaptic receptors generates a strong depolarization that induces the entire innervated motor unit (the motorneuron and muscle fibers it innervates) to generate an action potential, activating calcium release from internal stores in the muscle fibers. This triggers myosin binding to actin and initiates a contraction, usually referred to as a twitch. However, recent studies have demonstrated that each step of this process is more complex, and subject to extensive modulation (Hooper and Weaver 2000). In general, such non-classical

367

368

Neurobiology of Motor Control: Fundamental Concepts and New Directions

properties are referred to as “history-dependent” because muscle force depends not only on current levels of activation, length, and velocity, but also events in the recent past. 12.2.1.1 Motorneurons Are Subject to Neuromodulation and History-Dependence That Can Significantly Alter Their Output

Motorneurons cannot be considered simple relays or integrators of synaptic input; rather, they can strongly affect the degree to which muscle can be activated. The same synaptic input to a motorneuron can generate vastly different firing rates as a result of neuromodulatory effects on motorneurons. As a consequence, the voluntary activation of a muscle, even at maximal effort, may elicit a wide range of force levels depending upon the neuromodulatory state of the motorneuron. The existence of persistent inward currents (PIC) in motorneurons, carried by sodium and calcium ions, leads to bistability: sustained motorneuron firing can continue after synaptic input is removed (Heckman et al. 2008) (see also Chapter 8). Neuromodulators can act on metabotropic receptors and, through their actions on the PIC, increase motorneuron sensitivity to excitatory inputs by 6-fold in vertebrates (Lee and Heckman 2000). Furthermore, different patterns of descending inhibitory input can alter the ability of motorneurons to recruit motor units (Powers et al. 2012). Thus, the level and pattern of descending synaptic input onto motorneurons have a complex relationship to the final activity of the muscles. 12.2.1.2 Presynaptic Neurotransmitter Release at the Neuromuscular Junction Is History-Dependent

After a rapid volley of action potentials (a tetanus), the presynaptic terminal may release higher amounts of transmitter than it would have in a resting state, and this post-tetanic potentiation appears to be due to increased levels of calcium in the presynaptic terminal (Zucker and Regehr 2002). 12.2.1.3 Post-Synaptic Muscle Excitation Is History-Dependent and Subject to Modulation

In both vertebrate skeletal (Enoka and Pearson 2013) and invertebrate muscle, whether a train of impulses temporally summates depends, as explained above, on the difference between interspike interval and twitch duration. In non-spiking invertebrate muscle and vertebrate smooth muscle, excitatory junction potentials can summate spatially as well as temporally. Invertebrate and smooth muscles are also subject to neuromodulation both from dedicated modulatory neurons or circulated neuromodulators (extrinsic modulation), and modulatory substances released from the muscle’s own motorneurons (intrinsic modulation) (Katz and Frost 1996). Such modulatory compounds (often biogenic amines or peptides) may not induce force changes on their own, but rather change the strength and speed of response to the conventional transmitter released by the motorneuron, thus allowing the muscle to respond to both low and high rates of motorneuron firing (Brezina et al. 2000). A recent study in intact animals demonstrated that motorneuronal activity during one behavior (an attempt to grasp food) did not generate significant force, but prepared the muscle to generate much stronger forces in response to the same motorneuron firing at higher frequencies during a subsequent behavior (swallowing) (Lu et al. 2015).

Muscle, Biomechanics, and Implications for Neural Control

12.2.1.4 Contractile Dynamics of Cross-Bridge Interactions Are History Dependent

The properties of active muscle force generated through actin–myosin interactions are also history-dependent; some of these properties are discussed in more detail below. Examples include the level of muscle force depending on the history of muscle excitation, length, and lengthening or shortening velocity (Vandenboom et al. 2013). This history-dependence can result in skeletal muscle generating increasingly higher “staircase” force profiles in response to identical stimulation pulse trains. In smooth muscle, prior stretch can remodel the actin–myosin filaments, altering muscle force-length properties; resetting the force-length relation allows muscles that interact with soft organs to continue to generate force even as they are steadily stretched (e.g., as occurs during lung inflation in breathing, Gunst et al. 2003). These multiple modulatory and history-dependent features of the motor unit mean that defining the context and history of muscle activation is critical for determining the forces that muscles generate (Hooper and Weaver 2000; Perreault et al. 2003; Siebert et al. 2007). Furthermore, this list of history-dependent properties is not exhaustive, and in particular does not include the effects of muscle fatigue on the neuromuscular transform (Enoka et al. 2011). 12.2.1.5 The Molecular Motors of Muscles Give Rise to the Functional and History-Dependent Properties of Muscle Force Generation

The phenomenological properties of whole muscle described above all result, ultimately, from active and passive molecular processes in the muscle. The active properties arise from the dynamics of the force-generating actin–myosin interactions (Huxley 1957). Simulations of populations of actin–myosin interactions (Fig. 12.1), using mechanistic models of cross-bridge dynamics, can reproduce muscle force-frequency, force-length, and force-velocity properties (Zahalak 1986; Zahalak and Ma 1990). Several other history-dependent muscle properties are not included in phenomenological muscle models but have important functional roles in movement. Catch and Latch Catchlike properties of muscles may be important for force enhancement to maximize muscle performance, and have been found in a variety of species including humans (Binder-MacLeod and Kesar 2005). Some molluscan muscles, such as the anterior byssus retractor muscle of the mussel Mytilus edulis, show a powerful catch property in which a short neural activation results in sustained tension that is maintained without fatigue and using very little energy. Catch is due to binding of a thick filament protein, twitchin, to the thin filaments, thus locking the thick and thin filaments rigidly together, and is regulated by twitchin phosphorylation state (Funabara et al. 2005; reviewed in Hooper et al. 2008). Vertebrate and (non-catch) invertebrate smooth muscles show a phenomenologically similar latch state in which sustained contractions are maintained with relatively low levels of intracellular calcium and ATP consumption. This state arises via phosphorylation of a myosin regulatory light chain, which slows cross-bridge cycling, in particular cross-bridge detachment (Murphy and Rembold 2005). Yu et al. (1997) used a biophysically-based model of this process to create a non-isometric smooth muscle model that effectively captures the latch properties of a wide variety of vertebrate and invertebrate muscles. Short Range Stiffness The force a muscle produces when stretched after a prolonged rest

is higher than that predicted by the force-length relationship (Nichols and Cope 2004;

369

370

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Ca2+

Actin + Myosin

Actin + Calcium activated myosin

Ca2+

Nonactivated actomyosin (latch-bridge)

Activated actomyosin (regular cross-bridge)

Figure 12.1 Cross-bridge interactions underlying classical and history-dependent characteristics of muscle force generation. Formation of cross-bridges that generate muscle force. Top left: thin (actin) filament above thick (myosin) filament; unbound myosin head is shown. Top right: calcium exposes myosin binding sites (in vertebrate muscles) or directly induces contraction by binding to myosin (molluscan muscle). Bottom right: ATP induces the dissociation of the actin–myosin complex and ATP hydrolysis causes a conformational change that moves the myosin head, which binds to a new position on the actin. As the myosin head returns to its initial position, the actin filament slides relative to the myosin. Bottom left: Under appropriate circumstances, a latch cross-bridge forms, which can maintain the attachment and thus tension with very little requirement for energy. The mechanism of latch is still an area of active investigation (e.g., Yu et al. 1997). Catch, which occurs in invertebrate muscles and results in the ability to maintain tension with less or even no energy use, occurs via a conceptually similar connection of the thin and thick filaments by the giant sarcomere associated protein, twitchin (see Hooper and Thuma 2005, Hooper et al. 2008 for references).

Nishikawa et al. 2007). This force is present in activated and isolated muscle fibers, and is likely due to differences in the relative population of attached and unattached actin–myosin molecules in isometric vs. shortening or lengthening muscles (Getz et al. 1998; Campbell and Moss 2002). This property is referred to as short-range stiffness (Rack and Westbury 1974; Morgan 1977) and plays an important role in producing restoring forces in response to muscle stretch in isolation (Kirsch et al. 1994) and during limb movement (Joyce et al. 1974; Perreault et al. 2004). Short-range stiffness also alters sensory information encoding by proprioceptive afferents in response to muscle stretch, with proprioceptive response increasing when the muscle has zero (isometric) vs. non-zero velocity (Haftel et al. 2004). This history-dependence in proprioceptive response plays an important role in the motor control of standing balance (Loram et al. 2009; Welch and Ting 2009; Safavynia and Ting 2013) [See also Chapter 9.]. Force Enhancement and Depression Muscle forces during isometric contraction also

depend upon the past history of stretching, with higher forces being produced after active stretch (Abbott and Aubert 1952; Edman et al. 1982) and lower forces after active shortening (Abbott and Aubert 1952; Marechal and Plaghki 1979; Herzog et al. 2000).

Muscle, Biomechanics, and Implications for Neural Control

Although the mechanisms have been debated (Minozzo and Lira 2013), the current view is that contributions from actin–myosin interactions (Amemiya et al. 1988; Bartoo et al. 1997; Herzog et al. 2008), engagement of passive elements within muscle sarcomeres (Edman and Tsuchiya 1996; Herzog and Leonard 2002), and sarcomere non-uniformity may all play a role (Morgan 1990; Edman 2012). 12.2.2 Muscle Power Depends on Behavioral Context

Most of the work mentioned above examining muscle properties was performed by varying muscle activation, length, and velocity individually. In vivo, however, muscle activation, length, and velocity typically change independently and simultaneously. The work loop technique (Josephson 1985; Ahn 2012) was developed to allow muscle force production to be studied under the kinematic conditions that muscles experience in vivo during motor behaviors. In this technique an isolated muscle is exposed to the time-varying pattern of activation, length, and velocity that it would experience in vivo, demonstrating interactions between the force-frequency, force-length, and force-velocity properties that can radically change the function of a muscle. By plotting muscle force versus length (Fig. 12.2), the magnitude and sign of power generation in the muscle can be visualized. This work reveals that a muscle can act as a motor, strut, spring, or brake depending on when it is activated relative to when it shortens and lengthens. To produce energy, i.e., to act as a motor, a muscle must be active primarily during muscle shortening. To absorb energy, i.e., to act as a brake, a muscle must be active primarily during muscle lengthening. Without themselves producing (mechanical) energy, muscles can transmit energy, i.e., act as a strut, by being isometric, and store energy, i.e., act as a spring, by being elastic. Muscles acting in all these fashions are found in vertebrates and invertebrates alike (Dickinson et al. 2000). Because of these complex interactions among activation, length, and velocity, plots of force versus velocity during motor behaviors depend heavily on behavioral context (Kawakami and Fukunaga 2006; de Brito Fontana et al. 2014) and do not look like those identified when only one of activation, length, and velocity are allowed to vary (Kawakami et al. 2002; Finni et al. 2003). We use three examples to demonstrate these concepts: Two muscles that are innervated by the same motorneuron and have similar isolated muscle properties play opposite roles in cockroach locomotion: one acts as a motor and the other as a brake (Ahn and Full 2002; Ahn et al. 2006). The two muscle share similar activation patterns and cross the same joint. However, the collective effects of small differences in behavioral context and intrinsic muscle properties cause one muscle to produce energy, whereas the other absorbs it (Fig. 12.2C). The motor muscle is active as it shortens. The brake muscle is activated shortly later, causing it to be active late in shortening and, importantly, at the onset of lengthening. This difference in activation timing relative to when muscle length changes is compounded by the brake muscle having a longer muscle deactivation time constant, and the larger range of the muscle force-length relationship that the brake muscle visits, and the larger strains (length changes) it experiences, during lengthening. When subjected to identical experimental conditions, however, each muscle can be made to produce and absorb energy. Therefore, the functional difference between the two muscles arises from the different behavioral contexts in which they perform.

371

372

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B

C

D

E

F

Figure 12.2 Work loop analysis reveals function of muscles as motors, brakes, springs, and struts. The work generated by a muscle is indicated by the area within the loop (shaded areas). Counterclockwise arrows indicate mechanical energy generation (e.g., A, B, F), whereas clockwise loops indicate mechanical energy absorption (e.g., C). When a muscle acts as a spring or strut, no mechanical energy is generated (D, E, F). [The “mechanical” distinction being made because, even in cases in which mechanical energy is being absorbed or the muscle is acting as a strut, the muscle is generally activated during at least part of the movement cycle and hence ATP—chemical energy—is being consumed.] (A) Scallop swimming generates positive work. Starting at the bottom right with the shell fully open, muscle activation causes a rise in force that closes the shell. The muscle deactivates in the upper left, and force drops as the shell continues to close. The shell then opens using no muscle force through active recoil of elastic hinge elements. (B) The pectoralis muscle of birds generates positive power for flight. (C) In running cockroaches, muscles suited for shortening and power production instead absorb energy, and thus act as brakes. (D) In flies, an intrinsic wing muscle acts as a spring to steer and direct power production by flight muscles. (E) In some fish, muscle acts as both a motor and a strut during the locomotor cycle. Cranial muscle fibers first shorten and produce power, which is transmitted by more caudal muscle fibers acting as struts. The more caudal muscle then generates power that is transmitted caudally. (F) In vivo muscle force and length measurements in running turkeys indicate a dual role for the gastrocnemius muscle. It generates positive power during uphill running. During level running, it acts like a strut, allowing energy storage and recovery in spring-like tendons. Modified from Dickinson et al. 2000 with permission.

Muscle, Biomechanics, and Implications for Neural Control

In flight muscles of the moth Manduca sexta the function of flight muscles varies with the gradient of temperature along the wing (George et al. 2013). The proximal portion of the muscle is at a higher temperature and generates power, whereas more distal portions of the muscle are cooler and able to store elastic energy in linked cross-bridges due to slower activation and deactivation time constants. In human jumping, counter-movements take advantage of both classical and history-dependent muscle properties. To reach a maximal jump height, humans use a countermovement technique in which the center of mass is lowered before it is raised. Simulations have shown that the counter-movement increases the force applied during the propulsive phase by allowing muscle force to increase over time as muscles lengthen during the lowering phase (Pandy and Zajac 1991). The active stretching also takes advantage of muscle force enhancement, a history-dependent property in which active muscle force generation increases after stretch (McGowan et al. 2013). The change in posture may further increase force by changing muscle length, and thus position on the force-length curve. Energy storage in tendon and muscle also contribute to increased jump power (Bobbert and van Ingen Schenau 1988; Voigt et al. 1995). Therefore, the countermovement generates a behavioral context in which muscle force and power production are increased. 12.2.3 Muscle Specialization Reflects Behavioral Repertoire

The degree of muscle specialization, plasticity, and history-dependence likely depends upon the mechanical and functional context of movement. As described above, the force that results from a given descending neural command can be modified by modulators and history-dependent properties. As a consequence, muscles may be quite diverse and may lie on a continuum between being highly modifiable to having relatively stereotyped and predictable dynamic responses. We believe it reasonable to combine these observations into a principle: the degree of specialization of muscle properties depends on how multi-functional the muscle is throughout the natural history of the animal. The more singular the function of a muscle, the more likely specialization occurs peripherally, i.e., at the level of muscle contractile properties. In contrast, the more behaviors the muscle contributes to, the more likely it is that modulation occurs centrally, i.e., in motorneuron properties. Some examples may illustrate this principle. In muscles that interact with soft organs requiring sustained force production at varying lengths, changing muscle fiber properties (e.g., attachment points of the myosin head and the actin filament) may solve the problem. However, in skeletal muscles that play roles in both posture and movement, the ability to produce sustained or dynamic forces may be modulated at the level of the motorneuron (Heckman et al. 2008). Similarly, specialization in muscle force-frequency, force-length, and force-velocity properties are observed in muscles with more stereotyped functions, such as flight power muscles (Biewener 2011) or vocal muscles (Elemans et al. 2004). Human leg muscles show substantial specializations (Lieber and Ward 2011), and insect hindlimb muscles are more specialized than forelimb muscles (Ritzmann et al. 2004). The specific demands of propulsion vs. posture may, respectively, drive the differences in twitch rate found in human gastrocnemius, composed primarily of fast-twitch motor units, and soleus, composed primarily of slow-twitch motor units, calf muscles. Although soleus is important for standing, it cannot be rapidly activated

373

374

Neurobiology of Motor Control: Fundamental Concepts and New Directions

and deactivated, and is inhibited during rapid postural responses to perturbation (Macpherson 1988; Macpherson and Horak 2013). Arm and forelimb muscles tend to be more heterogeneous in their properties, and to have greater flexibility through modulation at both motorneuronal and higher levels. In summary, considering desired behavioral outputs can shed light on how muscle and muscle control systems are structured, and at what levels specialization and flexibility occur.

12.3 Organismal Structures Transform Muscle Force into Behavior Muscles do not function in isolation: both the structures in which they operate, and the environment in which an animal behaves, affect muscular function. This section focuses on the transformation from muscle force to the body forces and movements that generate motor behavior. An important general issue is that muscle function cannot be deduced from a simple read-out of motorneuron activity or muscle force alone, but rather requires consideration of all forces acting on the body structure. Furthermore, although muscle forces directly affect accelerations, some physical forces, e.g., viscosity, depend on velocity, and the forces of gravity and elasticity depend on position. These observations imply that animal configuration, movement speed, limb or animal scale, and environment properties (e.g., the viscosity of the medium through which an animal moves: for example, air versus water) are all important for understanding any given muscle’s role in generating a movement. Depending on whether inertial, viscous, or elastic forces dominate, muscle activity can appear to be directly related to movement acceleration, velocity, or position. We illustrate how the functional effects of the force of a single muscle cannot be interpreted in isolation, but depend upon mechanical interactions throughout the body from both active and passive structures. It is thus critical to focus on the overall behavioral context to understand the neuromechanical interactions that transform muscle forces to functional movements. Classical descriptions of muscle function are based on anatomical arrangements describing muscles as having singular functions. The structures of muscles observed within the body provide a useful first approximation to their function. These anatomical characteristics are important for building the lumped, phenomenological muscle models typically used in musculoskeletal models (Zajac 1989). The architecture of the muscles, tendons, joints, and soft tissue structures leads to a first-order functional description of muscles as cable-like force actuators that move the individual joints that they span (Delp and Loan 2000). While these models are very useful for understanding muscle control of movement, as a result of their simplifying assumptions, they cannot account for the complexities of muscle anatomy and function observed during the majority of real behaviors. Muscle Shape Muscles are often considered to be fusiform with distinct tendinous

attachments to a skeleton. Thus, muscle function is frequently characterized by describing the joints they span and how they generate torques at these joints. However, muscles have a wide range of forms and ways of attaching to both hard and soft tissues. Furthermore, the mass and deformation of the muscle itself must also be considered (Blemker et al. 2005; Pai 2010). This is particularly true for muscles in the tongue and in

Muscle, Biomechanics, and Implications for Neural Control

soft-bodied animals that have no skeletal attachments, in which muscles both generate support and act as a skeleton. Motor Unit Distribution Motor units are often assumed to be uniformly distributed within

a muscle such that the muscle is homogenously activated according to the size principle (Henneman 1957; Duchateau and Enoka 2011). However, in vivo work demonstrates that regions in single muscles can be differentially activated and have different effects on movement. Regionalized variations in muscle contractile properties and architecture are also observed. Properties of Surrounding Structures Muscle function is often considered in isolation based

on the forces and/or torques the muscle produces on the structures to which it attaches. However, the responses of these structures depend on the mechanical properties of the structure and the forces and accelerations the structure receives from other muscles and from the environment. It is impossible to describe a muscle’s function without considering these other factors. 12.3.1 Effects of Muscle Force Depend on the Properties of the Body and the Environment

Biomechanical affordances and constraints arising from interactions with other parts of the body and the environment affect the degree of neural control required to perform a movement. Biomechanical affordances refer to types of movements facilitated by the body structure, and how body structures define ways of moving that require little energy or neural control to produce. Biomechanical constraints refer to movements that are difficult or impossible to achieve with a given structure or refer to the neural input required to achieve a movement, e.g., precise timing or activity of a particular muscle. Again, the structural and material properties of the body affect how muscle activation alters body motion or shape to produce movement. For example, musculoskeletal systems can apply precise, concentrated forces to the environment that are useful for legged locomotion. Unlike soft-bodied structures, however, they cannot apply distributed forces along the body, nor assume complex shapes to conform to undulating terrain or find their way through tortuous crevices and curving structures. Ultimately, what a neural motor pattern “means” for a behavior depends on biomechanical affordances and constraints. 12.3.1.1 The Relative Importance of Inertial, Viscous, and Spring-Like Forces Affect the Role of Muscle Force

We first discuss the relative importance of different environmental forces for shaping motor output, and then give examples of how biomechanical structures affect the neural control of muscles for motor function. We examine different types of body structure that play a significant role in shaping movement: musculoskeletal systems, tendons and fascia, hydrostatic structures, and muscular hydrostats. Rather than focusing on anatomical descriptions, we discuss how different body structures and their interactions with the environment determine the kinds of motor functions that a muscle can produce. For all structures, the relative importance of inertial, viscous, gravitational, and elastic forces determine the dynamics of movement resulting from muscle activity. Important

375

376

Neurobiology of Motor Control: Fundamental Concepts and New Directions

determinants include the size of the animal or limb, movement speed, and the medium within which the movement occurs. As a rule of thumb, larger, terrestrial animals are dominated by acceleration-dependent inertial forces and position-dependent gravitational forces, whereas small terrestrial animals are dominated by velocity-dependent viscous forces and position-dependent elastic forces. Note that the dominating force within a large animal might not be inertial if it is moving a small part of its body (Charles and Hogan 2012); similarly, the dominating force within a small animal may not be viscous or positional if it is engaged in a highly rapid strike movement, dominated by inertial forces. The nature of the forces directly affects the dynamic equations of motion that determine whether a system is stable, oscillatory, or decaying in nature. For example, acceleration- and position-dependent force interactions are described by an oscillating system (i.e., a second order differential equation), whereas interactions between velocity and position-dependent forces are described by an exponentially decaying system (i.e., a first order differential equation). In general, large animals tend to activate their muscles in shorter durations relative to movement duration, relying on the inertial forces to complete the movements of the large masses of their limbs, whereas smaller animals must continuously activate muscle during movement, since the smaller masses of their limbs result in movements dominated by viscous or elastic forces (Hooper et al. 2009; Hooper 2012). Some animals can employ ballistic movements, in which an impulse of force can initiate a movement, such as a jump. In contrast, viscous forces are highly dissipative. For this reason, small and/or aquatic animals must generate power to overcome viscous forces within their environment. As a consequence, muscles must be continuously activated or the organism will cease to move (Hooper 2012). Movements in a viscous environment can be more ballistic and require less fine motor control, since the environment is inherently stabilizing and dampens oscillations. Such considerations are also important when moving different body parts or at different speeds. Inertial, viscous, and elastic forces are also determined by the properties of the tissues themselves. Inertial and gravitational forces are much more important in larger appendages such as the arms and leg, although movement through water versus air can also alter the relative importance of viscous versus gravitational forces. For example, simulations and passive dynamic walkers illustrate that little to no muscle activity is necessary to produce walking-like movements (Collins et al. 2005; Kuo 2007). This means that a transient impulse to a muscle can set in motion a complex movement that relies on kinetic and potential energy exchange of pendulum dynamics. It is likely, however, that muscles play more of a role in stabilization and control of walking, since gravitational forces are destabilizing when changing postures, and because perturbations can induce undamped oscillations (Ting et al. 2009). In contrast, human fingers have small mass and higher damping properties than larger limbs (Lin and Rymer 2001; Deshpande et al. 2012; Park et al. 2014). Insect limbs have low mass, a great deal of elasticity (Blickhan 1986; Sensenig and Shultz 2003) and can also have significant energy absorption through hysteretic damping (Dudek and Full 2006). In fact, in the stick insect, if swing muscle activity were to cease at any time during swing, swing leg movement would cease (Hooper et al. 2009). As a consequence, muscle activity is associated only with limb propulsion and is not required for braking limb movement. The specific dynamics of the body and environment determine whether muscles are required to power the body or limb in a stable environment, or to stabilize the body

Muscle, Biomechanics, and Implications for Neural Control

or limb in an unstable one. In the swimming lamprey, the inclusion of water viscosity is critical to generating the traveling wave necessary for propulsion (Bowtell and Williams 1994; McMillen et al. 2008). Moreover, neuromechanical modeling shows that changing body stiffness would alter the swimming motions caused by identical muscle activation patterns (Tytell et al. 2010). Similarly, when subjected to unpredictable forces, human arm stiffness and damping is increased in the direction of the applied forces, thus reducing the need for corrective neural control of arm position and desired movement trajectory (Franklin et al. 2004). 12.3.1.2 Muscle Function Depends on Behavioral Context and Environmental Forces

Behavioral context (body configuration and velocity) and environmental forces alter the relative sensitivity to, and hence the effects of, muscle force. As Bernstein (1967) stated: “…one and the same impulse…may produce completely different effects because of the interplay of external forces and because of variations in the initial conditions.” For example, computer simulations of human walking demonstrate a posture-dependent (crouched vs. upright) effect on whether a muscle generates extensor joint torques or center of mass acceleration (Hicks et al. 2008; Steele et al. 2010). A salient example of this difference can be demonstrated by introducing additional neural impulses to a cockroach limb muscle during two different behaviors: postural control and running (Sponberg et al. 2011b). Introducing artificial muscle action potentials during postural control caused graded linear effects on body rotation and velocity. In contrast, the same action potentials introduced during running had a wide range of complex effects—including no effect—on locomotion. More action potentials were generally required to elicit any measurable effects, which were generally non-linear. The effects of muscle stimulation varied dramatically depending on stimulation phase, increasing vertical velocity in one gait phase but turning the body in another. Furthermore, when investigated with the work loop technique, the muscle absorbed energy when stimulated as it would be during normal running (Full et al. 1998) but, due to interactions with the kinematics of the limb at the time of stimulation and consistent with the behavioral effects, produced positive work with added action potentials (Sponberg et al. 2011a). 12.3.1.3 Biomechanical Affordances and Constraints of Body Structures Affect Muscle Functions Skeletal Systems Musculoskeletal systems such as human arms and legs are familiar and

intensively studied structures. These systems are defined by a muscle attaching to hard skeletal elements, externally in vertebrates (an endoskeleton), internally in insects and other arthropods (a hard cuticle containing muscle, an exoskeleton). The articulations between bones (vertebrate) and body or limb segments (arthropod) define kinematic degrees of freedom for body and limb movements. Bones do not deform enough under load to contribute to movement control, but exoskeletal elements can sometimes deform and store energy necessary for movement. Muscles generate multiple joint torques defined by the complexity of skeletal structure. In vertebrates and invertebrates alike, some joints move in multiple degrees of freedom (multiple directions), and some muscles span multiple joints. Muscle moment arms relative to the articulated joint centers determine the leverage of the muscle force that

377

378

Neurobiology of Motor Control: Fundamental Concepts and New Directions

generates joint torque. Limb muscles are commonly referred to as uniarticular (spanning a single joint) or biarticular (spanning two joints). Articulated joints often have multiple kinematic degrees of freedom that define the different ways in which they can move, which must be considered in combination when a muscle is activated. For example, the ball and socket hip joint supports flexion-extension, adduction-abduction, and internal and external rotation. At a minimum, “uniarticular” hip muscles produce torques that affect three kinematic degrees of freedom at the hip, each of which must be coordinated with other muscles to produce a functional movement. The human knee joint also has multiple degrees of freedom, such as flexion–extension and varus–valgus rotation. Linear motion at both joints may also need to be considered. Thus, even simple line-of-action models (Delp et al. 1990) of the biarticular muscles crossing the hip and knee require specifying the joint torque contributions of five moment arms. In finger muscles, muscle tendons diverge and converge with the tendons of other muscles and span multiple joints (Valero-Cuevas et al. 1998, 2007; Valero-Cuevas 2005). Abdominal muscles have even further complexity due to their sheet-like structure and their ability to generate torques across multiple vertebrae via both hard and soft-tissue connections. Muscle moment arms also change with posture due to geometric considerations that alter the joint torques they produce (Young et al. 1992; Murray et al. 2000). Moment arms are often thought to decrease as joints move away from a neutral position, resulting in constant muscle force producing less torque. However, in cat hindlimb, the moment arms of ankle muscles in the medial-lateral direction increase with displacement from the neutral configuration (Young et al. 1992), generating restorative torque even with constant muscle force. As a result, frontal plane motion of the cat hindlimb can be intrinsically stable to external perturbations, but sagittal plane motion may require more neural control (Bunderson et al. 2008). Moment arms of human shoulder muscles change substantially with arm posture, and do so in a systematic way that may simplify the transformation between joint postures and joint torques (Buneo et al. 1997). Particularly ingenious musculoskeletal interactions are present in insects. In addition to joint motions, some insect muscles produce sufficient deformation of the exoskeleton to store and release energy that contributes to movement (Gronenberg 1996). In the proximal hind-tibia of the locust, a specialized region composed of resilin acts as a shock absorber as it buckles, preventing the leg from damaging itself when the animal kicks against a substrate (Bayley et al. 2012). Nymphs of the planthopper Issus have interdigitating cogs—i.e., a gear—that ensure exact coordination of both legs during jumping, synchronizing their movements within milliseconds and minimizing yaw movements (Burrows and Sutton 2013). Both mechanisms limit the degree of precise neural timing of muscle activation required to control and coordinate movements. Tendinous and Fascial Connections The tendons that connect muscles to bone have a wide

range of mechanical properties that play an important role in muscle function. The interactions between muscle and tendon architecture determine the capacity of the muscle to generate force, to lengthen over large excursions, and to generate power (Wilson and Lichtwark 2011). Muscle architecture refers to the arrangement of muscle fibers and tendon. At one extreme, long parallel muscle fibers attach to a tendon at the end of the muscle, allowing for a large range of motion. Short muscle fibers, alternatively, are typically pennate, inserting along the length of a tendon or aponeurosis that runs

Muscle, Biomechanics, and Implications for Neural Control

through the muscle, which provides high force and power generation but limits the capacity for length change. Tendon elasticity allows muscle fiber kinematics to differ substantially from musculotendon kinematics based on the relative motion between the skeletal insertions of the muscle. As a consequence, lengthening or shortening of the musculotendon unit cannot be directly related to the direction of muscle fiber lengthening and shortening. Tendons can be actively stretched through muscular contraction and the interaction with environmental forces such that energy is stored. In vivo measurements of muscle fascicle length have revealed that gastrocnemius muscles are largely isometric both during highly dynamic tasks such as running (birds: Roberts et al. 1997; humans: Wilson and Lichtwark 2011) and in less dynamic tasks such as standing balance (humans: Loram et al. 2009). The gastrocnemius remaining at near constant length as the length of its musculotendon unit changes in, say, running, means that the tendon elastically stretches (energy storage) during ankle flexion and lengthens (energy release) during ankle extension, and hence stores and releases mechanical energy across the step cycle. This is an example of a general mechanism of mechanical energy storage in running, in which elastic elements in the stiffened (by extensor muscle contraction) leg store energy through the beginning of stance and return it during the end of stance (Dickinson et al. 2000). It is important to note that this ability to store mechanical energy comes at a metabolic cost. In the case at hand, as the gastrocnemius tendon lengthens it exerts force on the gastrocnemius muscle, whose activation must therefore increase for the muscle to maintain a constant length. Moreover, maximum muscle efficiency is not at isometric contraction, but in contractions about one-quarter of maximum muscle velocity (see Lichtwark and Wilson 2008 for references). Total efficiency of the musculotendon unit, alternatively, depends on contraction velocity, tendon compliance, muscle fiber length, and muscle volume. Theoretical work on human legs shows that no single combination of these parameters is optimal for both running and walking, and that human values are intermediate between the optima for the two gaits (Lichtwark and Wilson 2008). Consequently, the isometric gastrocnemius data, and the braking and spring-like activities shown in Fig. 12.2, are an example of a repeating theme in biomechanics, that it is the entire ensemble (neurons, muscles, tendons, bones) that is selected for, not the maximum efficiency (or any other aspect) of any single member of the ensemble, and that this selection is made across some sort of weighted sum of the entirety of the animal’s behavioral repertoire. The sheets of fascia around muscle and muscle compartments also have connections between them that can transmit substantial force (Huijing 2003; Maas and Sandercock 2010). These connections allow muscles to generate torques about joints they do not span. For example, the rectus femoris, a quadricep muscle, has a knee extensor moment arm. When the muscle insertion is surgically relocated to produce a knee flexion moment arm, activation of the muscle nonetheless generates knee extensor torque. This presumably occurs because the muscle belly remains connected via fascia with other quadriceps muscles that still have extensor moment arms (Riewald and Delp 1997; Asakawa et al. 2002). Torques about distant muscles can also be generated through fascia. A striking example is the crural fascia, a thick band of fascia that transmits force from thigh muscles to the ankle joint (Stahl and Nichols 2014) and is particularly pronounced in cheetah (van Ingen Schenau 1994). In decerebrate preparations, in which motorneuron output can be completely controlled, disrupting the crural fascia

379

380

Neurobiology of Motor Control: Fundamental Concepts and New Directions

causes ankle and foot lateral motions to become more variable. This demonstrates that additional neural control of the ankle would be necessary without the forces transmitted by the crural fasci. Hydrostatic Structures Several structures in vertebrates, and many entire bodies in

invertebrates, contain no hard tissues, but instead only muscle surrounding a central fluid-filled cavity. Such systems are called hydrostatic skeletons (Kier 2012). Familiar examples are worms and caterpillars (Trimmer and Lin 2014) and human and turtle penises. Because biological tissues and fluids are mostly water and water is nearly incompressible, contraction of one part of these structures leads to expansion of other parts. Appropriate arrangements of muscles around the central fluid cavity can allow the structure to contract along its length (longitudinal muscles), contract strongly at particular locations along its length (circumferential muscles), or twist (helical muscles). These systems have an essentially infinite number of degrees of freedom (as opposed to musculoskeletal systems, in which the degrees of freedom are constrained to specific joints) and consequently can readily conform to complex, irregular terrains and assume very complex shapes. Models of hydrostatic skeletal structures have provided important insights into the relationship between neural control and biomechanics. An early attempt to model the leech body applied forces to the model elements and determined the equilibrium shape that minimized potential energy (Wadepuhl and Beyn 1989). Kristan and Skalak subsequently developed a model incorporating more details of the passive and active properties of leech musculature, again determining equilibrium shape by minimizing potential energy (Skierczynski et al. 1996). They demonstrated that only patterns of neural activity that were observed in vivo produced model activity resembling that observed in normal animals. Using neural patterns from semi-intact preparations or isolated nerve cords, in both of which sensory feedback is altered or absent, resulted in the model producing abnormal body shapes (Kristan et al. 2000). Cohen and colleagues have developed a neuromechanical model of C. elegans (Boyle et al. 2012) that accounts for transitions from swimming to crawling on the basis of the viscosity of the environment and the effect of changes in viscosity on sensory feedback. These studies illustrate the vital importance of understanding the structural and environmental context in which muscles function. Muscular Hydrostatic Structures Muscular hydrostats (tongues, trunks, tentacles) are struc-

tures completely occupied by muscle with no central fluid-filled cavity, and which thus have hydrostatic skeletons (Kier and Smith 1985). Longitudinal, circumferential, and helical muscles in these structures allow them to generate shortening, lengthening, stiffening, bending, and twisting motions. These complex shapes alter the effect of a muscle’s contraction on the structure based on the geometry of the muscles. For example, shortening the helical fibers while maintaining a constant volume cylinder can either lengthen or shorten the structure, depending on whether the helical muscles are at an angle greater or less than 54∘ , respectively, predictions confirmed by measurements in squid tentacles. Length change speed in muscular hydrostats depends on muscle geometrical arrangement and changes in muscle properties. For example, squid tentacles, which strike and capture prey, can elongate in 20 to 40 ms, reaching peak velocities of 2 m/s and peak accelerations of 250 m/sec2 (Kier and Leeuwen 1997). Comparison of the muscle fibers in the tentacle (which can elongate rapidly) versus the arms (which cannot) suggest that

Muscle, Biomechanics, and Implications for Neural Control

the tentacle’s ability to lengthen rapidly is due to the short lengths and oblique arrangements of tentacle muscle fibers (Kier and Schachat 2008). A computational model of the squid strike (Van Leeuwen and Kier 1997) suggests that the remarkably high velocity of shortening of these fibers is due to the cross-striations of the muscle fibers and their unusually short thick filaments. Kinematics and kinetics must both be considered when analyzing tongue function. The tongue of the lizard, Tupinambus nigropunctatis, consists of two adjacent longitudinal muscles that shorten the tongue wrapped in circumferential muscles that elongate it. A model of tongue function (Chiel et al. 1992) showed that the relative effectiveness of the circumferential and longitudinal muscles was determined both by kinematics (the constant volume constraint) and kinetics (the relative forces in the muscles). The forces generated by each muscle depend on the shape of the muscular hydrostat and determine the forces that protrude and retract the tongue. As a consequence, the effects activating the muscles depend on tongue length. For example, to generate lapping behavior, strong activation of the circumferential muscle combined with low activation of the longitudinal muscle generates a large protrusion, with a small increase of activation of the longitudinal muscle rapidly retracting the tongue. The model also suggested that the low-pass filtering properties and mechanics of the muscles would require significant transformations of neural inputs to create fast lapping movements. In contrast, when animals use pharyngeal tamping to swallow prey, the model predicts that the tongue should be relatively short to make it stiff and able to cope with large mechanical loads. In general, the model demonstrated that, to be properly understood, neural control and mechanics must be analyzed together. 12.3.2 Muscles Are Multi-Functional Different Regions in Single Muscles Can Have Different Functions during Motor Behaviors Because

the structural and contractile properties of single muscles can be highly heterogeneous, a muscle’s motorneuron pool cannot be treated as a single entity. Muscles are comprised of many fiber types and, in addition to their distinct neural innervations, motor units can have different physical arrangements. These different arrangements, acting through both hard and soft-tissue connections, can result in parts of the muscle serving different functions, including generating forces at multiple locations. In vivo observations of muscle activity have revealed that different muscle subregions and compartments can be differentially activated during motor behaviors. For example, the cat biceps femoris is a sheet-like muscle with an origin at a tendinous insertion on the pelvis that spreads out to a thin membrane-like attachment that lies over the musculature of the hip, knee, and ankle (Fig. 12.3). The anterior portion of the muscle consequently produces only hip extension torque, whereas the posterior portion produces both hip extension and knee flexion torque. The internal architecture of muscle fascicles and tendons also varies across the muscle. During behavior, a continuum of activity occurs across the muscle. The anterior region is active in slow walking, with more posterior regions being recruited as speed increases (Chanaud et al. 1991). However, the anterior and posterior regions can have opposite activity patterns, one being excited when the other is inhibited (Chanaud and Macpherson 1991). Thus this single muscular structure can generate a continuum of hip and knee torque combinations. As mentioned above, fascial connections also transmit forces to other

381

382

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Hip

Non-homogeneous fascicle orientation

Anterior Bicep Knee Middle Bicep

Posterior Bicep Crural Fascia

Soft tissue insertion

Ankle

Figure 12.3 Complex muscle and connective structure in cat biceps femoris muscle. The biceps femoris muscle has a “classic” distinct origin on the pelvis. The muscle then fans out in a sheet-like organization and inserts into the soft tissue overlying other muscles. The lines of action of the various muscle fibers are complex because of the differing muscle architecture, but, in general, the joint torques generated when the muscle is activated can be estimated from the moment arms about the joints they span. As a consequence, the anterior biceps has a large moment arm at the hip allowing hip torque generation, but its line of action is coincident with the knee joint center, and so it generates little knee torque. In contrast, the posterior biceps generates hip torque as well as large knee flexion torque because of its large moment arm about the knee. Although the biceps femoris does not span the ankle joint, the crural fascia transmits force from it to the calcaneus bone, and thus it also generates ankle torque.

muscles and joints, contributing to both medial-lateral and sagittal ankle torque (Stahl and Nichols 2014). In more distal muscles such as the gastrocnemius, even more extreme intra-muscle variations in muscle contractile properties, architecture, and tendon arrangements are found (reviewed in Higham and Biewener 2011). Thus, differential activation of muscles can depend on the different torques they generate about joints (Higham and Biewener 2008), but also on different contractile properties and architectural arrangements that interact differently with movement mechanics to facilitate both power generation and absorption (Wakeling 2009; Daley and Biewener 2011). Muscle properties can also contribute to rapid stabilization of the body in the face of perturbation without immediate changes in neural activity (Jindrich and Full 2002; Daley and Biewener 2006; Bunderson et al. 2010).

Muscle, Biomechanics, and Implications for Neural Control

Differential muscle compartment activation is also important in invertebrate muscle. Independent activation of motorneurons projecting to the same muscle can be related to distinct motor functions. For example, the marine mollusk, Aplysia californica, feeds on strips of seaweed by grasping the strip, pulling some into the buccal cavity, releasing the grasper from the seaweed but still holding the seaweed in place in the cavity, repositioning the grasper further along the strip, and then again pulling more into the cavity. To hold the seaweed in place when the grasper is repositioned (protracted) to pull in more, the identified motorneuron, B38, fires to contract only the anterior part of the retractor (I3) muscle (Fig. 12.4). Once the grasper has again closed on the seaweed, other motorneurons fire to contract the entire I3 muscle and thus pull the grasper towards the buccal cavity (retraction) (McManus et al. 2014). Unlike all other I3 motorneurons, B38 is active during protraction rather than retraction. Effectively, the different regions of the muscle are used as if one were pulling a bucket out of a well by a rope and used one hand to pull the rope and the other to hold the rope in place when the first hand was repositioned forward on the rope. Muscle Function Depends on Adjacent and Distant Muscles In addition to the joint torques that

muscles produce about the joints that they span, they can also contribute to accelerations, torque, and power throughout the body. The Newtonian dynamics of a skeleton idealized as a system of interconnected rigid links reveals that a muscle can accelerate all joints in the skeleton because forces are transmitted through the rigid connections between segments (Zajac and Gordon 1989). Thus, acceleration of one skeletal segment accelerates the segments attached to it in a manner dependent on segment inertias and the connections between the segments. The net behavioral effect of a muscle therefore depends on the simultaneous actions of muscle acting at the same and other joints. For example, simulation of the cat hindlimb demonstrated that posterior biceps femoris force either accelerates the toe or, if toe movement is resisted by an external force, generates a force to counter that force (van Antwerp et al. 2007). The direction of toe

Figure 12.4 The dual roles of the I3 muscle increase the efficiency of swallowing. A1) A schematic of the feeding apparatus—the buccal mass—of Aplysia californica (left panel); the plane of section (right panel) indicates the level at which the schematics in A2 and A3 are drawn. The jaw muscles consist primarily of the I3 muscle (dark gray structure with asterisk in left panel; two dark gray parallelogram-like figures in right panel); when the I3 muscles completely contract, they push the grasper (gray ball-like structure to the left in left panel; light gray oval in right panel) towards the esophagus (i.e., the I3 muscles retract the grasper). A2 and A3, b and d) Activation of two I3 motorneurons, B6 and B3, induce the I3 muscle to retract the closed grasper, pulling seaweed into the buccal cavity. A2 and A3, a and c) Activation of B38 during the protraction phase enhances the ingestion of seaweed by pinching the anterior of I3 and allowing I3 to hold the seaweed in place as the grasper protracts for the next swallow. As a consequence, the total inward amount that the seaweed has translated in panel d, A3, is greater than in panel d, A2 (compare right end of seaweed to vertical dashed lines in A2, A3). In all panels in A2 and A3, black arrows are grasper retractions and protractions; open arrows are seaweed movements towards and away from esophagus, and white cylindrical object with black border is the seaweed. Grasper closing on seaweed (b, d in A2, A3) is represented by “pinching” of grasper onto the seaweed. In A2 a, c, even though the grasper is not strongly pinching the seaweed, it is nonetheless somewhat egested during each grasper protraction (short open arrows). Modified from McManus et al. 2014.

383

384

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A1

A2

B6 B3

a

A3 B38

B38

B6 B3

B38

B3

b

B6

B38

B6 B3

b

B38

B3

c

B6

d

B3

a

B6

b

B6

B38

B6 B3

B38

d

B3

B38

Retraction (towards esophagus, ingestion) Protraction (away from esophagus, egestion)

acceleration depends on the torque about the ankle joint, whether achieved actively or passively (Fig. 12.5). By varying ankle torque from zero to the magnitude necessary to immobilize the joint, endpoint acceleration direction can be altered by over 90∘ in some cases. The whole-limb function of identical activations of the biceps femoris muscle therefore depends on the level of activation of ankle muscles. Similarly, power generated by a muscle at one joint often needs to be directed by a muscle at another joint to achieve a functional goal (Zajac 2002; Zajac et al. 2002).

Muscle, Biomechanics, and Implications for Neural Control

Hip posterior biceps femoris (PBF)

PBF + Ankle free

Knee

PBF + Ankle immobilized

PBF + Ankle free

Ankle torque 0

PBF + Ankle immobilized

100%

Figure 12.5 Direction of toe acceleration induced by activation of posterior biceps femoris (PBF) muscle differs across varying levels of ankle torque. In a musculoskeletal model of the cat hindlimb, the PBF was maximally activated simultaneously with ankle flexion/extension and ad/abduction torques that canceled 0 to 100% of their induced acceleration at the ankle, respectively. Total ankle torque levels required to functionally immobilize the joint were relatively low, and never exceeded 20% of the muscle torques produced by the PBF. Toe acceleration directions varied by over 90∘ in both the sagittal and dorsal planes. These results demonstrate how multi-muscle coordination or even the contributions of passive joint torques can dramatically alter the endpoint action of the limb when muscles are activated. Adapted from van Antwerp et al. 2007 with permission.

In soft tissue structures, expansion or contraction of one muscle can significantly affect surrounding muscles, as was described above for muscular hydrostats. For instance, changing one muscle’s shape can alter the mechanical advantage of other muscles. A model of the Aplysia feeding grasper predicted that as the grasper closed, its shape would elongate and stretch the thin protractor muscle (I2) in the posterior of the feeding apparatus. This stretch would change I2’s position on its length-tension curve, and its mechanical advantage, enhancing its ability to produce protraction (Novakovic et al. 2006). Experimental tests verified this prediction (Ye et al. 2006b) (Fig. 12.6). An important implication of these results is that activity of the grasper closer motorneurons (the B8a/b motorneurons) may strongly modulate the forces protractor motorneurons (B31/B32 and B61/B62) produce in I2 by altering I2 mechanics in the periphery. 12.3.3 Specialization of Biomechanical Structures Reflect Behavioral Repertoire

The data presented above suggest that the degree of specialization in organismal structures likely depends on behavioral context and repertoire. This observation allows us to expand our earlier principle to the degree of specialization of muscle architecture and contractile properties may depend on the specialization of the biomechanical structure that contains it. Structures that produce a limited repertoire of highly stereotyped and rapid motions are more likely to be associated with highly specialized muscle properties and biomechanical structures. In multi-functional biomechanical structures

385

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B

Distance (mm)

Angle (degrees)

50

50

40

40

30

30

20

20

10

10

0

0

10

10

8

8

6

6

4

4

2

2

0

0

1

2

3

0

4

Base of radular cleft

Radular surface

Prow Radular stalk Odontophore

Spherical shape

0

1

2

3

4

20 Hz, 4 s I2 stimulation

10 Hz, 3 s I2 stimulation

Time (s)

Angle (degrees)

C

Distance (mm)

386

D

50

50

40

40

30

30

20

20

10

10

0

0

10

10

8

8

6

6

Elongated shape

4

4

2

2 0

rotation

0 0

1

2

3

10 Hz, 3 s I2 stimulation

4

0

1

2

3

4

20 Hz, 4 s I2 stimulation Time (s)

Figure 12.6 The shape of the grasper in Aplysia californica determines the effectiveness of the I2 protractor muscle. (A) When the grasper is open, it assumes a spherical shape. If I2 is activated at 10 Hz for 3 seconds, some translation but no rotation of the grasper is observed. (B) Higher and longer activation (20 Hz for 4 s) induces stronger translation, but no rotation, of the grasper. (C) When the grasper is closed, it assumes an elongated shape. If I2 is now activated at 10 Hz for 3 seconds, the resulting grasper translation is as large as that obtained in panel B when I2 was stimulated at twice the frequency and for a longer duration. Elongating the grasper thus significantly enhanced I2’s ability to protract the grasper. (D) When the grasper is closed and I2 is activated at 20 Hz for 4 s, the resulting translation is much stronger and the grasper also rotates substantially. From Ye et al. 2006b with permission.

Muscle, Biomechanics, and Implications for Neural Control

that produce a wide range of behaviors (slow, fast, multiple movement types such as different gaits or reaching trajectories), alternatively, muscle is likely less specialized, allowing more flexibility in the ways it can be used by the neural control system to produce this behavioral generality. For example, the squid tentacle, specialized for extremely rapid striking, has unusual features: short thick filaments, oblique arrangements of muscle fibers, a precise arrangement of the muscles that elongate and shorten the tentacle. In contrast, in the same animals, the arms, which are used for grasping and manipulating objects, show none of these specializations. Similar specialization of muscle properties is observed in mantis shrimp depending on whether prey capture involves spearing or smashing (Blanco and Patek 2014). In terrestrial runners, variable muscle architecture specialization is present along the proximal–distal axis of the limb (Biewener and Daley 2007). To help mediate interactions with unpredictable variations in terrain, distal leg muscles have long elastic tendons with short muscle fibers; furthermore, the muscles tend to contract isometrically, allowing the tendon to absorb and return energy upon impact. More proximal leg muscles tend to have long fibers for operating at high velocities (Biewener and Daley 2007; Lieber and Ward 2011; Wilson and Lichtwark 2011). Similar specialization of leg muscles for posture versus locomotion is evident in the differences in the contractile properties in two ankle extensors: the gastrocnemius, which is almost all fast muscle, versus the soleus, which is all slow (Burke 1981; Kaya et al. 2003). Moreover, the weight-bearing ankle muscles have increased contraction relaxation time constants compared to ankle dorsiflexors and arm muscles (Burke 1981; Belanger and McComas 1985).

12.4 Biomechanics Defines Meaningful Patterns of Neural Activity What patterns of neural activity are needed to generate a particular motor behavior? We have focused above on how behavioral context determines the mechanical effects of single muscles. We focus here on the coordination of multiple muscles to generate functional movements and address the broad question of motor control from the perspective of biomechanics. We do not attempt a comprehensive review, but provide an introduction to some of the key questions currently under study. How do nervous systems harness the complex interactions between the motor system and the environment to produce flexible and robust motor outputs? In the previous sections, we discussed how identical patterns of neural activity or muscle activation can generate qualitatively different motor outputs (e.g., because of differences in body posture or environmental forces). But within a given context, one must also understand how ensembles of motorneurons can be activated in spatial and temporal patterns to produce meaningful behavior. One must also understand the many ways in which different patterns of neural activity or muscle activation can generate essentially identical motor outputs. With this understanding, one can begin to analyze variability in and between individuals, and the ways in which sensory feedback, exploration, and learning shape the variability in and across individuals to generate effective behavior (Ting et al. 2015).

387

388

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A classical approach to studying motor behavior is to identify a canonical pattern of muscle activity that produces the forces and torques necessary to achieve a specific movement. The use of biomechanical models and optimization has been extremely important in identifying efficient patterns of movement that often resemble those observed experimentally. However, these approaches are based on three assumptions that do not account for the diversity of motor solutions and individual differences observed experimentally. First, classical approaches assume there is a “correct” or “optimal” way of generating a movement, with optimality based on minimizing movement time, energy, or some other feature (Todorov and Jordan 2002; Scott 2004, 2008; Todorov 2004; Shadmehr and Krakauer 2008). However, biomechanical redundancy provides an abundance of motor solutions that may all be “good enough” to produce a desired action rapidly and robustly. These solutions may only be locally optimal, reflecting motor experience and competing costs and constraints (Loeb 2012; Tiel et al. 2015). Second, computing the “correct” trajectory requires an internal model or representation of the periphery for the nervous system to use. Finding the solution involves a computation of inverse kinematics or dynamics (see Chapter 11), similar to the computations needed to direct a robot limb (see Chapter 14). However, exploration and directed search guided by prior experience, biomechanical affordances, and the immediately prior movement, rather than exhaustive computation, may underlie motor movements and motor learning (Smith and Thelen 2003; Huang et al. 2008; Loeb 2012; Herzfeld and Shadmehr 2014; Wu et al. 2014). Third, it is also assumed that complex biomechanical computations are done quickly and efficiently and are readily instantiated by neural circuits. Recent studies, however, suggest that subjects prefer to use habitual movement patterns (Cohen and Sternad 2009) even when they have experienced more “optimal” ones (Ganesh et al. 2010; de Rugy et al. 2012). Optimization approaches consider motor redundancy to be a problem for which a single solution must be found. In contrast, more recent approaches consider biomechanical motor abundance as providing a rich source of variation in how movements are learned, remembered, and recalled. Motor exploration and variability are considered essential to discovering novel useful movement patterns and do not necessarily follow rules of engineering approaches (Smith and Thelen 2003; Huang et al. 2008; Loeb 2012; Herzfeld and Shadmehr 2014; Wu et al. 2014). Moreover, when animals must rapidly respond, as when tracking prey or avoiding predators, there is very little time for elaborate computation. Rapid and reliable responses of “good enough” solutions allow animals to transition rapidly and seamlessly among different behaviors. Movement pattern variation is also observed in “don’t care” regions that do not affect critical aspects of motor performance. Individual patterns or styles of movement may emerge from interactions between the large motor solution space and individual differences in physical properties (e.g., size variations among animals of the same species) or how the motor space is explored (Loeb 2012; Furuya and Altenmuller 2013; Cullins et al. 2014). Indeed, sensory feedback may act to regulate the level of variability as an aid to motor movement. A recent study demonstrated that, in the presence of sensory feedback, animals varied less from one another, but showed more variation within their own behavior, suggesting that all of them were moved into a common solution space (Cullins et al. 2015; Hooper 2015).

Muscle, Biomechanics, and Implications for Neural Control

12.4.1 Organismal Structures Are Multi-Functional Coordination of Multiple Muscles Reconfigures the Body to Produce Different Behaviors Limb and

body multi-functionality is critical to create a large motor repertoire, but requires understanding how muscles work together to generate movements. The ability to rapidly reconfigure a peripheral structure, flexibly adjusting motor responses as the environment changes, may confer selective advantages on animals. How are the same set of muscles and structures coordinated in different ways to produce different behaviors? Because, as discussed above, a muscle’s function cannot be determined in isolation, movements cannot be constructed simply by adding the individual actions of each muscle. The coordinated actions of muscles over space and time instead need to be considered holistically. In thinking of the ways that muscles can work together to generate different actions of a limb or the body, the number of possible behaviors is immense. If one considers just simple on/off combinations of muscle activation across n muscles, one obtains 2n possible joint torque patterns. This number of possibilities increases dramatically when one allows for different levels and timings of muscle activation, and becomes extremely large when the effects of changes in posture, environment, and movement that can modulate muscle function are factored in. The potential behavioral repertoire that a set of muscle can generate is thus much greater than the total number of muscles or even motor units. In these complex systems, biomechanical constraints and affordances may reduce the set of viable combinations of muscle activity. For example, a neural pattern that would force the knee past its range of motion is clearly not functional. Certain patterns may harness the intrinsic biomechanical dynamics of the system and require less energy to produce, such as passive dynamic walking, and thus presumably be selected for. However, even with biomechanical constraints, the number of possible muscle coordination patterns that can achieve even a simple isometric force are extremely high (Bunderson et al. 2008, 2010). This number becomes much greater in more complex motor activities: it is possible to walk with many different gaits: slow, quick, smooth, sideways, skipping, backwards, or lurching (Cleese 1970). All these very different gaits, for each of which there will be many “good enough” patterns of muscle coordination, will reflect the biomechanical dynamics of the limb and environment. Indeed, a variety of multi-muscle patterns are recruited for walking and balance in animals and humans (Collins 1995; Ivanenko et al. 2004; Ting and Macpherson 2005; Cappellini et al. 2006; Torres-Oviedo et al. 2006; Clark et al. 2010; Yakovenko et al. 2011; Zelik et al. 2014). Multi-functionality is not limited to muscle coactivation, but also applies to the timing and sequence of muscle activation. For example, the Aplysia feeding systems uses different muscle coordination patterns in various forms of swallowing and in rejection. During weak swallows, a grasper closer muscle (I4) acts purely to hold the food as other muscles push the grasper back into the buccal cavity; during stronger swallows, because of a change in grasper position, I4 both grasps food and pulls it inward, and a second muscle (the hinge) plays a critical role in grasper retraction (Ye et al. 2006a). During strong rejections, alternatively, the hinge muscle rotates the grasper ventrally and then dorsally; both capabilities are “unmasked” by a larger protraction phase (Ye et al. 2006b). These data suggest that multi-functionality emerges from mechanical structures in which flexible coalitions of muscles perform different functions in different

389

390

Neurobiology of Motor Control: Fundamental Concepts and New Directions

mechanical contexts and from neural circuitry that reorganizes itself to exploit these coalitions by changes in phasing, duration, and intensity of motorneuron activation. Multi-Functionality May Be Mediated by Motor Modules Multiple lines of evidence indicate

that muscle activation patterns at a given instant in time are constrained, reflecting consistent motor patterns that produce meaningful motor outputs (Ting et al. 2015). Several computational methods have demonstrated consistent structure underlying muscle activation patterns across multiple muscles and motor behaviors (d’Avella et al. 2003; Giszter et al. 2007; Ting 2007; Ting and McKay 2007; Bizzi et al. 2008). Using signal processing methods, such as principal components analysis, independent components analysis, and nonnegative matrix factorization (Lee and Seung 1999; Tresch et al. 1999, 2006; Ting and Chvatal 2010), motor signals can be decomposed into underlying motor modules, or muscle synergies that reflect consistent patterns of multi-muscle coordination that generate specific actions (see also Chapter 11). Motor modules have been associated with biomechanical outputs in both experimental and modeling studies (Neptune et al. 2009; Clark et al. 2010; Chvatal et al. 2011; Allen and Neptune 2012; Ting and Macpherson 2005; Safavynia and Ting 2013), and can be recruited across a variety of motor behaviors (Tresch et al. 1999; Torres-Oviedo et al. 2006; Chvatal et al. 2011; Roh et al. 2011; Chvatal and Ting 2013) suggesting that they form a repertoire of whole limb actions for movement (Fig. 12.7). However, this underlying structure does not mean that actions are necessarily stereotyped. Motor modules may facilitate rapid adaptation by allowing meaningful motor actions to be flexibly combined, producing a wide range of different muscle activation patterns for movement. Variability observed across different types of behaviors, and trial-by-trial variability, can be accounted for by varying combinations of motor modules (Fig. 12.7) (Tresch et al. 1999; Hart and Giszter 2004; Cheung et al. 2005; Torres-Oviedo and Ting 2007; Roh et al. 2011). Variability across instances of movement may thus reflect differences in descending drive to stored movement patterns (Churchland et al. 2006) that could facilitate motor exploration (Huang et al. 2008; Wu et al. 2014) rather than random noise in individual muscles or trajectories. Indeed, learning to perform novel tasks is faster if it can be achieved by altering the recruitment of motor modules versus requiring activity incompatible with motor module coordination (Berger et al. 2013). The computational methods used to analyze motor patterns have many limitations and may not be easily instantiated as neural mechanisms (Tresch and Jarc 2009; Ting and Chvatal 2010; Burkholder and van Antwerp 2013; Steele et al. 2013; Zelik et al. 2014), although they may still be useful in understanding motor coordination. Consistent structures in motor patterns are a hallmark of coordinated movement, whether they arise from specific neural structures (Saltiel et al. 2001; Lemay and Grill 2004; Hart and Giszter 2010; Overduin et al. 2012), reflect optimal coordination of biomechanics (Li et al. 2005; Kurtzer et al. 2006; Berniker et al. 2009; Kargo et al. 2010; Kutch and Valero-Cuevas 2012; Steele et al. 2013) or emerge from complex neural and biomechanical interactions (Ting and McKay 2007; Ting et al. 2009; McKay and Ting 2012; Giszter and Hart 2013). The necessity of appropriate structure in neuromechanical interaction is highlighted in motor disorders, where current techniques can be used to evaluate different types and potential mechanisms of impairment (Cheung et al. 2009, 2012; Safavynia et al. 2011;

Muscle, Biomechanics, and Implications for Neural Control

Motor Modules

Possible Biomechanical Functions

Descending Commands

weight acceptance propulsion flexion

Descending commands

leg swing m1 m2 m3 m4 m5 m6 m7

time

m1

Motor Modules

m2 m2 m4

Individual Muscle Activations

Neuro m5 m7 m6

Mechanics

Figure 12.7 Multi-functionality of musculoskeletal systems is facilitated by motor modules. Motor modules reflect consistent, individual-specific patterns of multi-muscle activation that produce actions necessary for behavior. Each motor module can be recruited as a unit to modulate a given biomechanical output such as weight support, propulsion, and limb flexion. For example, the timing and amplitude of the recruitment, or descending command, to each module can be varied to generate a continuum of different muscle activation patterns appropriate for different gait speeds, as well as responses to perturbation during human walking.

Giszter and Hart 2013; Roh et al. 2013; Ting et al. 2015). For example, impairments in descending corticospinal drive in stroke are associated with a merging of motor modules in hemiplegic gait and a reduced ability to modulate motor module recruitment that limits leg motor capacity (Clark et al. 2010; Allen et al. 2013, 2014; Routson et al. 2014). In contrast, after spinal cord injury, motor patterns can lose all structure and become indistinguishable from random variation (Chvatal et al. 2013). In both cases, where motor patterns do exist, they may not be modulated appropriately in response to sensory feedback (Hayes et al. 2014; Routson et al. 2014). Motor modules imply a reduction in dimensionality, and thus in the number of possible patterns. However, if one considers the very large number of different tasks that an animal or human may engage in over a lifetime, many different patterns are required (Zelik et al. 2014). It is thus not surprising that the number of possible muscle coordination patterns may far exceed the number of muscles (Chiel et al. 2009). This

391

392

Neurobiology of Motor Control: Fundamental Concepts and New Directions

combinatorial problem is also present in the neural coding of sensory information, in which large sets of so-called overcomplete representations are proposed to account for the ability to recognize visual features (Olshausen and Field 2004). As all existing algorithms necessarily reduce dimensionality, they are inherently limited in their ability to clarify the full complexity of motor systems. Biomechanical models can be useful in identifying motor patterns that are compatible with and harness body affordances. Motor modules that reflect the natural dynamics of the peripheral motor system are similar to those measured experimentally and can be used to reproduce essential features of movement in frog leg (Berniker et al. 2009) and to drive simulations of human walking (Neptune et al. 2009; Allen and Neptune 2012; Allen et al. 2013). As a consequence, modules reflect an interaction between the neural and motor systems, and often align with coordination patterns that optimize energetic efficiency given biomechanical constraints (McKay and Ting 2012; Steele et al. 2013). Nonetheless, even among such solutions, variations in patterns exist. Including individual-specific motor structure constrains coordination patterns from among the many possible, and improves the accuracy of computer simulations of movement (Walter et al. 2014). 12.4.2 Many Functionally-Equivalent Solutions Exist for Sensorimotor Tasks

For any given motor behavior, there is an abundance of ways in which a wide variety of motor commands can generate similar or functionally-equivalent behaviors. Substantial redundancy exists at many levels, whether one examines joint torques, movement kinematics, or different behaviors that achieve the same goal (see also Chapter 11). While redundancy is usually considered a “problem” from the perspective of reverse-engineering a solution, the ability to choose from many solutions underlies the adaptability and robustness of biological systems. Indeed, the concept of motor abundance (Latash 2012) is one that should be celebrated and used to understand the many different ways in which variation in movement solutions (Loeb 2012; Ting et al. 2015) and variability in movements arises (Scholz and Schoner 1999; Scholz et al. 2000; Valero-Cuevas et al. 2009). Within these “motor equivalent” solutions, there may be some that are less desirable than others for any number of reasons, including energetics, stability, and generalizability across tasks. There may be solutions that are not optimal, but “good enough” to achieve the motor function. In the sections that follow, we provide evidence suggesting that biomechanics shapes the feasible ranges for variation in motor commands that generate a given motor task. Motor Equivalents Exist at Many Levels of Organization Even at the level of generating joint

torques, substantial variations in possible muscle activation patterns exist. Consider the problem of generating an isometric force with a simple arm model (Fig. 12.8). Even with only two antagonistic muscles spanning the joint, there are nonetheless an infinite number of muscle activation patterns that will generate a given magnitude of joint flexion torque at the elbow. A classical approach to find the minimum energy solution would predict that only the flexor muscle (m1 ) would be activated, and its activation would increase with endpoint force magnitude (Fig. 12.8 m1 plot, lower line labeled “Necessary”). If both muscles are activated, a feasible range of activity for each muscle can be found (shaded gray areas in m1 and m2 plots); the upper limit is limited by the

Muscle, Biomechanics, and Implications for Neural Control

m1

y ar

0

elbow

0 0

0.5 Normalized Force Magnitude

1

d

ss ce e N

em2

C

1

ined

tra ons

ine

Muscle Excitation em1

1

a str

m2

F

n Co

m1

m2 Unconstrained

Optional

0

0.5

1

Normalized Force Magnitude

Figure 12.8 Muscle feasible ranges determine biomechanical constraints on goal-equivalent muscle activity. In a simple example of an elbow joint with an antagonistic muscle pair—flexor m1 and extensor m2 —the muscle activation level that can generate a force vector at the hand is illustrated. The line marked “Necessary” in the m1 plot shows the minimum level of activation (that is, m1 ’s activation when m2 ’s activation is zero) necessary to generate normalized forces from 0 to 1. When producing any given level of normalized force (the “x”s on the plots show an example for producing a normalized force of 0.5) except for maximum normalized force, for which m1 excitation must be 1 and m2 excitation must be 0 (open circles on m1 and m2 plots), the normalized force can be produced across a range of m2 activations by activating m1 to a greater degree so as to cancel the opposing elbow torque caused by m2 . In some cases, such co-activation of m1 and m2 may be desirable to provide limb stability. The upper limit of m1 excitation (line marked “Constrained” on m1 plot) is bounded by m2 ’s strength, which in this case generates much less elbow torque. For example, to produce a normalized force of 0.5, m1 amplitude varies along the dashed line in the m1 plot as m2 amplitude varies along the dashed line in the m2 plot; the x’s in the m1 plot show what m1 amplitude must be to counteract the m2 amplitudes marked by the x’s in the m2 plot (upper m1 plot x corresponds to upper m2 plot x; middle m1 to middle m2 ; bottom m1 to middle m2 ). Because m2 is the weaker muscle, for all normalized force values except 1 (for which m2 excitation must be zero), m2 has a much wider possible excitation range than m1 . Only at maximal force is there a unique solution to the force generation problem, although this need not always be the case, particularly when there are multiple agonist and antagonists crossing a joint (Sohn et al. 2013).

strength of the opposing muscle. In this example, agonist (m1 ) variations are relatively small at any given normalized force level, whereas the antagonist muscle (m2 ) can be activated from 0 to 1 at all normalized force levels ≤0.5. Only at the maximal force (open circles on m1 and m2 plots) can a unique solution for each muscle be defined. Using such techniques shows that the variation in muscle activity for isometric force production in the finger is relatively constrained, allowing little variability (Valero-Cuevas et al. 1998; Kutch and Valero-Cuevas 2012), whereas the possible variation is much greater in cat hindlimb (Sohn et al. 2013) and in human walking (Simpson et al. 2015). These appear to match the variability in muscle activity measured experimentally. However, different patterns that produce the same force may endow the limb with other characteristics that may or may not matter to the movement, such as limb stability (Franklin et al. 2004; Bunderson et al. 2008; Sohn and Ting 2013). Considering a movement trajectory over time, Bernstein (1967) realized that the range of possible forces that could produce the trajectory depended on the initial conditions of the movement (Fig. 12.9). Variations in range indicate time points where the forces generating the movement could be highly variable. As predicted, different patterns of joint torques have been shown to produce similar kinematic outputs in arm movement (Gottlieb et al. 1995), and substantial variations in muscle activity that deviate from an optimal pattern have been observed in many movements, including human walking.

393

394

Neurobiology of Motor Control: Fundamental Concepts and New Directions

force

time

Figure 12.9 Range of possible forces that will generate a given kinematic trajectory. At only two times (asterisks) is a unique force required. Measurement of kinematic outputs thus does not uniquely define the forces necessary to generate the task. This lack of unique definition arises because forces must be integrated twice to determine the displacement of an organism or body and thus the required force will, in general, vary as a result of the prior displacement and velocity of the system. Further, due to biomechanical constraints and affordances, there may be times during a movement in which force variations have no impact on motor output, and thus can assume any value the muscles can produce. Figure is new but based on a classic, analogous figure in Bernstein (1967).

At other times, when the input force must be very precise, there is no room for variability. In a modeling study of single-legged locomotion, lowering and pushing backward the leg needed to be precisely coordinated to generate efficient locomotion, creating a biomechanical “bottleneck”. The highest fitness pattern generators for this model found by a genetic algorithm consequently produced precise timing at this phase of the pattern. In contrast, late in stance, the model leg continued to move backwards but could no longer exert force. During this “don’t care” region of the behavior, evolved pattern generators showed high variability (Beer et al. 1999). Ultimately, the motor goal of an animal can rarely be characterized by a single kinematic or kinetic pattern. Rather, the fundamental question is “Did the animal meet a functionally-relevant goal?” Furthermore, movement kinematics and kinetics may vary and yet have the same adaptive fitness. If a squirrel monkey succeeds in plucking a breadfruit from various vantage points while climbing a tree, very different hand trajectories or force profiles may be adaptively equivalent. Limb stability may be increased by coactivating muscles (Franklin et al. 2004; Selen et al. 2009), or by changing arm configuration (Trumbower et al. 2009). Variations in responses and individuals have been found in animals ranging from insects (Hooper et al. 2006) to humans (Nussbaum and Chaffin 1997; Borzelli et al. 1999; Welch and Ting 2008; Torres-Oviedo and Ting 2010). People can walk with different gaits or recover balance using a wide range of strategies (Macpherson and Horak 2013). 12.4.3 Structure and Variability in Motor Patterns Reflect Biomechanics

A hallmark of biological behaviors is that they vary. Sources of noise that may cause variability exist at all levels of the neuromuscular transform. These include stochastic processes associated with synaptic communication between neurons, variation in neuron and muscle properties, and the effect of body mechanics and environmental forces discussed earlier. As a consequence, movements vary at every level studied: at the level of motor signals within a given motor task, during repetitions of the same task in the same

Muscle, Biomechanics, and Implications for Neural Control

subject, across motor tasks, and across individuals. Such variations have been observed in insects, vertebrates, and humans, including those highly trained in movement such as musicians and athletes. But variability in motor control is not random. Several different approaches use a biomechanical perspective to understand the structure and variability of motor behaviors. The goals of a task, and the biomechanical affordances and constraints defined by body structure, environment, and behavioral context all play a role in shaping variability. The effect of variations in muscle activity on motor output highly depends on biomechanical interactions. Many different biomechanical approaches have been used to understand “don’t care” regions of motor variability where there is little to no effect on motor output, versus defining sets of “task equivalent” outputs that all achieve the same goal “well enough” because their differences make little difference in task fulfillment. Taking an evolutionary perspective, how one learns to move is based on innate motor mechanisms, prior motor experience, and different motor goals. Not surprisingly, then, the need to rapidly recall reliable motor actions may lead to the individual movement styles that are found in invertebrates and vertebrates alike. We will continue to use the simple example in Fig. 12.9 of combining two forces to generate a total net force to explain some key ideas and theories in the field related to variability, movement strategies, motor learning, and the role of experience in individual differences in movement. Assume two equivalent muscles that generate forces F1 and F2 and sum to generate a total desired force. To generate 4 N of force, any solution satisfying the equation F1 + F2 = 4 N is acceptable (Fig. 12.10, bold diagonal line). Thus, one muscle can be activated to 4 N while the other is kept off, or they can share the load, with the symmetric case being each muscle producing 2 N. This example can be directly mapped to questions of how muscles spanning the same joint should be activated to generate a desired joint torque (Herzog and Leonard 1991), or how two fingers together generate forces on an object (Latash et al. 2001; Scholz et al. 2002). However, it is also useful to think about this example in the context of controlling multiple muscles as described above, where muscle interaction might vary non-linearly, and is subject to movement context, environmental forces, and so on. The Structure in Variability Depends on Biomechanical Task Relevance Evidence suggests that

the nervous system allows greater variability in task-irrelevant “don’t care” dimensions, where large fluctuations in muscle activity do not affect motor goals. In our example, each of the two forces may fluctuate over time with similar dynamics, e.g., with a mean value of 2 N and variance of ±1 N (Fig. 12.10, bars along axes). Total force will thus also fluctuate such that Ftot = (F1 + ΔF1 ) + (F2 + ΔF2 ). If the fluctuations in F1 and F2 are independent, then the fluctuations in total force should be the vector sum of the √

F1 and F2 fluctuations, ΔFtot = ΔF12 + ΔF22 , and the Ftot fluctuations will appear as a circle when F1 and F2 are plotted against each other. Nonetheless, if the fluctuations in F1 and F2 are structured such that F1 increases as F2 decreases, i.e., ΔF1 + ΔF2 = 0, then the goal of generating Ftot = 4 N force can be perfectly maintained even as F1 and F2 fluctuate. Thus, different combinations of F1 and F2 , along the diagonal line of equivalent solutions, can be used. If the fluctuations in F1 and F2 do not precisely cancel each other out, then the variability in Ftot will be an ellipsoid oriented along the manifold

395

Force2

Neurobiology of Motor Control: Fundamental Concepts and New Directions

5N

Isoline for Total Force Output

Task-irrelevant Variability

4N

F2 variability

396

Task-relevant Variability

3N

2N

S1

1N S2

F1 variability

Force1

Figure 12.10 Simple example to illustrate different types of motor variability structures using a system where the total desired force Ftot = F1 + F2 . Due to noise in biological systems, the levels of F1 and F2 can vary, in a range indicated by the bars along each axis. Dashed diagonal lines indicate combinations of F1 and F2 that generate a consistent level of Ftot , and denote goal-equivalent manifolds, “don’t care” dimensions, where fluctuations in F1 and F2 are task-irrelevant, i.e., do not change the total force. If F1 and F2 vary independently, the fluctuations will lie in the gray circular area, and total force will vary as a result. Structuring the fluctuations in F1 and F2 to reduce Ftot variation (i.e., making F1 + F2 ≈ 0) results in greater variation along the task-irrelevant direction. In changing total force level, many different combinations of F1 and F2 can be selected (arrows going from 4 N isoline to 5 N isoline). The solution an individual chooses may depend on the initial state of force generation, e.g., S1 and S2, and a preferred ratio of changes in F1 and F2 used to increase Ftot (lines with different slopes leaving S1 and S2).

of equivalent solutions, with the variability being greater in the task-irrelevant versus task-relevant direction. There are several theories that use biomechanics to explain structured variability in motor control, such as the uncontrolled manifold, theory of minimum intervention, and optimal feedback control (Bernstein 1967; Scholz and Schoner 1999; Todorov and Jordan 2002). Lower variability in task-relevant versus task-irrelevant dimension are predicted based on the idea that the nervous system only makes corrections to motor outputs that are relevant to task goals. Task-relevant and task-irrelevant variations can be identified in situations where explicit models of the biomechanical relationships between the goal and the measured components can be generated. Structured variability has been identified at the level of joint angles, joint torques, and muscles in a range of different experimental tasks. For example, the variability of finger location in space during pointing depends on the variability of arm joint angles and arm geometry and comparing the measured variability of finger location to that predicted by uncorrelated variations at each joint allows the degree of structure in the variability to be measured (Scholz et al. 2000). Leg joint torques similarly demonstrate structured variability in maintaining total vertical ground-reaction force in hopping (Yen et al. 2009). Muscle activity during isometric force production by the index finger also exhibits structured variability in maintaining endpoint force (Valero-Cuevas et al. 2009).

Muscle, Biomechanics, and Implications for Neural Control

The computational metrics used may or may not reflect the actual goals and processes used by the nervous system. Representing variables in coordinate frames and units is useful for our understanding, but can influence the interpretation of variability (Sternad et al. 2010). Biomechanical models may not include all the constraints or affordances relevant to the behavior. However, the insights gained have real relevance to solving motor control problems. For example, in tennis serves, training that focuses on tight control of certain critical features of the movement, while allowing variability in others, is more effective than methods that emphasize consistency and repeatable movements (Handford 2006). Overconstraining movements in “don’t care” regions may actually be detrimental to expert performance. Learning Is Simplified by an Abundance of Equivalent Motor Solutions Biomechanical redun-

dancy also predicts that many different patterns in motor coordination can equivalently achieve the same task-level goal. In our example, consider the goal of increasing total force from 4 N to 5 N. In this case, the changes in F1 and F2 must be coordinated to achieve a net increase of 1 N. The most direct path is to increase forces perpendicular to the manifold of equivalent solutions, thus increasing both F1 and F2 by 0.5 N so that ΔF1 + ΔF2 = 1 N. This solution would be predicted by a minimum muscle stress criteria in which maximal muscle forces are avoided (Crowninshield and Brand 1981), or by minimizing signal-dependent noise in muscles (Harris and Wolpert 1998). However, F1 and F2 can be coordinated along any line that intersects with the 5 N solution manifold. In the extreme, either F1 or F2 would increase by 1 N, but all intermediate solutions are also viable (Fig. 12.10 arrows), each achieving 5 N in a slightly different way. How does one select a particular “good enough” solution to achieve a task? Given the need to coordinate multiple muscles to control a limb, very specific coordination structures may be necessary to reliably move a limb in desired directions or to achieve a given force level (Ting and McKay 2007; Ting et al. 2015). Recent work suggests that movement variability is essential to motor learning (Huang et al. 2008; Shadmehr et al. 2010; Herzfeld and Shadmehr 2014; Wu et al. 2014), as it may help individuals explore the landscape of possible movement patterns (Loeb 2012). The starting point of an individual and the exploration and refinement process can all affect the pattern an individual ultimately selects. In our example, consider two individuals, S1 and S2 , who use different F1 and F2 combinations to generate 4 N. In generating 5 N, a random search strategy would most likely result in solutions close to the starting point (Fig. 12.10). Because S2 ’s solution for generating 4 N relies on high levels of F1 and low levels of F2 , S2 ’s solution for generating 5 N would likely rely more on F1 as well. In contrast, S1 relies more equally on F1 and F2 . This example demonstrates how S1 and S2 may end up reaching the same solution starting from different initial conditions and using different movement strategies. Evolutionary, developmental, and learning processes help identify “good enough” solutions that are critical for animal survival (Lacquaniti et al. 2013). Rather than searching exhaustively for optimal solutions, animals must rapidly and reliably generate movements in novel situations. Default movement patterns are established in the embryonic stage, where spontaneous motor activity such as kicking and flailing are observed (Bekoff 2001). These patterns, available at birth, rapidly adapt, allowing a fawn, for example, to run minutes after birth. Human infants are born with the capacity for stepping and kicking (Yang et al. 2004), and, through exploration

397

398

Neurobiology of Motor Control: Fundamental Concepts and New Directions

(Smith and Thelen 2003), existing movement patterns are refined and new ones created throughout development (Dominici et al. 2011), along with the ability to recruit them in task-specific manners (Angulo-Kinzler et al. 2002). Models of spinal circuitry and biomechanics suggest that stable “good enough” solutions can be found in just a few iterations of random searching (Tsianos et al. 2014). Movement strategies in adults may appear to rapidly optimize because they have been refined over time. Data suggest, however, that subjects do not optimize “online” but instead employ a range of different strategies leading to suboptimal performance (Ganesh et al. 2010; Snaterse et al. 2011; de Rugy et al. 2012). Similar challenges are faced in sports or the classroom in which stable, but non-optimal, and difficult-to-change, solutions are often learned (Chi and Roscoe 2002; Handford 2006). Individuals Have Their Own Motor Styles The effects of movement history and experience

lead to individual differences in movement and movement styles (Ting et al. 2015). A general principle is that individual—not averaged—solutions solve neuromotor problems (see also Fig. 5.1). In both invertebrates and vertebrates (including humans), individuals may have their own “motor program styles”, i.e., they show significant individual variations in motor outputs that are both consistent within a given animal and differ from one individual to another (Golowasch et al. 2002; Prinz et al. 2004; Marder and Goaillard 2006; Calabrese et al. 2011; Nussbaum and Chaffin 1997; Borzelli et al. 1999; Welch and Ting 2008; Torres-Oviedo and Ting 2010), although this variability can be reduced by training and is less in motor output components critical to producing the behavior. For example, when all feeding-related motorneurons in Aplysia are examined, the distribution of the durations of motorneuron activations varies across individuals (Fig. 12.11). However, the distributions of the motorneurons that play a critical role in feeding are similar across individuals (Cullins et al. 2014). In humans, regulation of individual temporal variability in motor output enhances learning speed (Wu et al. 2014). Individual variation in the temporal patterns of motor response in reactive balance varies systematically in humans naive to the task (Welch and Ting 2008), but becomes smaller and closer to an energetically optimal solution in cats trained daily on the same task (Lockhart and Ting 2007). Individual-specific patterns of muscle activity associated with generating leg forces are also found in both cats (Torres-Oviedo et al. 2006, Fig. 12.12) and humans (Torres-Oviedo and Ting 2010). Consideration of individual movement patterns (Ting et al. 2012) also improves simulations of human movement (Walter et al. 2014). Experience, social learning, and training play a strong role in shaping individual movement styles. Observing the actions of others activates neural circuitry similar to that used in self motion—so-called “mirror neurons”—and may help animals learn faster through mimicry (Rizzolatti and Strick 2013). Different movement patterns for grasping can be identified in musicians, shaped by their specific training (Gentner et al. 2010), and different musicians display different movement styles (Furuya and Altenmuller 2013). Context and interaction define how humans produce and perceive language, resulting in characteristic speech sounds specific to different languages, cultures, and individuals (Kuhl 2004). Variability in movement styles may be subject to evolutionary selection. In Drosophila melanogaster, fast and slow larval feeding rates can be inherited, and adults from slower feeding larva live longer than adults from faster feeding larva

Muscle, Biomechanics, and Implications for Neural Control

1

RN (% of Behavior Duration)

0.8 Swallows

0.6

0.4 Bites

0.2

0

0

0.2

0.4 0.6 0.8 I2 (% of Behavior Duration)

1

Figure 12.11 Individuality of motor pattern features is behavior-dependent. Durations of activity in the protractor muscle (I2) and the muscle closing the feeding grasper (I4) as a percentage of behavior duration are shown for bites and swallows in two different animals (light, dark ellipses). The distributions of the behavior percentages are represented by ellipses whose edges are iso-density contours bounding 50% of the density for bivariate Gaussian distributions fitted to each group of data. RN duration varies between individuals in bites but not swallows (bite ellipses are not aligned along the vertical axis but swallow ellipses are) because the duration of grasper closure is important for effectiveness in swallowing. In contrast, I2 duration correlates with motor output in both behaviors—bites have long I2 durations (stronger protractions), swallows have short I2 durations (weaker protractions). Consistent with I2’s importance in both behaviors, in each behavior I2 duration is also similar in both animals. From Cullins et al. 2014 with permission.

(Foley and Luckinbill 2001). More generally, variation in behavior duration can distinguish individuals. For example, great tits (Parus major) vary in how quickly they explore their environment. These differences are correlated with other behavioral traits (e.g., aggressiveness), are heritable, and have differential effects on fitness depending on environmental factors (Dingemanse and Réale 2005). 12.4.4 Specialization of Neuromechanical Systems Reflect Behavioral Repertoire Motor System Multi-Functionality Determines the Complexity of its Neural Control System The

data presented above demonstrate that the relative importance of biomechanical specialization versus neural control complexity varies within and across animals. It thus seems reasonable to assert that the more multi-functional the motor system, the less the biomechanical specialization and the greater the complexity of the neural control. Highly specialized biomechanical systems can produce very precise, rapid, and specialized behaviors, as in insect jumping or flight. Specialization of the motor periphery can tune interactions with the environment without need for complex neural control mechanisms. For example, the command neurons that cause the most rapid and

399

400

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Cat 1

Cat 2

Cat 3

A Motor module structure SRTA SRTM GRAA GRAP BFMP REFM VMED

hip E + knee F hip F + knee E knee E

B Motor module force direction

180

90

270

0.5 N

0

C Motor module recruitment by perturbation direction 0

180

360

0 0 180 360 Perturbation direction (deg)

180

360

Figure 12.12 Individual-specific motor modules for achieving similar biomechanical function. (A) Motor modules that correspond to extensor force generation for postural control in cats vary across individuals. Open bars are muscles recorded in three animals. The motor modules for each animal include extensor muscles, reflecting biomechanical constraints on force generation. But the level of antagonistic muscle activity varies across individuals in each module. These modules were characteristic for each individual animal across a range of postural configurations and perturbation types. The different modules resulted in similar directions of ground-reaction forces (B) and the recruitment of motor modules with respect to the direction of postural disturbance was similar across individuals (C). Motor modules encoding common biomechanical functions across individuals may nonetheless reflect individual motor styles. From Torres-Oviedo et al. 2006 with permission.

invariant crayfish escape responses are highly reliable and may be critical to optimize performance for survival (Edwards et al. 1999). Similar specialization is seen in Mauthner neuron initiated escape responses in fish, in which tail stimuli rapidly generate a complex whole body movement (Korn and Faber 2005). The limited biomechanical affordances of specialized systems are key to their efficiency and reliability, and may help to optimize an essential motor behavior in evolution. Even these systems, however, are embedded in slower neural systems that can alter the rapid reflex responses through modulation, learning, or incorporation of social context (e.g., the effects of social dominance in crayfish on the escape response (Edwards et al. 1999), or of immediate environmental context on escape swim direction (Korn and Faber 2005)). In contrast, structures serving multiple purposes have less specialized biomechanics, and more complex neural control systems, that allow the system to be multi-functional.

Muscle, Biomechanics, and Implications for Neural Control

Increasing the biomechanical degrees of freedom and affordances allows for greater movement variability and richness in behavioral repertoire. For multi-functional systems, the key to survival is the ability to generate a variety of movements that rapidly and reliably reconfigure the body for different motor tasks, and this may be more important than optimal performance of any one task. Consider the highly developed nervous system in extremely reconfigurable animals such as octopuses. Not only do the neural control systems work to reconfigure the body to achieve desired motor goals (Hochner 2012), but additional motor functions can be achieved through tool use, communication, and social interactions (Hochner et al. 2006).

12.5 Conclusions Throughout this chapter we have illustrated how biomechanical considerations are critical for understanding the neural control of movement at three levels: how neural activity is transformed into force in individual muscles, how organismal structure and environmental forces affect individual muscle function, and how motorneuron activity is structured to create, and the effects of motorneuron variability on, motor behavior. At each level, a biomechanical perspective is necessary to understand the extent to which a muscle’s activity contributes to motor performance and whether different patterns of activity can equally well fulfill motor goals, which in turn can give rise to individual movement styles (Ting et al. 2015). The effects of neural signals on motor outputs is highly non-linear and context dependent, relying on the specific biomechanical constraints and affordances of the motor periphery. Consequently, there is similarly no direct mapping between a desired motor output and the neural signals necessary to generate the behavior. Given that a given neural activity can generate multiple motor outputs, and that a given motor output can be generated by multiple neural activities, it seems unlikely that motor solutions are generated through explicit neural computation and optimization. Rather, they may arise from initial “default behaviors” created through evolutionary and developmental processes. During the lifetime of an individual, motor movements are adapted through exploration, guided by neuromechanical constraints. Individual motor styles are made possible by motor abundance, and are equally effective in achieving motor goals. As a general principle, we have seen that the level of specialization of the periphery has an inverse relation to the multi-functionality, neural control complexity, and thus the behavioral repertoire of an animal. While there remain many open questions related to the structure and variability of motor systems, there are exciting prospects for future work as the ability to measure, manipulate, and simulate neuromechanical systems improves (Roth et al. 2014). Such approaches may facilitate the discovery of the essential features and components of neuromechanical systems, and make it possible to address the many open questions that remain. For example, how does ongoing sensory feedback shape motor activity during behavior (Shaw et al. 2014)? How are neuromechanical dynamics actively shaped by motor processes (Ting et al. 2009)? How can variability be used to enhance motor function (Wu et al. 2014)? What actual neural dynamics underlie motor behavior (Shaw et al. 2014)? Answering these challenging questions will provide deep insights into motor control across phylogeny.

401

402

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Physical modeling and robotic approaches may also shed light on the complex and non-linear interactions that underlie movement, and provide principles for the development of autonomous robots with more complex motor capabilities (Ijspeert et al. 2007; Scrivens et al. 2008; Boxerbaum et al. 2012; Daltorio et al. 2013; Horchler et al. 2015). A biomechanical perspective also has translational implications for understanding motor deficits and the mechanisms of neural plasticity and for developing motor rehabilitation practices to help treat sensorimotor deficits. Understanding interactions between neural control and biomechanics may also provide insights and principles to guide the effective design of devices, such as artificial limbs, that interact with humans, and of assistive and rehabilitative robots. Finally, the debates within the field of motor control about structure vs. variability and the roles of representation, environment, and context parallel more general debates in cognitive studies about how animals and humans interact with the world (Rosch 1999). Elenor Rosch stated that “Mind and world occur together in a succession of situations which are somewhat lawful and predictable. We want to be able to find those laws and to find a level of description which neither turns human action into something mechanical like engineering nor something mental like fantasy.” The interactions between animals and their environment create specific contexts in which animals need to generate reliable solutions. As a consequence, specific interactions between brain, body, and environment cannot be studied in isolation (Chiel and Beer 1997; Chiel et al. 2009) and shape how different individuals solve problems for moving, thinking, and learning. Engineering approaches have great power, but for this work to have relevance to biological systems it is important that it not be too prescriptive and take into account the natural variability, adaptability, and creativity in movement that are critical to survival in an instant, over a lifetime, and across evolutionary time.

Acknowledgements We thank M. Hongchul Sohn and Yun Seong Song for assistance in preparing the references and Kyle Blum and M. Hongchul Sohn for preparing the figures. We are also grateful for the comments and suggestions of SL Hooper. Supported by NIH HD46922, HD075612, and NSF EFRI-1137229 (LHT) and NSF IIS-1065489 and NIH NS087249 (HJC).

References Abbott BC, Aubert XM (1952) The force exerted by active striated muscle during and after change of length. J Physiol 117:77–86. Ahn AN (2012) How muscles function—the work loop technique. J Exp Biol 215:1051–1052. Ahn AN, Full RJ (2002) A motor and a brake: two leg extensor muscles acting at the same joint manage energy differently in a running insect. J Exp Biol 205:379–389. Ahn AN, Meijer K, Full RJ (2006) In situ muscle power differs without varying in vitro mechanical properties in two insect leg muscles innervated by the same motorneuron. J Exp Biol 209:3370–3382.

Muscle, Biomechanics, and Implications for Neural Control

Allen JL, Kautz SA, Neptune RR (2013) The influence of merged muscle excitation modules on post-stroke hemiparetic walking performance. Clin Biomech 28:697–704. Allen JL, Kautz SA, Neptune RR (2014) Forward propulsion asymmetry is indicative of changes in plantarflexor coordination during walking in individuals with post-stroke hemiparesis. Clin Biomech 29:780–786. Allen JL, Neptune RR (2012) Three-dimensional modular control of human walking. J Biomech 45:2157–2163. Amemiya Y, Iwamoto H, Kobayashi T, Sugi H, et al. (1988) Time-resolved X-ray diffraction studies on the effect of slow length changes on tetanized frog skeletal muscle. J Physiol 407:231–241. Angulo–Kinzler RM, Ulrich B, Thelen E (2002) Three-month-old infants can select specific leg motor solutions. Motor Control 6:52–68. Asakawa DS, Blemker SS, Gold GE, Delp SL (2002) In vivo motion of the rectus femoris muscle after tendon transfer surgery. J Biomech 35:1029–1037. Bartoo ML, Linke WA, Pollack GH (1997) Basis of passive tension and stiffness in isolated rabbit myofibrils. Am J Physiol 273:C266–276. Bayley TG, Sutton GP, Burrows M (2012) A buckling region in locust hindlegs contains resilin and absorbs energy when jumping or kicking goes wrong. J Exp Biol 215:1151–1161. Beer RD, Chiel HJ, Gallagher JC (1999) Evolution and analysis of model CPGs for walking: II. General principles and individual variability. J Comput Neurosci 7:119–147. Bekoff A (2001) Spontaneous embryonic motility: an enduring legacy. Int J Dev Neurosci 19:155–160. Belanger AY, McComas AJ (1985) A comparison of contractile properties in human arm and leg muscles. Eur J Appl Physiol Occup Physiol 54:326–330. Berger DJ, Gentner R, Edmunds T, Pai DK, d’Avella A (2013) Differences in adaptation rates after virtual surgeries provide direct evidence for modularity. J Neurosci 33:12384–12394. Berniker M, Jarc A, Bizzi E, Tresch MC (2009) Simplified and effective motor control based on muscle synergies to exploit musculoskeletal dynamics. Proc Natl Acad Sci USA 106:7601–7606. Bernstein N (1967) The Coordination and Regulation of Movements. Pergamon Press: New York. Biewener AA (2011) Muscle function in avian flight: achieving power and control. Philos T Roy Soc B 366:1496–1506. Biewener AA, Daley MA (2007) Unsteady locomotion: integrating muscle function with whole body dynamics and neuromuscular control. J Exp Biol 210:2949–2960. Binder–MacLeod S, Kesar T (2005) Catchlike property of skeletal muscle: recent findings and clinical implications. Muscle Nerve 31:681–693. Bizzi E, Cheung VC, d’Avella A, Saltiel P, Tresch M (2008) Combining modules for movement. Brain Res Rev 57:125–133. Blanco MM, Patek SN (2014) Muscle trade-offs in a power-amplified prey capture system. Evolution 68–5:1399–1414. Blemker SS, Pinsky PM, Delp SL (2005) A 3D model of muscle reveals the causes of nonuniform strains in the biceps brachii. J Biomech 38:657–665. Blickhan R (1986) Stiffness of an arthropod leg joint. J Biomech 19:375–384.

403

404

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Bobbert MF, van Ingen Schenau GJ (1988) Coordination in vertical jumping. J Biomech 21:249–262. Borzelli G, Cappozzo A, Papa E (1999) Inter- and intra-individual variability of ground reaction forces during sit-to-stand with principal component analysis. Med Eng Phys 21:235–240. Bowtell G, Williams TL (1994) Anguilliform body dynamics: a continuum model for the interaction between muscle activation and body curvature. Journal Math Biol 32:83–91. Boxerbaum AS, Shaw KM, Chiel HJ, Quinn RD (2012) Continuous wave peristaltic motion in a robot. Int J Robot Res 31:302–318. Boyle JH, Berri S, Cohen N (2012) Gait modulation in C. elegans: an integrated neuromechanical model. Front Comput Neurosci 6:10. Brezina V, Orekhova IV, Weiss KR (2000) Optimization of rhythmic behaviors by modulation of the neuromuscular transform. J Neurophysiol 83:260–279. Bunderson NE, Burkholder TJ, Ting LH (2008) Reduction of neuromuscular redundancy for postural force generation using an intrinsic stability criterion. J Biomech 41:1537–1544. Bunderson NE, McKay JL, Ting LH, Burkholder TJ (2010) Directional constraint of endpoint force emerges from hindlimb anatomy. J Exp Biol 213:2131–2141. Buneo CA, Soechting JF, Flanders M (1997) Postural dependence of muscle actions: implications for neural control. J Neurosci 17:2128–2142. Burke RE (1981) Motor units: anatomy, physiology and functional organization. In Handbook of Physiology. Brooks VB (ed). Washington, DC: Am Physiol Soc. Burkholder TJ, van Antwerp KW (2013) Practical limits on muscle synergy identification by non-negative matrix factorization in systems with mechanical constraints. Med Biol Eng Comput 51:187–196. Burrows M, Sutton G (2013) Interacting gears synchronize propulsive leg movements in a jumping insect. Science 341:1254–1256. Calabrese RL, Norris BJ, Wenning A, Wright TM (2011) Coping with variability in small neuronal networks. Integr Comp Biol 51:845–855. Campbell KS, Moss RL (2002) History-dependent mechanical properties of permeabilized rat soleus muscle fibers. Biophys J 82:929–943. Cappellini G, Ivanenko YP, Poppele RE, Lacquaniti F (2006) Motor patterns in human walking and running. J Neurophysiol 95:3426–3437. Chanaud CM, Macpherson JM (1991) Functionally complex muscles of the cat hindlimb III. Differential activation within biceps femoris during postural perturbations. Exp Brain Res 85:271–280. Chanaud CM, Pratt CA, Loeb GE (1991) Functionally complex muscles of the cat hindlimb V. The roles of histochemical fiber-type regionalization and mechanical heterogeneity in differential muscle activation. Exp Brain Res 85:300–313. Charles SK, Hogan N (2012) Stiffness, not inertial coupling, determines path curvature of wrist motion. J Neurophysiol 107:1230–1240. Cheung VC, d’Avella A, Tresch MC, Bizzi E (2005) Central and sensory contributions to the activation and organization of muscle synergies during natural motor behaviors. J Neurosci 25:6419–6434. Cheung VC, Piron L, Agostini M, Silvoni S, et al. (2009) Stability of muscle synergies for voluntary actions after cortical stroke in humans. Proc Natl Acad Sci USA 106:19563–19568.

Muscle, Biomechanics, and Implications for Neural Control

Cheung VCK, Turolla A, Agostini M, Silvoni S, et al. (2012) Muscle synergy patterns as physiological markers of motor cortical damage. Proc Natl Acad Sci USA 109:14652–14656. Chi MTH, Roscoe RD (2002) Reconsidering conceptual change: issues in theory and practice. In The Processes and Challenges of Conceptual Change. Limon M, Mason L (eds). Alphen aan den Rijn, The Netherlands: Kluwer. Chiel HJ, Beer RD (1997) The brain has a body: adaptive behavior emerges from interactions of nervous system, body and environment. Trends Neurosci 20:553–557. Chiel HJ, Crago P, Mansour JM, Hathi K (1992) Biomechanics of a muscular hydrostat: A model of lapping by a reptilian tongue. Biol Cyber 67:403–415. Chiel HJ, Ting LH, Ekeberg Ö, Hartmann MJ (2009) The brain in its body: motor control and sensing in a biomechanical context. J Neurosci 29:12807–12814. Churchland MM, Afshar A, Shenoy KV (2006) A central source of movement variability. Neuron 52:1085–1096. Chvatal SA, Macpherson JM, Torres-Oviedo G, Ting LH (2013) Absence of postural muscle synergies for balance after spinal cord transection. J Neurophysiol 110:1301–1310. Chvatal SA, Ting LH (2013) Common muscle synergies for balance and walking. Front Comput Neurosci 7:48. Chvatal SA, Torres-Oviedo G, Safavynia SA, Ting LH (2011) Common muscle synergies for control of center of mass and force in non-stepping and stepping postural behaviors. J Neurophysiol 106:999–1015. Clark DJ, Ting LH, Zajac FE, Neptune RR, Kautz SA (2010) Merging of healthy motor modules predicts reduced locomotor performance and muscle coordination complexity post-stroke. J Neurophysiol 103:844–857. Cleese J (1970) Memorandum from the Ministry of Silly Walks, BBC TV. Cohen RG, Sternad D (2009) Variability in motor learning: relocating, channeling and reducing noise. Exp Brain Res 193:69–83. Collins JJ (1995) The redundant nature of locomotor optimization laws. J Biomech 28:251–267. Collins S, Ruina A, Tedrake R, Wisse M (2005) Efficient bipedal robots based on passive-dynamic walkers. Science 307:1082–1085. Crowninshield RD, Brand RA (1981) A physiologically based criterion of muscle force prediction in locomotion. J Biomech 14:793–801. Cullins MJ, Gill JP, McManus JM, Lu H, et al. (2015) Sensory feedback reduces individuality by increasing variability within subjects. Curr Biol 25:2672–2676. Cullins MJ, Shaw KM, Gill JP, Chiel HJ (2014) Motorneuronal activity varies least among individuals when it matters most for behavior. J Neurophysiol 113:981–1000. d’Avella A, Saltiel P, Bizzi E (2003) Combinations of muscle synergies in the construction of a natural motor behavior. Nat Neurosci 6:300–308. Daley MA, Biewener AA (2006) Running over rough terrain reveals limb control for intrinsic stability. Proc Natl Acad Sci USA 103:15681–15686. Daley MA, Biewener AA (2011) Leg muscles that mediate stability: mechanics and control of two distal extensor muscles during obstacle negotiation in the guinea fowl. Philos T Roy Soc Lond B 366:1580–1591. Daltorio KA, Boxerbaum AS, Horchler AD, Shaw KM, et al. (2013) Efficient worm-like locomotion: slip and control of soft-bodied peristaltic robots. Bioinspir Biomim 8:035003.

405

406

Neurobiology of Motor Control: Fundamental Concepts and New Directions

de Brito Fontana H, Roesler H, Herzog W (2014) In vivo vastus lateralis force–velocity relationship at the fascicle and muscle tendon unit level. J Electromyogr Kinesiol 24:934–940. de Rugy A, Loeb GE, Carroll TJ (2012) Muscle coordination is habitual rather than optimal. J Neurosci 32:7384–7391. Delp SL, Loan JP (2000) A computational framework for simulating and analyzing human and animal movement. Comput Sci Eng 2:46–55. Delp SL, Loan JP, Hoy MG, Zajac FE, et al. (1990) An interactive graphics-based model of the lower extremity to study orthopaedic surgical procedures. IEEE T Biomed Eng 37:757–767. Deshpande AD, Gialias N, Matsuoka Y (2012) Contributions of intrinsic visco-elastic torques during planar index finger and wrist movements. IEEE T Biomed Eng 59:586–594. Dickinson MH, Farley CT, Full RJ, Koehl MAR, et al. (2000) How animals move: an integrative view. Science 288:100–106. Dingemanse NJ, Réale D (2005) Natural selection and animal personality. Behavior 142:1165–1190. Dominici N, Ivanenko YP, Cappellini G, d’Avella A, et al. (2011) Locomotor primitives in newborn babies and their development. Science 334:997–999. Duchateau J, Enoka RM (2011) Human motor unit recordings: origins and insight into the integrated motor system. Brain Res 1409:42–61. Dudek DM, Full RJ (2006) Passive mechanical properties of legs from running insects. J Exp Biol 209:1502–1515. Edman KA, Elzinga G, Noble MI (1982) Residual force enhancement after stretch of contracting frog single muscle fibers. J Gen Physiol 80:769–784. Edman KA, Tsuchiya T (1996) Strain of passive elements during force enhancement by stretch in frog muscle fibres. J Physiol 490:191–205. Edman KAP (2012) Residual force enhancement after stretch in striated muscle. A consequence of increased myofilament overlap? J Physiol (London) 590:1339–1345. Edwards DH, Heitler WJ, Krasne FB (1999) Fifty years of a command neuron: the neurobiology of escape behavior in the crayfish. Trends Neurosci 22:153–161. Elemans CP, Spierts IL, Müller UK, van Leeuwen JL, Goller F (2004) Bird song: superfast muscles control dove’s trill. Nature 431:146–146. Enoka RM, Baudry S, Rudroff T, Farina D, et al. (2011) Unraveling the neurophysiology of muscle fatigue. J Electromyogr Kinesiol 21:208–219. Enoka RM, Pearson KG (2013) The motor unit and muscle action. In Principles of Neural Science. Kandel ER, Schwartz JH, Jessell TM, Siegelbaum SA, Hudspeth AJ (eds). New York, NY: McGraw-Hill Medical. Finni T, Ikegawa S, Lepola V, Komi PV (2003) Comparison of force–velocity relationships of vastus lateralis muscle in isokinetic and in stretch-shortening cycle exercises. Acta Physiol Scand 177:483–491. Foley PA, Luckinbill LS (2001) The effects of selection for larval behavior on adult life-history features in Drosophila melanogaster. Evolution 55:2493–2502. Franklin DW, So U, Kawato M, Milner TE (2004) Impedance control balances stability with metabolically costly muscle activation. J Neurophysiol 92:3097–3105. Full RJ, Stokes DR, Ahn AN, Josephson RK (1998) Energy absorption during running by leg muscles in a cockroach. J Exp Biol 201:997–1012.

Muscle, Biomechanics, and Implications for Neural Control

Funabara D, Kanoh S, Siegman MJ, Butler TM, et al. (2005) Twitchin as a regulator of catch contraction in molluscan smooth muscle. J Muscle Res Cell Motil 26:455–460. Furuya S, Altenmuller E (2013) Flexibility of movement organization in piano performance. Front Hum Neurosci 7:173. Ganesh G, Haruno M, Kawato M, Burdet E (2010) Motor memory and local minimization of error and effort, not global optimization, determine motor behavior. J Neurophysiol 104:382–390. Gentner R, Gorges S, Weise D, aufm Kampe K, et al. (2010) Encoding of motor skill in the corticomuscular system of musicians. Curr Biol 20:1869–1874. Getz EB, Cooke R, Lehman SL (1998) Phase transition in force during ramp stretches of skeletal muscle. Biophys J 75:2971–2983. Giszter S, Patil V, Hart C (2007) Primitives, premotor drives, and pattern generation: a combined computational and neuroethological perspective. Prog Brain Res 165:323–346. Giszter SF, Hart CB (2013) Motor primitives and synergies in the spinal cord and after injury—the current state of play. Ann NY Acad Sci 1279:114–126. George NT, Irving TC, Williams C, Daniel, TL (2013) The cross-bridge spring: can cool muscles store elastic energy? Science 340:1217–1220. Golowasch J, Goldman MS, Abbott LF, Marder E (2002) Failure of averaging in the construction of a conductance-based neuron model. J Neurophysiol 87:1129–1131. Gordon AM, Huxley AF, Julian FJ (1966) The variation in isometric tension with sarcomere length in vertebrate muscle fibres. J Physiol 184:170–192. Gottlieb GL, Chen CH, Corcos DM (1995) Relations between joint torque, motion, and electromyographic patterns at the human elbow. Exp Brain Res 103:164–167. Gronenberg W (1996) Fast actions in small animals: springs and click mechanisms. J Comp Physiol A 178:727–734. Gunst SJ, Tang DD, Opazo Saez A (2003) Cytoskeletal remodeling of the airway smooth muscle cell: a mechanism for adaptation to mechanical forces in the lung. Respir Physiol Neurobiol 137:151–168. Haftel VK, Bichler EK, Nichols TR, Pinter MJ, Cope TC (2004) Movement reduces the dynamic response of muscle spindle afferents and motoneuron synaptic potentials in rat. J Neurophysiol 91:2164–2171. Handford C (2006) Serving up variability and stability. In Movement System Variability. Davids K, Bennet S, Newell K (eds). Champaign, IL: Human Kinetics. Harris CM, Wolpert DM (1998) Signal-dependent noise determines motor planning. Nature 394:780–784. Hart CB, Giszter SF (2004) Modular premotor drives and unit bursts as primitives for frog motor behaviors. J Neurosci 24:5269–5282. Hart CB, Giszter SF (2010) A neural basis for motor primitives in the spinal cord. J Neurosci 30:1322–1336. Hayes HB, Chvatal SA, French MA, Ting LH, Trumbower RD (2014) Neuromuscular constraints on muscle coordination during overground walking in persons with chronic incomplete spinal cord injury. Clin Neurophysiol 125:2024–2035. Heckman CJ, Hyngstrom AS, Johnson MD (2008) Active properties of motoneurone dendrites: diffuse descending neuromodulation, focused local inhibition. J Physiol 586:1225–1231. Henneman E (1957) Relation between size of neurons and their susceptibility to discharge. Science 126:1345–1347.

407

408

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Herzfeld DJ, Shadmehr R (2014) Motor variability is not noise, but grist for the learning mill. Nat Neurosci 17:149–150. Herzog W, Leonard TR (1991) Validation of optimization models that estimate the forces exerted by synergistic muscles. J Biomech 24:31–39. Herzog W, Leonard TR (2002) Force enhancement following stretching of skeletal muscle: a new mechanism. J Exp Biol 205:1275–1283. Herzog W, Leonard TR, Joumaa V, Mehta A (2008) Mysteries of muscle contraction. J Applied Biomech 24:1–13. Herzog W, Leonard TR, Wu JZ (2000) The relationship between force depression following shortening and mechanical work in skeletal muscle. J Biomech 33:659–668. Hicks JL, Schwartz MH, Arnold AS, Delp SL (2008) Crouched postures reduce the capacity of muscles to extend the hip and knee during the single-limb stance phase of gait. J Biomech 41:960–967. Higham TE, Biewener AA (2008) Integration within and between muscles during terrestrial locomotion: effects of incline and speed. J Exp Biol 211:2303–2316. Higham TE, Biewener AA (2011) Functional and architectural complexity within and between muscles: regional variation and intermuscular force transmission. Philos T Roy Soc Lond B 366:1477–1487. Hill AV (1938) The heat of shortening and the dynamic constants of muscle. Proc Roy Soc Lond B 126:136–195. Hill AV (1953) The mechanics of active muscle. Proc R Soc Lond B Sci 141:104–117. Hochner B (2012) An embodied view of octopus neurobiology. Curr Biol 22:R887–892. Hochner B, Shomrat T, Fiorito G (2006) The octopus: a model for a comparative analysis of the evolution of learning and memory mechanisms. Biol Bull 210:308–317. Hooper SL (2012) Body size and the neural control of movement. Curr Biol 22:R318–322. Hooper SL (2015) Sensory-motor integration: more variability reduces individuality. Curr Biol 20:R991–993 Hooper SL, Guschlbauer C, Blumel M, Rosenbaum P, et al. (2009) Neural control of unloaded leg posture and of leg swing in stick insect, cockroach, and mouse differs from that in larger animals. J Neurosci 29:4109–4119. Hooper SL, Guschlbauer C, von Uckermann G, Buschges A (2006) Natural neural output that produces highly variable locomotory movements. J Neurophysiol 96:2072–2088. Hooper SL, Guschlbauer C, von Uckermann G, Büschges A (2007) Different motor neuron spike patterns produce contractions with very similar rises in graded slow muscles. J Neurophysiol 97:1428–1444. Hooper SL, Hobbs KH, Thuma JB (2008) Invertebrate muscles: thin and thick filament structure; molecular basis of contraction and its regulation, catch and asynchronous muscle. Prog Neurobiol 86:72–127. Hooper SL, Thuma JB (2005) Invertebrate muscles: muscle specific genes and proteins. Physiol Rev 85:1001–1060. Hooper SL, Weaver AL (2000) Motorneuron activity is often insufficient to predict motor response. Curr Opin Neurobiol 10:676–682. Horchler AD, Daltorio KA, Chiel HJ, Quinn RD (2015) Designing responsive pattern generators: stable heteroclinic channel cycles for modeling and control. Bioinspir Biomim 10:026001. Huang VS, Shadmehr R, Diedrichsen J (2008) Active learning: learning a motor skill without a coach. J Neurophysiol 100:879–887.

Muscle, Biomechanics, and Implications for Neural Control

Huijing PA (2003) Muscular force transmission necessitates a multilevel integrative approach to the analysis of function of skeletal muscle. Exerc Sport Sci Rev 31:167–175. Huxley AF (1957) Muscle structure and theories of contraction. Prog Biophys and Biop Ch 7:255–318. Ijspeert AJ, Crespi A, Ryczko D, Cabelguen JM (2007) From swimming to walking with a salamander robot driven by a spinal cord model. Science 315:1416–1420. Ivanenko YP, Poppele RE, Lacquaniti F (2004) Five basic muscle activation patterns account for muscle activity during human locomotion. J Physiol 556:267–282. Jindrich DL, Full RJ (2002) Dynamic stabilization of rapid hexapedal locomotion. J Exp Biol 205:2803–2823. Josephson RK (1985) Mechanical power output from striated-muscle during cyclic contraction. J Exp Biol 114:493–512. Joyce G, Rack PM, Ross H (1974) The forces generated at the human elbow joint in response to imposed sinusoidal movements of the forearm. J Physiol 240:351–374. Kargo WJ, Ramakrishnan A, Hart CB, Rome LC, Giszter SF (2010) A simple experimentally based model using proprioceptive regulation of motor primitives captures adjusted trajectory formation in spinal frogs. J Neurophysiol 103:573–590. Katz PS, Frost WN (1996) Intrinsic neuromodulation: altering neuronal circuits from within. Trends Neurosci 19:54–61. Kawakami Y, Fukunaga T (2006) New insights into in vivo human skeletal muscle function. Exerc Sport Sci Rev 34:16–21. Kawakami Y, Muraoka T, Ito S, Kanehisa H, Fukunaga T (2002) In vivo muscle fibre behaviour during counter-movement exercise in humans reveals a significant role for tendon elasticity. J Physiol 540:635–646. Kaya M, Leonard T, Herzog W (2003) Coordination of medial gastrocnemius and soleus forces during cat locomotion. J Exp Biol 206:3645–3655. Kier W, Leeuwen J (1997) A kinematic analysis of tentacle extension in the squid Loligo pealei. J Exp Biol 200:41–53. Kier WM (2012) The diversity of hydrostatic skeletons. J Exp Biol 215:1247–1257. Kier WM, Schachat FH (2008) Muscle specialization in the squid motor system. J Exp Biol 211:164–169. Kier WM, Smith KK (1985) Tongues, tentacles and trunks: the biomechanics of movement in muscular hydrostats. Zool J Linnean Soc 83:307–324. Kirsch RF, Boskov D, Rymer WZ (1994) Muscle stiffness during transient and continuous movements of cat muscle: perturbation characteristics and physiological relevance. IEEE T Biomed Eng 41:758–770. Korn H, Faber DS (2005) The Mauthner cell half a century later: a neurobiological model for decision-making? Neuron 47:13–28. Kristan WB, Skalak R, Wilson RJA, Skierczynski BA, et al. (2000) Biomechanics of hydroskeletons: studies of crawling in the medicinal leech. In Biomechanics and Neural Control of Posture and Movement. Winters JM, Crago PE (eds). New York, NY: Springer. Kuhl PK (2004) Early language acquisition: cracking the speech code. Nat Rev Neurosci 5:831–843. Kuo AD (2007) The six determinants of gait and the inverted pendulum analogy: a dynamic walking perspective. Hum Mov Sci 26:617–656.

409

410

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Kurtzer I, Pruszynski JA, Herter TM, Scott SH (2006) Primate upper limb muscles exhibit activity patterns that differ from their anatomical action during a postural task. J Neurophysiol 95:493–504. Kutch JJ, Valero-Cuevas FJ (2012) Challenges and new approaches to proving the existence of muscle synergies of neural origin. PLoS Comput Biol 8:e1002434. Lacquaniti F, Ivanenko YP, d’Avella A, Zelik KE, Zago M (2013) Evolutionary and developmental modules. Front Comput Neurosci 7:61. Latash ML (2012) The bliss (not the problem) of motor abundance (not redundancy). Exp Brain Res 217:1–5. Latash ML, Scholz JF, Danion F, Schoner G (2001) Structure of motor variability in marginally redundant multifinger force production tasks. Exp Brain Res 141:153–165. Lee DD, Seung HS (1999) Learning the parts of objects by non-negative matrix factorization. Nature 401:788–791. Lee RH, Heckman CJ (2000) Adjustable amplification of synaptic input in the dendrites of spinal motoneurons in vivo. J Neurosci 20:6734–6740. Lemay MA, Grill WM (2004) Modularity of motor output evoked by intraspinal microstimulation in cats. J Neurophysiol 91:502–514. Li W, Todorov E, Pan X (2005) Hierarchical optimal control of redundant biomechanical systems. Proc 26th Annu Int Conf IEEE EMBS 2:4618–4621. Lichtwark GA, Wilson AM (2008) Optimal muscle fascicle length and tendon stiffness for maximising gastrocnemius efficiency during human walking and running. J Theor Biol 252: 662–673. Lieber RL, Ward SR (2011) Skeletal muscle design to meet functional demands. Philos T Roy Soc Lond B 366:1466–1476. Lin DC, Rymer WZ (2001) Damping actions of the neuromuscular system with inertial loads: human flexor pollicis longus muscle. J Neurophysiol 85:1059–1066. Lockhart DB, Ting LH (2007) Optimal sensorimotor transformations for balance. Nat Neurosci 10:1329–1336. Loeb GE (2012) Optimal isn’t good enough. Biol Cybern 106:757–765. Loram ID, Lakie M, Di Giulio I, Maganaris CN (2009) The consequences of short-range stiffness and fluctuating muscle activity for proprioception of postural joint rotations: the relevance to human standing. J Neurophysiol 102:460–474. Lu H, McManus JM, Cullins MJ, Chiel HJ (2015) Preparing the periphery for a subsequent behavior: motorneuronal activity during biting generates little force but prepares a retractor muscle to generate larger forces during swallowing in Aplysia. J Neurosci 35:5051–5066. Maas H, Sandercock TG (2010) Force transmission between synergistic skeletal muscles through connective tissue linkages. J Biomed Biotechnol 2010:575672. Macpherson JM (1988) Strategies that simplify the control of quadrupedal stance. II. Electromyographic activity. J Neurophysiol 60:218–231. Macpherson JM, Horak FB (2013) Posture. In Principles of Neural Science. Kandel ER, Schwartz JH, Jessell TM, Siegelbaum SA, Hudspeth AJ (eds). New York, NY: McGraw-Hill Medical. Marder E, Goaillard JM (2006) Variability, compensation and homeostasis in neuron and network function. Nat Rev Neurosci 7:563–574. Marechal G, Plaghki L (1979) Deficit of the isometric tetanic tension redeveloped after a release of frog muscle at a constant velocity. J Gen Physiol 73:453–467.

Muscle, Biomechanics, and Implications for Neural Control

McGowan C, Neptune R, Herzog W (2013) A phenomenological muscle model to assess history dependent effects in human movement. J Biomech 46:151–157. McKay JL, Ting LH (2012) Optimization of muscle activity for task-level goals predicts complex changes in limb forces across biomechanical contexts. PLoS Comput Biol 8:e1002465. McManus JM, Lu H, Cullins MJ, Chiel HJ (2014) Differential activation of an identified motorneuron and neuromodulation provide Aplysia’s retractor muscle an additional function. J Neurophysiol 112:778–791. McMillen T, Williams T, Holmes P (2008) Non-linear muscles, passive viscoelasticity and body taper conspire to create neuromechanical phase lags in anguilliform swimmers. PLoS Comput Biol 4:e1000157. Meinertzhagen IA, Lemaire P, Okamura Y (2004) The neurobiology of the ascidian tadpole larva: recent developments in an ancient chordate. Annu Rev Neurosci 27:453–485. Minozzo FC, Lira CA (2013) Muscle residual force enhancement: a brief review. Clinics (Sao Paulo) 68:269–274. Morgan D (1977) Separation of active and passive components of short-range stiffness of muscle. Am J Physiol 232:C45–C49. Morgan DL (1990) New insights into the behavior of muscle during active lengthening. Biophys J 57:209–221. Morris LG, Hooper SL (1997) Muscle response to changing neuronal input in the lobster (Panulirus interruptus) stomatogastric system: spike number vs. spike frequency dependent domains. J Neurosci 17:5956–5971. Murphy RA, Rembold CM (2005) The latch-bridge hypothesis of smooth muscle contraction. Can J Physiol Pharmacol 83:857–864. Murray WM, Buchanan TS, Delp SL (2000) The isometric functional capacity of muscles that cross the elbow. J Biomech 33:943–952. Neptune RR, Clark DJ, Kautz SA (2009) Modular control of human walking: a simulation study. J Biomech 42:1282–1287. Nichols TR, Cope TC (2004) Cross-bridge mechanisms underlying the history-dependent properties of muscle spindles and stretch reflexes. Can J Physiol Pharmacol 82:569–576. Nishikawa K, Biewener AA, Aerts P, Ahn AN, et al. (2007) Neuromechanics: an integrative approach for understanding motor control. Integr Comp Biol 47:16–54. Novakovic VA, Sutton GP, Neustadter DM, Beer RD, Chiel HJ (2006) Mechanical reconfiguration mediates swallowing and rejection in Aplysia californica. J Comp Physiol A 192:857–870. Nussbaum MA, Chaffin DB (1997) Pattern classification reveals intersubject group differences in lumbar muscle recruitment during static loading. Clin Biomech (Bristol, Avon) 12:97–106. Olshausen BA, Field DJ (2004) Sparse coding of sensory inputs. Curr Opin Neurobiol 14:481–487. Overduin SA, d’Avella A, Carmena JM, Bizzi E (2012) Microstimulation activates a handful of muscle synergies. Neuron 76:1071–1077. Pai DK (2010) Muscle mass in musculoskeletal models. J Biomech 43:2093–2098. Pandy MG, Zajac FE (1991) Optimal muscular coordination strategies for jumping. J Biomech 24:1–10. Park J, Pažin N, Friedman J, Zatsiorsky VM, Latash ML (2014) Mechanical properties of the human hand digits: age-related differences. Clin Biomech (Bristol, Avon) 29:129–137.

411

412

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Perreault EJ, Heckman CJ, Sandercock TG (2003) Hill muscle model errors during movement are greatest within the physiologically relevant range of motor unit firing rates. J Biomech 36:211–218. Perreault EJ, Kirsch RF, Crago PE (2004) Multijoint dynamics and postural stability of the human arm. Exp Brain Res 157:507–517. Powers RK, Elbasiouny SM, Rymer WZ, Heckman CJ (2012) Contribution of intrinsic properties and synaptic inputs to motoneuron discharge patterns: a simulation study. J Neurophysiol 107:808–823. Prinz AA, Bucher D, Marder E (2004) Similar network activity from disparate circuit parameters. Nat Neurosci 7:1345–1352. Rack PM, Westbury DR (1969) The effects of length and stimulus rate on tension in the isometric cat soleus muscle. J Physiol 204:443–460. Rack PM, Westbury DR (1974) The short range stiffness of active mammalian muscle and its effect on mechanical properties. J Physiol 240:331–350. Riewald SA, Delp SL (1997) The action of the rectus femoris muscle following distal tendon transfer: does it generate knee flexion moment? Dev Med & Child Neurol 39:99–105. Ritzmann RE, Quinn RD, Fischer MS (2004) Convergent evolution and locomotion through complex terrain by insects, vertebrates and robots. Arthropod Struct Dev 33:361–379. Rizzolatti G, Strick PL (2013) Cognitive functions of the premotor system. In Principles of Neural Science. Kandel ER, Schwartz JH, Jessell TM, Siegelbaum SA, Hudspeth AJ (eds). New York, NY: McGraw-Hill Medical. Roberts TJ, Marsh RL, Weyand PG, Taylor CR (1997) Muscular force in running turkeys: the economy of minimizing work. Science 275:1113–1115. Roh J, Cheung VC, Bizzi E (2011) Modules in the brain stem and spinal cord underlying motor behaviors. J Neurophysiol 106:1363–1378. Roh J, Rymer WZ, Perreault EJ, Yoo SB, Beer RF (2013) Alterations in upper limb muscle synergy structure in chronic stroke survivors. J Neurophysiol 109:768–781. Rosch E (1999) Reclaiming concepts. J Consciousness Stud 6:61–77. Roth E, Sponberg S, Cowan NJ (2014) A comparative approach to closed-loop computation. Curr Opin Neurobiol 25:54–62. Routson RL, Kautz SA, Neptune RR (2014) Modular organization across changing task demands in healthy and poststroke gait. Physiol Rep 2:e12055. Safavynia SA, Ting LH (2013) Sensorimotor feedback based on task–relevant error robustly predicts temporal recruitment and multidirectional tuning of muscle synergies. J Neurophysiol 109:31–45. Safavynia SA, Torres-Oviedo G, Ting LH (2011) Muscle synergies: implications for clinical evaluation and rehabilitation of movement. Top Spinal Cord Inj Rehabil 17:16–24. Saltiel P, Wyler-Duda K, D’Avella A, Tresch MC, Bizzi E (2001) Muscle synergies encoded within the spinal cord: evidence from focal intraspinal NMDA iontophoresis in the frog. J Neurophysiol 85:605–619. Scholz JP, Danion F, Latash ML, Schoner G (2002) Understanding finger coordination through analysis of the structure of force variability. Biol Cybern 86:29–39. Scholz JP, Schoner G (1999) The uncontrolled manifold concept: identifying control variables for a functional task. Exp Brain Res 126:289–306. Scholz JP, Schoner G, Latash ML (2000) Identifying the control structure of multijoint coordination during pistol shooting. Exp Brain Res 135:382–404.

Muscle, Biomechanics, and Implications for Neural Control

Scott SH (2004) Optimal feedback control and the neural basis of volitional motor control. Nat Rev Neurosci 5:534–546. Scott SH (2008) Inconvenient truths about neural processing in primary motor cortex. J Physiol 586:1217–1224. Scrivens JE, Deweerth SP, Ting LH (2008) A robotic device for understanding neuromechanical interactions during standing balance control. Bioinspir Biomim 3:026002. Selen LP, Franklin DW, Wolpert DM (2009) Impedance control reduces instability that arises from motor noise. J Neurosci 29:12606–12616. Sensenig AT, Shultz JW (2003) Mechanics of cuticular elastic energy storage in leg joints lacking extensor muscles in arachnids. J Exp Biol 206:771–784. Shadmehr R, Krakauer JW (2008) A computational neuroanatomy for motor control. Exp Brain Res 185:359–381. Shadmehr R, Smith MA, Krakauer JW (2010) Error correction, sensory prediction, and adaptation in motor control. Annu Rev Neurosci 33:89–108. Shaw KM, Lyttle DN, Gill JP, Cullins MJ, et al. (2015) The significance of dynamical architecture for adaptive responses to mechanical loads during rhythmic behavior. J Comput Neurosci 38:25–51. Siebert T, Rode C, Herzog W, Till O, Blickhan R (2007) Non-linearities make a difference: comparison of two common Hill-type models with real muscle. Biol Cybern 98:133–143. Simpson CS, Sohn MH, Allen JL, Ting LH (2015) Feasible muscle activation ranges based on inverse dynamics analyses of human walking. J Biomech 48:2990–2997. Skierczynski BA, Wilson RJ, Kristan WB, Jr., Skalak R (1996) A model of the hydrostatic skeleton of the leech. J Theor Biol 181:329–342. Smith LB, Thelen E (2003) Development as a dynamic system. Trends Cogn Sci 7:343–348. Snaterse M, Ton R, Kuo AD, Donelan JM (2011) Distinct fast and slow processes contribute to the selection of preferred step frequency during human walking. J Appl Physiol 110:1682–1690. Sohn MH, McKay JL, Ting LH (2013) Defining feasible bounds on muscle activation in a redundant biomechanical task: practical implications of redundancy. J Biomech 46:1363–1368. Sohn MH, Ting LH (2013) The cost of being stable: a quantitative approach to examine trade-off between effort and stability. American Society of Biomechanics, Omaha, NB. http://www.asbweb.org/conferences/2013/abstracts/369.pdf. Sponberg S, Libby T, Mullens CH, Full RJ (2011a) Shifts in a single muscle’s control potential of body dynamics are determined by mechanical feedback. Philos T Roy Soc B 366:1606–1620. Sponberg S, Spence AJ, Mullens CH, Full RJ (2011b) A single muscle’s multi-functional control potential of body dynamics for postural control and running. Philos T Roy Soc Lond B 366:1592–1605. Stahl VA, Nichols TR (2014) Short-term effect of crural fasciotomy on kinematic variability and propulsion during level locomotion. J Mot Behav 46:339–349. Steele KM, Seth A, Hicks JL, Schwartz MS, Delp SL (2010) Muscle contributions to support and progression during single-limb stance in crouch gait. J Biomech 43:2099–2105. Steele KM, Tresch MC, Perreault EJ (2013) The number and choice of muscles impact the results of muscle synergy analyses. Front Comput Neurosci 7:105.

413

414

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Sternad D, Park SW, Muller H, Hogan N (2010) Coordinate dependence of variability analysis. PLoS Comput Biol 6:e1000751. Thuma JB, Hooper SL (2010) Direct evidence that stomatogastric (Panulirus interruptus) muscle passive responses are not due to background actomyosin cross-bridges. J Comp Physiol A 196:649–657. Ting LH (2007) Dimensional reduction in sensorimotor systems: a framework for understanding muscle coordination of posture. Prog Brain Res 165:299–321. Ting LH, Chiel HJ, Trumbower RD, Allen JL, et al. (2015) Neuromechanical principles underlying movement modularity and their implications for rehabilitation. Neuron 86:38–54. Ting LH, Chvatal SA (2010) Decomposing muscle activity in motor tasks: methods and interpretation. In Motor control: Theories, Experiments, and Applications. Danion F, Latash ML (eds). Oxford, UK: Oxford University Press. Ting LH, Chvatal SA, Safavynia SA, Lucas McKay J (2012) Review and perspective: neuromechanical considerations for predicting muscle activation patterns for movement. Int J Numer Method Biomed Eng 28:1003–1014. Ting LH, Macpherson JM (2005) A limited set of muscle synergies for force control during a postural task. J Neurophysiol 93:609–613. Ting LH, McKay JL (2007) Neuromechanics of muscle synergies for posture and movement. Curr Opin Neurobiol 17:622–628. Ting LH, van Antwerp KW, Scrivens JE, McKay JL, et al. (2009) Neuromechanical tuning of non–linear postural control dynamics. Chaos 19:026111. Todorov E (2004) Optimality principles in sensorimotor control. Nat Neurosci 7:907–915. Todorov E, Jordan MI (2002) Optimal feedback control as a theory of motor coordination. Nat Neurosci 5:1226–1235. Torres-Oviedo G, Macpherson JM, Ting LH (2006) Muscle synergy organization is robust across a variety of postural perturbations. J Neurophysiol 96:1530–1546. Torres-Oviedo G, Ting LH (2007) Muscle synergies characterizing human postural responses. J Neurophysiol 98:2144–2156. Torres-Oviedo G, Ting LH (2010) Subject-specific muscle synergies in human balance control are consistent across different biomechanical contexts. J Neurophysiol 103:3084–3098. Tresch MC, Cheung VC, d’Avella A (2006) Matrix factorization algorithms for the identification of muscle synergies: evaluation on simulated and experimental data sets. J Neurophysiol 95:2199–2212. Tresch MC, Jarc A (2009) The case for and against muscle synergies. Curr Opin Neurobiol 19:601–607. Tresch MC, Saltiel P, Bizzi E (1999) The construction of movement by the spinal cord. Nat Neurosci 2:162–167. Trimmer BA, Lin HT (2014) Bone-free: soft mechanics for adaptive locomotion. Integr Comp Biol 54:1122–1135. Trumbower RD, Krutky MA, Yang B-S, Perreault EJ (2009) Use of self-selected postures to regulate multi-joint stiffness during unconstrained tasks. PloS One 4:e5411. Tsianos GA, Goodner J, Loeb GE (2014) Useful properties of spinal circuits for learning and performing planar reaches. J Neural Eng 11:056006.

Muscle, Biomechanics, and Implications for Neural Control

Tytell ED, Hsu CY, Williams TL, Cohen AH, Fauci LJ (2010) Interactions between internal forces, body stiffness, and fluid environment in a neuromechanical model of lamprey swimming. Proc Natl Acad Sci USA 107:19832–19837. Valero-Cuevas FJ (2005) An integrative approach to the biomechanical function and neuromuscular control of the fingers. J Biomech 38:673–684. Valero-Cuevas FJ, Jae-Woong Y, Brown D, McNamara RV, et al. (2007) The tendon network of the fingers performs anatomical computation at a macroscopic scale. IEEE T Biomed Eng 54:1161–1166. Valero-Cuevas FJ, Venkadesan M, Todorov E (2009) Structured variability of muscle activations supports the minimal intervention principle of motor control. J Neurophysiol 102:59–68. Valero-Cuevas FJ, Zajac FE, Burgar CG (1998) Large index-fingertip forces are produced by subject-independent patterns of muscle excitation. J Biomech 31: 693–703. van Antwerp K, Burkholder T, Ting L (2007) Inter-joint coupling effects on muscle contributions to endpoint force and acceleration in a musculoskeletal model of the cat hindlimb. J Biomech 40:3570–3579. van Ingen Schenau GJ (1994) Proposed action of bi-articular muscles and the design of hindlimbs of bi- and quadrupeds. Hum Mov Sci 13:665–681. Van Leeuwen J, Kier WM (1997) Functional design of tentacles in squid: linking sarcomere ultrastructure to gross morphological dynamics. Philos T Roy Soc Lond B 352:551–571. Vandenboom R, Gittings W, Smith IC, Grange RW, Stull JT (2013) Myosin phosphorylation and force potentiation in skeletal muscle: evidence from animal models. J Muscle Res Cell Motil 34:317–332. Vogel S (1988) Life’s Devices: the Physical World of Animals and Plants. Princeton University Press: Princeton, NJ. Voigt M, Simonsen EB, Dyhre–Poulsen P, Klausen K (1995) Mechanical and muscular factors influencing the performance in maximal vertical jumping after different prestretch loads. J Biomech 28:293–307. Wadepuhl M, Beyn W (1989) Computer simulation of the hydrostatic skeleton. The physical equivalent, mathematics and application to worm-like forms. J Theor Biol 136:379–402. Wakeling JM (2009) The recruitment of different compartments within a muscle depends on the mechanics of the movement. Biol Lett 5:30–34. Walter JP, Kinney AL, Banks SA, D’Lima DD, et al. (2014) Muscle synergies may improve optimization prediction of knee contact forces during walking. J Biomech Eng 136:021031. Welch TD, Ting LH (2008) A feedback model reproduces muscle activity during human postural responses to support-surface translations. J Neurophysiol 99:1032–1038. Welch TD, Ting LH (2009) A feedback model explains the differential scaling of human postural responses to perturbation acceleration and velocity. J Neurophysiol 101:3294–3309. Wilson A, Lichtwark G (2011) The anatomical arrangement of muscle and tendon enhances limb versatility and locomotor performance. Philos T Roy Soc Lond B 366:1540–1553. Wu HG, Miyamoto YR, Gonzalez Castro LN, Olveczky BP, Smith MA (2014) Temporal structure of motor variability is dynamically regulated and predicts motor learning ability. Nat Neurosci 17:312–321.

415

416

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Yakovenko S, Krouchev N, Drew T (2011) Sequential activation of motor cortical neurons contributes to intralimb coordination during reaching in the cat by modulating muscle synergies. J Neurophysiol 105:388–409. Yang JF, Lam T, Pang MY, Lamont E, et al. (2004) Infant stepping: a window to the behaviour of the human pattern generator for walking. Can J Physiol Pharmacol 82:662–674. Ye H, Morton DW, Chiel HJ (2006a) Neuromechanics of coordination during swallowing in Aplysia californica. J Neurosci 26:1470–1485. Ye H, Morton DW, Chiel HJ (2006b) Neuromechanics of multi-functionality during rejection in Aplysia californica. J Neurosci 26:10743–10755. Yen JT, Auyang AG, Chang YH (2009) Joint-level kinetic redundancy is exploited to control limb-level forces during human hopping. Exp Brain Res 196:439–451. Young RP, Scott SH, Loeb GE (1992) An intrinsic mechanism to stabilize posture—joint-angle-dependent moment arms of the feline ankle muscles. Neurosci Lett 145:137–140. Yu SN, Crago PE, Chiel HJ (1997) A nonisometric kinetic model for smooth muscle. Am J Physiol 272:C1025–C1039. Zahalak GI (1986) A comparison of the mechanical behavior of the cat soleus muscle with a distribution-moment model. J Biomech Eng 108:131–140. Zahalak GI, Ma SP (1990) Muscle activation and contraction: constitutive relations based directly on cross-bridge kinetics. J Biomech Eng 112:52–62. Zajac FE (1989) Muscle and tendon: properties, models, scaling, and application to biomechanics and motor control. Crit Rev Biomed Eng 17:359–411. Zajac FE (2002) Understanding muscle coordination of the human leg with dynamical simulations. J Biomech 35:1011–1018. Zajac FE, Gordon ME (1989) Determining muscle’s force and action in multi-articular movement. Exerc Sport Sci Rev 17:187–230. Zajac FE, Neptune RR, Kautz SA (2002) Biomechanics and muscle coordination of human walking. Part I: introduction to concepts, power transfer, dynamics and simulations. Gait Posture 16:215–232. Zaniolo G, Lane NJ, Burighel P, Manni L (2002) Development of the motor nervous system in ascidians. J Comp Neurol 443:124–135. Zelik KE, La Scaleia V, Ivanenko YP, Lacquaniti F (2014) Can modular strategies simplify neural control of multidirectional human locomotion? J Neurophysiol 111:1686–1702. Zucker RS, Regehr WG (2002) Short-term synaptic plasticity. Annu Rev Physiol 64:355–405.

417

13 Plasticity and Learning in Motor Control Networks John Simmers 1 and Keith T. Sillar 2 1 Institut de Neurosciences Cognitives et Intégratives d’Aquitaine, CNRS UMR 5287, Université de Bordeaux, Bordeaux, France 2 School of Psychology and Neuroscience, University of St Andrews, Fife, Scotland, UK

13.1 Introduction The ability to produce movements that are not only well coordinated, but also adaptable, is a fundamental necessity of life. Without such a capacity, animals may fail to avoid predation, feed, or find a mate—three sure-fire ways of exiting the gene pool unproductively. Getting into the gene pool in the first place involves the complex realm of motor system development. This process follows species-specific (though largely conserved) rules to create animal populations whose constituents possess fundamentally similar motor repertoires, but with subtly different and individually styled motor skills. Individual differences in motor behavior emerge from a developmental template that initially sets up the underlying motor circuits, in combination with how this template interacts with, and is sculpted by, experience and environmental contingencies. The ecological arenas in which the life or death struggles for existence are enacted are brutal, unforgiving, and not entirely level playing fields. However, this stark reality also ensures that the evolutionary arms race is guaranteed to succeed because the least capable participants in the game are weeded out—slow prey are caught, sluggish predators starve—while the swiftest and strongest survive to pass on their genes (Sillar et al. 2016). However, because even within a single generation the game can change, players need the ability to modify their game plan to find new strategies in order to survive. For example, the maturation process is instrumental in configuring a player’s profile, metaphorically and literally. During development it is important to reach peak motor performance sooner rather than later because thereafter individuals enter a prolonged decline, not least in terms of motor performance, to which the authors of this chapter can certainly testify. Animals must also cope with injuries or diseases that affect movement. Motor networks must therefore continue to be adaptable to enable the behaviors they generate to remain versatile. What mechanisms define motor networks initially and then allow them to display short- and long-term plasticity? More precisely, what neural processes confer on such networks a capacity for short-term memory that tracks motor performance, and long-term motor learning that adapts behavior in response to experience and Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

418

Neurobiology of Motor Control: Fundamental Concepts and New Directions

external contingencies? Motor behavior arises in two general ways: (1) sensorimotor reflexes in which motor output is the response to a defined sensory input from the periphery; and (2) autonomous processes whereby the motor output arises from internally driven mechanisms within the central nervous system. We know much about how the neuronal processes that generate reflexes are dynamically adjusted in an experience-dependent manner as an animal’s external environment changes. Notable examples include synaptic plasticity in pain pathways of the dorsal horn (reviewed in Luo et al. 2014), memory formation in the cerebellar pathways mediating aversive eye blink reflexes in mammals, including humans (reviewed in Thompson 2013), and motor learning in defensive gill, siphon, and tail withdrawal behaviors of the marine mollusk Aplysia. In the latter especially, the cellular and molecular substrates of non-associative (habituation and sensitization) and associative (classical and operant conditioning) learning have been widely explored and are now well understood (reviewed in Glantzman 2006; Hawkins et al. 2006; Byrne and Hawkins 2015; Hawkins and Byrne 2015). Unlike reflex driven behavior, whose occurrence strictly depends on the (unpredictable) timing of extrinsic triggering stimuli, the motor commands for relatively stereotyped rhythmic acts, such as locomotion or respiration, or for more variable, goal-directed acts, such as feeding or sexual behavior, are driven at rates determined intrinsically by the autogenic properties of central nervous circuitry. How sensory feedback and modulatory input acutely alter the activity of the “central pattern generator” (CPG) networks responsible for stereotyped motor rhythms has been intensely studied. But, how CPGs alter their output according to previous behavioral experience and memory storage is much less understood. Moreover, in the case of goal-directed (or motivated) behaviors, in which the spontaneous urge of when and how to act can be adjusted by sensory information through non-associative and associative learning, the adaptive interaction between external stimuli and the dynamics of internal “decision-making” neural circuitry is still poorly understood. This chapter reviews recent studies that have provided insights into experiencedependent plasticity and memory formation in CPG networks with the goals of highlighting general principles and stimulating future exploration. We discuss two examples of behaviorally-relevant motor learning and plasticity expressed by CPGs: 1) a form of short term motor memory in a vertebrate spinal circuit controlling swimming in frog (Xenopus) tadpoles; and 2) longer term motor learning in the feeding circuit of Aplysia. In so doing we skim over, or even expressly avoid, material on motor system development, neuromodulation, and CPG-extrinsic mechanisms of learning and memory. The reader is directed to reviews covering these topics (motor system development, Sillar et al. 2014; intrinsic short term network plasticity, El Manira 2014; neuromodulation, Miles and Sillar 2011; Marder 2012; Marder et al. 2014).

13.2 Homeostatic Motor Network Assembly Spinal motor networks differentiate very early in development, in utero or in the egg, shortly after the ectoderm is induced to develop into neural tissue. Neuronal differentiation results from a largely pre-programmed transcriptional coding system in which local concentrations of the protein Sonic Hedgehog (SHH), emanating from the ventral

Plasticity and Learning in Motor Control Networks

floor plate, blend with Bone Morphogenic Factor (BMF), arising dorsally from the roof plate, to determine the fate of progenitor cells (reviewed in Jessell 2000). This highly conserved vertebrate code directs the differentiation of different populations of spinal neurons whose anatomy, function, and connectivity are becoming increasingly understood (Grillner and Jessell 2009; Grossmann et al. 2010). The position of progenitor cells on the dorso-ventral axis of the spinal cord determines the ratio of SHH to BMP to which they are exposed, which in turn determines whether they differentiate into, for example, motorneurons (MNs) or interneurons. This coding system thus determines the proportions of neurons that comprise CPG networks. On a network level, CPGs are composed of MNs that innervate skeletal muscles and which in turn are driven by excitatory (E) and inhibitory (I) interneurons. On a functional level it is important that the transcriptional code described above apportions appropriate numbers of each type of neuron and properly weights the synaptic strengths between the different neuron types. The reason is relatively simple, at least conceptually: MN numbers determine the strength and fine control of muscle contractions while the balance of E and I components set the operational limits of CPG output frequency and intensity (Miles and Sillar 2011). It is thus critically important that development gets it right. This is especially true because the entire process is under strict homeostatic control; if the number of one neuronal phenotype changes during development there will be consequences for other neuron classes in the network. Electrical activity plays an instrumental role in the specification of neuronal identity and transmitter phenotype in CPG neurons (Borodinsky et al. 2004, 2012; reviewed in Marder and Goaillard 2006). Too low a level of spontaneous electrical activity during development (relative to an internal set point) homeostatically up-regulates the number of E interneurons and decreases I interneuron numbers. In contrast, activity above the internal set point lowers E, and increases I, interneuron numbers. Spontaneous movements are a signature of early motor system development. It is easy to imagine that the spontaneous neural activity that drives these movements plays a fundamental role in motor network ontogeny and neural circuit specification. Neuromodulation plays an equally formative role, with several neuromodulators being implicated, including, in particular, the biogenic amines serotonin (e.g., Sillar et al. 1995; Vinay et al. 2002; Hilaire et al. 2010) and dopamine (Reimer et al. 2013). In the case of dopamine, in zebrafish a phylogenetically conserved descending diencephalic dopaminergic tract (DDT), which develops in the first day post-fertilization, orchestrates MN and interneuron differentiation. At this critical developmental stage the early innervation by DDT axons and their release of dopamine activates D4a receptors on progenitor cells, which in turn alters SSH signaling to generate MNs. Experimental hyper-activation of this signaling pathways increases MN numbers above normal, with a compensatory and homeostatic reduction in interneuron numbers as outlined above. To summarize and set the scene for the cases studies that follow, early development of motor networks results from two interacting processes: (1) pre-programmed molecular genetic definition of neuron subtypes regulated by transcription factors; and (2) activity-dependent homeostatic tuning of neuronal bioelectrical properties and transmitter phenotype. Altering either process causes plastic changes in the cellular and synaptic make-up of otherwise similar motor networks in different individuals. On the surface the basic motor repertoires of two individuals of the same species may be qualitatively comparable. However, probing beneath the surface may reveal subtle

419

420

Neurobiology of Motor Control: Fundamental Concepts and New Directions

differences in underlying network components. Furthermore, since motor network plasticity may reside with activity-dependent mechanisms, these changes may not necessarily be reflected in day-to-day motor behaviors, but instead only at performance extremes, at which times plasticity may be particularly beneficial.

13.3 Short-Term Motor Learning Conferred by Sodium Pumps A general feature of rhythmic motor systems, including those controlling locomotion, is that behavior can occur over a wide range of frequencies and intensities. For example, humans transition within a single gait from a slow amble when taking a walk in the park to brisker paces when rushing to make an appointment or catch a bus. A similar capability is a feature of axial swimming in fish or tadpoles. These extremes of the same behavior are driven by largely common elements within one neural network, with additional network components being recruited to accommodate speed and intensity increases (e.g., Menelaou and McLean 2012; El Manira 2014; Zhang et al. 2011). One consequence of motor systems operating at performance extremes is that the neuromuscular components producing the behavior are susceptible to fatigue. Animals are thus forced to balance a trade-off between high endurance (low risk of muscle fatigue) and high performance (high risk of muscle fatigue and exhaustion). Following a bout of high speed, high intensity locomotion, for instance after escaping from a predator or chasing and catching prey, it is important to rest in order for muscles to recover and lactic acid levels to subside. Consider, for example, the 100 m final at the London Olympics in 2012. Seven of the eight finalists ran the race in under 10 seconds (for the record the eighth runner, Asafa Powell, suffered a groin strain before crossing the line). Now imagine the race was re-run a minute later. It is hard to envisage any of the finalists clocking a sub-10 second time. This is because there is an inverse relationship between inter-effort interval and subsequent performance. A similar relationship applies to all animals. A potentially huge problem would arise if an attack by a predator coincided with the rest period between episodes of behavioral effort. To deal with this contingency, many animals retain a residual, emergency muscle capacity during seeming exhaustion. Do the central networks, the CPGs, that drive the muscles, also fatigue? Nervous systems are highly energy consuming; a cursory internet search suggests that the human brain comprises ∼2% of body weight but accounts for ∼20% of energy consumption. Growing evidence suggests that motor control networks, when pushed to performance extremes, indeed both “run down” and, similar to muscles, retain an emergency residual capacity for re-activation even when seemingly exhausted, although the output of the re-solicited network is slower and weaker than that produced by a rested system. 13.3.1 Swimming CPG Network Plasticity in Xenopus Frog Tadpoles

Recent research demonstrates how the relatively simple CPG controlling swimming in Xenopus tadpoles achieves, in an experience-dependent manner, the trade-off between endurance and performance. In immobilized, minimally dissected preparations the tadpole spinal cord generates a rhythmic, centrally generated motor output (Fig. 13.1A) in

A

B

30s 1s

5s

30s

C

Episode duration (sec)

Plasticity and Learning in Motor Control Networks

20 ms

VR 20 sec

Ei

10 mV

usAHP

dIN MN

cIN left

6 4 2 0

0

0

5 10 15 20 25 30 35 Inter-swim interval (sec)

Action potential number 50 100 150 200

−1 −2 −3 −4 −5 −6

E ii

midline

to muscle

8

1

usAHP amplitude (mV)

20 mV

Spinal neuron

D

10

MNs = 39/67 cINs = 18/28

dIN

to muscle

(aINs = 8/16)

MN

cIN

dINs = 0/31 right

Figure 13.1 Na+ pump-based mechanism for short term memory of locomotor activity in Xenopus tadpoles. (A) The duration of episodes of rhythmic fictive swimming, recorded from left and right spinal ventral roots, depends on the interval (stated in seconds at left of traces) from the preceding bout; short episodes follow short inter-swim intervals. (B) Graph showing linear relationship between swim episode interval and swim episode duration. (C) Patch-clamp recording from a spinal neuron during a fictive swim episode. Note that resting potential after swim bout termination is ∼10 mV more hyperpolarized than before swimming began due to an ultraslow (us) AHP. Inset is expanded excerpt of four cycles of neuron firing in time with ventral root activity. (D) usAHP amplitude correlates linearly with the number of action potentials that precede it. (Ei) Neurons in swim CPG. Circles depict populations of neurons; boxes depict left and right CPG half-centers. (Eii) Distribution of usAHP in spinal neuron classes within swim CPG network. dIN, descending interneuron; cIN, commissural interneuron; aIN, ascending interneuron; MN, motorneuron. B-D from Zhang and Sillar (2012) with permission; Ei schematic kindly prepared by L. Picton.

421

422

Neurobiology of Motor Control: Fundamental Concepts and New Directions

which MN output to the axial muscles alternates across the spinal cord and propagates from head to tail. This CPG rhythm, so-called “fictive swimming”, would produce the tail oscillations that normally propel the animal forward during actual behavior. In addition to possessing the cellular and synaptic components necessary for rhythmogenesis, the network is also endowed with an intrinsic mechanism that confers activity-dependent plasticity on swimming output; the duration of sensory-evoked swimming episodes is directly proportional to the interval from the end of the last swim bout (Fig. 13.1A, B; Zhang and Sillar 2012). Short inter-swim intervals produce short subsequent episodes with a progressive gradation up to prolonged episodes following long rest intervals. Importantly, this short term memory operates over durations of one to two minutes. If the interval from the end of one episode to the start of the next is greater than two minutes there is no influence of the first episode and from then on the locomotor network is effectively completely rested. A minute or two may not seem a long time to us but since tadpoles swim at up to 20 Hz, this time period is equivalent to up to 2400 cycles of locomotion. The mechanism for this short term memory involves the plasma membrane Na+ /K+ exchange pump, referred to here as the Na+ pump. That this ubiquitously expressed protein should contribute to short term memory is initially surprising given that the textbook role of Na+ pumps is to homeostatically maintain cation gradients across cell membranes. The initial evidence that Na+ pumps play a role in regulating CPG output was that inhibiting pump activity breaks down the linear relationship between episode interval and duration (Zhang and Sillar 2012). Thus, the presence of low concentrations of the pump inhibitor ouabain leads to prolonged swim bouts that do not self-terminate, and a similar result is obtained in nominally zero K+ saline, which removes the driving force for pump function. Na+ Pumps Induce a Prolonged Post-Swim After-Hyperpolarization, the usAHP Patch clamp

recordings from spinal neurons reveal that fictive swimming episodes often terminate with their membrane potential dropping to a resting level substantially more hyperpolarized, by up to 10 mV, than before the episode began (Fig. 13.1C). The membrane potential then slowly recovers towards pre-swim levels over the course of a minute, on average. This post-swim event is termed the ultraslow afterhyperpolarization (usAHP). The usAHP is a cellular rather than a network property because it can be triggered by high frequency firing induced by direct current injection into a recorded neuron. Furthermore, usAHP amplitude is directly related to spike number (Fig. 13.1D), and when sodium spikes are blocked with tetrodotoxin (TTX), no amount of sustained depolarization will produce a usAHP. The usAHP is thus a very accurate spike-counting device. The possible involvement of the Na+ pump in the usAHP was initially inspired by an earlier report of a similar phenomenon in Drosophila larval MNs in which their post-firing hyperpolarization was shown not to be associated with a change in membrane conductance (Pulver and Griffith 2010). Similarly, in tadpole neurons no conductance change is observed. This observation largely (though not completely) rules out membrane ion channel opening or closing during the hyperpolarization because this should cause a change in input resistance. An alternative explanation is that an ion pump is involved, since this would not change membrane conductance (Na+ pumps generate a pump current, but do not alter membrane input resistance). Importantly, this would in

Plasticity and Learning in Motor Control Networks

turn mean that there is no shunting of membrane currents during the change in membrane potential. The TTX experiments showing that sodium spikes are required to generate a usAHP suggested that CPG neurons were detecting the increased intracellular sodium concentration this firing would cause. Na+ pumps act as sodium detectors as a central part of their role as homeostatic devices that maintain transmembrane ion gradients. Na+ pumps keep intracellular sodium levels low and intracellular potassium levels high by extruding three Na+ ions and importing two K+ ions per transport cycle. Cycle rates vary as a function of ion concentration gradients and reportedly can be as high as 200 Hz. The increased intracellular Na+ following high frequency firing could therefore increase Na+ pump cycle rate (Fig. 13.2). Because pump activity results in a net extrusion of

long episode

A

CPG neuron

RP K+ K+ + K+K

sensory stim.

ext int

20 mV

Na+ Na+ Na+ + Na+ Na + Na

B

5 sec

short episode

CPG neuron

RP K+ K+ + + + K+ K +K K K K+

sensory stim.

Na+ Na+ Na+ + + Na+ Na Na + Na+ Na Na+ Na+ Na+ Na+

Figure 13.2 Role of Na+ /K+ pump-mediated usAHP in recalling past locomotor network performance. (A) Following an episode of fictive swimming, a spinal CPG neuron expresses a usAHP that decays before the onset of a subsequent sensory stimulus-evoked bout of swimming. The usAHP arises from Na+ /K+ pump activation due to spike-mediated sodium entry during the preceding network activity. Neuron membrane potential eventually returns to rest level (dashed line; RP, resting potential). (B) After an intense bout of swimming, pump operation is enhanced by the elevated levels of intracellular Na+ , resulting in a larger usAHP that has insufficient time to decay before the start of the next evoked swim episode. The persistent usAHP memory trace reduces CPG network excitability and subsequent episode duration is decreased. From Simmers (2012) with permission.

423

424

Neurobiology of Motor Control: Fundamental Concepts and New Directions

positive change (3 Na+ out for only 2 K+ in per cycle), the increased pump activity would drive the membrane potential to a more negative level, creating, la voila, a post firing usAHP. This elegantly simple mechanism thus promotes the ubiquitous Na+ pump from a mere housekeeping enzyme to an efficient spike counting mechanism that sets a neuron’s resting membrane potential according to the intensity of immediately preceding firing (compare Fig. 13.2A and B). Targeted Distribution of usAHP in Spinal Neurons A notable advantage of the tadpole CPG

network is its relative simplicity and accessibility, which has allowed all its main spinal neuron classes and their synaptic interactions to be characterized by paired patch recordings (reviewed in Roberts et al. 2010). The basic rhythm generating circuitry, schematized in Fig. 13.1Ei, relies on the firing of excitatory glutamatergic interneurons with descending projections on the same side as their somata. These dINs excite all ipsilateral neurons of the motor system. The left and right sides of the spinal cord are reciprocally coupled by glycinergic inhibitory commissural interneurons, the cINs. The axial MNs on each side of the cord are thus driven to fire in each cycle by inputs from ipsilateral dINs and prevented from firing by cINs when the other side of the cord is active. A fourth class of neurons that are rhythmically active during locomotion are the ascending interneurons, aINs, which are inhibitory and provide an ipsilateral signal that gates sensory information during swimming. Cataloguing the incidence of the usAHP among these four classes of CPG neurons has revealed an interesting and presumably functionally important differentiation within and between cell types, pointing to a hitherto unknown cell property heterogeneity linked to pump-driven network plasticity. The dINs, the neurons responsible for excitatory driving in swimming, are uniquely identifiable on the basis of their anatomy and physiology. They have depolarized resting membrane potentials (∼ −50 mV) compared to other CPG neurons (∼ −60 mV). They have broad action potentials and, both during swimming and in response to current injection, due to Na+ channel inactivation, fire only single spikes upon depolarization. Thus, during swimming, the dINs require the mid-cycle cIN inhibitory synaptic potentials to hyperpolarize them and remove Na+ channel inactivation before they can rebound fire. To date, no dIN has ever displayed a usAHP after swimming or repetitive firing induced by current injection (Fig. 13.1Eii). This lack presumably stems from dINs having a different Na+ pump complement than other spinal CPG neurons, which could also explain their relatively depolarized resting membrane potential. With respect to the other CPG neurons, based on more than 100 recordings so far, the majority (∼58%) of, but not all, MNs, aINs, and cINs display usAHPs (Fig. 13.1Eii). A likely explanation for this heterogeneity is the subunit composition of the pumps, especially with respect to the α subunit (Aydemir-Koksoy 2002; Bøttger et al. 2011). α1-containing sodium pumps have a high affinity for intracellular sodium and are maximally activated in the resting state (Dobretsov and Stimers 2005) allowing them to contribute continuously to the resting membrane potential. However, pumps containing the α3 subunit have a very high threshold for activation and only large increases in intracellular sodium, for example resulting from high frequency action potential firing, are sufficient to engage them (Blanco and Mercer 1998; Dobretsov and Stimers 2005; Azarias et al. 2013). Once activated, the additional extrusion of 3 Na+ for 2 K+ hyperpolarizes the membrane potential in an activity-dependent manner.

Plasticity and Learning in Motor Control Networks

A working hypothesis therefore is that neurons displaying a dynamic usAHP contain the α3 subunit while those that do not possess only α1. Another possible contributory factor is the involvement of auxiliary pump subunits called FXYD proteins, which span the membrane in close association with the pump protein and which when bound inhibit it (Geering 2006; Pavlovic et al. 2013). FXYD proteins can be phosphorylated at one or more of three phosphorylation sites located on the intracellular side of the protein, at which time FXYD uncouples from the main protein, causing dis-inhibition. Although speculative, intense spiking activity could change the FXYD phosphorylation state, for example by neuromodulation or via Ca++ -activated processes stimulated by Ca++ entry through voltage-dependent channels. If this FXYD pathway were differentially present in selected CPG neurons it could explain the heterogenous expression of the usAHP in different neuron types. Interaction between usAHP and Swim Properties It seems logical that hyperpolarization of

some spinal CPG neurons by usAHPs would negatively impact network rhythm generation, but two aspects of this process remain unexplained. First, this impact is never sufficient to completely abolish rhythm generating capability. No matter how short the interval between swim bouts, swimming can always be reactivated, albeit for only a few cycles at the shortest inter-swim intervals (e.g., Fig. 13.1A, middle panel). This retention of a residual swim capacity is likely because of the lack of usAHPs in dINs, which therefore remain at the same level of excitability and can continue to contribute to rhythm generation. The second aspect is why hyperpolarization of the other neuron types reduces the extent to which dINs can drive the rhythm when swimming is too quickly re-initiated. One possibility relates to the cINs. Their residual hyperpolarization will likely cause them (like other non-dIN CPG neurons) to gradually reduce their firing as the new episode proceeds and so reduce the amplitude of their mid-cycle dIN inhibition. dIN firing relies on rebound from cIN inhibition, and silencing the cINs on one side of the cord using an optogenetic approach abruptly terminates swimming on the opposite side of the cord (Moult et al. 2013). The reduced firing of cINs and other CPG neurons during the usAHP is linked to the deinactivation of an A-type potassium channel (caused by the usAHP hyperpolarization) which then delays the onset of firing, slowing swimming in subsequent cycles and thereby reducing the duration of activity (Zhang et al. 2015). A reasonable working hypothesis is thus that usAHPs in a portion of spinal neurons leads to cycle-by-cycle drop-out of cINs from the network. This progressive reduction in mid-cycle dIN inhibition eventually falls below the level required to trigger rebound dIN firing, and so network activity ceases. These ideas need experimental examination, and are also a situation in which computer simulation could be extremely useful in guiding the experimental work. 13.3.2 Comparative Aspects of Na+ Pump Contribution to Neural Network Function

The role of a Na+ pump in Xenopus tadpole swimming was largely inspired by a remarkably similar finding made in Drosophila larval MNs (Pulver and Griffith 2010). In this system a Na+ pump-dependent, long-lasting MN hyperpolarization impacts the larval crawling motor pattern. Na+ pumps also appear to play similar functional roles in mammalian spinal locomotor networks (Ballerini et al. 1977; Picton et al. 2017) and the leech

425

426

Neurobiology of Motor Control: Fundamental Concepts and New Directions

heartbeat CPG (Tobin and Calabrese 2005). These data suggest that Na+ pumps may be widely involved in motor network control and plasticity. Pump-mediated control of neuron electrical properties has also been described in, for example, sensory neurons in lamprey spinal cord (Parker et al. 1996), dopaminergic neurons in mouse brain (Johnson et al. 1992), cerebellar Purkinje cells (Forrest et al. 2012), and cortical pyramidal neurons (Gulledge et al. 2013). In addition to its originally defined role in ionic homeostasis, the ubiquitous Na+ pump thus appears to be also widely used for integrative functions. We predict that many more examples of hitherto unsuspected Na+ pump roles will be discovered in the years to come, most likely when, as in the Xenopus swimming example given here, system performance limits are being tested (and hence neuronal firing is most changing), as in anoxia or hyperthermia. Defects in Na+ pump function are also involved in aging and several diseases (Amaiz and Ordiers 2014), including Parkinsonian-like dystonia (Bøttger et al. 2012).

13.4 CPG Network Plasticity and Motor Learning Conferred by Operant Conditioning Goal-directed behaviors, such as eating and sexual activity, are also generated by intrinsically driven motor commands that are adjusted through behavioral experience and learning. One such process is operant conditioning, an associative learning procedure by which an animal learns from the consequences of its own behavior. In operant conditioning, a contingent association is made between a specific goal-directed act (the operant) and a reinforcing (rewarding) or aversive (punishing) outcome (Skinner 1981). As a result of positive reinforcement, operant conditioning leads to the rewarded behavior being generated more frequently and in a stereotyped recurrent manner, and ultimately can result in a habitual or even compulsive-like behavioral expression that persists after removal of the rewarding stimulus (Balleine 2005; Everitt and Robbins 2005, 2016). Conversely in an operant-negative reinforcement relationship, an animal learns through unfavorable behavioral consequences to diminish the expression of a particular act, thereby decreasing the likelihood of further confrontation with the punishing stimulus (Reynolds 1975). Although the neurobiological basis for appetitive and aversive classical conditioning have been extensively studied, the way in which reinforcing or punishing stimuli influence the central auto-active networks responsible for operant behavior has proven much more elusive. Organisms with simpler and more tractable nervous systems, such as the marine mollusk Aplysia, have provided important insights into the cellular and network mechanisms underlying associative learning (reviewed in Brembs 2003; Hawkins and Byrne 2015). In its search for food, the herbivorous Aplysia spontaneously performs exploratory biting movements involving cyclic protractions and retractions of its tongue-like radula (see also Fig. 12.4). In the absence of ingestible food these radula actions are produced relatively infrequently and with highly variable, unpredictable inter-bite intervals (Horn et al. 2004; Hooper 2004; Nargeot et al. 2007; Cullins et al. 2015). This autonomous goal-directed behavior can be sustainably modified by different forms of appetitive learning, including operant conditioning (Brembs et al. 2002; Lorenzetti et al. 2006; Nargeot et al. 2007). After a 30 to 40-min period of contingent

Plasticity and Learning in Motor Control Networks

A

in vivo

in vitro

Training

Neuron identification

B

Protractor Retr. Closure

Test

C

I2 n. n.2,1 R n.

Control (no food reward) B51

B30

Contingent food reward

B30 B63

B64

B63

B52

Non-contingent food reward B65

D

Control

B4/5

Contingent

B65 5 sec

Non-contingent

I2 n. n.2,1 R n. B63 B30 B65 1 min

Figure 13.3 Neural correlates of in vivo operant conditioning of Aplysia feeding behavior. (A) Experimental protocol. Three groups of animals were trained for 30 min: a naive (control) group in which animals were not presented with food during spontaneous radula biting (indicated by vertical bars), a contingent group in which animals were delivered food (arrowheads) at each bite, and a non-contingent group where food was delivered at regular intervals independent of bite occurrences. (B) Schematic of the buccal CPG that drives protraction (Protractor), retraction (Retr), and closure (Closure) phases of each bite cycle. Rectangle offsets indicate when each act occurs in the motor pattern. Electrical synapse, resistor symbol; chemical inhibition/excitation, filled circles/triangles. (C) Extracellular recordings of a single buccal motor pattern (BMP; fictive bite (protraction, retraction and closure phases recorded from motor nerves I2 n, n.2,1 and Rn, respectively)) and associated bursts recorded intracellularly from the B63/B30/B65 neurons. Note that burst onset in all three neurons preceded BMP onset (vertical dotted line). (D) Repetitive BMPs and B63/B30/B65 bursting in control, contingent, and non-contingent preparations. Note faster, regularized bursting and BMP genesis in the contingent preparation. Vertical scale bars in C and D represent 25 mV and 30 mV, respectively. From Nargeot et al. (2009) and Sieling et al. (2014), both with permission.

reinforcement training during which an animal receives a food reward (such as a piece of seaweed) in association with each spontaneous radula bite (Fig. 13.3A), the variability of bite occurrences is drastically reduced, leading for several hours to an accelerated, more stereotyped rhythmic biting behavior. These behavioral changes do not develop when seaweed delivery is uncorrelated (i.e., non-contingent) with spontaneous biting movements. The plasticity thus emerges from a specific operant-reward association in which Aplysia memorizes the positive consequences (i.e., access to food) of its own actions. This behavioral transition from sporadic to faster, regularized autonomous actions is strikingly similar to reward-induced changes in motivated behaviors of more complex organisms, including

427

428

Neurobiology of Motor Control: Fundamental Concepts and New Directions

humans, where internally-driven urges (e.g., for food, sex, drugs) can become compulsive through contingent reinforcement in operant conditioning (Pitman 1989; Cardinal et al. 2004; Everitt and Robbins 2005; Balleine 2005). Cellular and Network Correlates of Aplysia Motor Learning The Aplysia feeding/biting CPG is

distributed between the bilaterally-paired buccal ganglia. Many of the network’s key components, along with their bioelectrical properties and synaptic connectivity, have been identified and characterized (Fig. 13.3B; Susswein and Byrne 1988; Hurwitz et al. 1997; Kabotyanski et al. 1998; Elliott and Susswein 2002). Two operational features of the buccal CPG have been pivotal in identifying the neuronal substrates of behavioral plasticity in this system. First, the network continues spontaneously to produce the motor output patterns for radula biting in vitro. Second, in vitro fictive biting conserves essential features of the behavioral modifications previously induced in vivo by food-related classical or operant conditioning (Susswein et al. 1986; Nargeot et al. 2007; Brembs et al. 2002; Lechner et al. 2000) and similar changes in biting motor activity can be induced by appropriate sensory nerve stimulation in vitro (Nargeot et al. 1997, 1999a,b; Brembs et al. 2004; Lorenzetti et al. 2006, 2008; Mozzachiodi et al. 2008; also see below). Combining associative learning paradigms with in vitro electrophysiological approaches has thus enabled determining the sites and actions of sensory signaling on membrane and synaptic properties of identified buccal CPG neurons and relating these specific celland circuit-wide influences to changes in the behavior the network produces. The radula biting motor program consists of repetitive sequences of burst activity in motor nerves controlling radula protraction, retraction, opening, and closure, with each phase controlled by specific subsets of buccal CPG interneurons. A cycle of such bursting, called a buccal motor pattern (BMP), corresponds to the motor drive for a single radula bite (Fig. 13.3C). During normal ingestive biting, which transports edible material through the buccal mass into the foregut, the left and right halves of the radula protract out of the buccal cavity to grasp the food which is then withdrawn into the cavity during retraction. When something inedible enters the buccal cavity, the radula closes around the morsel and ejects it by concomitant radula protraction. Thus, by altering the phase relationships between protraction/retraction and opening/closing, the buccal CPG can switch radula biting from ingestion to egestion depending on food nature and quality (Susswein et al. 1986; Morton and Chiel 1993). Aplysia feeding consequently manifests two fundamental facets of behavioral choice and decision-making seen in both simpler (Kristan 2008; Nargeot and Simmers 2012) and more complex (Schall 2005) animals: deciding whether to repeat a specific act (an ingestive bite) and selecting between alternative behaviors (ingestion versus egestion). Moreover, in Aplysia, the “choice” of radula action is intrinsic to the feeding CPG network itself since buccal ganglia continue in vitro to spontaneously and interchangeably produce the motor patterns underlying ingestion and egestion in vivo (Nargeot et al. 1997; Horn et al. 2004). Thus, the buccal CPG not only generates radula motor activity, but also determines which behavioral output will be expressed. Motor Learning-Induced Plasticity of Network Output Selection A key “decision-maker” in

selecting Aplysia biting acts is the identified CPG network neuron B51, which, depending on its biophysical state, switches the network between producing ingestive or egestive BMPs (Nargeot et al. 1999a). An intrinsic plateau-generating capability (see

Plasticity and Learning in Motor Control Networks

Chapter 8.5) allows B51 to shift, in a regenerative, all-or-none manner, between silence and an active state in which the cell fires intensely. Through its excitatory connectivity in the CPG (Fig. 13.3b), and particularly with cells of the retractor generating subset, when B51 plateaus it promotes retraction and closure MN co-activation, and thus the production of ingestion-type BMPs (Fig. 13.3C). Alternatively, B51 silence results in closure MNs being driven principally by the protraction CPG subset, preventing ingestive, and promoting egestive, BMPs (Nargeot et al. 1999b). Thus, although B51 doesn’t contribute to cycle-by-cycle initiation of feeding BMPs, switching between the neuron’s two functional states determines the type of output the buccal CPG produces. B51’s ability to select between biting programs is modified by associative learning processes. This capability was discovered with in vitro and subsequent in vivo operant appetitive learning paradigms in which electrical stimulation of an esophageal sensory nerve (En2 ), believed to convey information about the presence of ingested food, was used to mimic contingent food-reward reinforcement in actual feeding (Nargeot et al. 1999a,b; Brembs et al. 2002). Both in vitro operant training, in which En2 stimulation was timed to each spontaneous BMP produced by the isolated buccal CPG (Nargeot et al. 1999a), and in vivo training, in which nerve stimulation was timed to radula bites in freely behaving animals (Brembs et al. 2002), produced long-lasting (several hours post-training) enhancement of ingestion-related biting. This contingent reinforcement was associated with an increase in input resistance and a decrease in plateau threshold of the B51 neuron, which together increased the cell’s excitability, the probability of plateau production, and hence the likelihood of the buccal CPG producing ingestive rather than egestive BMPs. In an experimental tour de force, the learning-induced excitability changes in B51 were then expressed in an isolated cell analogue of appetitive operant conditioning in which spontaneous plateau potentials in cultured B51 neurons were contingently reinforced with iontophoretically-applied pulses of dopamine (Mazzachiodi et al. 2008), the transmitter mediating food input signals from En2 in vivo. Furthermore, cAMP injection into naive isolated B51 neurons mimicked the excitability changes induced by single cell operant conditioning (Lorenzetti et al. 2008; Mozzachiodi et al. 2008) and cAMP injected into B51 neurons in situ increased their bursting activity and the frequency of ingestive BMPs (Mozzachiodi et al. 2008). These data, together with evidence for an associative convergence (upstream of cAMP production and acting through adenylyl cyclase) of activity and dopamine receptor-induced signaling cascades involving, respectively, PKC and PKA (Lorenzetti et al. 2008), have begun to reveal the molecular processes by which memory storage can be inscribed in the bioelectrical properties of key decision-making components of central motor control circuitry (Mazzachiodi et al. 2010). Motor Learning-Induced Plasticity in Network Output Initiation Much is thus known at the neu-

ronal and molecular levels about how operant learning affects what type of radula bite Aplysia produces. We now cover learning-induced plasticity of when each bite occurs. Since appetitive operant conditioning increases bite regularity and frequency (above and Nargeot et al. 2007), the first step was to identify the buccal CPG neurons that trigger the initial protraction phase of each bite. This work revealed that BMPs are triggered by spontaneous bursts in any of the bilaterally-paired and electrically-coupled

429

430

Neurobiology of Motor Control: Fundamental Concepts and New Directions

B63, B30, and B65 interneurons (Nargeot et al. 2009) (Fig. 13.3B). Which of these neurons fires first in any bite is arbitrary. The first firing neuron excites the other two types via the group’s electrical coupling, and once both B63 neurons fire, on which the actual pattern-initiating process strictly depends, a bite occurs (Nargeot et al. 2009). Thus, randomly distributed burst activity in this kernel of decision-making neurons underlies the slow and erratic motor output commands that drive a hungry Aplysia’s trial and error sampling of its environment in search of food. The decision-making B63/B30/B65 subset in vitro continues to express a memory trace of associative experiences of the intact animal. In ganglia isolated from untrained control or non-contingently trained animals, the three cell types produce erratic and weakly coordinated bursts, with a given BMP being instigated by burst onset in any of the three cell types (Fig. 13.3D, left, right). However, in buccal ganglia from contingently-rewarded animals, the three neuron types produce faster, stereotyped, and tightly coordinated rhythmic bursts with B63 invariably bursting first in each fictive bite cycle (Fig. 13.3D, middle). Thus, cellular correlates of the increased frequency and regularization of biting induced by operant conditioning in vivo are expressed in a subset of electrically-connected buccal CPG neurons that are necessary (B63) or sufficient (B30, B65) for bite initiation. Moreover, pharmacological (Nargeot et al. 1997; Reyes et al. 2005) and electrophysiological (Bédécarrats et al. 2013) data revealed that, as with B51, the reinforcing food-reward signal to B30, B63, and B65 is dopamine release from esophageal En2 sensory nerve fibers. These findings have wider implications for understanding decision-making processes and identifying convergence points of operant behavior and dopamine-mediated reward (Balleine 2005), since they show that learning-induced neural plasticity underlying motor act selection (ingestion vs. egestion by B51) and motor act initiation (by B63/B30/B65) can be encoded at different cellular loci. It is instructive to compare reward and decision-making in Aplysia to operant learning in more complex systems such as the corticostriatal brain circuitry of rodents and primates. Here, decision-making processes involving variably expressed behavioral acts that are goal-directed and sensitive to the outcome of rewarding feedback are mediated by cortical pathways that are distinct from striatal circuitry mediating the stimulus-response associations in the formation of relatively automatic, habitual actions (Balleine 2005; Balleine and O’Doherty 2010; Everitt and Robbins 2016). This is in direct contrast to Aplysia, where persistent transformation of an unpredictable, reward-reinforced act into stereotyped, habitual-like rhythmic expression is accomplished by functional plasticity within a single neural network. In this case, moreover, the operant learning occurs in a discrete neuronal subset of the feeding CPG itself, rather than arising from distributed excitability changes across more extensive central circuitry (Kemenes et al. 2001) or in specialized command-like neurons upstream to the CPG (Kristan 2008). Cellular and Synaptic Mechanisms for Aplysia Motor Learning Further electrophysiological

exploration has pinpointed the cellular sites at which the operant memory in the B63/B30/B65 neurons is induced and maintained (Nargeot et al. 2009). First, neuron excitability, defined experimentally as the minimum amount of injected depolarizing current required to reach spike threshold, is higher in all three neuron types after contingent training because of an increase in their membrane input resistances. Second,

Plasticity and Learning in Motor Control Networks

the electrical coupling between the neurons becomes stronger due to an increase in the actual junctional coupling between neuron pairs (rather than being an indirect consequence of their increased membrane resistances). These two correlative changes thus explain how the learning-induced transition from weakly to strongly coordinated bursting among the bite-initiating neuron subset could be achieved. The synchronized neuronal and regularized CPG activity of the Aplysia buccal motor system after operant learning shares features with CPGs responsible for more conventional rhythmic behaviors (see Chapter 8). In the pyloric network of the crustacean stomatogastric nervous system, rhythmogenesis arises from the endogenous oscillatory activity of a tightly electrically-coupled cell subset consisting of the AB and 2 PD pacemaker neurons (Selverston and Miller 1980). The fastest oscillator, AB, entrains synchronized bursting in the 2 PD neurons and together they drive the rest of the network in a stereotyped triphasic motor pattern (Miller and Selverston 1982; Bal et al. 1988). In the absence of modulatory or sensory input, pyloric pacemaker neuron rhythmicity, and thus pyloric CPG output, is very stable. When initially searching for food, Aplysia biting is instead highly irregular. This irregularity derives from network elements that are able to choose whether to instigate individual cycles of motor activity (neurons B63/B30/B65) and, if so, which output pattern to produce (neuron B51), with none of these neurons, in this network state, being regular bursting neurons. However, once food is detected, operant conditioning results in highly rhythmic (due to accelerated, regularized, and coincident B63/B30/B65 bursting) and primarily ingestive (due to increased B51 excitability) biting, effectively transforming feeding network operation into that of a more typical rhythmogenic CPG. The correlation between neuronal and behavioral changes following Aplysia operant conditioning does not prove the two are causally related. Therefore, computer-simulated ionic currents generated by the dynamic clamp hybrid technique (Prinz et al. 2004) was used to test causality by artificially reproducing or suppressing the membrane and synaptic plasticity associated with the motor learning (Fig. 13.4A, B; Sieling et al. 2014). With in vitro preparations from untrained animals, artificially increasing B63/B30/B65 membrane excitability (Fig. 13.4A, right) increased the frequency of fictive biting but not its regularity. Dynamic clamp-imposed increases in electrical coupling strength alone (Fig. 13.4B, right) increased BMP regularity but not frequency. However, simultaneously-imposed changes in both membrane and synaptic properties reproduced both these aspects of the motor plasticity induced by in vivo operant conditioning (Fig. 13.4C). Conversely, in preparations from trained animals, suppression of membrane excitability and synaptic plasticity using dynamic clamp abolished the learning-induced increases of both frequency and regularity (Fig. 13.4D). These findings are important because they demonstrate that plasticity in two independent cellular properties—voltage-independent membrane conductances that determine cell excitability and electrical synapse junctional conductivity—are separately responsible for complementary facets of changes in a goal-directed action resulting from dopamine-mediated operant learning. Gaining such a mechanistic understanding of operant conditioning in a simpler mollusk’s feeding behavior provides potentially important insights into the neuronal basis of experience-related plasticity in general, including how decision-making processes about action selection and initiation are regulated by learning, or even deregulated in the development of certain behavioral disorders.

431

432

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

Gleak: 0

V I

-60 nS

B

Gleak

Gjunc.: 0

V I

Gjunc.

+20 nS

B65 B63

V

B63 i

C

I 2

3

i

2 3 nA

–10

–10 nA

Control Gleak / Gjunc.:

0 / 0 nS

-60 / -10 nS

B63 B30 B65

D

Contingent Gleak / Gjunc.:

0 / 0 nS

+60 / –10 nS

B63 B30 B65

Figure 13.4 Use of dynamic clamp methodology to determine whether changes in specific conductances that govern cell excitability and electrical coupling are causal to associative learning-induced changes in Aplysia buccal CPG output and feeding behavior. (A) Left, experimental protocol for adding a dynamic clamp-defined leak conductance to a single B63 neuron of a naive control preparation. Introducing an artificial negative Gleak of –60 nS decreased cell input conductance thereby reproducing the increased excitability (evidenced by a decrease in spike threshold) occurring after operant conditioning. Horizontal and vertical scale bars represent 2 s and 20 mV, respectively. (B) Left, protocol for adding a dynamic clamp-defined junctional conductance (Gjunc ) to the natural coupling conductance between two CPG neurons. Introducing a positive Gjunc (20 nS) into a pre-junctional B63 strengthened its electrical coupling with post-junctional B65 to a level found after operant conditioning. Horizontal and vertical scale bars represent 2 s and 2 mV (B65) or 20 mV (B63), respectively. (C) Concomitant experimental increases in both the excitability of the BMP initiating neurons and their electrical coupling replicated conditioning-induced increases in burst frequency and regularity in a naive preparation. (D) Selectively decreasing both the excitability and coupling of these neurons in a trained preparation reverted spontaneous burst generation to an irregular and infrequent, untrained phenotype. From Sieling et al. (2014) with permission.

13.5 Discussion and Conclusions Our aim in this chapter was to summarize current knowledge on plasticity and learning in motor control networks, focusing specifically on two case studies: the circuits that control Xenopus tadpole swimming and feeding in the sea hare, Aplysia. The two

Plasticity and Learning in Motor Control Networks

examples contrast in that the plasticity we describe in the tadpole spinal CPG underpins relatively short-term memory, leading to experience-related alterations in the duration of locomotor bouts. In Aplysia on the other hand, the plasticity involves longer-term motor learning in which the temporal structure of an ongoing, spontaneously generated motor program is modified by external contingencies and the animal’s gustatory experiences. The relatively simple behaviors and learning plasticity produced by the defined and accessible nervous systems of invertebrate and simpler vertebrate models such as Aplysia and Xenopus offer ideal models for investigation. The small and/or segmentally reiterated central circuits of such animals not only enable detailed characterization of their cellular and network properties, but also understanding how these properties are acutely adjusted by sensory and/or modulatory influences, and more persistently by experience and learning. Our focus here has deliberately overlooked the substantial literature on motor learning, especially in mammals, in which extrinsic supraspinal circuitry, particularly of the motor cortex and cerebellum, play major roles. However, with the increasing recognition that neural plasticity within the spinal cord, or invertebrate equivalent, is also an important contributor to the acquisition and maintenance of motor learning (for reviews, see Wolpaw 2010; Edgerton et al. 2008; Thompson and Wolpaw 2014; Brownstone et al. 2015; also see below), we wished to focus on how the CPG circuitry directly responsible for rhythmic behavioral output is itself capable of playing a major role. Our two case studies emphasize how network plasticity depends on the precise molecular make-up and resulting heterogeneity of individual neurons within the network. Furthermore, although the CPG networks controlling rhythmic behaviors share similar general design principles (e.g., mutual excitation between synergists, reciprocal inhibition between antagonists, intrinsic oscillatory properties), even in different individuals of the same species, differences in genetic background and experience render each individual’s networks unique. These differences may not be apparent during normal network operation (a walk in the park), but instead are revealed only at performance extremes (running while being chased by a hungry predator). Motor Learning in Maturing and Adult Networks How motor networks assemble during

development is fundamental to the way they function in adulthood. At birth, animals already possess a motor circuit infrastructure that allows a range of vital behavioral functions, such as respiration, feeding movements, and, in many species (obviously excluding humans), self-supporting posture and locomotion. Subsequently, the motor programs present at each developmental stage are calibrated and refined by maturational and learning processes acting in combination. In this way, the imprecise grasping reflex of the human baby’s hand is refined to dexterous independent finger manipulation, and the rudimentary walking program of the toddler becomes shaped by experience. Importantly, however, learning-induced improvement or modification of a given motor task can only occur once the underlying motor infrastructure is sufficiently mature (Forssberg 1999; Grillner and Wallén 2004). A baby can be trained to walk prematurely, but not to skip or tie its shoelaces, and further maturation is required before training can add new actions to a child’s behavioral repertoire such as climbing trees or playing the piano.

433

434

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Neural plasticity associated with motor learning is not constrained only to a maturational timetable. Long-term adaptive changes in CNS network structure and function instead continue to occur throughout normal adult life in response to behavioral experience and learning. In adulthood, motor learning continuously enables humans to acquire and retain the capacity to execute new tasks, and even to perform them concurrently. For example, some of us can ride a uni-cycle, juggle balls, and sing the national anthem simultaneously. The acquisition of such motor skills might seem far removed from the ability of Xenopus tadpoles or Aplysia to adjust their respective behaviors according to past experience, but the underlying neurobiological principles could be essentially similar. CPG networks in general are not immutable entities; depending on acute sensory or modulatory influences, individual neurons or microcircuits can be dynamically recruited into, or excluded from, a given network in order to generate different forms of the same behavior, or even distinct behaviors (for reviews see Marder and Bucher 2007; Harris-Warrick 2011; see also Chapter 8.9). Thus, when specific movements are required (such as running from a predator rather than merely walking) descending corticospinal pathways can call upon different subsets of spinal circuitry with functional synergies appropriate for achieving that specific task. Presumably, therefore, just as the hungry Aplysia learns through positive reinforcement to specify the feeding circuit configuration necessary for reliable injestive biting, when we have acquired a different behavioral skill (juggling, for example), we recall an appropriate combination of limb motor control subcircuits that endures in the absence of continued training. In other words, animals learn and remember new motor acts through a memorized ability to dynamically reconfigure a pre-existing network ensemble, rather than merely to activate hardwired circuitry that had been previously created to perform the new task. How the memory trace for such motor circuit orchestration is encoded by the CNS remains poorly understood, although a variety of experience-specific changes have been identified in supraspinal motor regions (basal ganglia, cerebellum, motor cortex) as well as in the spinal cord. In learning related to voluntary motor tasks such as limb reaching, fine digit movement or postural control, memory storage processes range from reorganization of synaptic connectivity and transmitter properties to experience-dependent changes in membrane conductances and neuronal excitability (for reviews see Adkins et al. 2006; Wolpaw 2010). Presumably such multisite processes, as well as hitherto unsuspected mechanisms like ubiquitous Na+ pumps, are also involved in encoding the memory trace of adaptive plasticity in CPG-driven motor behaviors where extrinsic sensory and brain inputs must integrate with the intrinsic activity of rhythmogenic circuitry. Neural Plasticity and Motor System Adaptation Following Injury A major neurological issue

is to try to restore important motor functions that have been impaired by injury or disease. Here again, the motor functions that rehabilitation attempts to restore depend on CNS plasticity, particularly in the spinal cord as the essential, final common pathway for most motor behaviors. Current therapeutic methods employed to improve or restore a locomotor capability, for example after spinal cord injury, commonly include repeated exercise, such as treadmill walking (Edgerton et al. 2008; Rossignol et al. 2009; Harkema et al. 2012; Martinez et al. 2013). Such rehabilitative motor training, also in conjunction with pharmacological (Chau et al. 1998; Antri et al. 2003) or electrical

Plasticity and Learning in Motor Control Networks

(Ichiyama et al. 2008) stimulation procedures can promote activity-dependent changes in sensorimotor circuitry below the spinal injury site (Petruska et al. 2007; Côté and Gossard 2004; Rossignol and Frigon 2011) and lead to specific improvements in stepping patterns. In the complete absence of supraspinal inputs to the lumbosacral cord region in adult rats, coordinated, weight-bearing locomotor movements that resemble normal stepping can be expressed by otherwise paralyzed hindlimbs through locomotor training with a combination of serotonergic agonist administration and epidural electrical stimulation (Courtine et al. 2009). Evidently, the successful recovery of hindlimb locomotor movements relies on both the inherent rhythm generating capability of lumbosacral CPGs and the ability of sensory inputs to remodel this circuitry to produce appropriate stepping patterns in the absence of brain inputs. Thus, the stepping recovery derives from a post-lesional induction of new use-dependent functional states in which the spinal circuits learn the specific motor tasks that are being reinforced by training. In a similar vein, following a spinal hemisection several weeks earlier, cats are able to re-express hindlimb treadmill walking within hours of a further complete spinal section, a lesion that normally requires weeks before any locomotor recovery occurs (Barriére et al. 2008). Again this suggests that the lumbar CPG itself participates in the recovery process after the initial partial lesion when the remaining descending inputs are able to facilitate independent CPG operation, allowing it to function in an already autonomous state after the subsequent complete cord transection (Rossignol et al. 2009). In animals and humans alike, operant conditioning of spinal reflex pathways can also induce spinal neural plasticity (Chen and Wolpaw 1995; Thompson et al. 2009) with consequent implications for improving locomotor performance after spinal cord injury. For example, in rats in which a lateralized cord injury produces an asymmetry in walking gait, appropriate operant conditioning of the soleus H-reflex on the injured side eliminates the asymmetry and restores relatively normal locomotion (Chen et al. 2006). The H reflex is the electrical analogue of the spinal stretch reflex (SSR) elicited by direct electrical stimulation of muscle spindle 1a afferents that synapse with α-MNs (see also Chapter 9). Reflex operant conditioning experiments have shown that rats, monkeys, and humans can gradually increase or decrease the SSR or the H reflex (reviewed in Thompson and Wolpaw 2014). Since the conditioned changes persist for several days after complete spinal cord transection (Wolpaw and Lee 1989), the adaptive plasticity must reside within the spinal cord itself. Underlying mechanisms include reflex-conditioning changes in GABAergic inhibition of spinal MNs as well as long-lasting alterations in MN firing threshold and axonal conduction velocity (Carp and Wolpaw 1994; Carp et al. 2001) due to changes in membrane sodium channels (Halter et al. 1995; reviewed in Wolpaw 2010). From a therapeutic perspective, it is relevant that such spinal reflex conditioning protocols have been recently found to improve the locomotor performance of humans with walking impairment caused by chronic incomplete spinal cord injury (Thompson et al. 2013). There is also growing evidence that nervous systems may respond to injury or insult by re-engaging earlier developmental mechanisms to compensate for loss of network function in adulthood. In the adult rat, chloride homeostasis is affected by spinal cord injury such that a long-term down regulation of the chloride co-transporter, KCC2, leads to a depolarizing (excitatory) rather than a hyperpolarizing (inhibitory) actions of the synaptic transmitters GABA and glycine (Payne et al. 2003; Vinay and Jean-Xavier 2008). This switch could participate in restoration of spinal CPG network function since it reflects

435

436

Neurobiology of Motor Control: Fundamental Concepts and New Directions

a recapitulation of an early developmental state of chloride homeostasis that may be a pre-requisite for recovery processes to occur (Rivera et al. 2005; Vinay and Jean-Xavier 2008). Similarly, the membrane properties of deafferented spinal MNs in the adult turtle gradually revert to a phenotype normally only found in embryonic MNs (Perrier and Hounsgard 2000; Perrier et al. 2000). This suggests that extrinsic inputs normally serve in both the acute regulation and the long-term maintenance of the biophysical properties of their target spinal neurons in the mature nervous system. Finally, in the zebrafish locomotor system described earlier, dopamine regulates spinal network differentiation, but the modulator can also play a role in adult motor circuitry after lesion by promoting regenerative plasticity and functional recovery (Reimer et al. 2013). Following complete spinal transection, a procedure from which even the adult fish eventually recovers and resumes apparently normal swimming behavior, the spinal cord generates additional MNs by activating the same dopaminergic system that was pivotal during initial development. The increased dopamine acts on the D4a receptor subtype on spinal progenitor cells that have lain dormant in the spinal cord, but can be subsequently re-activated to replace MNs lost following spinalization and thereby promote behavioral recovery of locomotor CPG function. In addition to providing new major insights into the capacity for neural plasticity within the adult nervous system, these findings have obvious important therapeutic ramifications for improving rehabilitation treatments for a variety of CNS lesions and diseases.

References Adkins DL, Boychuk J, Remple MS, Kleim JA (2006) Motor training induces experience-specific patterns of plasticity across motor cortex and spinal cord. J Appl Physiol 101:1776–1782. Amaiz GRL, Ordieres MGL (2014) Brain Na+ , K+ -ATPase activity in aging and disease. Int J Biomed Sci 10:85–102. Antri M, Mouffle C, Orsal D, Barthe JY (2003) 5-HT1A receptors are involved in short- and long-term processes responsible for 5-HT-induced locomotor function recovery in chronic spinal rat. Eur J Neurosci 18:1963–1972. Aydemir-Koksoy A (2002) Na+ , K+ -ATPase: a review. J Ankara Med Sch 24:73–82. Azarias G, Kruusmägi M, Connor S, Akkuratov EE, et al. (2013) A specific and essential role for Na,K-ATPase α3 in neurons co-expressing α1 and α3. J Biol Chem 288:2734–2743. Bal T, Nagy F, Moulins M (1988) The pyloric central pattern generator in crustacea: a set of conditional neuronal oscillators. J Comp Physiol 163:715–727. Balleine BW (2005) Neural bases of food-seeking: affect, arousal and reward in corticostriatolimbic circuits. Physiol Behav 86:717–730. Balleine BW, O’Doherty JP (2010) Human and rodent homologies in action control: corticostriatal determinants of goal-directed and habitual action. Neuropsychopharm 35:48–69. Ballerini L, Bracci E, Nistri A (1997) Pharmacological block of the electrogenic sodium pump disrupts rhythmic bursting induced by strychnine and bicuculline in the neonatal rat spinal cord. J Neurophysiol 77:17–23.

Plasticity and Learning in Motor Control Networks

Barrière G, Leblond H, Provencher J, Rossignol S (2008) Prominent role of the spinal central pattern generator in the recovery of locomotion after partial spinal cord injuries. J Neurosci 28:3976–3987. Bédécarrats A, Cornet C, Simmers J, Nargeot R (2013) Implication of dopaminergic modulation in operant reward learning and the induction of compulsive-like feeding behavior in Aplysia. Learn Mem 20:318–327. Blanco G, Mercer RW (1998) Isozymes of the Na-K-ATPase: heterogeneity in structure, diversity in function. Am J Physiol 275:F633–F650. Borodinsky LN, Belgacem YH, Swapna I (2012) Electrical activity as a developmental regulator in the formation of spinal cord circuits. Curr Opin Neurobiol 22:624–630. Borodinsky LN, Root CM, Cronin JA, Sann SB, et al. (2004) Activity-dependent homeostatic specification of transmitter expression in embryonic neurons. Nature 429:523–530. Bøttger P, Doganli C, Lykke-Hartmannet K (2012) Migraine- and dystonia-related disease-mutations of Na+ /K+ -ATPases: relevance of behavioral studies in mice to disease symptoms and neurological manifestations in humans. Neurosci Biobehav Rev 36:855–871. Bøttger P, Tracz Z, Heuck A, Nissen P, et al. (2011) Distribution of Na/K-ATPase alpha 3 isoform, a sodium-potassium P-type pump associated with rapid-onset of dystonia parkinsonism (RDP) in the adult mouse brain. J Comp Neurol 519:376–404. Brembs B (2003) Operant conditioning in invertebrates. Curr Opin Neurobiol 13:710–717. Brembs B, Baxter DA, Byrne JH (2004) Extending in vitro conditioning in Aplysia to analyze operant and classical processes in the same preparation. Learn Mem 11:412–420. Brembs B, Lorenzetti FD, Reyes FD, Baxter DA, Byrne JH (2002) Operant reward learning in Aplysia: neuronal correlates and mechanisms. Science 296:1706–1709. Brownstone RM, Bui TV, Stifani N (2015) Spinal circuits for motor learning. Curr Opin Neurobiol 33:166–173. Byrne JH, Hawkins RD (2015) Nonassociative learning in invertebrates. Cold Spring Harb Perspect Biol 7:a021675. Cardinal RN, Everitt BJ (2004) Neural and psychological mechanisms underlying appetitive learning: links to drug addiction. Curr Opin Neurobiol 14:156–162. Carp JS, Chen XY, Sheikh H, Wolpaw JR (2001) Operant conditioning of rat H-reflex affects motoneuron axonal conduction velocity. Exp Brain Res 136:269–273. Carp JS, Wolpaw JR (1994) Motoneuron plasticity underlying operantly conditioned decrease in primate H-reflex. J Neurophysiol 72:431–442. Chau C, Barbeau H, Rossignol S (1998) Early locomotor training with clonidine in spinal cats. J Neurophysiol 79:392–409. Chen XY, Wolpaw JR (1995) Operant conditioning of H-reflex in freely moving rats. J Neurophysiol 173:411–415. Chen Y, Chen XY, Jakeman LB, Chen L, et al. (2006) Operant conditioning of H-reflex can correct a locomotor abnormality after spinal cord injury in rats. J Neurosci 26:12537–12543. Côté MP, Gossard JP (2004) Step training-dependent plasticity in spinal cutaneous pathways. J Neurosci 24:11317–11327. Courtine G, Gerasimenko Y, van den Brand R, Yew A, et al. (2009) Transformation of nonfunctional spinal circuits into functional states after the loss of brain input. Nat Neurosci 12:1333–1342.

437

438

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Cullins MJ, Shaw KM, Gill JP, Chiel HJ (2015) Motor neuronal activity varies least among individuals when it matters most for behavior. J. Neurophysiol 113:981–1000. Demarque M, Spitzer NC (2010) Activity-dependent expression of Lmx1b regulates specification of serotonergic neurons modulating swimming behavior. Neuron 67:321–334. Dobretsov M, Stimers JR (2005) Neuronal function and 𝛼3 isoform of the Na/K-ATPase. Front Biosci 10:2373–2396. Edgerton VR, Courtine G, Gerasimenko YP, Lavrov I, et al. (2008) Training locomotor networks. Brain Res Rev 57:241–254. Elliott CJH, Susswein AJ (2002) Comparative neuroethology of feeding control in mollusks. J Exp Biol 205:877–896. El Manira A (2014) Dynamics and plasticity of spinal locomotor circuits. Curr Opin Neurobiol 29:133–141. Everitt BJ, Robbins TW (2005) Neural systems of reinforcement for drug addiction: from actions to habits to compulsion. Nat Neurosci 8:1481–1489. Everitt BJ, Robbins TW (2016) Drug addiction: updating actions to habits to compulsions ten years on. Annu Rev Psychol 67:23–50. Forrest MD, Wall MJ, Press DA, Feng J (2012) The sodium–potassium pump controls the intrinsic firing of the cerebellar purkinje neuron. PLoS One 7:e51169. Forssberg H (1999) Neural control of human motor development. Curr Opin Neurobiol 9:676–682. Geering K (2006) FXYD proteins: new regulators of Na-K-ATPase. Amer J Physiol – Renal Physiol 209:F241–F250. Glanzman DL (2006) The cellular mechanisms of learning in Aplysia: of blind men and elephants. Biol Bull 210:271–279. Grillner S, Jessell TM (2009) Measured motion: searching for simplicity in spinal motor networks. Curr Opin Neurobiol 19:572–586. Grillner S, Wallén P (2004) Innate versus learned movements—a false dichotomy? Prog Brain Res 143:3–12. Grossmann KS, Giraudin A, Britz O, Zhang J, Goulding M (2010) Genetic dissection of rhythmic motor networks in mice. Prog Brain Res 187:19–37. Gulledge AT, Dasari S, Onoue K, Stephens EK, et al. (2013) A sodium-pump-mediated afterhyperpolarization in pyramidal neurons. J Neurosci 33:13025–13041. Halter JA, Carp JS, Wolpaw JR (1995) Operantly conditioned motoneuron plasticity: possible role of sodium channels. J Neurophysiol 73:867–871. Harkema SJ, Hillyer J, Schmidt-Read M, Ardolino E, et al. (2012) Locomotor training: as a treatment of spinal cord injury and in the progression of neurologic rehabilitation. Arch Phys Med Rehabil 93:1588–1597. Harris-Warrick RM (2011) Neuromodulation and flexibility in central pattern generator networks. Curr Opin Neurobiol 21:685–692. Hawkins RD, Byrne JH (2015) Associative learning in invertebrates. Cold Spring Harb Perspect Biol 7:a021709. Hawkins RD, Kandel ER, Bailey CH (2006) Molecular mechanisms of memory storage in Aplysia. Biol Bull 210:174–191. Hilaire G, Voituron N, Menuet C, Ichiyama RM, et al. (2010) The role of serotonin in respiratory function and dysfunction. Respir Physiol Neurobiol 174:76–88. Hooper SL (2004) Variation is the spice of life. J Neurophysiol 92:40–41.

Plasticity and Learning in Motor Control Networks

Horn CC, Zhurov Y, Orekhova IV, Proekt A, et al. (2004) Cycle-to-cycle variability of neuromuscular activity in Aplysia feeding behavior. J Neurophysiol 92:157–180. Hurwitz I, Kupfermann I, Susswein AJ (1997) Different roles of neurons B63 and B34 that are active during the protraction phase of the buccal motor programs in Aplysia californica. J Neurophysiol 78:1305–1319. Ichiyama RM, Gerasimenko YP, Zhong H, Roy RR, Edgerton VR (2005) Hindlimb stepping movements in complete spinal rats induced by epidural spinal cord stimulation. Neurosci Lett 383:339–344. Jessell TM (2000) Neuronal specification in the spinal cord: inductive signals and transcriptional codes. Nat Rev Genet 1:20–29. Johnson SW, Seutin V, North RA (1992) Burst firing in dopamine neurons induced by N-methyl-D-aspartate: role of electrogenic sodium pump. Science 258:665–667. Kabotyanski EA, Baxter DA, Byrne JH (1998) Identification and characterization of catecholaminergic neuron B65, which initiates and modifies patterned activity in the buccal ganglia of Aplysia. J Neurophysiol 79:605–621. Kemenes G, Staras K, Benjamin PR (2001) Multiple types of control by identified interneurons in a sensory-activated rhythmic motor pattern. J Neurosci 21:2903–2911. Kong JH, Butler SJ, Novitch BG (2013) My brain told me to do it. Dev Cell 25:436–438. Kristan WB (2008) Neuronal decision-making circuits. Curr Biol 18:R928–R932. Lechner HA, Baxter DA, Byrne JH (2000) Classical conditioning of feeding in Aplysia: II. Neurophysiological correlates. J Neurosci 20:3377–3386. Lorenzetti FD, Baxter DA, Byrne JH. (2008) Molecular mechanisms underlying a cellular analog of operant reward learning. Neuron 59:815–828. Lorenzetti FD, Mozzachiodi R, Baxter DA, Byrne JH (2006) Classical and operant conditioning differentially modify the intrinsic properties of an identified neuron. Nat Neurosci 9:17–19. Luo C, Kuner T, Kuner R (2014) Synaptic plasticity in pathological pain. Trends Neurosci 37:343–355. Marder E (2012) Neuromodulation of neural circuits: back to the future. Neuron 76:1–11. Marder E, Bucher D (2007) Understanding circuit dynamics using the stomatogastric nervous system of lobsters and crabs. Annu Rev Physiol 69:291–316. Marder E, Goaillard J-M (2006) Variability, compensation and homeostasis in neuron and network function. Nat Rev Neurosci 7:563–574. Marder E, O’Leary T, Shruti S (2014) Neuromodulation of circuits with variable parameters: single neurons and small circuits reveal principles of state-dependent and robust neuromodulation. Annu Rev Neurosci 37:329–346. Martinez M, Delivet-Mongrain H, Rossignol S (2013) Treadmill training promotes spinal changes leading to locomotor recovery after partial spinal cord injury in cats. J Neurophysiol 109:2909–2922. Menelaou E, McLean DL (2012) A gradient in endogenous rhythmicity and oscillatory drive matches recruitment order in an axial motor pool. J Neurosci 32:10925–10939. Miles GB, Sillar KT (2011) Neuromodulation of vertebrate locomotor control networks. Physiology 26:393–411. Miller JP, Selverston AI (1982) Mechanisms underlying pattern generation in lobster stomatogastric ganglion as determined by selective inactivation of identified neurons. II. Oscillatory properties of pyloric neurons. J Neurophysiol 48:1378–1391.

439

440

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Morton DW, Chiel HJ (1993) In vivo buccal nerve activity that distinguishes ingestion from rejection can be used to predict behavioral transitions in Aplysia. J Comp Physiol A 172:17–32. Moult PR, Cottrell GA, Li W (2013) Fast silencing reveals a lost role for reciprocal inhibition in locomotion. Neuron 77:129–140. Mozzachiodi R, Byrne JH (2010) More than synaptic plasticity: role of nonsynaptic plasticity in learning and memory. Trends Neurosci 33:17–26. Mozzachiodi R, Lorenzetti FD, Baxter DA, Byrne JH. (2008) Changes in neuronal excitability serve as a mechanism of long-term memory for operant conditioning. Nat Neurosci 11:1146–1148. Nargeot R, Baxter DA, Byrne JH (1997) Contingent-dependent enhancement of rhythmic motor patterns: An in vitro analog of operant conditioning. J Neurosci 17:8093–8105. Nargeot R, Baxter DA, Byrne JH (1999a) In vitro analog of operant conditioning in Aplysia. I. Contingent reinforcement modifies the functional dynamics of an identified neuron. J Neurosci 19:2247–2260. Nargeot R, Baxter DA, Byrne JH (1999b) In vitro analogue of operant conditioning in Aplysia. II. Modifications of the functional dynamics of an identified neuron contribute to motor pattern selection. J Neurosci 19:2261–2272. Nargeot R, Le Bon-Jego M, Simmers J (2009) Cellular and network mechanisms of operant learning-induced compulsive behaviour in Aplysia. Curr Biol 19:975–984. Nargeot R, Petrissans C, Simmers J (2007) Behavioral and in vitro correlates of compulsive-like food-seeking induced by operant conditioning in Aplysia. J Neurosci 27:8059–8070. Nargeot R, Simmers J (2012) Functional organization and adaptability of a decision-making network in Aplysia. Front Neurosci 6:113. Parker D, Hill R, Grillner S (1996) Electrogenic pump and a Ca2+ -dependent K+ conductance contribute to a posttetanic hyperpolarization in lamprey sensory neurons. J Neurophysiol 76:540–553. Pavlovic P, Fuller W, Shattock MJ (2013) Novel regulation of cardiac Na pump via phosphorylation. J Mol Cell Cardiol 61:83–93. Payne JA, Rivera C, Voipio J, Kaila K (2003) Cation-chloride co-transporters in neuronal communication, development and trauma. Trends Neurosci 26:199–206. Perrier JF, Hounsgaard J (2000) Development and regulation of response properties in spinal cord motoneurons. Brain Res Bull 53:529–535. Perrier JF, Noraberg J, Simon M, Hounsgaard J (2000) Dedifferentiation of intrinsic response properties of motoneurons in organotypic cultures of the spinal cord of the adult turtle. Eur J Neurosci 12:2397–2404. Petruska JC, Ichiyama RM, Jindrich DL, Crown ED, et al. (2007) Changes in motoneuron properties and synaptic inputs related to step training after spinal cord transection in rats. J Neurosci 27:4460–4471. Picton LD, Nascimento F, Sillar KT, Miles GB (2017) Sodium pump currents mediate activity-dependent changes in mammalian motor networks. J Neurosci 37:906–921. Pitman RK (1989) Animal models of compulsive behavior. Biol Psychiatry 26:189–198. Prinz AA, Abbott LF, Marder E (2004). The dynamic clamp comes of age. Trends Neurosci 27:218–24. Pulver SR, Griffith LC (2010) Spike integration and cellular memory in a rhythmic network from Na+ /K+ pump current dynamics. Nat Neurosci 13:53–59.

Plasticity and Learning in Motor Control Networks

Rehm KJ, Deeg KE, Marder E (2008) Developmental regulation of neuromodulator function in the stomatogastric ganglion of the lobster, Homarus americanus. J Neurosci 28:9828–9839. Reimer MM, Norris A, Ohnmacht J, Patani R, et al. (2013) Dopamine from the brain promotes spinal motor neuron generation during development and adult regeneration. Dev Cell 25:478–491. Reyes FD, Mozzachiodi R, Baxter DA, Byrne JH (2005) Reinforcement in an in vitro analog of appetitive classical conditioning of feeding behavior in Aplysia: blockade by a dopamine antagonist. Learn Mem 12:216–220. Reynolds GS (1975) A primer of operant conditioning, 2nd ed., Glenview, IL:Scott Foresman. Rivera C, Voipio J, Kaila K (2005) Two developmental switches in GABAergic signalling: the K+ -Cl– cotransporter KCC2 and carbonic anhydrase CAVII. J Physiol 562:27–36. Roberts A, Li WC, Soffe SR (2010) How neurons generate behavior in a hatchling amphibian tadpole: an outline. Front Behav Neurosci 4:16. Rossignol S, Barrière G, Alluin O, Frigon A (2009) Re-expression of locomotor function after partial spinal cord injury. Physiology 24:127–139. Rossignol S, Frigon A (2011) Recovery of locomotion after spinal cord injury: some facts and mechanisms. Annu Rev Neurosci 34:413–440. Schall JD (2005) Decision making. Curr Biol 15:R9–R11. Selverston AI, Miller JP (1980) Mechanisms underlying pattern generation in lobster stomatogastric ganglion as determined by selective inactivation of identified neurons. I. Pyloric system. J Neurophysiol 44:1102–1121. Sieling F, Bédécarrats A, Simmers J, Prinz AA, Nargeot R (2014) Differential roles of nonsynaptic and synaptic plasticity in operant reward learning-induced compulsive behavior. Curr Biol 24:941–950. Sillar KT, Combes D, Simmers J (2014) Neuromodulation in developing motor circuits. Curr Opin Neurobiol 29:73–81. Sillar KT, Picton LD, Heitler WJ (2016) The Neuroethology of Predation and Escape. John Wiley & Sons. Sillar KT, Woolston A-M, Wedderburn JFS (1995) Brainstem serotonergic interneurones control the development of vertebrate spinal locomotor circuitry. Proc Biol Sci 259:65–70. Simmers J (2012) Motor control: learning new moves with old pumps. Curr Biol 22:R194–R196. Skinner BF (1981) Selection by consequences. Science 213:501–504. Susswein AJ, Byrne JH (1988) Identification and characterization of neurons initiating patterned neural activity in the buccal ganglia of Aplysia. J Neurosci 8:2049–2061. Susswein AJ, Schwartz M, Feldman E (1986) Learned changes of feeding behavior in Aplysia in response to edible and inedible food. J Neurosci 6:1513–1527. Thompson AK, Chen XY, Wolpaw JR (2009) Acquisition of a simple motor skill: task-dependent adaptation plus long-term change in the human soleus H-reflex. J Neurosci 29:5784–5792. Thompson AK, Pomerantz FR, Wolpaw JR (2013) Operant conditioning of a spinal reflex can improve locomotion after spinal cord injury in humans. J Neurosci 33:2365–2375. Thompson AK, Wolpaw JR (2014) Operant conditioning of spinal reflexes: from basic science to clinical therapy. Front Integr Neurosci 8:25.

441

442

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Thompson RF (2013) An essential memory trace found. Behav Neurosci 127:669–675. Tobin A-E, Calabrese RL (2005) Myomodulin increases Ih and inhibits the Na/K pump to modulate bursting in leech heart interneurons. J Neurophysiol 94:3938–3950. Vinay L, Brocard F, Clarac F, Norreel JC, et al. (2002) Development of posture and locomotion: an interplay of endogenously generated activities and neurotrophic actions by descending pathways. Brain Res Brain Res Rev 40:118–129. Vinay L, Jean-Xavier C (2008) Plasticity of spinal cord locomotor networks and contribution of cation–chloride cotransporters. Brain Res Rev 57:103–110. Wolpaw JR (2010) What can the spinal cord teach us about learning and memory? Neuroscientist 16:532–549. Wolpaw JR, Lee CL (1989) Memory traces in primate spinal cord produced by operant conditioning of H–reflex. J Neurophysiol 61:563–572. Zhang H-Y, Issberner JP, Sillar KT (2011) Development of a spinal locomotor rheostat. Proc Natl Acad Sci USA 108:11674–11679. Zhang H-Y, Picton L, Li WC, Sillar KT (2015) Mechanisms underlying the activity-dependent regulation of locomotor network performance by the Na+ pump. Sci Rep 5:16188. Zhang H-Y, Sillar KT (2012) Short-term memory of motor network performance via activity-dependent potentiation of Na+ /K+ pump function. Curr Biol 22:526–531.

443

14 Bio-inspired Robot Locomotion Thomas Buschmann 1 and Barry Trimmer 2 1 2

Institute of Applied Mechanics, Technische Universität München, Garching, Germany Department of Biology, Tufts University, Medford, MA, USA

14.1 Introduction Wheeled vehicles such as cars and trains have greater mechanical efficiency, speed, and load carrying capacity than animals. These advantages, however, require infrastructure such as roads or tracks. Without these structures wheeled locomotion is slow and restricted to sufficiently flat and level ground. This prevents using wheeled vehicles on vast parts of the earth’s surface, parts that legged and worm-like animals easily traverse. The same is true for many artificial structures, e.g., stairs, which humans easily traverse but wheeled vehicles cannot. These limitations have motivated efforts to build bio-inspired robots capable of animal-like locomotion. Most of these robots mimic locomotion by animals with articulated stiff skeletons. This design format can produce fast, powerful, efficient, and precise movements. The potential of this approach has been demonstrated experimentally in a large number of one-, two-, four-, six- and eight-legged walking and running machines. Recent work has been directed at increasing robot capabilities to match the adaptability and robustness of animal locomotion by using soft materials, many actuators and sensors, and “neuromechanical” controllers. For robots inspired by quadruped and biped mammals, stiff skeletal structures are reasonable from both engineering and biological points of view. The actuators for these machines are mostly made from rigid materials and are directly connected to the moving structures by stiff linkages. This arrangement has some technological advantages (see below). However, the “soft” nature of muscle not only provides viscoelastic dynamical properties, which have their own advantages (see below), but also allows a wider range of motion, as muscle can deform at extreme joint positions. It is therefore reasonable to believe that many biological concepts, soft actuator technologies, and control systems designed for highly deformable structures, will be incorporated into some robots. An example of this approach is the recent development of a cheetah-like robot powered by an electric motor. The remarkable performance of this robot was made possible only by realizing that in running quadrupeds much of the impact loading of the foot and lower limb is borne by tensile elements. Using this design principle allowed making the robot’s legs extremely light (Ananthanarayanan et al. 2012). This in turn helped in designing an electric motor capable of variable working stiffness (active compliance control) (Hyun et al. 2014), which maximized mechanical performance. Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

444

Neurobiology of Motor Control: Fundamental Concepts and New Directions

The rest of this chapter contains four parts. Section 14.2 provides background and context to help non-mechanical engineers understand the more detailed information presented in subsequent sections. Section 14.3 reviews walking machines with stiff skeletal structures and Section 14.4 soft robots. Section 14.5 concludes the chapter and gives an outlook on possible future developments.

14.2 Mechanical Engineering Background and a Biological Example It is first useful to define several concepts and terms used in mechanical engineering. An actuator is a device for producing force that is in turn used to achieve a desired behavior by applying suitable control methods. Actuators are often rigidly connected to the bodies of a robot they are intended to move, which are themselves often rigid, meaning they do not deform significantly (typical strains ≪1%) under normal operating loads. In this case the motion of the bodies induced by an actuator can be calculated using standard equations of rigid body dynamics (Shabana 2005). In the presence of external forces not accounted for in the control, even with high-fidelity force actuators, the motion of the robot will differ from the intended one because the actual trajectory arises from the sum of actuator and external forces. Obtaining the desired trajectory in this case can be achieved by high-gain speed or position feedback that measures the activity (in this example, the object’s motion), compares it to the desired one, and computes a strong corrective actuator force (see also Chapter 9). When actuators instead drive non-rigid (soft) objects, the equations relating actuator force to object movement are much more complicated, and planning and control much more difficult (Section 14.4). In mechanics, the number of degrees of freedom (DoF) is the minimum number of position variables which, together with fixed system parameters, uniquely defines the system’s geometrical configuration. This is equal to the number of equations of motion that describe the system’s dynamics. If a rigid part of the system connecting two bodies with non-negligible mass is replaced by a compliant element (e.g., a spring), this increases the number of DoFs because describing the geometric configuration of the two bodies now requires introducing additional position variables (in this example, spring length). Many robotic applications use control systems based on mathematical models of robot kinematics and dynamics to infer suitable actuator inputs from desired system outputs. This approach becomes increasingly challenging as the number of DoFs increases and is difficult to apply to systems having more DoFs than actuators (underactuated systems), e.g., robots with unactuated joints or elastic components. A large body of work exists on compensating for undesired elasticity in robots so as to approximate the performance of an ideal rigid system. How to use elasticity to improve system performance, alternatively, is in general unresolved. Most robots therefore have rigid links, as this allows a principled approach to control system design. Animals, on the other hand, have, in either the muscle-tendon system for mammals or the whole body for soft-bodied animals, non-negligible compliance. This observation suggests that the impressive motor abilities of animals arise in part from motor plant compliance, and has led to the design of robots with compliant (soft) elements and biologically motivated control systems.

Bio-inspired Robot Locomotion

A Biological Example Before delving into the details of robotic locomotion, it is useful to

provide an overview of motor control using three examples of human posture control. Consider standing erect and then leaning forward by rotating only about the ankles, with the legs and trunk remaining in a straight line (i.e., with the knees and hips fully extended). At the instant ankle rotation begins, the feet apply a backwards horizontal force and a moment to the ground to oppose the force moving the body center of mass forward; at the end of the leaning forward movement the feet apply a similar but reversed force and moment to stop body movement. How quickly the body center of mass can be accelerated and decelerated at movement beginning and ending is limited by the friction between feet and ground (which limits the horizontal force) and the fact that adhesive forces are usually not present (which limits the applied moment); this is the reason the same movement on ice skates or on stilts must be made very slowly. How far the body can lean forward without moving the feet is limited to positions at which body center of mass remains inside the footprint; leaning further forward triggers a step to block the incipient fall. The angles that can be achieved when leaning backward are much smaller because the feet do not point backward and the base of support over which the center of mass can be kept is thus much smaller. This range could be extended by grasping the floor with the toes to produce adhesive forces, but in most mammals this is not an option. Now consider bending forward at the waist from a standing erect original position. There is no danger of falling no matter how far forward the trunk rotates. The difference is that, as the trunk rotates forward, the legs angle back (by a backwards rotation at the ankles), hence maintaining body center of mass over the footprint. The neural mechanisms underlying these movements are many and hierarchical, but have large overlap. The decision to lean or bend is made at very high levels. The fundamental motor patterns (which muscles to activate with how much force and in what order) are carried out by motor cortex and spinal networks. Vestibular and peripheral sensory afferents alert higher centers how far one is leaning and trigger a step (again instantiated on the muscle level by motor cortex and spinal network activity) if the lean is excessive. Multiple sensory afferents signal something going awry (foot slip, insufficient muscle activation to achieve the desired movement, being bumped). For small errors these afferents typically correct the difficulty via low-level reflex changes in muscle activation, but larger errors require more substantial changes requiring the involvement of higher neural centers. There are thus also mechanisms that decide at what level corrective action must be taken. Robotic locomotion faces exactly the same problems and constraints. Feet push down but do not pull up. How forces can be applied depends in part on foot structure. Horizontal forces must not be so great as to cause foot slippage. Stability, either static as above or dynamic (see below), must be maintained, if not instant to instant, over one (normal walking) or a few (a stumble correction) step cycles. Movements of different parts of the robot relative to one another must not destabilize either the internal workings of the system or the system as a whole. Gaits must be chosen. Which patterns of joint movement will result in gait-appropriate foot trajectories must be calculated. How much force each actuator must apply to achieve these joint movements, which will depend, moment by moment, on the state of the other joints of the limb, must be calculated. Deviations from the desired trajectory will inevitably occur, and thus sensory feedback is required. Decisions must be made whether a deviation can be corrected by local, reflex-like changes

445

446

Neurobiology of Motor Control: Fundamental Concepts and New Directions

in actuator activation or require alterations in whole leg movement. And, for a robot to be truly functional, all this must be accomplished in real time.

14.3 Legged Robots with Skeletal Structures This section reviews design and control approaches in legged robots with skeletal structures. Emphasis is put on approaches using models and classical engineering methods. Section 14.3.1 reviews mechanism design, sensing, and actuation; Section 14.3.2 basic aspects of walking dynamics from a mechanical point of view; Section 14.3.3 trajectory-oriented approaches to walking control; Section 14.3.4 methods that put more emphasis on the walking cycle. 14.3.1 Mechanism Design, Sensing, and Actuation

Common walking robots have two, four, six, or eight legs, as do their biological counterparts. One-legged hopping robots, minimal implementations of two-legged hopping animals such as kangaroos, are also common. Legs are usually stiff segments connected by hinge joints. Fully actuated bipedal robots usually have six (3 hip, 1 knee, 2 ankle) joints per leg and flat feet for surface contact (Hirose and Ogawa 2006; Pfeiffer 2006; Nelson et al. 2012). Planar (meaning that the joints are arranged such that the leg segments all rotate in the same plane) bipeds often have point feet and only two DoFs per leg, one in hip and one in knee (e.g., Chevallereau et al. 2003). Quadruped and hexapod robots typically have point feet and three or four DoFs per leg (most animals walk on their toes: horse hip is human hip, horse stifle human knee, horse hock human ankle, horse fetlock the (fused) human joints in the ball of the foot, hence four joints), but the variety of kinematic designs and actuation mechanisms is larger for multi-legged robots than for bipeds. Electric motors with reduction gears are the most common actuators. Older designs mostly relied on brushed DC motors, more recent ones often use brushless motors with higher power density (Hirai et al. 1998; Nishiwaki et al. 2000; Lohmeier et al. 2009; Tsagarakis et al. 2011). Hydraulic and pneumatic actuators are also used, especially in multi-legged robots. Well-known machines using hydraulic actuation are the quadrupeds and bipeds developed by Boston Dynamics (Raibert et al. 2008; Nelson et al. 2012) and the quadruped HyQ (Semini 2010). The advantage of hydraulic actuation is higher power distally with a large portion of actuator mass located in the torso. This arrangement simplifies stabilization through step position control because it reduces leg mass (Section 14.3.3). Pneumatic actuators share the advantage of centralized pressure generation, but their main benefit is intrinsic compliance. Although how to effectively exploit joint compliance for walking remains an open question, its use is motivated by the compliance of biological muscle-tendon systems. Compliance has also been added using antagonistic mechanical springs in a bio-inspired setup (e.g., Yamaguchi and Takanishi 1997). An alternative approach is a series elastic actuator, which adds a spring between the output side of a geared electric motor and the joint (Pratt and Williamson 1995). Here, however, the primary intent was not to exploit passive elastic effects, but to provide high-quality force control by measuring and controlling spring deflection (Pratt and Pratt 1998; Hutter et al. 2012).

Bio-inspired Robot Locomotion

All walking robots have joint position sensing, which is usually also used to infer joint velocity. Due to the importance of foot loads in maintaining balance, most robots, especially bipeds, have contact force sensing. Contact switches, which only report whether a position or force threshold has been exceeded, are often used, particularly in multi-legged machines. Contact switches are cheaper and lighter than force/torque sensors and can be used to implement Finite State Machine based control approaches. They cannot, however, be used to implement contact force or active compliance control. Joint force or hydraulic pressure sensors are seldom used, but have the advantage of enabling additional control approaches. Inertial measurement units reporting body posture and angular velocity are used in most bipedal and many quadrupedal robots, where maintaining balance is a primary difficulty. 14.3.2 Basic Dynamics of Legged Locomotion

Walking machines are hybrid dynamical systems that have both continuous (e.g., joint angles) and discrete (e.g., sticking or sliding contacts) states. This leads to two views of system dynamics: (1) a continuous evolution of position, velocity, and force over time and (2) discrete changes of step length, step duration, etc., from one step cycle to the next. These are equivalent descriptions for analyzing overall system dynamics and stability, as step-to-step dynamical models are obtained by integrating the continuous dynamics. Nevertheless, different approaches to building and controlling walking machines put more emphasis on either the continuous or the discrete step-to-step transition approach. For bipedal robots, the different approaches are exemplified by fully actuated systems (Fig. 14.1) and semi-passive limit cycle walkers (Fig. 14.2, Section 14.3.4). Strictly speaking, every legged robot is underactuated, since, due to either ground or foot compliance, the foot-ground contact always has finite stiffness, and there is no actuator to control contact elastic deformation. Nevertheless, robots with one actuator per joint are commonly called fully actuated. Underactuation means that not all DoFs can be controlled to follow a particular trajectory. Especially in two- and four-legged systems, actuated joint motions can excite unstable dynamics in unactuated DoFs, making the robot fall. Maintaining stability of global system dynamics is thus a major concern for such robots, dominating lower level details such as actuation method or the details of individual joint control. Figure 14.1 Minimal model of a biped balance control system based on force and inertial measurements. The inertial data are used to calculate modified contact force setpoints, which are tracked via a low-level joint position control loop. Note that this is a simplified drawing emphasizing a basic principle. An actual implementation must add at least load distribution during double support and account for swing-stance and stance-swing transitions (see, e.g., Buschmann et al. 2009 or Takenaka et al. 2009b). From Buschmann et al. (2014) with permission.

Joint setpoints

Inertial data Balance control

Force setpoint

Joint angles Force

Force control

447

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Figure 14.2 Semi-passive limit cycle walker with typical design elements. Round feet with a (passive or active) knee-locking mechanism enable a passive rocking motion during stance. Limited actuation is included to compensate for energy losses and control is performed by a finite state machine. The degree of actuation and control may vary from fully passive (McGeer 1990) to some (Collins and Ruina 2005) or almost full actuation with step position control (Hobbelen and Wisse 2009). From Buschmann et al. (2014) with permission.

Finite state machine

Round feet

Contact state

Knees Energy supply

448

Because of the unilateral foot–substrate contact noted above (that feet only push down), rigorous mathematical treatment of stability is difficult for walking systems. Locomotion has been traditionally classified into statically or dynamically stable. In statically stable walking, inertial forces and dynamic effects are negligible and the necessary condition for stability is keeping the projected center of gravity within the base of support. In statically stable walking leg movement can cease at any time and the animal will not fall. The term dynamic stability is used in two ways. The first, used in much early biological literature, refers to the fact that in many gaits there are phases where, if leg motion suddenly ceased, the animal would fall. Returning to the leaning example in Section 14.2, walking could be defined in this sense as a repeated series of leaning forward movements that trigger steps to break the incipient falls. This definition is difficult to define mathematically. In mechanics dynamically stable walking is therefore instead defined as all stable walking patterns in which dynamic effects are non-negligible (Full et al. 2002; Buschmann et al. 2014). Even with this definition, despite much work devoted to it, a practical and mathematically sound “stability measure” for walking robots remains elusive. 14.3.3 Trajectory-Oriented Walking Control

Walking controllers must at all times ensure that (1) the unilateral constraint that feet cannot pull upward on the substrate is always obeyed, (2) unactuated dynamics are stabilized, and (3) the desired walking behavior is produced. These controllers must also function despite the underactuated and hybrid nature of robot locomotion. Due to these complexities, straight-forward solutions remain elusive. Experimentally successful systems instead rely on manually defined hierarchical control architectures with interacting model-based planning and control modules. Details of these architectures vary and an in-depth description of any such controller is beyond the scope of this chapter. Figure 14.3 shows a simplified layout: desired behavior is planned in an ideal manner, the forces and positions corresponding to the ideal behavior are modified using feedback from position and force sensors, and the final result is mapped to a joint space representation that controls the actuators. Since there is often feedback to the planning layers, as in animals, the distinction between planning and control can blur (Tajima et al. 2009; Nishiwaki and Kagami 2006).

Bio-inspired Robot Locomotion

Figure 14.3 Example of a hierarchical control architecture with real-time planning and balance control for fully actuated robots, as is used in many state of the art walking machines (e.g., Buschmann et al. 2009). Each level can have multiple sublevels, e.g., step-cycle and trajectory planning within the “Planning” level. Planning and control are typically formulated in terms of foot and center of gravity positions and contact forces. Joint space variables for low-level control are then obtained from inverse kinematics. Most early and some current implementations use offline planning, but online methods now dominate since they facilitate more reactive behavior. Modified from Buschmann et al. (2014).

Desired behavior Planning Ideal force & movement Stabilizing control Adapted force & movement Inverse kinematics Joint space commands Robot (with low-level control) Sensor data

Essentially, the planning layer is distinguished by a longer control horizon, with usually longer control cycles and more predictive components than in the stabilizing feedback control system. This distinction is analogous to the control horizon differences in animals between the fast, in-single-step-cycle corrective responses that small variations in ground height elicit vs. the sometimes several-cycle-long corrections that occur in stumbles. Planning in fully actuated, model-based systems combines kinematic trajectory planning of individual legs and dynamic planning for the whole system based on simplified lumped-mass models (Kajita and Tani 1995; see Buschmann 2010 for discussion). Reduced models are sets of differential equations with contact forces (or center of pressure) as inputs. Solving the planning problem in real-time is complicated by the inequality constraints for contact forces described above. That is, one cannot simply choose a leg movement pattern, but rather must find a pattern that satisfies both the system’s differential equations and the inequalities in the set of possible contact forces. Most solutions rely on first planning a feasible contact force reference trajectory and then finding, approximately, a set of leg movements that will generate it (Takenaka et al. 2009a; Buschmann et al. 2007; Kajita et al. 2003; reviewed in Buschmann et al. 2014). Since there are usually non-negligible modeling and sensing errors as well as unstable dynamics, a stabilizing controller is required. The most widespread approaches are (1) modifying contact force, (2) horizontally accelerating the center of mass, and (3) controlling footstep location. In all cases, contact forces are modified so as to maintain stability of unactuated DoFs. These mechanisms are roughly equivalent to the ankle, hip, and stepping stabilization strategies found in humans (Horak 1987). This similarity is not coincidence, but stems instead from using studies of animal locomotion to guide design of robotic control approaches and models; for instance, common lumped-mass models originated in biomechanical gait analysis.

449

450

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Contact force control, is used (among others) in the Honda, HRP-2, and Lola robots (Kajita et al. 2005; Takenaka 2009b; Buschmann et al. 2009). Foot force is measured by force/torque sensors and controlled with high-gain position controllers. The major component of this approach is contact moment control, in which force is applied in the correct direction (remember the ankle joint typically has two DoFs) through the moment arm of the foot to rotate the ankle and hence right the body. This approach is analogous to the increased force applied to the ball of the foot when rotating upright from the leaning forward position in the biological example given earlier. Note that if the lean were to the side, because the foot is thinner than it is long, even more force would be necessary for righting. As appropriate, normal forces may also be controlled. The second strategy, accelerating the upper body, is adopted in Tajima et al. (2009), Nishiwaki and Kagami (2006), and Takenaka et al. (2009b). Because it acts through the leg joints and segments, upper body acceleration requires changes in foot horizontal contact force. As such, it is analogous to the bending at the waist example given earlier, in that changes in one part of the body (the torso) require changes in other parts, with the feet acting as the ultimate intermediary through which (provided the feet do not slip) the forces act to change body position. In both strategies, contact force is controlled by modifying a position variable (e.g., center of mass position) to prevent divergence from the desired movement, with the second mapping underactuation from one DoF to another (e.g., from the upper body to the ankle joint). The stepping strategy is analogous to the biological example of excessive lean triggering a step to prevent falling, and is the most powerful. It has been used with great success in robots with heavy bodies and relatively light legs (Raibert 1986, 2008; Gehring et al. 2013), in which reduced order models can predict the effect of modifying foothold position with good accuracy. For robots with heavy articulated legs, determining appropriate footholds is more difficult, as modifying foot motion alters body motion. There have been successful demonstrations with heavy-legged robots in some situations (Urata et al. 2011; Wittmann et al. 2014), but the general case remains an open problem. 14.3.4 Limit Cycle Walkers

Global system quantities such as center of gravity trajectory or contact force patterns during walking can be reproduced with surprising accuracy using very simple lumped-mass models (Blickhan 1989). These models appear to be dynamically stable without active control, a property called self-stability or open-loop stability. Such models, however, often collect the overall behavior of the biomechanical system and parts of the neural control into a black box. It is thus unclear whether they are truly un-actuated on a fundamental level. Nonetheless, these and similar observations have motivated the design of purely or mainly passive mechanisms with no or minimal active control. An extreme example are the purely “passive dynamic walkers” pioneered by McGeer (1990). Similar to long-produced mechanical toys, these devices locomote by rocking from one foot to the other, one leg swinging forward as a pendulum while the body mass pivots up and forward over the stance leg as an inverted pendulum, rhythmically converting potential and kinetic energy back and forth across a step cycle. Because of frictional losses, completely passive walkers can only move down inclined planes. This limitation can be overcome by providing limited energy inputs in the cycle, much less than is required for fully actuated movements. Indeed, in large-limbed

Bio-inspired Robot Locomotion

animals, swing muscles contract only at swing beginning and then fall silent, allowing limb momentum to passively carry swing to completion (Hooper et al. 2009). Some recent robotic work has therefore focused on using actuation to supply energy on level ground and influence walking direction and speed while exploiting “passive dynamics” to increase efficiency (see Fig. 14.2). Collins and Ruina (2005) incorporated relatively soft springs in series with ankle joint actuators to provide energy via push-off after initial contact of the contralateral leg. Control was implemented as a finite state machine with input from ground contact and leg extension switches. More recent robots developed in Delft combine classical design elements of passive walkers with more actuation. This approach yields machines that consume very little energy, but they are quite limited in ability compared to fully actuated robots and much less robust against disturbances (Hobbelen et al. 2008). 14.3.5 CPG-Based Control and Step-Phase Control

A complete model of the neural control of legged locomotion is unavailable in any animal. This is especially true for high-level aspects such as balance control, essential for biped and quadruped walking (Buschmann et al. 2014). It is nonetheless well known that neural control of walking includes muscle activation by intrinsically active central neural networks, central pattern generators (CPGs), which generate the basic rhythmic pattern (see Chapter 8). Muscle length and force, joint angle, pressure being applied to the joints and to the foot, and a variety of cutaneous sensory feedback signals, modify CPG activity (Buschmann et al. 2014) (see Chapter 9). CPG-based walking control attempts to mimic this structure, often using a neural oscillator model with two mutually inhibiting neurons as the CPG (Matsuoka 1985; Endo et al. 2008). This model instantiates the half-center oscillator concept, the most important element of neural networks creating alternating left-right locomotory activity in vertebrates. A key CPG feature is an ability to synchronize their intrinsic oscillations with input signals (entrainment), thereby adapting the intrinsic activity to the environment. CPG-based methods introduce a large number of model parameters that have no obvious relationship to global locomotion output characteristics (cycle period, swing vs. stance duration, etc.). These parameters must be chosen appropriately to achieve even basic walking, but the lack of closed-form solutions makes this selection difficult. A possible solution is gradient descent methods with reinforcement learning (Sugimoto and Morimoto 2011). Most CPG-based controllers reset the neural oscillator when the swing leg makes ground contact, which improves stability (Shinya and Tsuchiya 2006; Nakanishi et al. 2004). This stabilizing mechanism is not limited to CPG-based controllers, and is analogous to the event-based step phase switching implemented in many legged robots. Many hopping robots also switch control laws according to step phase (Raibert 1986). Sato et al. (2012) use the same controller during flight and ground phases, but switch gains and desired variables. A phase-reset mechanism is also used in the planar biped MABLE (Sreenath et al. 2011) and the quadruped Scout II (Hawker and Buehler 2000). Event-based phase switching for the life-sized 3D biped Lola increases robustness to uneven ground (Buschmann et al. 2012). Stability in CPG-based controllers is further improved by direct sensory feedback (“reflexes”). The design typically follows biological prototypes or is performed

451

452

Neurobiology of Motor Control: Fundamental Concepts and New Directions

intuitively, leading to involved systems that are not straight-forward to setup and are difficult to tune. Typical biologically-motivated reflexes are vestibular responses (which correct for body tilt), reinforcement of force and load (increasing activation to assist the stance phase), and movement reinforcement (e.g., the Tekken series of quadrupeds and Kimura and Fukuoka 2000 and Fukuoka et al. 2003; 2010).

14.4 Soft Robots Although there has been tremendous progress in developing bio-inspired articulated robots, it is becoming increasingly clear that to achieve robustness and adaptability future robots must include soft components. The special challenges inherent in building and controlling soft materials have fostered the new field of soft robotics. In addition to incorporating highly deformable materials into traditional robots, many researchers are trying to develop robots constructed almost entirely from soft materials. Through the recent development of flexible and stretchable electronics, and by exploiting widely available durable elastomeric materials, it is now possible to build functional soft devices. These are being used to explore scientific and engineering challenges associated with robots that can change their size and shape dynamically. A Definition of Soft “Soft” is not a rigorously defined material property. Unlike stiffness

(expressed as Young’s modulus) and hardness (measured by indentation and other standardized tests), soft is an elastic property associated with large deformations. Interestingly, this property is scale-dependent; steel beams undergo large deformations when large forces are applied but steel is not typically considered to be soft. A good working description is that soft materials deform greatly under the loads they normally encounter. For a given structure this can be formally described by measuring the specific stiffness (force normalized by weight, length by the structural dimensions) (Lin et al. 2011a). 14.4.1 Limitations and Advantages of Soft Materials

In contrast to many stiff materials traditionally used in engineering such as steel or aluminum, soft materials typically have significant material damping and thus dissipate energy. It is therefore difficult to transmit forces in preferred directions and accelerations are dramatically affected by the damping characteristics of the material. This can result in substantial mechanical inefficiencies and unpredictability if soft materials are used as substitutes for stiff materials. The underlying assumptions of rigid body mechanics that most robotic control paradigms are based on are invalid when structural materials deform beyond certain limits (typically strains of a few percent). This issue is further complicated by the presence of nonlinear mechanical properties such as anisotropy, pseudo-elasticity, work softening and work hardening (see below), creep, yield, viscoelasticity, and other dynamic state transitions (Fung 1993) in soft materials. The fabrication of soft robots is also still in its infancy. Most engineering labs and workshops are tooled for working with hard materials and it is challenging to create long-lasting interfaces between soft and hard components. Despite these challenges, there are many practical advantages to using soft materials in robots. Soft robots are resistant to damage by sudden impact, can change shape and

Bio-inspired Robot Locomotion

size to access complex spaces, and can be made intrinsically safe (in contrast to “control” safe) for use in human and natural environments. Soft materials are typically an order of magnitude less dense than traditional engineering materials and can be manufactured at low temperatures through energy efficient processes. Soft robots therefore have the potential to be lightweight and cheap. By exploiting soft materials in machine design, engineers also have access to an entirely new range of properties. These include novel chemistries; new thermal, electrical, mechanical, and chemical specifications; and the opportunity to use biopolymers and biodegradable or biocompatible compounds. Such materials have much greater potential for interfacing with living tissues and for better diagnostic and surgical robots and prosthetics. 14.4.2 The Challenges 14.4.2.1 Actuators

Most robots are driven either directly by electromechanical devices (motors, servos, electromagnets) or indirectly by motors transmitting their force through hydraulics and pneumatics. A recent generation of walking robots is powered by internal combustion engines driving hydraulic pumps (Raibert et al. 2008). These mechanisms are generally heavy and made of hard materials. In contrast, animal movements are produced by converting chemical energy into displacements at the molecular scale inside muscles. This process depends on a highly structured arrangement of proteins acting in synchrony to convert nanoscale displacements into macroscopic force production. Muscles can only generate tensile forces and, if they shorten, they must be re-extended by other forces to work again as an actuator. Although man-made actuators are typically more powerful and forceful than muscles, they are not as versatile and cannot be as easily scaled or replicated throughout the robot. Animals often contain hundreds of muscles, varying from less than a millimeter to hundreds of centimeters in size. This ability to deploy many actuators at different scales allows animals to produce and control movements in remarkably complex ways (see also Chapter 12). Muscles are often used as a benchmark to describe the most desirable features of soft actuators; they have a typical blocked stress (force) of 200 kPa, strains (length changes) of 20%, scaled speed of a few lengths per s, and typical peak power of 100 W/kg. More importantly, they are deformable, multifunctional, capable of continuous activity for decades, and powered by locally available energy-rich hydrocarbons. One of the greatest challenges in building practical soft robots is the lack of suitable synthetic actuators. A large variety of deformable actuators (artificial muscles) have been developed to fill this need, of which several have been incorporated into robots (Fig. 14.4). Electroactive Polymers The most deformable are chemically reactive gels (still far from

practical as actuators) and electroactive polymers (Carpi et al. 2011; Keplinger et al. 2013). Electroactive polymers consist of two major classes, the ionic polymer-metal composites (IPMCs) (Bahramzadeh and Shahinpoor 2014) and dielectric polymers. These materials can produce relatively large stresses or strains, but are difficult to interface with other materials and even with the conducting surfaces necessary to activate them. IPMCs must be kept wet and dielectric polymers require very high electrical fields. These properties have limited their applications but their continued development could have a major impact on the field.

453

454

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A

B

C

D

E

F

G

H

I

Figure 14.4 Soft robots representing the primary actuation technologies that have been deployed. A, B, and C are pneumatic devices with embedded structures allowing them to deform in predefined ways. A is a tethered pneumatic robot that exhibits multiple gaits, B is a large-scale untethered pneumatic robot capable of operating in a variety of harsh environments, and C shows PneuFlex actuators assembled into a gripping hand. D and E use cable-like tendons and shape memory alloy actuators to move an octopus-inspired manipulator arm (D) or to locomote underwater (E). F is an untethered radio-controlled, caterpillar-inspired robot that can crawl and inch using SMA coils as its primary actuator. G is a worm-inspired robot powered by an SMA mesh. H and I are pneumatic devices that can jump by exploding combustible gas. A and I provided by R Shepherd, B and H by M Tolley, G by S Kim, all with permission. C from Deimel and Brock (2013), D from Laschi et al. (2012), E from Arienti et al. (2013), all with permission.

Crystalline Transition Materials Other materials that undergo state changes have also been

exploited as actuators. A common approach is to use the crystalline transition properties of shape-memory alloys (SMAs, e.g., Nitinol) (An et al. 2012). These metals transition from one packing structure to another at different temperatures and are often controlled by using current sources to produce resistive heating. Although SMA wires are intrinsically hard, when drawn to diameters of less than 200 microns and tightly coiled, they are macroscopically as soft as fabrics and can contract to less than half their passively stretched length. A key biomimetic advantage of these actuators is that they produce linear tensile forces in one direction and must be re-extended in the passive state to re-shorten, properties reminiscent of muscle. They have been used successfully in many soft robots (Kim et al. 2005; Lin et al. 2011b, 2013; Margheri et al. 2012; Mazzolai et al. 2012; Menciassi et al. 2004; Seok et al. 2013; Umedachi and Trimmer, 2014; Umedachi et al. 2013; Yuk et al. 2011). Unfortunately, they are relatively slow (most operate below 1 Hz) and, because they depend upon the dynamics of heating and cooling, their performance is quite variable.

Bio-inspired Robot Locomotion

There are now a large variety of polymeric systems that work similar to SMAs (Haines et al. 2014; Hu, 2007; Ogden et al. 2014). These hold great promise as soft actuators once their state transitions can be accurately controlled. Piezoelectric materials can also be used as actuators. These can be extremely hard ceramics or more compliant polymers. Although they are very fast and accurate (and have been used to drive flying microrobots, Ma et al. 2013), they produce relatively small displacements and require very high voltages. Pneumatics Other soft actuators use compressed gases to produce movement. These devices were originally called pneumatic artificial muscles (PAMs) or McKibbon actuators (Zhang and Philen 2012); the movements of an inflatable bladder are constrained by an external woven mesh to produce a shortening force. These devices have been used as actuators in several articulated robots (Narioka and Hosoda 2008; Nelson and Quinn 1999). PAMs are extremely reliable and durable, leading to their use in a variety of commercial applications. The primary challenge with PAM actuators in soft robots is the need for relatively stiff, large diameter gas delivery lines, powerful (and usually rigid) valve systems, and the difficulty of producing and storing compressed gases on board (which requires non-yielding containers). A new variety of inflatable actuators (embedded pneumatic networks, EPNs) have similar challenges but have other properties that make them attractive in soft robots (Shepherd et al. 2011). Instead of separating the inflatable bladder from the constraining mesh, EPNs produce displacement by differentially changing the stiffness of inflatable compartments. Effective stiffness can be changed by designing structural features that deform in preferred ways (Morin et al. 2014; Sun and Chen 2014), by fabricating chambers from a variety of elastomers, or by using anisotropic composites (Mosadegh et al. 2014). This technology has been incorporated into robotic grippers (Deimel and Brock 2013), manipulator arms (Martinez et al. 2013), and a crawling limbed robot that can carry its own gas compression pump (Tolley et al. 2014). Because compressed gases do not have high specific energy (typical values are similar to rechargeable lead acid batteries), alternative methods of energy transfer are needed for practical autonomous pneumatic robots (Messner et al. 2014). Since combustible materials such as hydrocarbons have high energy density, it may be possible to develop EPNs powered by the heat and gas expansion of explosions. This has been explored in a variety of contexts (Loepfe et al. 2014; Shepherd et al. 2013) and could lead to the development of fast-moving soft devices. Motor-Tendon Systems Extreme miniaturization would allow using existing hard motor

components as actuators without having a large impact on the bulk material properties of the robot. This strategy would allow using linear motors or spooling motors pulling flexible cables or compliant “tendons”. Such methods are commonly used in large articulated research robots (e.g., ECCEROBOT-2, Diamond et al. 2012) but have not been attempted in an entirely soft robot, which would require motors an order of magnitude smaller. 14.4.2.2 Sensors

Animals have remarkable sensory abilities. In addition to long-range senses such as vision, hearing (including sonar), and olfaction, locomotion is strongly affected by inertial sensing (gravity and acceleration detection), proprioception (measurement of

455

456

Neurobiology of Motor Control: Fundamental Concepts and New Directions

internal forces and displacements), and touch (see also Chapter 9). Three features of animal sensory reception are distinctly different from current implementations in robots. First, mechanosensing is typically highly distributed; skin has hundreds of thousands of individual sensors and much of the information about the environment derives from combining and comparing the input of multiple sensors. Replicating this type of sensing in robots is very difficult because of the need to provide distributed power and to communicate multiple channels of information to the central processor. By comparison, the multiplication of animal sensors appears to be relatively low “cost”. Second, the different types of sensor information needed for accurate locomotion control are tightly integrated. For example, self-movement is easily distinguished from those in the environment by rapid comparison of inertial, visual, and proprioceptive input. Third, in the absence of changing input, most primary sensory receptors adapt very quickly and cease to provide information. Much animal sensing is therefore active. For example, touch requires continuously changing interactions with the environment (both microscopic skin movements and larger limb or body movements) and the visual system continually scans the environment. In animals sensory system actuation and sensation are thus tightly coupled in ways not traditionally employed in robot design. A driving force in soft robotics is to develop “smart skins”; enveloping active elastomers that would endow soft robots with the highly distributed sensory input seen in animals. Furthermore, because most actuators can work in reverse as sensors (Bahramzadeh and Shahinpoor 2014), these skins could allow animal-like active sensing. There are two primary thrusts in developing such sensors. The first is to design and fabricate materials that are intrinsically reactive to stimuli such as light, temperature, or applied force (Ates 2013). An ongoing challenge with these materials is the difficulty of interfacing them with conventional electronics. The second is to fabricate elastomeric systems with embedded flexible electronics. This includes serpentine wiring arrays of relatively stiff materials patterned onto elastomers (Lu and Kim 2014). These systems provide flexibility and stretchability by minimizing strains of the stiff materials (Fan et al. 2014). A key advantage of this approach is that conventional micro-electronic devices such as transistors, amplifiers, and light emitting diodes can be incorporated into the designs (Rogers et al. 2010). Another approach is incorporating intrinsically elastic electric conducting materials such as carbon nanotubes or eutectic liquid metals into microchannels (Cheng and Wu 2012). Because these fluids can move and easily change their resistance, capacitance, and inductance, they are well suited to sensor fabrication (Gozen et al. 2014; Park et al. 2010). 14.4.2.3 Control of Soft Robots

One of the greatest challenges in successfully developing soft robots is movement control. Although all animals and robots ultimately control movements by managing mechanical energy distribution, the strategies for accomplishing this differ substantially in stiff and soft systems. In particular, without a stiff skeleton, the material properties of the physical plant will play a much more prominent role in movement control (Fig. 14.5). Existing control strategies are not well suited for soft robots, largely because soft robots are high dimensional dynamic systems with an essentially infinite number of DoFs. This precludes using traditional inverse-dynamics approaches (Craig 1989) for trajectory planning.

Bio-inspired Robot Locomotion

Figure 14.5 (A) General representation of interactions controlling behavior in animals and robots. Large dark arrows represent forward information from the central nervous system (or commands from a central computer) operating through structures (skeleton, muscles, limbs, etc.) made of particular materials leading to motor output (behavior). These movements interact with the environment to provide feedback to the peripheral nervous system which, in turn, modulates the central commands. Small light arrows represent additional embedded interactions. Materials and structures themselves can directly affect the operation of the central nervous system (e.g., sensory neurons are mechanically sensitive), the central nervous system controls the activity of the peripheral nervous system and, in some cases, the peripheral nervous system can directly control movement. (B, C) Graphical representations of how these components interact differently in soft animals and those with stiff skeletons. (B) The dominant view of articulated animals is that central commands drive movements by operating on muscles attached to the skeleton, that feedback from the peripheral nervous system helps fine-tune these movements, and that the material properties of structures are critical only in limited cases. (C) In soft animals it is likely that control is more distributed and that tissue material properties and the peripheral nervous system play much more important roles in behavior.

Furthermore, most soft materials, and structures built from them, are mechanically nonlinear. For example, embedding particles in linear elastomers changes the material’s loading and unloading characteristics and creates pseudo-elastic behaviour; repeated loading cycles often cause work hardening or softening (Dorfmann and Ogden 2004). Soft materials are also typically viscoelastic, and thus their mechanical properties depend on rate (loading frequency) and time (constant loading causes progressive deformation) (Dorfmann et al. 2008). An additional complication is that incorporating

457

458

Neurobiology of Motor Control: Fundamental Concepts and New Directions

fibres or composite materials into structures makes them anisotropic and extremely difficult to model mechanically (Lin et al. 2009). Not only is it difficult to create good mechanical descriptions of soft materials, there is no generally accepted framework linking soft material modeling and mechanical control theory. Simulation approaches have been developed for soft robots, but they have been designed for special geometries (Hannan and Walker 2003) or for cases in which reduced-order models are readily obtained (Saunders et al. 2009). More recently, a variety of dynamical approaches have been proposed for worm- and snake-like devices (Daltorio et al. 2013; Kano et al. 2013; Sato et al. 2012). These schemes are a significant step forward because they can generate different “gaits” and allow for locomotion in a variety of contexts. Another promising approach is using geometric transformations (Neˇcasová et al. 2011; Shapere and Wilczek 1987; Shapere and Wilczek 1989). This approach was developed for slow moving, shape-changing structures in fluids at low Reynolds numbers (negligible inertia), but has been adapted for oscillatory locomotion in granular media (Hatton et al. 2013). Geometric mechanics has also been used to model snake-like undulatory locomotion, and has been proposed as a unifying theory of robot locomotion in which dynamics are parameterized in terms of momenta, reference frame motion, and internal shape (Ostrowski and Burdick 1998). A key element for terrestrial locomotion will be to find appropriate constraints for elastic deformations and for discrete, rather than continuous, environmental contacts (Chirikjian and Burdick 1995). At present, most truly soft robots are either controlled by simple empirically discovered open loop sequences (Lin et al. 2013; Tolley et al. 2014), or the robots are directed in real time by an external observer. For limbless soft robots such as GoQBot these methods have generated multiple gaits including caterpillar-like crawling, inching, and ballistic rolling (Lin et al. 2011b, 2013; Umedachi et al. 2013). The four limbed EPN-based robot has many of the same gaits and can also exploit asynchrony in side-to-side limb movements to produce crab-like walking (Shepherd et al. 2011; Tolley et al. 2014). These bio-inspired solutions are mostly CPG-based (Marder and Bucher 2001) using limit cycle controllers and arrays of coupled oscillators (Ijspeert 2008; Li et al. 2014). Although these systems generate robust cyclic movements and even self-modify in different conditions (Adamatzky et al. 2009), they are not necessarily the best approach for moving in complex, varied environments (which includes most terrestrial locomotion). There is a pressing need of control methods for episodic and transient movements. Towards this goal, there is increasing recognition that CPGs are only one component of a larger control system linking central control systems (Iwasaki et al. 2014), sensory feedback (Paoletti and Mahadevan 2014; Schuldt et al. 2015), and morphological computation (Füchslin et al. 2013) (see Chapter 12). In the absence of a well-developed theory for the control of soft robots, many insights can be gained by exploring the very large design space with simulation and evolutionary robotics (Nolfi and Floreano 2001). This method has produced robust gaits in a variety of limbed articulated robots (Bongard 2011; Bongard et al. 2006) and is likely to produce novel solutions for soft robots as soft material simulators become more computationally efficient (Rieffel et al. 2009, 2013; Saunders et al. 2011a). In particular, using multi-objective fitness criteria has the potential to discover general principles of

Bio-inspired Robot Locomotion

robot locomotion, ones not necessarily linked to the neuromechanics of animal systems (Cheney et al. 2013; Lipson 2013). 14.4.3 Bioinspired Locomotion in Soft Robots

In contrast to the long history of building articulated bioinspired robots, soft robots are relatively new. The two most commonly studied biomechanical model systems, mollusks and worms, are hydrostatic in operation (Trueman 1975). In essence, hydrostatic control is very similar to the use of levers. Instead of mechanical advantage provided by different mechanical moments (Force × distance) acting across a fulcrum, hydrostatic systems gain mechanical advantage through Pascal’s law. This states that pressure exerted on an incompressible fluid is equal at all positions, and thus forces and displacements can be interchanged by connecting fluid compartments of different dimensions. This is the mechanism used in hydraulic jacks; a large displacement created by a small force acting on fluid in a small diameter cylinder will produce a large force (and small displacement) in a connected larger cylinder, because in both cylinders pressure must be equal. Muscular Hydrostats Cephalopods are particularly inspiring because of their extraor-

dinary versatility and ability to change shape and size. Each tentacle-like arm is densely packed with highly organized and complicated layers of muscle working together as a muscular hydrostat (Kier and Stella 2007; Mazzolai et al. 2007; see also Chapter 12). When reaching towards objects and bringing them towards the body, the arms employ a stereotypical set of movements based on the generation of traveling waves and pseudo-joints (Gutfreund et al. 1996; Sumbre et al. 2001, 2006; Yekutieli et al. 2005). This strategy is thought to reduce the number of DoF that need to be controlled when generating movements. However, it is unlikely that the vast majority of octopus movements rely on reducing DoF number in this way (Wells 1976). A deeper understanding of how complex cephalopod movements are generated requires more detailed information about how the neural and mechanical systems interact. This is a substantial technical challenge because it very difficult to record and decipher the neural activity controlling octopus movements (Gutfreund et al. 1998; Matzner et al. 2000; Sumbre et al. 2005), and the arms have an extensive peripheral nervous system that probably processes much of the sensorimotor information locally. Because of these challenges, attempts to replicate octopus-like arm movements in robots have concentrated on the structural features of the arm rather than mimicking the animal’s neuromechanical control system (Walker et al. 2005). For example, the octopus inspired soft robotic manipulator (Fig. 14.6) uses a soft elastomeric skin actuated by a combination of lengthwise tendon-like cables and circumferential SMA coils that change segment diameter and stiffness (Calisti et al. 2011; Laschi et al. 2009; Margheri et al. 2012; Mazzolai et al. 2012; Renda et al. 2012). When interacting with a fluid medium, this arm can replicate many of the movements seen in the animal. Research on octopus-like manipulators has stimulated interest in continuum and geometric control methods that could be applied to other arm-like devices (Cowan and Walker 2008; Füchslin et al. 2013; Hannan and Walker 2003; Neppalli et al. 2008). Hydrostats Worms use a much simpler mechanical hydrostat (see also Chapter 12). Each body segment is a fixed volume cylinder. Consequently, reducing segment

459

460

Neurobiology of Motor Control: Fundamental Concepts and New Directions

B

A

Octopus vulgaris

C

D

SMA artificial muscle

Figure 14.6 Octopus-inspired robot. (A) Octopus (Octopus vulgaris) grasping a human finger with one arm. (B) An octopus-like robot arm wrapping around a human wrist, in water. (C) Details of octopus-like robot arm. The external braid represents the mechanical structure of the arm, allowing for local and global deformations while maintaining arm shape. (D) Details of the SMA springs that generate local diameter reductions. From Kim et al. (2013) with permission.

diameter increases segment length and reducing segment length (by contracting longitudinal muscles) increases segment diameter (Wainwright 1988). The relationship between force and displacement is further modified by the anisotropic nature of the body wall, in particular a meshwork of fibers that increases the effects of muscle contraction (Kier 2012). This simple arrangement is exploited to produce a variety of wavelike locomotions. Similar principles have been used to build wormlike robots (Seok et al. 2013) (Fig. 14.7) using SMA meshes around an elastomeric body shell and segment length sensors to regulate oscillator commands to the mesh. Tension-Based Locomotion Caterpillar locomotion has often been described as peristaltic

and wormlike. However, because insect larvae have extensive internal air tubes (trachea), fluid pressure does not transmit forces and displacements in the same way as in true hydrostats. In the model system Manduca sexta (the tobacco hornworm), combined data from kinematics (Trimmer and Issberner 2007; van Griethuijsen and Trimmer 2009, 2010), ground reaction forces (Lin and Trimmer 2010a, 2012), tissue material characterization and modeling (Dorfmann et al. 2008; Lin et al. 2009; Paetsch et al. 2012;

Bio-inspired Robot Locomotion

A Circular muscle fibers Longitudinal muscle fibers Coelom

B

Oligochaeta

Longitudinal muscle

Circumferential muscle

C

Figure 14.7 Earth worm-inspired robot. (A) Muscular structure of Oligochaeta, which forms antagonistic pairs without skeleton or joints. (B) A mesh structure containing longitudinal and circumferential artificial muscles, creating an antagonistic pairing similar to the pairing in Oligochaeta. (C) Demonstration of various actuation modes. From Kim et al. (2013) with permission.

Woods et al. 2008), motorneuron recordings (Metallo et al. 2011; Simon et al. 2010a), body modeling (Lin et al. 2011a; Rieffel and Trimmer 2010; Saunders et al. 2011b), receptive field mapping and sensor characterization (Simon and Trimmer 2009; Tamarkin and Levine 1996; van Griethuijsen et al. 2013), and high-energy X-ray imaging (Simon et al. 2010b), have resulted in a new model for soft-bodied terrestrial locomotion (including climbing) called the “Environmental Skeleton” (Lin and Trimmer 2010b). This model predicts that conformable animals move by controlling the release of internal body tension; their locomotion therefore relies on stiff substrates to transmit most compressive forces. The motor activity that underlies a variety of gaits and complex behavior in caterpillars can be recorded with single neuron resolution using custom-built, flexible, multi-electrode arrays (Metallo et al. 2011). These and similar observations are expected to provide important insights for the development of hyperelastic control strategies. The model has been tested in a family of soft material robots (Softworms) (Fig. 14.8). The robots are monolithic, 10–15 cm long, and weigh 4 to 30 g. They are fabricated from silicon elastomers by casting or from a soft rubbery polymer using a multi-material 3D printer. Softworms are actuated with SMA microcoils controlled by pulse-width

461

462

Neurobiology of Motor Control: Fundamental Concepts and New Directions

A Power cables for SMAs and tracking LEDs) Stabilizers

Silicone rubber composite body

SMA coil actuators

5 cm

B

Figure 14.8 (A) A caterpillar inspired soft robot (GoQBot) fabricated from silicon elastomers and powered by a pair of SMA coils. The robot is capable of inching, crawling, and ballistic rolling. See Lin et al. (2011a). (B) A new generation of SMA powered soft robots 3-D printed from a soft polymer. These robots can be configured in any shape and their contact with the ground modified using modular “feet” printed from sticky or slippery materials. See Umedachi et al. (2013). A from Trimmer and Lin (2014) with permission.

modulated current (Lin et al. 2011b, 2013; Umedachi and Trimmer 2014; Umedachi et al. 2013), or with back-drivable motors coupled to the body using flexible “tendons”. Computer aided design allows making robots in any shape, including caterpillars, worms, fins, or wings. They can crawl, inch, roll (Lin et al. 2011a, 2013), climb steep inclines, swim, and are steered by differential activation of the SMAs (Umedachi and Trimmer 2014). The most inexpensive and robust versions are powered and controlled through fine wire tethers, but untethered versions can be fabricated that carry their own lithium polymer battery and a small four-channel radio receiver. Pneumatic Locomotion Although most EPN-based robots have lifelike movements and shapes reminiscent of various invertebrates, they are not based on any particular animal. The use of pressurized air as their primary motive force further distinguishes them from biological systems. Nonetheless, as bio-inspired structures, many exciting possibilities remain to be explored. Because they are not based on a single animal, pneumatic robots can be designed to fulfill diverse roles. They are particularly attractive

Bio-inspired Robot Locomotion

for use as manipulator arms and assistive devices in rehabilitation and prosthetics (Polygerinos et al. 2015).

14.5 Conclusion and Outlook Bio-inspired robots are driving robot engineering in a variety of new directions and are expected to find particular success operating in natural environments. This is a critical development because conventional robots have been designed for use in predefined and unchanging surroundings and are unsuitable for most home, office, and hospital applications. The most robust and adaptable robots are not only bio-inspired but also “bio-informed” in that they exploit fundamental biological principles of locomotion and control. An example of this design approach is BigDog, a dynamically stable quadruped robot (http://www.bostondynamics.com/robot_bigdog.html). Although the commercial nature of the project means that many details are not in the public domain, it appears that inverted pendulum mechanics played a central role in the mechanical design of BigDog legs (Lee and Biewener 2011), and thus presumably of BigDog control strategies. This replicates the well-known presence of inverted pendulum dynamics in many terrestrial biological legged walk systems (Raibert 1986). Another major development is increased incorporation of compliant materials into robots. In conventional robot engineering, compliance has been mainly used to diminish the effects of sudden and unexpected loading. With a better understanding of animal locomotion, compliance and self-stabilizing structural elements are now seen as ways to reduce the need for high central control precision and for increasing robustness. These ideas find their extreme expression in soft robots, where the material properties themselves are being exploited to make more capable devices. A major challenge in both rigid and soft robotics is better understanding of how control tasks should be distributed between centrally generated commands and peripheral mechanical responses. This is an area where more detailed studies of animal neuromechanics are likely to have the greatest impact. In addition to the use of CPG-based control systems, bio-inspired robotics would benefit from an intensive effort to understand the deep organizational principles that make neural commands, mechanical responses, and sensory feedback so well integrated in animals. Like animals, the next generation of bio-inspired robots will have control systems in which the central computer and the physical plant are equal partners in directing movement and adapting to complex environments. Once there is an established framework for designing such systems, engineered devices will quickly outperform their biological counterparts.

References Adamatzky A, Komosinski M, Crespi A, Ijspeert AJ (2009) Salamandra robotica: a biologically inspired amphibious robot that swims and walks. In Artificial Life Models in Hardware, pp. 35–64: Springer: New York, NY. An S-M, Ryu J, Cho M, Cho K-J (2012) Engineering design framework for a shape memory alloy coil spring actuator using a static two-state model. Smart Mater Struct 21:16.

463

464

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Ananthanarayanan A, Azadi M, Kim S (2012) Towards a bio-inspired leg design for high-speed running. Bioinspir Biomim 7:046005. Arienti A, Calisti M, Giorgio-Serchi F, Laschi C (2013) PoseiDRONE: design of a soft-bodied ROV with crawling, swimming and manipulation ability. Oceans – San Diego, 2013 1:23–27. Ates M (2013) A review study of (bio)sensor systems based on conducting polymers. Mater Sci Eng 33C:1853–1859. Bahramzadeh Y, Shahinpoor M. (2014) A review of ionic polymeric soft actuators and sensors. Soft Robotics 1:38–52. Blickhan R (1989) The spring mass model for running and hopping. J Biomech 22:1217–1227. Bongard J (2011) Morphological change in machines accelerates the evolution of robust behavior. Proc Natl Acad Sci USA 108:1234–1239. Bongard J, Zykov V, Lipson H (2006) Resilient machines through continuous self-modeling. Science 314:1118–1121. Buschmann T (2010) Simulation and Control of Biped Walking Robots. Ph.D. thesis, Technische Universität München. Buschmann T, Ewald A, Twickel A, Büschges A (2014) Controlling legs for locomotion—insights from robotics and neurobiology. Bioinspir Biomim 10:041001. Buschmann T, Ewald A, Ulbrich H, Büschges A (2012) Event-based walking control—from neurobiology to biped robots. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1793–1800. Buschmann T, Lohmeier S, Bachmayer M, Ulbrich H, Pfeiffer F (2007) A collocation method for real-time walking pattern generation. In Proceedings of the IEEE–RAS International Conference on Humanoid Robots (Humanoids), pp. 1–6. Buschmann T, Lohmeier S, Ulbrich H (2009) Biped walking control based on hybrid position/force control. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 3019–3024. Calisti M, Giorelli M, Levy G, Mazzolai B, et al. (2011) An octopus-bioinspired solution to movement and manipulation for soft robots. Bioinspir Biomim 6:036002. Carpi F, Kornbluh R, Sommer-Larsen P, Alici G (2011) Electroactive polymer actuators as artificial muscles: are they ready for bioinspired applications? Bioinspir Biomim 6:045006. Cheney N, MacCurdy R, Clune J, Lipson H (2013) Unshackling evolution: evolving soft robots with multiple materials and a powerful generative encoding. In Proceeding of the 15th Annual Conference on Genetic and Evolutionary Computation Conference, pp. 167–174. Cheng S, Wu Z (2012) Microfluidic electronics. Lab Chip 12:2782–2791. Chevallereau C, Abba G, Aoustin Y, Plestan F, et al. (2003) Rabbit: a testbed for advanced control theory. Control Systems, IEEE 23:57–79. Chirikjian GS, Burdick JW (1995) The kinematics of hyper-redundant robotic locomotion. IEEE T Robotic Autom 11:781–793. Collins S, Ruina A (2005) A bipedal walking robot with efficient and human-like gait. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 1983–1988. Cowan LS, Walker ID (2008) “Soft” continuum robots: the interaction of continuous and discrete elements. Artif Life XI 11, 126–133.

Bio-inspired Robot Locomotion

Craig JJ (1989) Introduction to Robotics: Addison-Wesley: Boston MA. Daltorio K, Horchler A, Shaw K, Chiel H, Quinn R (2013) Stable heteroclinic channels for slip control of a peristaltic crawling robot. In Biomimetic and Biohybrid Systems, vol. 8064 (N. Lepora N, Krapp AMH, Verschure PMJ, Prescott TJ, eds), pp. 59–70. Springer: New York, NY. Deimel R, Brock O (2013) A compliant hand based on a novel pneumatic actuator. In 2013 IEEE International Conference on Robotics and Automation (ICRA), pp. 2047–2053. Diamond A, Knight R, Devereux D, Holland O (2012) Anthropomimetic robots: concept, construction and modelling. Int J Adv Robotic Sys 9:209:2012. Dorfmann A, Ogden R (2004) A constitutive model for the Mullins effect with permanent set in particle-reinforced rubber. Int J Solids Struct 41:1855–1878. Dorfmann AL, Woods WA, Trimmer BA (2008) Muscle performance in a soft-bodied terrestrial crawler: constitutive modelling of strain-rate dependency. J R Soc Interface 5:349–62. Endo G, Morimoto J, Matsubara T, Nakanishi J, Cheng G (2008) Learning CPG-based biped locomotion with a policy gradient method: Application to a humanoid robot. Int J Robot Res 27:213–228. Fan JA, Yeo W-H, Su Y, Hattori Y, et al. (2014) Fractal design concepts for stretchable electronics. Nature Comm 5:3266. Füchslin RM, Dzyakanchuk A, Flumini D, Hauser H, et al. (2013) Morphological computation and morphological control: Steps toward a formal theory and applications. Artif Life 19:9–34. Fukuoka Y, Katabuchi H, Kimura H (2010) Dynamic locomotion of quadrupeds Tekken3&4 using simple navigation system. J Robot Mechatron 22:36–42. Fukuoka Y, Kimura H, Cohen A (2003) Adaptive dynamic walking of a quadruped robot on irregular terrain based on biological concepts. Int J Robot Res 22:187–202. Full R, Kubow T, Schmitt J, Holmes P, Koditschek D (2002) Quantifying dynamic stability and maneuverability in legged locomotion. Integr Comp Biol 42:149–157. Fung YC (1993) Biomechanics: mechanical properties of living tissues: Springer: New York, NY. Gehring C, Coros S, Hutter M, Bloesch M, et al. (2013) Control of dynamic gaits for a quadrupedal robot. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA). https://doi.org/10.3929/ethz-a-01002 3052. Gozen BA, Tabatabai A, Ozdoganlar OB, Majidi C (2014) High-density soft-matter electronics with micron-scale line width. Adv Mater 30:5211–5216. Gutfreund Y, Flash T, Fiorito G, Hochner B (1998) Patterns of arm muscle activation involved in octopus reaching movements. J Neurosci 18:5976–5987. Gutfreund Y, Flash T, Yarom Y, Fiorito G, et al. (1996) Organization of octopus arm movements: a model system for studying the control of flexible arms. J Neurosci 16:7297–307. Haines CS, Lima MD, Li N, Spinks GM, et al. (2014) Artificial muscles from fishing line and sewing thread. Science 343:868–872. Hannan MA, Walker ID (2003) Kinematics and the implementation of an elephant’s trunk manipulator and other continuum style robots. J Robotic Sys 20:45–63. Hatton RL, Ding Y, Choset H, Goldman DI (2013) Geometric visualization of self-propulsion in a complex medium. Phys Rev Lett 110:078101.

465

466

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Hawker G, Buehler M (2000) Quadruped trotting with passive knees: design, control, and experiments. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) 3:3046–3051. Hirai K, Hirose M, Haikawa Y, Takenaka T (1998) The development of Honda humanoid robot. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) 2:1321–1326. Hirose M, Ogawa K. (2006) Honda humanoid robots development. Philos T R Soc A 365:11–19. Hobbelen D, de Boer T, Wisse M (2008) System overview of bipedal robots Flame and Tulip: tailor-made for limit cycle walking. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 2486–2491. Hooper SL, Guschlbauer C, Blümel M, Rosenbaum P, et al. (2009) Neural control of unloaded leg posture and of leg swing in stick insect, cockroach, and mouse differs from that in larger animals. J Neurosci 29:4109–4119. Horak F, Nashner L (1986) Central programming of postural movements: adaptation to altered support-surface configurations. J Neurophysiol 55:1369–1381. Hu J (2007) Shape Memory Polymers and Textiles, pp. 360: Woodhead Publishing Limited: Cambridge, England. Hutter M, Gehring C, Bloesch M, Hoepflinger M, et al. (2012) StarlETH: A compliant quadrupedal robot for fast, efficient, and versatile locomotion. In Proceedings of the International Conference on Climbing and Walking Robots, pp. 1–8. Hyun DJ, Seok S, Lee J, Kim S (2014) High speed trot-running: implementation of a hierarchical controller using proprioceptive impedance control on the MIT Cheetah. Int J Robot Res 33:1417–1445. Ijspeert AJ (2008) Central pattern generators for locomotion control in animals and robots: a review. Neural Networks 21:642–653. Iwasaki T, Chen J, Friesen WO (2014) Biological clockwork underlying adaptive rhythmic movements. Proc Natl Acad Sci USA 111:978–983. Kajita, S, Kanehiro F, Kaneko K, Fujiwara K, et al. (2003) Biped walking pattern generation by using preview control of zero-moment point. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) 2:1620–1626. Kajita S, Nagasaki T, Kaneko K, Yokoi K, Tanie K (2005) A running controller of humanoid biped HRP-2LR. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 616–622. Kajita S, Tani K (1995) Experimental study of biped dynamic walking in the linear inverted pendulum mode. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) 3:2885–2891. Kano T, Kobayashi R, Ishiguro A (2013) Decentralized control scheme for adaptive earthworm locomotion using continuum-model-based analysis. Adv Robotics 28:197–202. Keplinger C, Sun J-Y, Foo CC, Rothemund P, et al. (2013) Stretchable, transparent, ionic conductors. Science 341:984–987. Kier WM (2012) The diversity of hydrostatic skeletons. J Exp Biol 215:1247–1257. Kier WM, Stella MP (2007) The arrangement and function of octopus arm musculature and connective tissue. J Morphol 268:831–843.

Bio-inspired Robot Locomotion

Kim B, Lee S, Park JH, Jong-Oh P (2005) Design and fabrication of a locomotive mechanism for capsule-type endoscopes using shape memory alloys (SMAs) IEEE/ASME Transactions on Mechatronics 10:77–86. Kim S, Laschi C, Trimmer B (2013) Soft robotics: a bioinspired evolution in robotics. Trends Biotechnol 31:287–294. Kimura H, Fukuoka Y (2000) Biologically inspired dynamic walking of a quadruped robot on irregular terrain—adaptation at spinal cord and brain stem. In Proceedings of the First International Symposium on Adaptive Motion of Animals and Machines, http:// adaptivemotion.org/AMAM2000/amam2000.html. Laschi C, Cianchetti M, Mazzolai B, Margheri L, et al. (2012) Soft robot arm inspired by the octopus. Adv Robot 26:709–727. Laschi C, Mazzolai B, Mattoli V, Cianchetti M, Dario P (2009) Design of a biomimetic robotic octopus arm. Bioinspir Biomim 4:015006. Lee DV, Biewener AA (2011) BigDog-inspired studies in the locomotion of goats and dogs. Integr Comp Biol 51:190–202. Li G, Zhang H, Zhang J, Bye RT (2014) Development of adaptive locomotion of a caterpillar-like robot based on a sensory feedback CPG model. Adv Robotics 28:389–401. Lin HT, Dorfmann AL, Trimmer, BA (2009) Soft-cuticle biomechanics: a constitutive model of anisotropy for caterpillar integument. J Theor Biol 256:447–457. Lin HT, Leisk G, Trimmer BA (2011b) GoQBot: A caterpillar-inspired soft-bodied rolling robot. Bioinspir Biomim 6:026007–21. Lin HT, Leisk GG, Trimmer BA (2013) Soft robots in space: a perspective for soft robotics. In Acta Futura, vol. 6 (Barker DJ, ed), pp. 69–79. ESTEC, PPC-PF, 2201 AZ, Noordwijk, The Netherlands: European Space Agency, Advanced Concepts Team. Lin HT, Slate DJ, Paetsch, CR, Dorfmann, AL, Trimmer BA (2011a) Scaling of caterpillar body properties and its biomechanical implications for the use of a hydrostatic skeleton. J Exp Biol 214:1194–1204. Lin HT, Trimmer BA (2010a) The substrate as a skeleton: ground reaction forces from a soft-bodied legged animal. J Exp Biol 213:1133–1142. Lin HT, Trimmer BA (2010b) Caterpillars use the substrate as their external skeleton: a behavior confirmation. Commun Integr Biol 3:471–474. Lin HT, Trimmer BA (2012) A new bi-axial cantilever beam design for biomechanics force measurements. J Biomech 45:2310–2314. Lipson H (2013) Challenges and opportunities for design, simulation, and fabrication of soft robots. Soft Robotics 1:21–27. Loepfe M, Schumacher CM, Stark WJ (2014) Design, performance and reinforcement of bearing-free soft silicone combustion-driven pumps. Ind Eng Chem Res 53: 12519–12526. Lohmeier S, Buschmann T, Ulbrich H (2009) System design and control of anthropomorphic walking robot LOLA. IEEE/ASME Transactions on Mechatronics, 14:658–666. Lu N, Kim D-H (2014) Flexible and stretchable electronics paving the way for soft robotics. Soft Robotics 1:53–62. Ma KY, Chirarattananon P, Fuller SB, Wood RJ (2013) Controlled flight of a biologically inspired, insect-scale robot. Science 340:603–607. Marder E, Bucher D (2001) Central pattern generators and the control of rhythmic movements. Curr Biol 11:R986–R996.

467

468

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Margheri L, Laschi C, Mazzolai B (2012) Soft robotic arm inspired by the octopus. I. From biological functions to artificial requirements. Bioinspir Biomim 7:025004. Martinez RV, Branch JL, Fish CR, Jin L, et al. (2013) Robotic tentacles with three–dimensional mobility based on flexible elastomers. Adv Mater 25:205–212. Matsuoka K (1985) Sustained oscillations generated by mutually inhibiting neurons with adaptation. Biol Cybern 52:367–376. Matzner H, Gutfreund Y, Hochner B (2000) Neuromuscular system of the flexible arm of the octopus: physiological characterization. J Neurophysiol 83:1315–1328. Mazzolai B, Laschi C, Cianchetti M, Patane F, et al. (2007) Biorobotic investigation on the muscle structure of an octopus tentacle. Conf Proc IEEE Eng Med Biol Soc 2007, pp. 1471–1474. Mazzolai B, Margheri L, Cianchetti M, Dario P, Laschi C (2012) Soft-robotic arm inspired by the octopus. II. From artificial requirements to innovative technological solutions. Bioinspir Biomim 7:025005. McGeer T (1990) Passive dynamic walking. Int J Robot Res 9:62–82. Menciassi A, Gorini S, Pernorio G, Dario P (2004) A SMA actuated artificial earthworm. In Proceedings of the 2004 IEEE International Conference on Robotics and Automation (ICRA 2004). Messner WC, Paik J, Shepherd R, Kim S, Trimmer BA (2014) Energy for biomimetic robots: challenges and solutions. Soft Robotics 1:106–109. Metallo C, White RD, Trimmer BA (2011) Flexible parylene-based microelectrode arrays for high resolution EMG recordings in freely moving small animals. J Neurosci Meth 195:176–84. Morin SA, Kwok SW, Lessing J, Ting J, et al. (2014) Elastomeric tiles for the fabrication of inflatable structures. Adv Funct Mater 24:5541–5549. Mosadegh B, Polygerinos P, Keplinger C, Wennstedt S, et al. (2014) Pneumatic networks for soft robotics that actuate rapidly. Adv Funct Mater 24:2163–2170. Nakanishi J, Morimoto J, Endo G, Cheng G, et al. (2004) An empirical exploration of phase resetting for robust biped locomotion with dynamical movement primitives. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Narioka K, Hosoda K (2008) Designing synergistic walking of a whole-body humanoid driven by pneumatic artificial muscles: An empirical study. Adv Robotics 22:1107–1123. Neˇcasová Š, Takahashi T, Tucsnak M (2011) Weak solutions for the motion of a self-propelled deformable structure in a viscous incompressible fluid. Acta Appl Math 116:329–352. Nelson GM, Quinn RD (1999) Posture control of a cockroach-like robot. Control Systems, IEEE 19:9–14. Nelson G, Saunders A, Swilling B, Bondaryk J, et al. (2012) Petman: A humanoid robot for testing chemical protective clothing. J Robot Soc Japan, 30:372–377. Neppalli S, Csencsits MA, Jones BA, Walker I (2008) A geometrical approach to inverse kinematics for continuum manipulators. In IEEE/RSJ International Conference on Intelligent Robots and Systems, 2008. pp. 3565–3570. Nishiwaki K, Kagami S (2006) High frequency walking pattern generation based on preview control of ZMP. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 2667–2672.

Bio-inspired Robot Locomotion

Nishiwaki K, Sugihara T, Kagami S, Kanehiro F, et al. (2000) Design and development of research platform for perception–action integration in humanoid robot: H6. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 3:1559–1564. Nolfi S, Floreano D (2001) Evolutionary robotics. The biology, intelligence, and technology of self-organizing machines. MIT Press: Cambridge, MA. Ogden S, Klintberg L, Thornell G, Hjort K, Bodén R (2014) Review on miniaturized paraffin phase change actuators, valves, and pumps. Microfluid Nanofluid 17:53–71. Ostrowski J, Burdick J (1998) The geometric mechanics of undulatory robotic locomotion. Int J Robotic Res 17:683–701. Paetsch C, Trimmer BA, Dorfmann A (2012) A constitutive model for active–passive transition of muscle fibers. Int J Nonlin Mech 47:377–387. Paoletti P, Mahadevan L (2014) A proprioceptive neuromechanical theory of crawling. Proc R Soc B 281:20141092. Park Y-L, Majidi C, Kramer R, Bérard P, Wood RJ (2010) Hyperelastic pressure sensing with a liquid-embedded elastomer. J Micromech Microeng 20:125029. Pfeiffer F (2006) The TUM walking machines. Philos T Roy Soc A 365:109–131. Polygerinos P, Wang Z, Galloway KC, Wood RJ, Walsh CJ (2015) Soft robotic glove for combined assistance and at-home rehabilitation. Robotics and Autonomous Systems 73:135–143. Pratt G, Williamson M (1995) Series elastic actuators. Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 1:399–406. Pratt J, Pratt G (1998) Intuitive control of a planar bipedal walking robot. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) 3:2014–2021. Raibert M. (1986) Legged Robots that Balance (Artificial Intelligence) MIT Press, Cambridge. Raibert M, Blankenspoor K, Nelson G, Playter R, the BigDog Team (2008) BigDog, the rough-terrain quadruped robot. In Proceedings of the 17th International Federation of Automatic Control World Congress, pp. 10822–10825. Renda F, Cianchetti M, Giorelli M, Arienti A, Laschi C (2012) A 3D steady-state model of a tendon-driven continuum soft manipulator inspired by the octopus arm. Bioinspir Biomim 7:025006. Rieffel J, Knox D, Smith S, Trimmer B (2013) Growing and evolving soft robots. Artif Life 20:143–162. Rieffel J, Saunders F, Nadimpali S, Zhou H, et al. (2009) Evolving soft robotic locomotion in PhysX. In Proceedings of the 11th Annual Conference Companion on Genetic and Evolutionary Computation Conference Computational Intelligence on Consumer Games and Graphics Hardware (CIGPU) pp. 2499–2504. Rieffel J, Trimmer B (2010) Body/brain co-evolution in soft robots. In Artificial Life XII. Proceedings of the Twelfth International Conference on the Synthesis and Simulation of Living Systems, (Fellermann MDH, Hanczyc MM, Laursen LL, Maurer S, et al., eds), pp. 257–261. MIT Press: Cambridge, MA. Rogers JA, Someya T, Huang Y (2010) Materials and mechanics for stretchable electronics. Science 327:1603–1607. Sato T, Kano T, Ishiguro A. (2012) A decentralized control scheme for an effective coordination of phasic and tonic control in a snake-like robot. Bioinspir Biomim 7:016005.

469

470

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Saunders F, Golden E, White R, Rife J (2011a) Experimental verification of soft-robot gaits evolved using a lumped dynamic model. Robotica 29:823–830. Saunders F, Rieffel J, Rife J (2009) A method of accelerating convergence for genetic algorithms evolving morphological and control parameters for a biomimetic robot. In Proc. International Conference on Autonomous Robots and Agents. Saunders F, Trimmer BA, Rife J (2011b) Modeling locomotion of a soft-bodied arthropod using inverse dynamics. Bioinspir Biomim 6:016001. Schuldt D, Rife J, Trimmer BA (2015) A template for robust soft-body crawling with reflex-triggered gripping. Bioinspir Biomim 10:016018. Semini C (2010) HyQ—Design and Development of a Hydraulically Actuated Quadruped Robot. Ph.D. thesis, University of Genoa. Seok S, Onal CD, Cho K-J, Wood, RJ, et al. (2013) Meshworm: a peristaltic soft robot with antagonistic nickel titanium coil actuators. IEEE/ASME Trans Mechatronics, 18:1485–1497. Shabana AA (2005) Dynamics of Multibody Systems. 3rd Revised Edition. Wiley: New York, NY. Shapere A, Wilczek F (1987) Self-propulsion at low Reynolds number. Phys Rev Lett 58:2051–2054. Shapere A, Wilczek F. (1989) Geometry of self-propulsion at low Reynolds number. J Fluid Mech 198:557–585. Shepherd RF, Ilievski F, Choi W, Morin SA, et al. (2011) Multigait soft robot. Proc Natl Acad Sci USA 108:20400–20403. Shepherd RF, Stokes AA, Freake J, Barber J, et al. (2013) Using explosions to power a soft robot. Angew Chem Int Edit 125:2964–2968. Shinya A, Tsuchiya K (2006) Stability analysis of a simple walking model driven by an oscillator with a phase reset using sensory feedback. IEEE Transactions on Robotics 22:391–397. Simon MA, Fusillo SJ, Colman K, Trimmer BA (2010a) Motor patterns associated with crawling in a soft-bodied arthropod. J Exp Biol 213:2303–2309. Simon MA, Trimmer BA (2009) Movement encoding by a stretch receptor in the soft–bodied caterpillar, Manduca sexta. J Exp Biol 212:1021–1031. Simon MA, Woods Jr WA, Serebrenik YV, Simon SM, et al. (2010b) Visceral–locomotory pistoning in crawling caterpillars. Curr Biol 20:1458–1463. Sreenath K, Park H-W, Poulakakis I, Grizzle J (2011) A compliant hybrid zero dynamics controller for stable, efficient and fast bipedal walking on MABEL. Int J Robot Res 30:1170–1193. Sugimoto N, Morimoto J (2011) Phase-dependent trajectory optimization for CPG-based biped walking using path integral reinforcement learning. In Proc. IEEE–RAS Int. Conf. on Humanoid Robots (Humanoids), pp. 255–260. Sumbre G, Fiorito G, Flash T, Hochner B. (2005) Neurobiology: motor control of flexible octopus arms. Nature 433:595–596. Sumbre G, Fiorito G, Flash T, Hochner B (2006) Octopuses use a human-like strategy to control precise point–to–point arm movements. Curr Biol 16:767–72. Sumbre G, Gutfreund Y, Fiorito G, Flash T, Hochner B (2001) Control of octopus arm extension by a peripheral motor program. Science 293:1845–1848. Sun H, Chen X-P (2014) Towards honeycomb PneuNets robots. In Robot Intelligence Technology and Applications 2, pp. 331–340: Springer: New York, NY.

Bio-inspired Robot Locomotion

Tajima R, Honda D, Suga K (2009) Fast running experiments involving a humanoid robot. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 1571–1576. Takenaka T, Matsumoto T, Yoshiike T (2009a) Real time motion generation and control for biped robot—1st report: walking gait pattern generation. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1084–1091. Takenaka T, Matsumoto T, Yoshiike T, Hasegawa T, et al. (2009b) Real time motion generation and control for biped robot—4th report: integrated balance control. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1601–1608. Tamarkin DA, Levine RB (1996) Synaptic interactions between a muscle-associated proprioceptor and body wall muscle motor neurons in larval and adult Manduca sexta. J Neurophysiol 76:1597–1610. Tolley MT, Shepherd RF, Mosadegh B, Galloway KC, et al. (2014) A resilient, untethered soft robot. Soft Robotics 1. DOI: 10.1089/soro.2014.0008 Trimmer B, Issberner J (2007) Kinematics of soft-bodied, legged locomotion in Manduca sexta larvae. Biol Bull 212:130–142. Trueman ER (1975) The locomotion of soft-bodied animals. London: Edward Arnold. Tsagarakis N, Li Z, Saglia J, Caldwell D (2011) The design of the lower body of the compliant humanoid robot. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 2035–2040. Umedachi T, Trimmer BA (2014) Design of a 3-D printed soft robot with posture and steering control. In Proceedings of 2014 IEEE International Conference on Robotics and Automation (ICRA). Umedachi T, Vikas V, Trimmer B (2013) Highly deformable 3-D printed soft robot generating inching and crawling locomotions with variable friction legs. In 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 4590–4595. Urata J, Nishikawa K, Nakanishi Y, Okada K, et al. (2011) Online decision of foot placement using singular lq preview regulation. In Proceedings of the IEEE–RAS International Conference on Humanoid Robots (Humanoids). van Griethuijsen LI, Trimmer BA (2009) Kinematics of horizontal and vertical caterpillar crawling. J Exp Biol 212:1455–1462. van Griethuijsen LI, Trimmer BA (2010) Caterpillar crawling over irregular terrain: anticipation and local sensing. J Comp Physiol A196:397–406. van Griethuijsen LI, Banks KM, Trimmer BA (2013) Spatial accuracy of a rapid defense behavior in caterpillars. J Exp Biol 216379–387. Wainwright SA (1988) Axis and Circumference. Harvard University Press: Cambridge, Massachusetts. Walker ID, Dawson D, Flash F, Grasso F, et al. (2005) Continuum robot arms inspired by cephalopods. In Proceedings of the 2005 SPIE Conference on Unmanned Ground Vehicle Technology IV, pp. 303–314. Wells MJ (1976) Octopus: Physiology and Behaviour of an Advanced Invertebrate. Springer: New York, NY. Wittmann R, Hildebrandt A, Ewald A, Buschmann T (2014) An estimation model for footstep modifications of biped robots. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

471

472

Neurobiology of Motor Control: Fundamental Concepts and New Directions

Woods Jr WA, Fusillo SJ, Trimmer BA (2008) Dynamic properties of a locomotory muscle of the tobacco hornworm Manduca sexta during strain cycling and simulated natural crawling. J Exp Biol 211:873–882. Yamaguchi J, Takanishi A (1997) Development of a biped walking robot having antagonistic driven joints using nonlinear spring mechanism. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) 1:185–192. Yekutieli Y, Sagiv-Zohar R, Aharonov R, Engel Y, et al. (2005) Dynamic model of the octopus arm. I. Biomechanics of the octopus reaching movement. J Neurophysiol 94:1443–1458. Yuk H, Kim D, Lee H, Jo S, Shin JH (2011) Shape memory alloy-based small crawling robots inspired by C. elegans. Bioinspir Biomim 6:046002. Zhang Z, Philen M (2012) Pressurized artificial muscles. J Intell Mat Syst Str 23:255–268.

473

Index a

b

Acceleration sensitive afferent, see Sensory organs and neurons Acetylcholine, see Neurotransmitters/ neuromodulators Across-individual variation 108, 109, 113, 161, 162, 195, 394, 398–400, 417, 433 Action potential (see also Extracellular mono-and bipolar recording, Intracellular recording, sharp electrode, and Patch clamp) 8, 28 Antidromic spike 277 Conduction velocity 18, 19, 147, 435 Intracellular (transmembrane) shape bi or monophasic 33 Threshold 251, 435 Affordance 375, 377, 388, 389, 392, 394, 395, 397, 400, 401 Aggression, see Motor pattern All-trans-retinal 87, 88 AMPA, see Neurotransmitters/neuromodulators Amygdala 214, 215 Anomala 164 Antennal Johnston organ, see Sensory organs and neurons Aplysia californica 137, 161, 238 Aplysia feeding 120, 383–386, 389, 398, 418, 426–434 Aplysia tail withdrawal 161, 418 Archaerhodopsin 236 Assistance reflex 273, 274, 277, 278, 280, 282, 292, 293

Basal ganglia 178, 181–191, 214–216, 434 Conservation from lamprey to mammals 184, 185 Bat 138 Bat echolocation calls 136, 138, 149, 307 Bat flight 152, 153 Bat phylogeny 138 Bird, mammal, turtle, sauropsid forebrain organization 144, 145 Body/nervous system morphology Arthropod 274, 275 Axial (with respect to locomotion) 263 Human vs horse leg 446 Insect 196, 308, 320 Long-tail crustacea 308 Stomatogastric 309 Swimmeret 317 Multiple identically segmented Annelid 308, 315 Leech, lamprey 150, 196, 323 Terrestrial quadruped (limb CPG location) 323 Body shortening (leech), see Motor pattern Bombesin, see Neurotransmitters/ neuromodulators Bone morphogenic factor (BMF) 419 Bounding (locomotion gait), see Motor pattern Bötzinger complex 288, 328 Brachiation, see Motor pattern Bradykinin, see Neurotransmitters/ neuromodulators

Neurobiology of Motor Control: Fundamental Concepts and New Directions, First Edition. Edited by Scott L. Hooper and Ansgar Büschges. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc.

474

Index

Brain (invertebrate) 79, 82, 90, 92, 93, 135, 141, 162, 179, 195–199, 201–206, 210–216, 274, 275, 306, 314, 320, 321, 327 Brainstem 55, 159, 178, 181, 182–186, 188–190, 216, 227, 286, 321, 323, 324, 327–331 Bursicon, see Neurotransmitters/ neuromodulators

c C-start, see Motor pattern Ca indicators (to monitor neuron activity) 90–92, 236 Caenorhabditis elegans 76, 85, 137, 156, 157, 198, 202, 207, 208, 213, 380 Campaniform sensilla, see Sensory organs and neurons Canada geese respiration 329 Carbachol, see Neurotransmitters/ neuromodulators Cat 19, 59, 269, 283–288, 290–292, 378, 381, 382, 383, 385, 393, 400 Cat locomotion 225, 226, 233, 234, 249, 250, 264, 266, 323, 331 Cat paw shake, see Motor pattern Cat respiration 331 Cat swallowing 327 Central chemoreceptors, see Sensory organs and neurons Central complex 145, 179, 202–207, 213–216, 275 Central pattern generator (see also individual species entries, Motor pattern, and Rhythmogenesis and its mechanisms) Changes in extracellular ion concentrations and 243, 244, 246 Definition 178, 182, 183, 196, 198, 225–227, 264, 305, 418, 451 Development of 245, 247, 251, 252, 417–420, 433 Homeostasis in 245, 246, 252, 419, 435, 436 Electrical activity in 419 Spontaneous movements in 419

Fatigue of 420 Neuromodulation of 150, 159–161, 198, 201, 229, 230, 249, 250 Probabilistic output 197, 198, 430, 431 Reconfiguration of 227, 228, 249–252, 312, 313 Role of various membrane conductances in 241–246 Calcium-activated nonselective, ICAN 242, 243, 247 Calcium-activated potassium, IK(Ca) 242 Delayed rectifier potassium, IK(V) 242 Hyperpolarization-activated inward current, Ih 244, 245 Leak potassium, IK2P 242 L-type calcium current, ICaL 244 NMDA current 245 Persistent sodium, INaP 242–244, 246 Na/K ATPase 243 Sodium-activated potassium, IK(Na) 242 Transient potassium, IA 241, 244, 251 Transient receptor channel 242, 243 T-type calcium current, ICaT 230, 244 Summary of vertebrate 321–324 Locomotor 236–238, 322, 323 Respiratory 323 Feeding 324 Synaptic mechanisms 234, 239, 246–249 Electrical synapses 248 Excitatory 247–248 Inhibitory 246, 247 Multi-component 248 Cerci, see Sensory organs and neurons Cerebellum 59, 95, 97, 98, 119, 162–164, 182, 418, 426, 433, 434 Channelrhodopsin 77, 87, 88, 95, 97, 98, 236 Cheetah 379 Cheetah (robot) 443 Chemoreceptor, see Sensory organs and neurons Chewing, see Motor pattern Chlordimeform, see Neurotransmitters/ neuromodulators

Index

Chordotonal organ, see Sensory organs and neurons Cichlid swim speed:gape size correlation 136 Clione limacina 150 Clione limacine swimming 150 Cockroach 56, 199, 202, 204–208, 274, 276, 278, 282, 319, 321, 371, 372, 377 Command neurons, see Movement selection and higher order control Computer simulation Action potential 112 Basic concepts 29, 121–123 Differential equation approach 121, 122 Probabilistic models 122–123 Biomechanical models 128–129, 293 C. elegans neuromechanical model 380 Cat hindlimb 289, 290 Combined neural to biomechanical in posture and locomotion 293, 294 Crayfish walking network 280 Failure of averaging in 109 Ion conductances 124–125 Lizard tongue biomechanical model 381 Multi-compartment 108, 112, 125, 126 Muscle models 128 Hill 268, 269 Of crustacean limb 267–274 Neuron models 123–126, 293 Conductance based 123–126 Reduced dimensional models 126, 269 Optimization and verification 118–120 Pyloric network (see also Crustacean stomatogastric system) 111–114 Simulation of large networks 116–118 Synaptic models 127–128 Tritonia swim network 114–117 Control theory 263–267 Copulation, see Motor pattern Coordinating interneurons 275, 306, 313, 316–319, 331

Coordination mechanisms Central connections 306, 311–313, 316–319, 324–326, 328, 330–332 Changes in descending input 306, 312–314, 316, 328–332 Muscle temporal filtering 311, 314, 315, 332 Sensory feedback 306, 311, 313–314, 317, 319, 320, 321, 324–327, 329, 330–332 Summary 321 Coordination of legs 306 Biped and tetrapod As speed changes 229, 233, 234, 247, 250, 306 By central connections 326 By sensory feedback 326, 327 Flexor-extensor 95, 96, 233, 234, 238, 252, 267–273, 323, 326, 327 Joint 273, 306, 323, 325 Left-right 157, 158, 229, 232, 238, 249, 252, 306, 323, 326 Role of Renshaw cells 238, 326, 327 Stance-swing 282, 306 Six-legged 197–199, 204–208, 320, 321 Other insects 320, 321 Stick insect (completely by sensory feedback) 320 Tetrapod gait 319 Tripod gait 319 Coordination of neural oscillators 305 Coordination of respiration and locomotion 322, 328–331 Chemoreception and 329, 330 Integer coupling of 322 Peripheral nervous system and 329, 330 Feedforward mechanisms 330, 331 Coordination of respiration and swallowing 322, 327–329 Coordination of segmental oscillators General 307 Lamprey swim 150, 323–325 By central connections 324, 325 By sensory feedback 324, 325 Intersegmental 325 Left-right alternation 229, 231, 324, 325

475

476

Index

Coordination of segmental oscillators (contd.) Leech Crawling 316, 317 By central connections 316, 317 By changes in descending input 316 Heartbeat 229, 231 Swim 150 By central connections 316 By sensory feedback 317 Metachronal 315 Tadpole swim 323–325 Coordination of stomatogastric (crustacean) rhythms By changes in descending input 312–314 By network fusion 312 By sensory feedback 311, 313–314 Central coordination 311–313 Integer coupling of different rhythms 312 Neuron switching between networks 312 On muscle level 311, 314, 315 Coordination of swimmeret unit oscillators 317–319 By central connections 317–319 By sensory feedback 319 Corazonin, see Neurotransmitters/ neuromodulators Cortex (see also Motor cortex) 32, 47, 55, 57, 91, 120, 144, 145, 153, 181–191, 341, 342, 348, 353, 355, 360, 361, 426, 430, 433, 434, 445 Corticospinal system 183, 342, 391, 434 Cough, see Motor pattern Courtship (Drosophila), see Motor pattern Crab (see also Anomala, Hermit crab, Hippoidea, Sand crab, and Stomatogastric system (crustacean) 152, 164, 229, 282, 309–315, 458 Crawling, see Motor pattern Crayfish 10, 33, 164, 197, 199, 201, 202, 234, 274–279, 281, 282, 306, 313, 315, 317, 400

Cricket 195, 202, 203 Crustacean cardiac ganglion (heartbeat CPG) 234 Crustacean cardioactive peptide (CCAP) (see Neurotransmitters/ neuromodulators) Crustacean cardioactive peptide/myoinhibitory peptide (CCAP/MIP) (see Neurotransmitters/neuromodulators) Crustacean stomach, see Crustacean stomatogastric system and Motor pattern/Gut movements Crustacean stomatogastric system 11, 159, 161, 198, 199, 227, 229, 230, 234, 240, 248, 251, 252 Cardiac sac/cardiac sac network and activity 306–315 Esophagus/esophageal network and activity 306–315 Gastric mill/gastric mill network and activity 227–229, 306–315 Pylorus/pyloric network and activity 111–114, 227–229, 234, 235, 239, 241, 243–245, 247–251, 306–315, 431 Cutaneous mechanoreceptors, see Sensory organs and neurons Cuttlefish 149

d Degree of freedom (in biological motor control) 343, 344, 350, 377, 378, 380, 400 Degree of freedom (in robotics) 289, 444 Dendronotus iris swimming 140, 141 Diencephalic locomotor region (DLR) 183, 185–187, 191 Diphtheria toxin 77, 90, 96, 98, 236 Dog respiration 329 Dolabrifera dolabrifera tail withdrawal 161 Dopamine, see Neurotransmitters/ neuromodulators Dorsal cochlear nucleus 163 Dorsal light response (lamprey) 183 DMRT3 gene effect on locomotion 157

Index

DREADDS 82, 88, 89 Drosophila melanogaster 75–98, 145, 201–206, 213, 319, 320, 398, 422, 425

e Ecdysis, see Motor pattern Ecdysis triggering hormone, see Neurotransmitters/ neuromodulators Eclosion hormone, see Neurotransmitters/ neuromodulators Efferent copy 95 Electric fish 135, 147–149 Elephant 129, 341 Enkephalin, see Neurotransmitters/ neuromodulators Ephrin receptor effect on locomotion 157 Escape, see Motor pattern Evolutionary concepts and examples Convergent evolution, see Homoplasy in this section Correlation of mushroom body function with behavior, not phylogeny 139 Deep homology 142–145 Hox genes and 142 Of mammalian basal ganglia and insect central complex? 145, 179, 213–216 Of sauropsid (bird, turtle, reptile) and mammalian forebrain 144, 145 Divergence 164 In decapod crustacean tail flip 164 In nudibranch swim-related neurons 164 In origin of different nematode feeding behaviors 164 Dollo’s Law (untrue) 152 Evolvability 161–162 Homologous neurons 139–142 An identified serotonergic gastropod neuron 140 Mauthner (M) 140–143 Si1-Si3 neurons in Melibe and Dendronotus 140, 141 Homology 138

Gastropod dorsal-ventral and left-right swim networks not homologous 140 Of left-right swimming in Melibe and Dendronotus 140 Of vertebrate dorsal spinal cord and invertebrate ventral spinal cord? 179, 214 Of vertebrate respiration and its brain areas 139 Homoplasy 139, 145–149, 179 Convergent vs. parallel homoplasy 145, 146 In central pattern generators 150–152 Clione and Xenopus swimming 150 Leech and lamprey swimming 150–152 Of elongated body form in fish 149 Of fish electric organs 147–150 Of giant axons 147 Of insect and vertebrate legs 139, 142 Of protusible tongues (frog) 154–155 Of similar behaviors 139, 140 Bat-bird-pterosaur flight 146 Bat-whale sound perception 145 Bipedal hopping 146 Bird-insect flight 145 Dorsal-ventral swimming in Tritonia and Pleurobranchaea 140 Finger dexterity in primates 146 Levee building in ants 139 Octopus and human retrieval movements 147 Orb weaving in spiders 139 Stick insect flight 146 Vocal learning in birds 146 Loss 152 Of flight 152, 153 Of limbs 152 Re-evolution of lost traits 152 Neuron duplication 162–164 As origin of cerebellum 162–164 In acoustic communication 164 Orthology 139 Parallel evolution, see Homoplasy this section

477

478

Index

Evolutionary concepts and examples (contd.) Paralogy 139 Phylogenetic signal 136–138 Phylogenetics 136–138 Evolution of novel motor behaviors 152–161 Bat flight 152 By generalist neural circuitry 154–157 Protrusible frog tongues 154, 155 By neuromodulation 155, 159–161 Amphibian larvae swimming 159 Nudibranch swimming 159–161 Stomatogastric ganglion (crustacea) 159 By rewired network synaptic connectivity 155, 157, 158 Mammalian gaits 157, 158 Nematode feeding networks 156, 157 Extracellular mono-and bipolar recording 17–21 Axon stimulation 18 Compound action potential 18, 19, 22, 46 Effectively intracellular recording 25, 45, 46 Electromyography 20, 21, 56, 67, 95 History 21–25 Capillary electrometer 24 Cathode ray tube 25, 43 Injury current 21, 28, 46 Non-polarizable electrodes 21 Overcoming slow galvanometers 22, 24 Overshooting action potential 24 Triode vacuum tube 24, 43 Hook, suction, pin, and wire electrodes 19, 20, 43, 290 Monopolar vs. bipolar 44–46 Multi-unit recording 11, 18, 47, 206, 208, 309, 328 Number of action potential phases 18, 32, 42–46 Predicting intracellular action potential waveform from 25 Theory 32–44

Approximate solution based on current flow 38–40 Detailed model simulation 25, 36–38 Qualitative current flow explanation 40–42 Solid angle qualitative explanation 34–36 Eye, see Sensory organs and neurons Eye blink, see Motor pattern Eye movement, see Motor pattern

f Fan-shaped body 179, 206, 214–216 Fascia 375, 379, 381, 382 Feeding, see Motor pattern Flatworm 137, 149 Flight, see Motor pattern FMRFamide, see Neurotransmitters/ neuromodulators Force sensitive receptor, see Sensory organs and neurons Forebrain organization, sauropsid (bird, turtle, reptile), mammal 144, 145 Forelimb development 152 Frog 20, 135, 137, 142, 150, 154, 197, 323, 359, 360, 392, 418, 420 Frontal lobe 189 fruitless 203 Funnel canal organ, see Sensory organs and neurons Fura-2 91 𝛾-aminobutyric acid (GABA), see Neurotransmitters/ neuromodulators 𝛾 motor neuron, see Sensory organs and neurons

g GABA, see Neurotransmitters/ neuromodulators Gain (of amplifiers or feedback circuits and networks) 15, 29, 247, 273, 286, 444, 450, 459 Gait, see Motor pattern Gene expression, controlling 77–84, 236 Promoter bashing, enhancer trapping 77–81

Index

Cre/LOX P 79, 80, 95, 96 Driver lines 80–81 Drosophila 80–81 Mouse 81 Flp/FRT 79, 95 Gal4-UAS (Drosophila) 78, 79, 92–94 Targeting specific cell types 81 Temporally controlled activation 82–84 By RU-486 or estradiol application 82 By temperature 82 Tet and hybrid Tet-Cre systems 82–84 Genetic manipulation of neuron activity 87–90, 252 Ablation 90, 95, 236, 250, 326, 327 Chemogenetic 88, 89, 95 Optogenetic 87, 88, 213, 236, 247, 252, 253, 325 Thermogenetic 89, 90, 213 Genetically encoded calcium indicators 91, 252 Giant fiber (see also Medial giant fiber and Lateral giant fiber) 92, 93, 201–202 Gill withdrawal, see Motor pattern Globus pallidus externa 185, 186, 188, 215 Globus pallidus interna 185, 188, 215 Glutamate, see Neurotransmitters/ neuromodulators Glycine, see Neurotransmitters/ neuromodulators Gnathal ganglion, see Subesophageal/gnathal ganglion Golgi tendon organ, see Sensory organs and neurons Gradient Index Rod microendoscopes 91 Grasping, see Motor pattern Grasshopper 152, 202 Grooming, see Motor pattern Gymnotid (fish) 147–149

h H reflex, see Hoffmann (H) reflex Habituation 418 Halorhodopsin 77, 88, 89, 236 Hair cell (vestibular), see Sensory organs and neurons

Hair plate, see Sensory organs and neurons Hard-body robots 446–452 Actuators 446 Definition 444 Electric motor 446 Hydraulic 446 Pneumatic 446 CPG-based control of walking 451, 452, 463 Entrainment in 451 Half-center implementation 451 Sensory feedback in 451, 452 Compliance Addition 446, 463 Definition 444 Definition 443 Degrees of freedom 444, 449 Dynamics of legged locomotion 447, 448 Limit cycle walkers 376, 448, 450, 451, 463 Lumped mass models 450 Increased efficiency 451 Passive dynamic walker 450 Self-stability 450 Planar, definition 446 Posture example 445, 446 Rigid-body dynamics 444 Sensory feedback 444, 447, 451, 452 Stability 448 Dynamic 283, 448, 463 In biological literature 448 In mechanics 448 Static 283, 285, 348, 448 Trajectory oriented walking control 448–450 Contact force control 450 Stepping strategy 450 Upper body acceleration 450 Underactuated Definition 444 Presence 447, 449 Heartbeat, see Motor pattern Hermissenda crassicornis 159–161 Hermit crab 164 Hippocampus 55, 66, 88, 91, 144, 214, 215, 249

479

480

Index

Hippoidea 164 Histamine, see Neurotransmitters/ neuromodulators Hoffmann (H) reflex 286, 435 Hop, see Motor pattern Horse 138, 157, 158, 250, 446 Hox gene 142, 145, 152, 164, 237, 238 Hugin, see Neurotransmitters/ neuromodulators Human 44, 56, 82, 88, 116, 118, 138, 145, 283, 285, 322, 344, 349, 351, 373, 376–378, 380, 391, 397, 398, 402, 420, 433, 435, 445, 446 Human locomotion 266, 286, 373, 377, 379, 391–393 Human reaching 147, 341, 359 Human swallowing 327 Human respiration 329 Hydrostat (compare Muscular hydrostat, see also Soft-body robots) 375, 380, 459, 460

i Insect jump to flight transition, see Motor pattern Intracellular recording, sharp 9–17, 202 Cell damage 10, 19, 31 Diffusion of neuron fill into neuron cytoplasm 10, 31, 32 History 25–27 First intracellular recordings 25–27 First voltage clamp 25–26 Liquid junctional potentials 31–32 RC characteristics 13, 14, 29 Tip clogging 10 Voltage clamp 15–17 Single electrode discontinuous 14–16 Two electrode 15, 29, 30 Voltage recording (simple) 9–11, 29, 30 Voltage recording and current injection (“current clamp”) 12–15 Bridge balance 14 Single electrode bridge balance 14 Single electrode discontinuous 14, 15 Two electrode 14 Intrinsic cellular properties Bistability 239, 242, 244, 251, 252, 368

Endogenous oscillation 150, 197, 234, 238–240, 248, 249, 251, 309 Plateau potential 197, 239, 240, 243, 244, 251, 252, 309, 428, 429 Post-inhibitory rebound 150, 197, 234, 239, 240, 244, 247, 248, 309, 312 Spike afterhyperpolarization 242, 251 Spike frequency adaptation 12, 240, 247 Invariant movement features, see Movement selection and higher order control Ion conductances and pumps Calcium-activated nonselective, ICAN 230, 242, 243, 251, 252 Calcium-activated potassium, IK(Ca) 242, 250 Chloride 201, 207, 435, 436 Chloride/potassium exchanger 251 Delayed rectifier potassium, IK(V) 242 Hyperpolarization-activated inward current, Ih 13, 230, 234, 240, 244–246 Inward rectifier (potassium) Kir 186 Leak potassium, IK2P 242, 251 L-type calcium current, ICaL 230, 244 Na/K ATPase (see also Short term motor learning) 28, 243, 421–426 Persistent sodium, INaP 230, 240, 242–244, 251 Sodium-activated potassium, IK(Na) 242 Transient potassium, IA 234, 241 Transient receptor channel 89, 90, 242, 243 T-type calcium current, ICaT 230, 244

j Jerboa 146, 157 Jump, see Motor pattern

k Kangaroo rat

146, 157

l Lamprey 137, 142, 143, 162, 163, 182–185, 187–191, 426 Lamprey respiration 323, 331 Lamprey swimming 150–152, 189, 190, 227, 229–232, 239, 242, 306, 315, 323–326, 337

Index

Lapping, see Motor pattern Lateral accessory lobe 205–207, 215 Lateral giant (LG) neuron 164 Lateral habenula 187, 188, 190, 191 Lateral reticular nucleus 95, 98 Learning and plasticity (see also Operant learning and Short term motor learning) Following injury 197, 434–436 Homeostasis and 419, 435, 436 In maturing vs adult animals 433, 434 Leech 19, 120, 137, 150, 151, 162, 197, 199, 202, 207–213, 227, 229, 232, 234, 306, 308, 315–317, 319, 321, 380, 425 Leech body shortening 208–213 Leech crawling 162, 208–213, 316, 317 Leech heartbeat 197, 227, 229, 232, 234, 426 Leech swimming 150, 162, 208–213, 227, 306, 315, 316, 321 Leg coordination, see Coordination of legs Legged locomotion, see Motor pattern Length sensitive receptor, see Sensory organs and neurons Lizard 137, 152, 381 Load dependent receptor, see Sensory organs and neurons Lobster (see also Crustacean stomatogastric system) 11, 33, 112, 161, 164, 234, 235, 241, 275, 282, 308–315, 317 Locust 43, 67, 152, 196, 197, 199, 201–203, 227, 239, 240, 248, 249, 276, 277, 282, 378 Locust flight 152, 197, 199, 201, 203, 227, 239, 240, 248, 249

m M1 (vertebrate primary motor cortex, see also Motor cortex) 189, 341, 342, 353–355, 360 Mammal, turtle, bird, sauropsid forebrain organization 144, 145 Manduca sexta 199, 373, 460 Mauthner (M) neuron 140–143, 162, 201, 400

Mechanosensor/mechanoreceptor, see Sensory organs and neurons Medial giant (MG) neuron 164 Medulla 82, 86, 91, 95, 98, 232, 323, 324, 327, 331 Melibe leonine swimming 140, 141 Membrane properties Capacitance 13–17, 29, 31, 107, 123 Goldman-Hodgkin-Katz (GHK) equation 28 Membrane potential 8 Rest potential 26, 28, 239, 241, 244, 251, 423 Nernst potential 22, 27 That membrane currents are carried by protein channels 27 Mesencephalic locomotor region (MLR) 183–185, 187–189, 191, 330, 331 Mesothoracic ganglion 203, 320 Metabotropic, see Neurotransmitters/ neuromodulators Metathoracic ganglion 196, 199, 204, 320 Midbrain 181, 183, 190, 215 Midshipman fish 149, 164 Mirror neuron 398 Molting, see Motor pattern Monkey 55, 135, 353, 358, 394, 435 Mormyrid (fish) 147–149 Motor cortex 55, 91, 120, 341, 348, 355, 360, 433, 434, 445 Motor homunculus 189, 341, 342 Motor learning, short term (see also Operant learning) 420–426, 432–434 Dependence of swim bout duration on time from last swim 421, 422 Na pump conferred 28, 421, 422, 434 Auxiliary pump subunits (FXYD proteins) 425 Post-swim ultraslow afterhyperpolarization (usAHP) 422–424 How alters swim bout duration 423, 425 Not associated with change in membrane conductance 422

481

482

Index

Motor learning, short term (see also Operant learning) (contd.) Proportional to spike number 421–423 Spinal neuron distribution 424, 425 Presence of similar mechanisms in other species 422, 425, 426 Motor pattern (see also Motor pattern coordination and individual species entries) Aggression 195 Airway defensive (cough, sneeze) 321 Body shortening (leech) 208–213 Bounding (locomotion gait) 250 Brachiation 135 C-start 142 Cat paw shake 286, Chewing 181, 182, 246, 321, 324 Copulation 178 Courtship (Drosophila) 203 Crawling 162, 199, 207–213, 425 Ecdysis (molting) 201 Escape 201–202, 205 Escape in Drosophila 92, 93 Eye blink 418 Eye movement 178, 181, 182, 187, 189, 190 Feeding 178, 182, 189, 190, 199, 323, 341, 418, 433 Feeding in Apysia 418 Feeding in Drosophila 92, 94, 202 Feeding in gastropod mollusks 197 Feeding in locust 199 Feeding in nematodes 156, 157 Flight 93, 135, 146, 152, 197, 199, 202–204, 227, 239, 240, 248, 249, 343 Gait 154, 157, 158, 181, 198, 204, 250, 251, 293, 319, 326, 377, 391, 420, 435, 445, 449 Gill withdrawal (aplysiid mollusks) 418 Grasping 341, 342, 433 Grooming 196, 201, 207, 208, 341, 343 Gut movements (crustacean stomatogastric system, CPG driven) 159, 227–230, 234, 235, 239, 240, 306–315

Heartbeat (CPG driven, crustacean) 234 Heartbeat (CPG driven, leech) 197, 229, 230, 234, 426 Hop 326 Insect jump to flight transition (Drosophila, locust) 92, 93, 203 Jump 196, 373 Lapping (tongue) 381 Legged locomotion 95, 135, 146, 154, 157, 158, 178, 181, 182, 187, 196, 198–208, 226, 227, 229, 231–234, 236–239, 244–247, 252, 263, 264, 276–321, 323, 330, 331, 343, 425, 433 Molting (ecdysis) 201 Orienting 181, 183, 187, 189, 204, 205 Posture 178, 182–184, 186, 198, 239, 244, 252, 263, 264, 266, 267, 270, 273–276, 282–286, 292, 293, 342, 345–349, 352, 353, 355, 358–360, 373, 376–378, 387, 389, 400, 433, 434, 445, 447 Prehensile 182, 341–361 Prey capture 154 Reaching (see also Prehensile this section) 182, 342, 343 Reflex 181–183, 196, 264, 266, 270, 272–278, 280, 282–286, 290, 292, 293, 418, 433, 435 Respiration 178, 182, 239, 321, 433 Frog 323 Lamprey 323 Mammals 227–232, 238, 240, 242–245, 247, 249, 323, 327, 328, 330, 331 Retrieval 147 Scratching 197, 343 Sexual behavior 417, 418, 426 Siphon withdrawal (aplysiid mollusks) 418 Stridulation 202 Swallow (vertebrate) 182, 321, 324, 327, 328 Phases of 327 Swimmeret 120, 197, 201, 234, 315, 317–319

Index

Tail flip (decapod crustacea) 164, 201 Tail withdrawal (aplysiid mollusks) 161, 418 Traveling wave 150 Undulatory swimming 135, 140, 141, 149–152, 159–162, 164, 189, 190, 207–213, 239, 249, 321, 323, 325, 343, 418, 420–425 Vocal sound production 164 Motor module, see Movement production Motor primitive, see Movement production Motor synergy, see Movement production Mouse/rat 13, 59, 75–98, 152, 242, 247, 248, 252, 426, 435 Mouse/rat locomotion 158, 233, 234, 236–238, 242, 247, 248, 252, 435 Movement primitive, see Movement production Movement production 375–387 Context (contraction state of other muscles, limb posture) dependence 375, 383–386 Effects of surrounding medium 377 In hydrostats 380 In muscular hydrostats 381, 382 Effects of fiber geometry and movement 380, 381 In skeletal systems 377–380 Effects of tendons and facia 378–380 Limb/animal size and 376, 377 Optimal vs. ‘‘don’t care” solutions 388, 389, 392–397 Redundancy in motor control 343–349, 389–397 As aid to learning 397, 398 As source of motor abundancy 388, 392 Definition 343–345, 388 Ill-posed problem 345 Well-posed problem 346 Forward problem 345–346 Inverse problem 346, 348, 349 Joint angle level 347 Muscle level 347, 389 Trajectory level 346–347 Redundancy in motor control, overcoming 349–361

General considerations 349–350 Invariant motor features and 350–352 Donder’s law 351 Fitt’s law 351 Two-thirds power law 351 Increasing number of task conditions and 353–357 End-state comfort 355 Jerk 352 Minimum jerk model 352, 353 Minimum variance theory 355, 356 Neuron directional tuning 353–355 Population vector 55, 353–355 Summary 356, 357 Decreasing degrees of freedom and 357–361 Kinematic constraints 358, 359 Motor synergy (motor module, motor primitive, movement primitive, muscle synergy) 357, 359, 360, 390–392, 400 Parametric curve movement construction 358 Summary 360, 361 Relative importance of inertial, viscous, and spring-like forces in 375–377 Movement selection and higher order control Action selection 343, 344 Basal ganglia and 179, 184–188, 434 Central complex and 179, 202–207, 213–216 Command neurons 95, 120, 178, 201–204, 207, 208, 210, 213, 275, 321, 399, 400, 430 Courtship 203 Crawling 207, 210, 211 Escape 178, 201–202, 205, 399, 400 Food intake vs. locomotion 202 Insect forward and backward walking 275, 321 Locomotion 95, 178, 203, 204, 213 Stridulation 202 Swimming 207–208, 210

483

484

Index

Movement selection and higher order control (contd.) Conservation throughout vertebrates 184, 185, 187, 189–191 Control by cortex (higher vertebrates)/pallium (lower vertebrates) 189–191 Deep homology of central complex and basal ganglia? 179, 213–216 Descending control 178, 182–190, 202–207, 275, 342 Direct GO pathway 184, 187 Disinhibition and 184, 189 Dopamine and (vertebrate) 184–187 Feeding vs. locomotion in Drosophila 92, 94, 202 GO (direct) pathway 184, 186 Goal directed behavior 189, 190, 195, 196, 342, 418, 426 Indirect (NO GO) pathway 184, 186, 187 Internal state 195, 197, 213, 214, 277, 278, 332 Intrinsically probabilistic 197, 198 Lack of requirement of neo-cortex in 189, 190 Locomotion vs. feeding (Drosophila) 92, 94, 202 Motivational state 195, 197, 205 Mouse forelimb reaching 95, 97, 98 Mouse locomotion 95–98 Mushroom body and 202–208 NO GO indirect pathway 184, 186 Selection by neuromodulators/neurohormones (invertebrate) 198–201, 207–208, 210–213 Subesophageal ganglion and 178, 199, 203, 213 Swimming vs. crawling in C. elegans 207 Swimming vs. crawling vs. body shortening in leech 162, 207–213 Touch avoidance in C. elegans 207

Multifunctional networks 162, 181, 197, 202, 227, 228, 249–252, 321, 322, 420, 426–432, 434 Multi-electrode extracellular recording 202, 206, 252, 290 History 56, 63–67 Implementation 56–60, 204–207 Data acquisition chain 60 Electrodes 58–59 Microdrives 59–60 Jitter 66 Requirement of multiple recording sites 58 Unit identification in (spike sorting) 60–67, 207–208 Muscarine, see Neurotransmitters/ neuromodulators Muscle 366–374 Actomyosin 367, 369–371 Architecture of (arrangement of muscle fibers and tendon) 378 Catch 128, 369, 370 Dependence of action on contraction state of other muscles, “behavioral context” 371–373 Fatigue 346, 369, 420 Force-frequency 366, 367, 369, 371, 373 Force-length 128, 283, 290, 367, 369, 371, 373 Force-velocity 128, 283, 290, 367, 369, 371, 373 History-dependence 367–371, 373 Latch 128, 369, 370 Low-pass filtering 314, 315, 366, 367, 381 Multifunctionality of 381–384 Neuromodulation of 368 Neuromuscular junction 368 Passive force production in 367 Power production in (motors, springs, brakes, struts) 372–373 Regional differences in function 373, 375, 381–384 Shape (variation in) 374–375 Short range stiffness 266, 272, 283, 284, 369, 370 Temporal summation in 366, 367

Index

Viscoelastic properties, advantages 443 Work-loop technique 371 Muscle tendon organ, see Sensory organs and neurons Muscle receptor organ, see Sensory organs and neurons Muscle spindle, see Sensory organs and neurons Muscle synergy, see Movement production Muscular hydrostat (compare Hydrostat, see also Soft-body robots) 375, 380, 381, 385, 459 Mushroom body 139, 198, 202–208, 215

n Negative feedback 15, 29, 264, 266, 269, 272–276, 292 Delay and 266 Proportional-integral-derivative (PID) controller (and subsets thereof ) 265–267, 269, 270, 276, 283, 285, 292 Neuron anatomy 8 Axon 8 Dendrite 8 Effects of extended geometry 29 Nicotine, see Neurotransmitters/ neuromodulators Nitric oxide, see Neurotransmitters/ neuromodulators NMDA, see Neurotransmitters/ neuromodulators Noradrenaline, see Neurotransmitters/ neuromodulators Neocortex 55, 144, 145, 189, 190, 240, 244, 249 Neurotransmitters/neuromodulators (includes their receptors and agonists/antagonists) Acetylcholine 89, 199, 207, 230, 313 Muscarinic 88, 89, 279, 280 Nicotinic 201 Bombesin 90 Bradykinin 242 Bursicon 201 Carbachol 26, 201 Chlordimeform 201

Corazonin 201 Crustacean cardioactive peptide (CCAP) 201 Crustacean cardioactive peptide/myoinhibitory peptide (CCAP/MIP) 201 Dopamine 13, 145, 184–188, 201, 205, 207, 211, 212, 215, 243, 248, 251, 316, 317, 419, 426, 429–431, 436 Ecdysis triggering hormone 201 Eclosion hormone 201 Enkephalin 185 FMRFamide 201, 230 GABA 163, 184–186, 188, 201, 205, 214, 215, 230, 250, 251, 286, 435 Glutamate 113, 127, 163, 184–186, 188, 189, 232, 235, 236, 243, 245, 247, 249, 323, 326, 331, 424 AMPA 127, 230, 247 NMDA 127, 159, 230, 236, 245, 247 Glycine 230, 232, 250, 251, 324, 424, 435 Histamine 77, 88, 89, 185, 186, 230 Hugin 92, 94, 202 Metabotropic 207, 243, 247, 248, 253, 368 Nitric oxide 159, 195 Noradrenaline 159, 230, 251 Octopamine 195, 199–200, 205, 212, 230, 249 Opioid 205 Ort (histamine-gated chloride channel) 77, 88, 89 Ouabain 422 P2X purinoreceptor 88, 93 Picrotoxin 201 Pilocarpine 199, 201 Serotonin (5HT) 117, 141, 159, 161, 185, 186, 195, 201, 207, 211–213, 230, 236, 250, 251, 313, 419, 435 Substance P 90, 185, 186, 251, 331 Tetrodotoxin (TTX) 27, 248, 422, 423 Tyramine 199, 201, 207 𝛾-aminobutyric acid, see GABA this section NMDA, see Neurotransmitters/ neuromodulators

485

486

Index

NO GO indirect pathway, see Movement selection and higher order control Noradrenaline, see Neurotransmitters/ neuromodulators

o Oarfish 149 Ocelli, see Sensory organs and neurons Octavolateralis nuclei 163 Octopamine, see Neurotransmitters/ neuromodulators Octopus 129, 401, 454, 460 Octopus reaching 147, 341, 454, 459, 460 Olfactory bulb 187, 190 Olfactory system 67, 89, 139, 198, 205 Operant learning 426–432, 435 Definition 426 Hybrid biological-computer simulation (dynamic clamp) testing of 431, 432 Inherent to CPG itself in Aplysia feeding 428 Molecular, cellular, and synaptic basis of learning 428–430 cAMP and 429 Dopamine and 429 Increased cell excitability and 429–432 Protein kinase A (PKA) and 429 Protein kinase C (PKC) and 429 Strengthened electrical coupling and 431, 432 Negative reinforcement 426 Positive reinforcement 426, 429 Opioid, see Neurotransmitters/ neuromodulators Optic tectum 163, 187, 189–191 Oregon green bapta-1 91 Orienting, see Motor pattern Ort (histamine-gated chloride channel), see Neurotransmitters/ neuromodulators Ouabain, see Neurotransmitters/ neuromodulators

p P2X purinoreceptor, see Neurotransmitters/ neuromodulators

Pallidum 184 Pallium 144, 184–191 Parallel fiber 163 Parkinson’s disease 186, 244, 426 Patch clamp 201–202, 421, 422 Cell-detached patches 16, 17 Diffusion of neuron fill into neuron cytoplasm 9, 11 Electronic circuit 29–31 History (first single channel recordings) 26, 27 Liquid junctional potentials 31–32 Perforated 12, 13 Single channel recording (cell membrane under electrode intact) 9, 26, 27 Whole cell patch clamp (cell membrane intact) 16 Whole cell voltage clamp (cell membrane under electrode ruptured or perforated) 13, 17 Whole cell voltage recording (cell membrane under electrode ruptured or perforated) 9, 12 Whole cell voltage recording and current injection (cell membrane under electrode ruptured or perforated) 12, 13 Pedunculopontine nucleus 187 Pharynx 92, 94, 157, 327 Phase response curve (PRC) 114, 115 Phylogeny of Animalia 137 Phylogeny of fish 143 Phylogeny of Vertebrata 137 Picrotoxin, see Neurotransmitters/ neuromodulators Pilocarpine, see Neurotransmitters/ neuromodulators Pleurobranchaea californica swimming 140, 159–161 Position sensitive afferent, see Sensory organs and neurons Positive feedback 266, 273–275, 278, 292 Postural reflex 273, 275, 276, 282–285 Praying mantis 274, 341 Pre-Bötzinger complex (PreBötC) 227–229, 232, 236, 240, 242, 243, 246, 251, 323, 328

Index

Prehension, see Motor pattern Presynaptic inhibition 272, 277, 278, 286, 292 Prey capture, see Motor pattern Primary afferent depolarization 271–273, 277, 281, 293 Primate 56, 59, 85, 95, 146, 182, 185, 188, 189, 341, 342, 353, 430 Pristionchus pacificus 157 Proprioception, see Sensory organs and neurons Prothoracic gangion 199, 203, 320 Protocerebrum 94, 203–206, 214, 215, 275 Purkinje cell 119, 163, 426 Pyramidal cell 55, 66, 426

r Rabbit 33, 189 Rabbit respiration 331 Radula 386, 426–429 Rana temporaria feeding 155 Rana temporaria swimming 159 Rat, see Mouse/rat and Rodent Ray (fish) 137, 142, 149, 163 Reaching, see Motor pattern Reflex, see Motor pattern Reflex chain 225, 264 Reflex regulation during locomotion 285, 286, 292 Reflex reversal 273, 277–282, 292 Renshaw cell 238, 326, 327 Resistance reflex 270, 272, 273, 275–278, 280, 284, 292 Respiration, see Motor pattern Reticulospinal tract/neuron 183, 189, 191, 231 Retrieval, see Motor pattern Rhythmogenesis and its mechanisms (see also Central pattern generation and Intrinsic cellular properties) Endogenous bursting neurons 150, 197, 234, 238–241, 244, 251, 309 Half-center 150, 226, 232, 234, 241, 275, 309, 316, 317, 451 Half-center vs. unit burst generator 232–234

Reciprocal excitation 241, 433 Reciprocal inhibition 247, 433 Unit burst generator 198, 232–234, 320, 322, 323, 325 Robots, see Hard-bodied robots and Soft-bodied robots Rodent 56, 66, 85, 86, 146, 157, 189, 229, 233, 239, 242, 244, 247, 250, 251, 323, 327, 430

s Sag potential 13 Sand crab 164 Sauropsid, mammal, turtle, bird forebrain organization 144, 145 Scratching, see Motor pattern Sensitization 161, 418 Sensory homunculus 196, 341 Sensory organs and neurons Antennal Johnston organ 196 Campaniform sensilla 275, 276, 278, 282 Central chemoreceptors 329, 330 Cerci 213 Chemoreceptor 213 Chordotonal organ 196, 274, 276, 278, 280, 282, 294, 320 Cutaneous mechanoreceptors 210, 213, 284, 285, 291, 292 Eye 351 Force sensitive 290 Funnel canal organ (FCO) 282 𝛾 motor neuron 269–273 Length sensitive 286, 288, 290, 292 Load dependent 288, 289, 290 Mechanosensor/mechanoreceptor 196, 205, 210, 213, 284 Muscle (Golgi) tendon organ 275, 284, 288, 292 Muscle receptor organ 269, 274, 276, 278, 280, 282 Muscle spindle 269–273, 277, 284–286, 288–290, 292, 435 Ocelli 213 Position sensitive afferent 276 Proprioceptor/proprioception 95, 248, 249, 274–277, 330

487

488

Index

Sensory organs and neurons (contd.) Stretch receptor 274, 276, 278, 282, 283, 291, 324, 325 Tactile hair (hair plate) 153, 203, 275, 291, 292 Tegula 248, 249 Topographic projection of 196 Touch receptor 207 Type Ia afferent 269–273, 276, 284–286, 288–292, 326, 327, 435 Type Ib afferent 284, 288–292 Type II afferent 284, 286, 288–292 Velocity sensitive 276 Vestibular 153, 183, 282, 285, 452 Sensory feedback 120, 182, 183, 196–198, 210, 225–227, 233, 249, 263–294, 305–308, 311, 313, 316, 319–322, 324, 329, 380, 387, 388, 391, 401, 418, 445, 451, 458, 463 Serotonin, see Neurotransmitters/ neuromodulators Sexual behavior, see Motor pattern Shark 191 Shibire (shits ) 77, 89, 90 Siphon withdrawal, see Motor pattern Skate (fish) 149 Skeleton 374, 375, 383, 457 Sneeze, see Motor pattern Soft-body robots 452–463 Actuators 453–455 Crystalline transition materials 454, 455, 459, 460, 462 Definition 444 Electroactive polymers 453, 454, 461, 462 Pneumatic 454, 455, 462, 463 Tension based 455, 459–461 Advantages 452–453 Biologically inspired 454, 459–463 Caterpillar 454, 460, 462 Hydrostat 459, 460 Muscular hydrostat 459 Octopus 454, 459, 460 Worms 454, 459–462 Compliance, definition 444 Definition 443

Degrees of freedom 343, 357–361, 444, 459 Limitations 452 Movement control strategies 456–459 Sensory feedback 444, 455–456 ‘‘Soft”, definition 452 Underactuated, definition 444 Somatotopy 196 Sonic hedgehog (SHH) 418, 419 Spinal cord 81, 82, 87, 91, 95, 96, 98, 142, 154, 157–159, 162, 178, 179, 181–183, 189, 191, 196, 197, 214, 216, 225, 229, 232, 233, 236, 239, 242, 244, 246, 248–251, 264, 272, 275, 285, 321–325, 331, 344, 359, 360, 391, 419, 420, 422, 424, 426, 433–436 Springhare 146 Squat lobster 164 Squid 25, 26, 33, 137, 380, 381, 387 Squid giant axon 25, 26, 33 Stance/swing transition 225, 227, 274, 286 Stick insect 109, 146, 152, 153, 199, 202, 225, 227, 229, 276, 278, 282, 319–321, 376 Stick insect walking 199, 225, 227, 229, 276, 278, 282, 319–321, 376 Stomatogastric system (crustacean), see Crustacean stomatogastric system Stretch receptor, see Sensory organs and neurons Stretch reflex 266, 284–286, 290, 435 Striatum 144, 179, 184–189, 214–216 Matrix component 179, 187, 188, 214, 216 Striosomal compartment 179, 187, 188, 214, 216 Stridulation, see Motor pattern Strychnine, see Neurotransmitters/ neuromodulators Suboesophageal/gnathal ganglion 178, 179, 198, 199, 201, 203, 213, 216 Substance P, see Neurotransmitters/ neuromodulators Substantia nigra neurons 13, 244

Index

Substantia nigra pars compacta (SNc) 185, 187, 215 Substantia nigra pars reticulata (SNr) 13, 184–188, 191 Subthalamic nucleus (STN) 185–188, 191, 215 Superior colliculus 183, 185 Supraoesophageal ganglion 198 Swallow, see Motor pattern Swimmeret, see Motor pattern

t Tachykinin, see Neurotransmitters/ neuromodulators Tactile hair, see Sensory organs and neurons Tail flip, see Motor pattern Tail withdrawal, see Motor pattern Tectum 163, 183, 185, 187–191 Tegula, see Sensory organs and neurons Telson 164 Tendon 284, 286, 288, 290–292, 329, 373, 378, 379, 382, 387, 444, 446, 455, 459 Tetrapod (bat, bird, pterosaur) flight, forelimb use 146 Tetrodotoxin, see Neurotransmitters/ neuromodulators Thalamocortical neuron 244, 245 Thalamus 183–189, 215 Toadfish 164 Tobacco hawkmoth, see Manduca sexta Tongue 154, 155, 324, 327, 374, 380, 381 Touch receptor, see Sensory organs and neurons Tracing neuron projections and connections 84–87, 236, 252 Fluorescent reporters 84, 85, 236 Viral tracers 85–87, 236 Traveling wave, see Motor pattern Triggerfish 149 Tritonia diomedea swimming 114, 116, 117, 140, 159–162, 164, 227, 248

Tunicate 365 Turtle 137, 162, 197, 244, 380, 436 Turtle, sauropsid, mammal, bird forebrain organization 144, 145 Type Ia afferent, see Sensory organs and neurons Type Ib afferent, see Sensory organs and neurons Type II afferent, see Sensory organs and neurons Tyramine, see Neurotransmitters/ neuromodulators

u Undulatory swimming, see Motor pattern Uropod 164

v Velocity sensitive afferent, see Sensory organs and neurons Ventral nerve cord 92, 94, 178, 179, 196, 198, 199, 201, 211–214, 216, 272, 275, 308, 309, 315, 317 Vestibular system, see Sensory organs and neurons Vibration sensitive afferent, see Sensory organs and neurons Vocal sound production, see Motor pattern

w Wasp 205 Wheeled vs. biological locomotion

443

x Xenopus laevis 33 Xenopus laevis tadpole swimming 150, 159, 227, 323–325, 418, 420, 421, 425, 426, 432–434

z Zebrafish 76, 81, 84, 98, 162, 213, 419, 436 Zebrafish swimming 227, 247, 249, 436

489