Cyclic Peptide Design [1st ed.] 978-1-4939-9503-5;978-1-4939-9504-2

Table of contents :
Front Matter ....Pages i-x
Design Principles for Intestinal Permeability of Cyclic Peptides (Alan M. Mathiowetz)....Pages 1-15
Strategies to Enhance Metabolic Stabilities (Bhavesh Khatri, Venkateswara Rao Nuthakki, Jayanta Chatterjee)....Pages 17-40
Designing Cell-Permeable Macrocyclic Peptides (George Appiah Kubi, Patrick G. Dougherty, Dehua Pei)....Pages 41-59
Computational Methods for Studying Conformational Behaviors of Cyclic Peptides (Fan Jiang, Hao Geng)....Pages 61-71
Computational Opportunities and Challenges in Finding Cyclic Peptide Modulators of Protein–Protein Interactions (Fergal Duffy, Nikunj Maheshwari, Nicolae-Viorel Buchete, Denis Shields)....Pages 73-95
Design of Cyclic Peptides as Protein Recognition Motifs (Ye Che)....Pages 97-106
Design and Synthetic Strategies for Helical Peptides (Licheng Tu, Dongyuan Wang, Zigang Li)....Pages 107-131
Click Chemistry for Cyclic Peptide Drug Design (Adel Ahmed Rashad)....Pages 133-145
Frontier Between Cyclic Peptides and Macrocycles (Philipp Ermert, Anatol Luther, Peter Zbinden, Daniel Obrecht)....Pages 147-202
Building upon Nature’s Framework: Overview of Key Strategies Toward Increasing Drug-Like Properties of Natural Product Cyclopeptides and Macrocycles (Maria-Jesus Blanco)....Pages 203-233
Design of Oxytocin Analogs (Kazimierz Wiśniewski)....Pages 235-271
DNA-Encoded Macrocyclic Peptide Library (Zhengrong Zhu, Alex Shaginian, La Shadric C. Grady, Christopher P. Davie, Kenneth Lind, Sandeep Pal et al.)....Pages 273-284
Peptide Display Technologies (Anthony Pitt, Zeke Nims)....Pages 285-298
Discovery of Functional Macrocyclic Peptides by Means of the RaPID System (Christos Tsiamantas, Manuel E. Otero-Ramirez, Hiroaki Suga)....Pages 299-315
Genetic Selections with SICLOPPS Libraries: Toward the Identification of Novel Protein–Protein Interaction Inhibitors and Chemical Tools (Francisco Castillo, Ali Tavassoli)....Pages 317-328
Back Matter ....Pages 329-332

Citation preview

Methods in Molecular Biology 2001

Gilles Goetz Editor

Cyclic Peptide Design

METHODS

IN

MOLECULAR BIOLOGY

Series Editor John M. Walker School of Life and Medical Sciences University of Hertfordshire Hatfield, Hertfordshire, AL10 9AB, UK

For further volumes: http://www.springer.com/series/7651

Cyclic Peptide Design Edited by

Gilles Goetz Hit Discovery and Optimization, Pfizer R&D, Groton, CT, USA

Editor Gilles Goetz Hit Discovery and Optimization Pfizer R&D Groton, CT, USA

ISSN 1064-3745 ISSN 1940-6029 (electronic) Methods in Molecular Biology ISBN 978-1-4939-9503-5 ISBN 978-1-4939-9504-2 (eBook) https://doi.org/10.1007/978-1-4939-9504-2 © Springer Science+Business Media, LLC, part of Springer Nature 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Humana imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer Nature. The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.

Preface If you are used to the Methods in Molecular Biology series, you will notice right away that there are very few methods and virtually no molecular biology in this particular edition. This book will certainly stand out in your collection. Very few chapters of this book are written in the typical MiMB format, but this does not take anything away from the value of the works presented here! Interests in cyclic peptide design, synthesis, and applications keep growing since this class of chemicals has become a credible alternative source of new drug leads on par with traditional small molecules. Cyclic peptides are particularly suited to bridge the chemical space gap between said small molecules and proteins and antibodies. This book covers strategies to improve cell permeability, intestinal permeability, and metabolic stability, which are the typical liabilities associated with cyclic peptides, to enhance protein-protein recognition, and to build upon nature’s cyclic peptides and macrocycles. Chapters also cover key peptide screening and display strategies, as well as important synthetic approaches towards cyclic and helical peptides. Researchers within the pharmaceutical industry as well as scientists and students in the bioorganic, medicinal, and natural product chemistry fields will find this book a critical resource and a go-to reference. Groton, CT, USA

Gilles Goetz

v

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v ix

1 Design Principles for Intestinal Permeability of Cyclic Peptides . . . . . . . . . . . . . . . Alan M. Mathiowetz 2 Strategies to Enhance Metabolic Stabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bhavesh Khatri, Venkateswara Rao Nuthakki, and Jayanta Chatterjee 3 Designing Cell-Permeable Macrocyclic Peptides . . . . . . . . . . . . . . . . . . . . . . . . . . . . George Appiah Kubi, Patrick G. Dougherty, and Dehua Pei 4 Computational Methods for Studying Conformational Behaviors of Cyclic Peptides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fan Jiang and Hao Geng 5 Computational Opportunities and Challenges in Finding Cyclic Peptide Modulators of Protein–Protein Interactions. . . . . . . . . . . . . . . . . . . . . . . . . Fergal Duffy, Nikunj Maheshwari, Nicolae-Viorel Buchete, and Denis Shields 6 Design of Cyclic Peptides as Protein Recognition Motifs. . . . . . . . . . . . . . . . . . . . . Ye Che 7 Design and Synthetic Strategies for Helical Peptides. . . . . . . . . . . . . . . . . . . . . . . . . Licheng Tu, Dongyuan Wang, and Zigang Li 8 Click Chemistry for Cyclic Peptide Drug Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . Adel Ahmed Rashad 9 Frontier Between Cyclic Peptides and Macrocycles . . . . . . . . . . . . . . . . . . . . . . . . . . Philipp Ermert, Anatol Luther, Peter Zbinden, and Daniel Obrecht 10 Building upon Nature’s Framework: Overview of Key Strategies Toward Increasing Drug-Like Properties of Natural Product Cyclopeptides and Macrocycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maria-Jesus Blanco 11 Design of Oxytocin Analogs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kazimierz Wis´niewski 12 DNA-Encoded Macrocyclic Peptide Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Zhengrong Zhu, Alex Shaginian, La Shadric C. Grady, Christopher P. Davie, Kenneth Lind, Sandeep Pal, Praew Thansandote, and Graham L. Simpson 13 Peptide Display Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Anthony Pitt and Zeke Nims

1

vii

17

41

61

73

97 107 133 147

203 235 273

285

viii

14

15

Contents

Discovery of Functional Macrocyclic Peptides by Means of the RaPID System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299 Christos Tsiamantas, Manuel E. Otero-Ramirez, and Hiroaki Suga Genetic Selections with SICLOPPS Libraries: Toward the Identification of Novel Protein–Protein Interaction Inhibitors and Chemical Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 Francisco Castillo and Ali Tavassoli

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

329

Contributors GEORGE APPIAH KUBI  Department of Chemistry and Biochemistry, The Ohio State University, Columbus, OH, USA MARIA-JESUS BLANCO  Sage Therapeutics, Inc., Cambridge, MA, USA NICOLAE-VIOREL BUCHETE  School of Physics, University College Dublin, Dublin, Ireland FRANCISCO CASTILLO  School of Chemistry, University of Southampton, Southampton, UK JAYANTA CHATTERJEE  Molecular Biophysics Unit, Indian Institute of Science, Bangalore, India YE CHE  Discovery Sciences, Pfizer Inc., Groton, CT, USA CHRISTOPHER P. DAVIE  GlaxoSmithKline, Cambridge, MA, USA PATRICK G. DOUGHERTY  Department of Chemistry and Biochemistry, The Ohio State University, Columbus, OH, USA FERGAL DUFFY  School of Medicine and Medical Science, University College Dublin, Dublin, Ireland; UCD Conway Institute of Biomolecular and Biomedical Research, University College Dublin, Dublin, Ireland PHILIPP ERMERT  Polyphor Ltd., Allschwil, Switzerland HAO GENG  Laboratory of Computational Chemistry and Drug Design, State Key Laboratory of Chemical Oncogenomics, Peking University Shenzhen Graduate School, Shenzhen, China LA SHADRIC C. GRADY  GlaxoSmithKline, Cambridge, MA, USA FAN JIANG  Laboratory of Computational Chemistry and Drug Design, State Key Laboratory of Chemical Oncogenomics, Peking University Shenzhen Graduate School, Shenzhen, China BHAVESH KHATRI  Molecular Biophysics Unit, Indian Institute of Science, Bangalore, India ZIGANG LI  State Key Laboratory of Chemical Oncogenomics, Peking University, Shenzhen Graduate School, Peking, China KENNETH LIND  GlaxoSmithKline, Cambridge, MA, USA ANATOL LUTHER  Polyphor Ltd., Allschwil, Switzerland NIKUNJ MAHESHWARI  School of Medicine and Medical Science, University College Dublin, Dublin, Ireland; UCD Conway Institute of Biomolecular and Biomedical Research, University College Dublin, Dublin, Ireland ALAN M. MATHIOWETZ  Pfizer Worldwide Research and Development, Cambridge, MA, USA ZEKE NIMS  Orbit Discovery Ltd., Oxford, UK VENKATESWARA RAO NUTHAKKI  Molecular Biophysics Unit, Indian Institute of Science, Bangalore, India DANIEL OBRECHT  Polyphor Ltd., Allschwil, Switzerland MANUEL E. OTERO-RAMIREZ  Department of Chemistry, Graduate School of Science, The University of Tokyo, Bunkyo, Tokyo, Japan SANDEEP PAL  GlaxoSmithKline, Stevenage, UK DEHUA PEI  Department of Chemistry and Biochemistry, The Ohio State University, Columbus, OH, USA ANTHONY PITT  Orbit Discovery Ltd., Oxford, UK ADEL AHMED RASHAD  College of Medicine, Drexel University, Philadelphia, PA, USA

ix

x

Contributors

ALEX SHAGINIAN  GlaxoSmithKline, Cambridge, MA, USA DENIS SHIELDS  School of Medicine and Medical Science, University College Dublin, Dublin, Ireland; UCD Conway Institute of Biomolecular and Biomedical Research, University College Dublin, Dublin, Ireland GRAHAM L. SIMPSON  GlaxoSmithKline, Stevenage, UK HIROAKI SUGA  Department of Chemistry, Graduate School of Science, The University of Tokyo, Bunkyo, Tokyo, Japan ALI TAVASSOLI  School of Chemistry, University of Southampton, Southampton, UK PRAEW THANSANDOTE  GlaxoSmithKline, Stevenage, UK CHRISTOS TSIAMANTAS  Department of Chemistry, Graduate School of Science, The University of Tokyo, Bunkyo, Tokyo, Japan LICHENG TU  State Key Laboratory of Chemical Oncogenomics, Peking University, Shenzhen Graduate School, Peking, China DONGYUAN WANG  State Key Laboratory of Chemical Oncogenomics, Peking University, Shenzhen Graduate School, Peking, China KAZIMIERZ WIS´NIEWSKI  Ferring Research Institute Inc., San Diego, CA, USA PETER ZBINDEN  Polyphor Ltd., Allschwil, Switzerland ZHENGRONG ZHU  GlaxoSmithKline, Cambridge, MA, USA

Chapter 1 Design Principles for Intestinal Permeability of Cyclic Peptides Alan M. Mathiowetz Abstract One of the most exciting facets of cyclic peptides is that they have the potential to be orally bioavailable, despite having physical properties well beyond the traditional “Rule-of-5” chemistry space (Lipinski et al., Adv Drug Deliv Rev. 23(1): 3–25, 1997). An important component of meeting this challenge is to design cyclic peptides with good intestinal permeability. Here we discuss the design principles for intestinal permeability that have been developed in recent year. These principles can be subdivided into three regimes: physical property guidelines, design strategies for the macrocyclic ring, and design strategies for side chains. The most important overall aims are to minimize solvent-exposed polarity while keeping size, flexibility, and lipophilicity within favorable ranges, thereby allowing peptide chemists to achieve intestinal permeability in addition to other important properties for their compounds, such as solubility and binding affinity. Here we describe a variety of design strategies that have been developed to help peptide chemists in this endeavor. Key words Cyclic peptides, Oral bioavailability, Intestinal permeability, Physical properties, Intramolecular hydrogen bonding, N-methylation, Beyond Rule-of-5

1

Introduction Cyclic peptides are an exciting therapeutic modality, with the potential to inhabit an attractive “middle space” between traditional small molecules and biologicals [1]. An intriguing possibility, as exemplified by cyclosporine A [2], is that cyclic peptides can be discovered which have significant oral bioavailability, providing the convenience and compliance of oral dosing while retaining the ability to bind to larger sites such as protein-protein interfaces. Cyclic peptides often lie just beyond the “Rule-of-5” [3], which describes a physical property space with an improved likelihood of oral bioavailability, but recent work has identified a wide range of cyclic peptides with significant oral bioavailability [4–11], including those shown in Table 1. A recent analysis of the properties of orally bioavailable drugs [12] described the physical properties most important for oral

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_1, © Springer Science+Business Media, LLC, part of Springer Nature 2019

1

2

Alan M. Mathiowetz

Table 1 Representative cyclic peptides oral %F > 15%, with EPSA and MDCK-LE

Compound Structure

Rat Oral MDCK-LE Papp %F 10 6 cm/s

EPSA

References

1

28.1

4.9

78 [29]

[5]

2

21

9.6

3

20.7

4.2

57

[8]

4

17

5.4

61

[11]

5

18

6.2

77

[10]

[7]

Design Principles for Intestinal Permeability

3

bioavailability, divided into those most important for intestinal absorption and those most important for a compound to escape gut wall or hepatic metabolism. The properties important for absorption, namely, polarity, hydrogen-bond counts, and molecular weight, are especially problematic for peptides, which are constructed with multiple polar peptide bonds. These properties are especially detrimental for intestinal permeability, and this is likely the primary hurdle for cyclic peptides to achieve oral absorption. The intestinal permeability of cyclic peptides is the focus of this chapter. Oral absorption also depends on maintenance of good solubility and proteolytic stability, but these factors are not systematically addressed here. There has been significant progress in recent years in understanding the key structural and physicochemical properties necessary for cyclic peptides to achieve significant intestinal permeability, and design strategies have been developed to obtain these properties. Nevertheless, the field remains challenging, in part because the complex three-dimensional structures of cyclic peptides underlie the attainment of necessary properties and in part because other critical properties such as solubility, clearance, and potency may have very different requirements. The developing design principles for cyclic peptide intestinal permeability will be covered in this chapter, in addition to a discussion of the ongoing challenges.

2

High Throughput Assays of Permeability The direct measurement of intestinal permeability in vivo is complex and labor-intensive, so high-throughput methods have been developed as surrogate measurements. Most recent publications that measure the permeability of new cyclic peptides use one or more of these higher-throughput methods. Perhaps the two most widely used methods are the parallel artificial membrane permeability assay (PAMPA) [13] and the human colon carcinoma cell line (Caco-2) [14]. PAMPA is an artificial membrane system, while Caco-2 is a closer representation of the physiological system, having been derived from a human colon cell line. On the other hand, Caco-2 contains transporters which can cause measurements to be dominated by active uptake or efflux processes. An interesting study of five enantiomeric pairs of cyclic peptides showed significant asymmetry in Caco-2 for most of the pairs [15], indicating that permeability in Caco-2 may be carrier-mediated. In contrast, the PAMPA results were not asymmetric, consistent with its lack of transporters. While carrier-mediated transport may be representative of true physiological processes, the desire to measure passive transport in the intestine has led to the development of the MadinDarby canine kidney low-efflux (MDCK-LE) assay [16], also sometimes known as the Ralph Russ canine kidney (RRCK) assay. This

4

Alan M. Mathiowetz

assay is becoming commonly used in studies of the passive permeability of cyclic peptides. Recent publications utilizing PAMPA assays to analyze the permeability of large sets of cyclic peptides include a study of 39 cyclic peptide natural products [17], a study of 62 designed cyclic peptides [18], and even larger-scale studies of cyclic peptide or peptomer libraries synthesized by split-pool techniques [19, 20]. The study of 62 cyclic peptides [18] also utilized the Caco-2 assay and includes an analysis of the correlation between PAMPA and Caco-2 for this set of compounds. Other recent studies utilizing Caco-2 include an analysis of the impact of multiple N-methylations in 54 cyclic hexapeptides [21], a study of the role of cis peptide bonds [22], and an analysis of several design approaches for improving cell permeability [23]. The MDCK-LE (or RRCK) assay has been utilized in several targeted studies, including the study of cyclic hexapeptide N-methylation [5], side chain substitutions in cyclic hexapeptides [6], and the discovery of an orally bioavailable CXCR7 peptide-peptoid hybrid [10]. Table 1 shows several recent peptides with rat oral bioavailability greater than 15% and reported values for MDCK-LE. For these, the MDCK-LE Papp is typically greater than 4  10 6 cm/s.

3

Fundamental Physical Properties The design of cyclic peptides with good intestinal permeability would be greatly facilitated by simple property guidelines that could easily be assessed during the design process. Such guidelines for medicinal chemistry synthesis were popularized nearly 20 years ago by Lipinski’s “Rule-of-5” [3], which highlighted molecular weight, logP, and hydrogen-bond donors and acceptors as key properties for oral bioavailability. The succeeding work of Veber et al. [24] highlighted the importance of low polar surface area and rotatable bonds for oral bioavailability. More recently, Varma et al. [12] were able to distinguish the factors important for absorption from those important for escaping gut wall and hepatic metabolism. Most of the historical guidelines, notably those regarding molecular weight, polar surface area, donors and acceptors, and rotatable bonds pertain primarily to absorption, with the exception being high logP. High logP is primarily detrimental to bioavailability because it leads to an increase in gut wall and hepatic metabolism. The relationship between logP and absorption or permeability is more complex. High logP is typically detrimental to solubility, an important factor in absorption, yet the work of Varma et al. [12] did not find a significant impact on overall absorption. The significant oral bioavailability of cyclosporine A [2], which lies far outside the “Rule-of-5” property guidelines, raises the possibility that there may be an alternative property space that

Design Principles for Intestinal Permeability

5

defines guidelines for permeability and oral bioavailability of cyclic peptides. A recent analysis of drugs and clinical candidates with molecular weight > 500 identified a “possible to be oral” property space [25] with greatly extended property guidelines such as molec˚ 2. ular weight < 1000 and topological polar surface area < 250 A Consistent with this, a recent analysis of cyclic peptide libraries synthesized by split-pool methods showed a dramatic decrease in permeability for compounds with molecular weight > 1000 [20]. Interestingly, while the “possible to be oral” property space increased most property thresholds significantly relative to the Rule-of-5, the guideline for the number of hydrogen-bond donors (HBD  6) increased very little [25], highlighting the paramount importance of keeping the number of hydrogen-bond donors low in the quest for permeability and oral bioavailability. Another possibility that has been raised [26] is that threedimensional properties will be the fundamental measuring stick Beyond the Rule-of-5, and a thorough characterization of the conformations of molecules, such as cyclic peptides, can identify three-dimensional property guidelines for permeability: 3D polar surface area < 100 A˚2 and radius of gyration < 7 A˚. It is also possible that a more rigorous analysis is required and sophisticated computational methods have been developed to predict the permeability of cyclic peptides, such as the physics-based method of Leung et al. [27]. While computational models like this are promising, it remains challenging to model accurately the relevant three-dimensional conformations of complex molecules such as cyclic peptides. Experimental methods for measuring effective polarity are very important, and perhaps the most valuable to the field of cyclic peptide design has been EPSA. EPSA is an experimental chromatographic method with reported polarity values analogous to polar surface area [28]. The original guidance was that passive permeability for cyclic peptides is far more likely for EPSA < 90 [29]. Note that the orally bioavailable cyclic peptides in Table 1 are well below this guideline. This guideline was borne out more recently for a large set of cyclic peptide design for CXCR7 [10] (see Fig. 1 left). It should be noted that an EPSA value below 90 does not guarantee good permeability, but it does significantly improve a compound’s prospects. It is not entirely known why low polarity (EPSA < 90) does not guarantee good permeability. One apparent factor is logP values being outside an optimal range [10] (see Fig. 1 right). If logP values are too high, this may confound the permeability measurements as compounds remain bound in the initial well or have especially poor solubility. It might also represent a genuine decrease in physiological permeability as compounds remain stuck in the hydrophobic interior of the membrane. Values of logP that are too low are also detrimental to permeability as these represent relatively polar

6

Alan M. Mathiowetz 9

ClogP < 7.5 7.5-10 > 10

8

Papp [10–6 cm/s]

7 6

9 8 7 6

5

5

4

4

3

3

2

2

1

1

0

0 70

80

90 100 110 120 130 140 150 160 EPSA

3

4

5

6

7 8 ClogP

9

10

11

12

Fig. 1 Distribution of MDCK-LE permeability versus EPSA (left) and ClogP (right) for a collection of cyclic peptides and peptide-peptoid hybrids. From reference [10]. https://pubs.acs.org/doi/abs/10.1021/acs. jmedchem.7b01028. Included with permission from ACS. Further permissions related to the material excerpted should be directed to the ACS

compounds. For the CXCR7 study, some compounds with low EPSA also had ClogP below the optimal range and had poor permeability [10], perhaps indicating that EPSA underestimated the true polarity of these compounds. It may be a general principle that most cyclic peptide series will have optimal logP ranges and both high and low logP values will be detrimental to permeability [19]. One additional property of considerable interest is molecular flexibility. While the number of rotatable bonds has often been highlighted as an important property to minimize for the sake of permeability and oral bioavailability [12, 24, 25], the applicability to cyclic peptides is complex. This complexity is primarily due to the possibility that increased flexibility may allow cyclic peptides to adopt alternating conformations, some of which have minimal polarity and are favorable for permeation, while others dynamically expose polarity in a way beneficial to aqueous solubility. In one recent study [9], the rigidity of cyclic hexapeptides was found to be beneficial to permeability and oral bioavailability, whereas other studies [30, 31] have found that flexibility is beneficial, allowing cyclic peptide to adopt both permeable and soluble conformations. It is likely that the value of flexibility and rigidity will be highly case dependent, depending on the properties of the specific conformations that can be adopted. A design campaign to identify intestinally permeable and orally bioavailable cyclic peptides would probably benefit by an exploration of both flexible and rigid analogs. Overall, these property analyses highlight the value of minimizing the effective surface polarity of molecules, allowing them to partition into membranes, and minimizing their effective size, to reduce barriers to diffusion across the membrane. In addition, it is

Design Principles for Intestinal Permeability

7

valuable to explore a range of logPs and monitor its impact as the ideal range may be case dependent and it is likely to affect both experimental accuracy and complementary physiological properties such as clearance. Likewise, the impact of flexibility should be explored, as it is likely to be case dependent as to its impact on permeability and other important properties such as solubility and oral bioavailability.

4

Design Strategies for the Ring Many design strategies have been explored for cyclic peptides, with the aim of improving their intestinal permeability by achieving favorable properties, most notably reduced surface polarity, especially solvent-exposed hydrogen-bond donors. In the context of drug discovery, these properties need to be balanced against other competing properties such as potency and clearance [10]. These design strategies can be divided into those primarily involving changes to the macrocyclic ring (see Table 2) and those involving changes to the side chains (Table 3). It is worth noting that cyclization itself provides some intrinsic advantages, as compared to acyclic analogs. Cyclization helps with proteolytic stability [37], which is essential for oral absorption and bioavailability. Cyclization also helps with permeability itself. This has been seen most clearly in direct comparisons of linear and cyclic analogs [11, 33]. The decapeptide comparison [11] is instructive in that the cyclic (Compound 4 in Table 1) and acyclic analogs have very similar EPSA (61 versus 66) and computed polarity but very different MDCK-LE permeability (5.4  10 6 cm/s versus 0.6  10 6 cm/s). These data may indicate that the optimal low-polarity

Table 2 Cyclic peptide backbone features improving permeability, with literature examples Design approach

References

N-methylation (N-alkylation)

[4, 5, 21, 32]

Cis peptide bonds

[22]

Intramolecular hydrogen bonding

[5, 7, 9–11, 17, 33, 34]

Selection of optimal stereochemistry of the ring

[33]

Prolines

[5, 9]

Thiazoles/oxazoles

[7, 17]

Depsipeptide linkages

[17]

Statine hybrids

[8]

Cyclopropane tethers

[35]

8

Alan M. Mathiowetz

Table 3 Cyclic peptide side chain features improving permeability, with literature examples Design approach

Examples

Nonpolar side chains

[6]

Weakly polar side chains (Thr, Tyr)

[6]

Alkylation of polar side chains

[23]

Replacement of arginines with heterocyclic amino acids

[10]

Replacement of tryptophans

[6, 10]

Branched side chains

[7, 10]

Peptoids

[10, 36]

conformation, indicated by EPSA, is much more easily adopted by the cyclic system than the acyclic. The highly flexible acyclic molecule may be expected to adopt a much greater number of less compact and more polar conformations, providing an entropic barrier to adopting the permeating conformation. Among the most significant contributors to high polarity and poor permeability in peptides are the NH groups of peptide bonds, in particular those exposed to solvent. These groups interact favorably with water and cause the peptides to favor the aqueous solvated state over a membrane permeating state. Methylation of these peptide nitrogens was one of the earliest and best studied design approaches for improving permeability of cyclic peptides [4, 5, 21, 32], inspired in part by the presence of seven N-methyls in the undecapeptide cyclosporine A. The choice of which amides to methylate is complex, since solvent exposure depends on the three-dimensional structure of the peptide and a peptide might adopt multiple conformations [33, 34]. Most notably, cyclosporine A itself can adopt dramatically different conformations depending on its environment [38]. Further complications arise from the potential for N-methyls to change the backbone conformation relative to the desmethylated analogs, and different patterns of Nmethyls can produce different conformations [5]. An additional factor to consider is the possibility that a given NH might be involved in a key receptor interaction and methylation may be detrimental to receptor affinity (e.g., in CXR7 [10]). Several of the earliest studies involved systematic N-methylation [4, 21, 32] of key peptides, which were successful at finding particular N-methyls and/or N-methylation patterns that improved permeability. A more targeted approach was described in [5], describing both an experimental on-resin N-methylation and a computational virtual library analysis of the possible stereochemistry options and N-methylation patterns. This led to the

Design Principles for Intestinal Permeability

9

identification of Compound 1 (Table 1), which had a rat oral bioavailability of 28%. This work highlighted the interplay of conformation and N-methylation, as N-methylation was selective for several scaffolds, and the best studied selectively methylated peptides were more permeable than either their permethylated or unmethylated analogs. An unusual structural feature of N-methylated peptide bonds is that they are able to adopt cis conformations far more readily than standard unmethylated peptide bonds. A recent study of 13 Nmethylated cyclic pentaalanine peptides [22] found that cis peptide bonds were found in a majority of the more permeable peptides. One interesting possibility raised is that the cis peptide unit serves as a recognition element for carrier-mediated Caco-2 transport, as previously seen in a study of enantiomeric pairs [15]. Another important design principle for permeable cyclic peptide backbones is the formation of intramolecular hydrogen bonds (IMHBs). This has been proposed as a favorable feature for peptide permeability [33, 34] and even a general design strategy for permeability and absorption in Beyond Rule-of-5 chemical space [38], based in part on the presence of multiple IMHBs seen for cyclosporine A in nonpolar solvents and, presumably, its membranepermeating conformation. An analysis of nine cyclic peptide diastereomers designed to maximize IMHB counts [33] found that the most permeable one formed two strong transannular IMHBs. The NHs involved in these stable, transannular IMHBs were effectively shielded from water, enabling this peptide to have good passive permeability even though none of the peptide amides were N-methylated. The presence and importance of IMHB has further been noted in multiple publications including [5, 7, 9–11, 17, 34]. The stereochemistry of the macrocyclic ring is critical as different diastereomers will have different preferred backbone conformations, intramolecular hydrogen-bonding patterns, and permeabilities [33]. Likewise, because of their different conformational preferences, different diastereomers may have different patterns of favorable N-methylation [5]. Fortunately, computational models have been found to be effective at evaluating the different possible diastereomers, N-methylation patterns, and IMHB patterns [5, 33, 34] and can help organize, predict, and interpret these complex interrelationships. Proline residues can be especially valuable and are frequently seen in cyclic peptides with good oral bioavailability [5–8, 10]. Examples include Compounds 1, 2, 3, and 5 in Table 1. Prolines can be favorable for several reasons: they eliminate one NH, they can force a turn conformation and stabilize intramolecular hydrogen bonding, and they increase the rigidity of the peptide. A recent study [9] compared Compound 1 to analogs made by

10

Alan M. Mathiowetz

reducing or increasing the number of prolines. The analog with no prolines (D-Pro of 1 replaced by D-Leu) did not form IMHBs in solutions, while the analog with two prolines (D-Leu of 1 replaced by D-Pro) maintained intramolecular hydrogen bonding and even showed a small improvement in rat oral bioavailability [9]. In addition to methylation of the amide nitrogen, or having the amide NH involved in intramolecular hydrogen bonding, the peptide bond can be replaced by heterocycles such as thiazole or oxazoline, thereby eliminating one of the hydrogen-bond donors. Thiazoles are occasionally seen in natural product peptides such as sanguinamide A, which has a rat oral bioavailability of 7% [7]. Compound 2 in Table 1 is a follow-up compound to sanguinamide A, retaining the thiazole while including additional modifications that improve permeability and oral bioavailability further. Another natural product example with good permeability, as measured by PAMPA, is patellamide C [17], which contains two thiazoles and two oxazolines in the macrocyclic ring. A related strategy for reducing hydrogen-bond donors is to employ depsipeptides. Depsipeptides, in which the peptide amide bond is replaced by an ester, can be beneficial to permeability if a solvent-exposed H-bond donor (NH) is replaced by the H-bond acceptor (O). In a recent study of 39 cyclic peptide natural products, depsipeptides were frequently found among the more permeable peptides, including enniatin B and guongomide A [17]. Several other variations of the peptide ring have been successfully employed in recent years. Statines were incorporated into cyclic peptides in [8], and the compounds showed improved water solubility and microsomal stability compared to analogs without the statines. Furthermore, there was a high degree of intramolecular hydrogen bonding and especially low exposed polarity as measured by EPSA. An example is Compound 3 in Table 1, which has good rat oral bioavailability (20.7%) and a very low EPSA of 57 despite the presence of two hydroxyl groups. More recently, cyclopropane tethers were incorporated into cyclic peptides, with very promising results [35]. The cyclopropane tether greatly restricts conformations of the ring and differs greatly depending on the cis or trans stereochemistry of the cyclopropyl as well as the stereochemistry of the adjacent center, thereby giving chemists the ability to generate a variety of shapes. The series of compounds made showed a wide variety of permeability and EPSA values. The best of these contained a cis-NfCf [35] unit, which reduced the EPSA by 15 units compared to the parent cyclic peptide. A further step of N-methylation enabled the best peptide to achieve excellent permeability of 12.7  10 6 cm/s in a pig kidney epithelial cell permeability assay.

Design Principles for Intestinal Permeability

5

11

Design Strategies for Side Chains Although design strategies for cyclic peptide permeability have predominantly focused on the macrocycle ring itself, the choice of side chains can have a dramatic effect [6]. The two overarching strategies are to utilize nonpolar or weakly polar side chains and to use branched side chains where possible to shield some of the polarity of the backbone. Table 3 lists different strategies that have been developed. Permeable peptides often contain small nonpolar side chains such as alanine and leucine in addition to prolines as previously mentioned (see Compounds 1, 3, and 4 in Table 1). In the context of drug discovery, however, a wider variety of side chains may need to be utilized to meet other important goals such as binding affinity to the target (see Compound 5 and [10]). A study of 16 analogs of Compound 1, differing at a single amino acid side chain and including both natural and nonnatural amino acids, showed a great variety of permeability and clearance values [6]. For this set of analogs, microsomal clearance correlated well with logD as did RRCK permeability, though to opposite effect. Compounds with higher logD generally had better permeability but higher clearance. The two exceptions were tryptophan and cyclohexyl alanine, which had poor permeability despite high logD. These exceptions may be due to the poor recovery in the RRCK assay [6] or may indicate that these amino acids do reduce permeability, as has been seen for tryptophan elsewhere [10]. This study indicated that even a single ionizable group (Asp or Lys amino acids) dramatically reduced permeability and oral bioavailability, while moderately small hydrophobic side chains (Leu, Ile, Phe) were especially good for permeability. The weakly polar threonine side chain was notable in that it enabled a good balance of permeability and clearance and resulted in a compound with very good rat bioavailability (23.8%) despite the addition of a second polar group. It is relatively rare to see cyclic peptides containing multiple polar side chains have good oral bioavailability. The compounds in Table 1, for example, each have zero or one polar side chain. Another approach for achieving good permeability while retaining multiple polar side chains is to methylate these side chains [23]. Methylation of polar side chains such as serine, tyrosine, and lysine may be favorable because it eliminates H-bond donors but does not entirely eliminate polarity and does not dramatically add to the size of the compounds. This approach may be beneficial for drug discovery, if polarity needs to be retained in the side chains for the sake of binding affinity to the target. It is worth exploring other options as well. In [10], an ionizable amino acid (Arg) was replaced by moderately polar heterocyclic amino acids such as pyridylalanine and thiazolylalanine, dramatically improving EPSA while retaining

12

Alan M. Mathiowetz

good binding affinity. This study also showed the potential value of replacing tryptophan side chains. Replacement of tryptophan by a bulky, but nonaromatic, t-butyl alanine was especially beneficial to permeability and retained good target potency. In general, branched side chains such as t-butyl alanine and t-butyl glycine are often favorable for permeability since their steric bulk is able to shield backbone NH groups from solvent [7, 10]. Compounds 2 and 5 are examples. This effect may be extended to threonine as a beneficial replacement for serine, as seen in [6]. Despite its close proximity to the backbone, branching at the beta carbon, as in t-butyl glycine, might not improve permeability more than branching at the gamma carbon, as in leucine, isoleucine, or t-butyl alanine. A study of 17 sanguinamide A analogs [30] found that a leucine replacement for t-butyl glycine increased the flexibility of the backbone and enabled significantly improved solubility while maintaining or perhaps improving permeability. Another design option to be considered is the use of peptoids as a replacement for amino acids at one or more positions. A peptoid has its side chain on the nitrogen position of a glycine rather than at the alpha carbon, allowing for a vast array of primary amines to be utilized to diversify the structure. Peptoids eliminate an amide NH and increase backbone flexibility, both of which are potentially favorable. A recent systematic study [36] found that peptoids have very similar permeability to the equivalent peptides, but the greater diversity of peptoid side chains makes it possible to find analogs with improved permeability. This was utilized to great effect in [10], where the great diversity of peptoid side chains was used to identify compounds with dramatically improved binding affinity and led, eventually, to a potent and orally bioavailable peptide-peptoid hybrid (see Compound 5).

6

Summary Many design strategies are available for improving intestinal permeability of cyclic peptides. While the complexity of peptide conformation flexibility makes it difficult to predict with certainty which tactics will work, the overarching strategy of reducing polarity, especially H-bond donors, through backbone and side chain modifications is often effective. The reduction, or even elimination, of solvent-exposed H-bond donors in the macrocyclic ring is typically driven by intramolecular hydrogen bonding and N-alkylation of the peptide bonds. The use of nonpolar or moderately polar side chains is also usually beneficial for obtaining optimal permeability as is the use of branched side chains that can shield some of the polarity of the peptide backbone. New insights and new design strategies, such as use of peptoids [36] or cyclopropyl tethers

Design Principles for Intestinal Permeability

13

[35], are showing considerable promise as well. A few design principles such as incorporation of flexibility or selection of an optimal logP range have more subtle trade-offs and should be explored during the process of optimizing compounds for both intestinal permeability and other important properties. References 1. Terrett N (2013) Drugs in middle space. Med Chem Commun 4(3):474–475. https://doi. org/10.1039/C2MD90062A 2. Legg B, Gupta SK, Rowland M, Johnson RWG, Solomon LR (1988) Cyclosporin: pharmacokinetics and detailed studies of plasma and erythrocyte binding during intravenous and oral administration. Eur J Clin Pharmacol 34(5):451–460. https://doi.org/10.1007/ BF01046701 3. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ (1997) Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 23 (1):3–25. https://doi.org/10.1016/S0169409X(96)00423-1 4. Biron E, Chatterjee J, Ovadia O, Langenegger D, Brueggen J, Hoyer D, Schmid Herbert A, Jelinek R, Gilon C, Hoffman A, Kessler H (2008) Improving Oral bioavailability of peptides by multiple N-methylation: Somatostatin analogues. Angew Chem Int Ed 47(14):2595–2599. https://doi.org/10. 1002/anie.200705797 5. White TR, Renzelman CM, Rand AC, Rezai T, McEwen CM, Gelev VM, Turner RA, Linington RG, Leung SSF, Kalgutkar AS, Bauman JN, Zhang Y, Liras S, Price DA, Mathiowetz AM, Jacobson MP, Lokey RS (2011) On-resin N-methylation of cyclic peptides for discovery of orally bioavailable scaffolds. Nat Chem Biol 7:810. https://doi.org/10.1038/nchembio. 664 6. Rand AC, Leung SSF, Eng H, Rotter CJ, Sharma R, Kalgutkar AS, Zhang Y, Varma MV, Farley KA, Khunte B, Limberakis C, Price DA, Liras S, Mathiowetz AM, Jacobson MP, Lokey RS (2012) Optimizing PK properties of cyclic peptides: the effect of side chain substitutions on permeability and clearance. Med Chem Commun 3(10):1282–1289. https://doi.org/10.1039/C2MD20203D 7. Nielsen Daniel S, Hoang Huy N, Lohman RJ, Hill Timothy A, Lucke Andrew J, Craik David J, Edmonds David J, Griffith David A, Rotter Charles J, Ruggeri Roger B, Price David A, Liras S, Fairlie David P (2014)

Improving on nature: making a cyclic heptapeptide orally bioavailable. Angew Chem Int Ed 53(45):12059–12063. https://doi.org/ 10.1002/anie.201405364 8. Bockus AT, Lexa KW, Pye CR, Kalgutkar AS, Gardner JW, Hund KCR, Hewitt WM, Schwochert JA, Glassey E, Price DA, Mathiowetz AM, Liras S, Jacobson MP, Lokey RS (2015) Probing the physicochemical boundaries of cell permeability and oral bioavailability in lipophilic macrocycles inspired by natural products. J Med Chem 58(11): 4581–4589. https://doi.org/10.1021/acs. jmedchem.5b00128 9. Nielsen Daniel S, Lohman RJ, Hoang Huy N, Hill Timothy A, Jones A, Lucke Andrew J, Fairlie David P (2015) Flexibility versus rigidity for orally bioavailable cyclic hexapeptides. Chembiochem 16(16):2289–2293. https:// doi.org/10.1002/cbic.201500441 10. Boehm M, Beaumont K, Jones R, Kalgutkar AS, Zhang L, Atkinson K, Bai G, Brown JA, Eng H, Goetz GH, Holder BR, Khunte B, Lazzaro S, Limberakis C, Ryu S, Shapiro MJ, Tylaska L, Yan J, Turner R, Leung SSF, Ramaseshan M, Price DA, Liras S, Jacobson MP, Earp DJ, Lokey RS, Mathiowetz AM, Menhaji-Klotz E (2017) Discovery of potent and orally bioavailable macrocyclic peptide–peptoid hybrid CXCR7 modulators. J Med Chem 60(23):9653–9663. https://doi.org/ 10.1021/acs.jmedchem.7b01028 11. Price DA, Eng H, Farley KA, Goetz GH, Huang Y, Jiao Z, Kalgutkar AS, Kablaoui NM, Khunte B, Liras S, Limberakis C, Mathiowetz AM, Ruggeri RB, Quan J-M, Yang Z (2017) Comparative pharmacokinetic profile of cyclosporine (CsA) with a decapeptide and a linear analogue. Org Biomol Chem 15 (12):2501–2506. https://doi.org/10.1039/ C7OB00096K 12. Varma MVS, Obach RS, Rotter C, Miller HR, Chang G, Steyn SJ, El-Kattan A, Troutman MD (2010) Physicochemical space for optimum oral bioavailability: contribution of human intestinal absorption and first-pass elimination. J Med Chem 53(3):1098–1108. https://doi.org/10.1021/jm901371v

14

Alan M. Mathiowetz

13. Kansy M, Senner F, Gubernator K (1998) Physicochemical high throughput screening: parallel artificial membrane permeation assay in the description of passive absorption processes. J Med Chem 41(7):1007–1010. https://doi.org/10.1021/jm970530e 14. Hidalgo IJ, Raub TJ, Borchardt RT (1989) Characterization of the human colon carcinoma cell line (Caco-2) as a model system for intestinal epithelial permeability. Gastroenterology 96(3):736–749 15. Marelli Udaya K, Bezenc¸on J, Puig E, Ernst B, Kessler H (2015) Enantiomeric cyclic peptides with different Caco-2 permeability suggest carrier-mediated transport. Chem Eur J 21 (22):8023–8027. https://doi.org/10.1002/ chem.201501270 16. Di L, Whitney-Pickett C, Umland John P, Zhang H, Zhang X, Gebhard David F, Lai Y, Federico James J, Davidson Ralph E, Smith R, Reyner Eric L, Lee C, Feng B, Rotter C, Varma Manthena V, Kempshall S, Fenner K, El-kattan Ayman F, Liston Theodore E, Troutman Matthew D (2011) Development of a new permeability assay using low-efflux MDCKII cells. J Pharm Sci 100(11):4974–4985. https://doi. org/10.1002/jps.22674 17. Ahlbach CL, Lexa KW, Bockus AT, Chen V, Crews P, Jacobson MP, Lokey RS (2015) Beyond cyclosporine a: conformationdependent passive membrane permeabilities of cyclic peptide natural products. Future Med Chem 7(16):2121–2130. https://doi.org/10. 4155/fmc.15.78 18. Wang CK, Northfield SE, Swedberg JE, Colless B, Chaousis S, Price DA, Liras S, Craik DJ (2015) Exploring experimental and computational markers of cyclic peptides: charting islands of permeability. Eur J Med Chem 97:202–213. https://doi.org/10. 1016/j.ejmech.2015.04.049 19. Furukawa A, Townsend CE, Schwochert J, Pye CR, Bednarek MA, Lokey RS (2016) Passive membrane permeability in cyclic Peptomer scaffolds is robust to extensive variation in side chain functionality and backbone geometry. J Med Chem 59(20):9503–9512. https:// doi.org/10.1021/acs.jmedchem.6b01246 20. Pye CR, Hewitt WM, Schwochert J, Haddad TD, Townsend CE, Etienne L, Lao Y, Limberakis C, Furukawa A, Mathiowetz AM, Price DA, Liras S, Lokey RS (2017) Nonclassical size dependence of permeation defines bounds for passive adsorption of large drug molecules. J Med Chem 60(5):1665–1672. https://doi.org/10.1021/acs.jmedchem. 6b01483

21. Ovadia O, Greenberg S, Chatterjee J, Laufer B, Opperer F, Kessler H, Gilon C, Hoffman A (2011) The effect of multiple N-methylation on intestinal permeability of cyclic Hexapeptides. Mol Pharm 8(2):479–487. https://doi. org/10.1021/mp1003306 22. Marelli Udaya K, Ovadia O, Frank Andreas O, Chatterjee J, Gilon C, Hoffman A, Kessler H (2015) Cis-peptide bonds: a key for intestinal permeability of peptides? Chem Eur J 21 (43):15148–15152. https://doi.org/10. 1002/chem.201501600 23. Buckton LK, McAlpine SR (2018) Improving the cell permeability of polar cyclic peptides by replacing residues with alkylated amino acids, Asparagines, and D-amino acids. Org Lett 20 (3):506–509. https://doi.org/10.1021/acs. orglett.7b03363 24. Veber DF, Johnson SR, Cheng H-Y, Smith BR, Ward KW, Kopple KD (2002) Molecular properties that influence the Oral bioavailability of drug candidates. J Med Chem 45 (12):2615–2623. https://doi.org/10.1021/ jm020017n 25. Doak Bradley C, Over B, Giordanetto F, Kihlberg J (2014) Oral Druggable space beyond the rule of 5: insights from drugs and clinical candidates. Chem Biol 21(9):1115–1142. https://doi.org/10.1016/j.chembiol.2014. 08.013 26. Guimara˜es CRW, Mathiowetz AM, Shalaeva M, Goetz G, Liras S (2012) Use of 3D properties to characterize beyond rule-of-5 property space for passive permeation. J Chem Inf Model 52(4):882–890. https://doi.org/ 10.1021/ci300010y 27. Leung SSF, Mijalkovic J, Borrelli K, Jacobson MP (2012) Testing physical models of passive membrane permeation. J Chem Inf Model 52 (6):1621–1636. https://doi.org/10.1021/ ci200583t 28. Goetz GH, Farrell W, Shalaeva M, Sciabola S, Anderson D, Yan J, Philippe L, Shapiro MJ (2014) High throughput method for the indirect detection of Intramolecular hydrogen bonding. J Med Chem 57(7):2920–2929. https://doi.org/10.1021/jm401859b 29. Goetz GH, Philippe L, Shapiro MJ (2014) EPSA: a novel supercritical fluid chromatography technique enabling the Design of Permeable Cyclic Peptides. ACS Med Chem Lett 5 (10):1167–1172. https://doi.org/10.1021/ ml500239m 30. Bockus AT, Schwochert JA, Pye CR, Townsend CE, Sok V, Bednarek MA, Lokey RS (2015) Going out on a limb: delineating the

Design Principles for Intestinal Permeability effects of β-branching, N-methylation, and side chain size on the passive permeability, solubility, and flexibility of Sanguinamide a analogues. J Med Chem 58(18):7409–7418. https://doi. org/10.1021/acs.jmedchem.5b00919 31. Rossi Sebastiano M, Doak BC, Backlund M, Poongavanam V, Over B, Ermondi G, Caron G, Matsson P, Kihlberg J (2018) Impact of dynamically exposed polarity on permeability and solubility of chameleonic drugs beyond the rule of 5. J Med Chem 61(9):4189–4202. https://doi. org/10.1021/acs.jmedchem.8b00347 32. Chatterjee J, Gilon C, Hoffman A, Kessler H (2008) N-methylation of peptides: a new perspective in medicinal chemistry. Acc Chem Res 41(10):1331–1342. https://doi.org/10. 1021/ar8000603 33. Rezai T, Yu B, Millhauser GL, Jacobson MP, Lokey RS (2006) Testing the conformational hypothesis of passive membrane permeability using synthetic cyclic peptide Diastereomers. J Am Chem Soc 128(8):2510–2511. https:// doi.org/10.1021/ja0563455 34. Rezai T, Bock JE, Zhou MV, Kalyanaraman C, Lokey RS, Jacobson MP (2006) Conformational flexibility, internal hydrogen bonding, and passive membrane permeability: successful in Silico prediction of the relative Permeabilities of cyclic peptides. J Am Chem Soc 128 (43):14073–14080. https://doi.org/10. 1021/ja063076p

15

35. Matsui K, Kido Y, Watari R, Kashima Y, Yoshida Y, Shuto S (2016) Highly Conformationally restricted Cyclopropane tethers with three-dimensional structural diversity drastically enhance the cell permeability of cyclic peptides. Chem Eur J 23 (13):3034–3041. https://doi.org/10.1002/ chem.201604946 36. Schwochert J, Turner R, Thang M, Berkeley RF, Ponkey AR, Rodriguez KM, Leung SSF, Khunte B, Goetz G, Limberakis C, Kalgutkar AS, Eng H, Shapiro MJ, Mathiowetz AM, Price DA, Liras S, Jacobson MP, Lokey RS (2015) Peptide to Peptoid substitutions increase cell permeability in cyclic Hexapeptides. Org Lett 17(12): 2928–2931. https://doi.org/10.1021/acs. orglett.5b01162 37. March DR, Abbenante G, Bergman DA, Brinkworth RI, Wickramasinghe W, Begun J, Martin JL, Fairlie DP (1996) Substrate-based cyclic Peptidomimetics of Phe-Ile-Val that inhibit HIV-1 protease using a novel enzyme-binding mode. J Am Chem Soc 118(14):3375–3379. https://doi.org/10.1021/ja953790z 38. Alex A, Millan DS, Perez M, Wakenhut F, Whitlock GA (2011) Intramolecular hydrogen bonding to improve membrane permeability and absorption in beyond rule of five chemical space. Med Chem Commun 2(7):669–674. https://doi.org/10.1039/C1MD00093D

Chapter 2 Strategies to Enhance Metabolic Stabilities Bhavesh Khatri, Venkateswara Rao Nuthakki, and Jayanta Chatterjee Abstract Macrocyclic peptides are a unique class of molecules that display a relatively constrained peptidic backbone as compared to their linear counterparts leading to the defined 3-D orientation of the constituent amino acids (pharmacophore). Although they are attractive candidates for lead discovery owing to the unique conformational features, their peptidic backbone is susceptible to proteolytic cleavage in various biological fluids that compromise their efficacy. In this chapter we review the various classical and contemporary chemical and biological approaches that have been utilized to combat the metabolic instability of macrocyclic peptides. We note that any chemical modification that helps in providing either local or global conformational rigidity to these macrocyclic peptides aids in improving their metabolic stability typically by slowing the cleavage kinetics by the proteases. Key words Cyclic peptides, Proteolytic stability, Peptide macrocycles, Oral availability, Cyclotides, Peptide stapling, Peptide hormones, Bicyclic peptides, Epitope grafting

1

Introduction Peptide therapeutics have gained significant attention in drug discovery and development over the past few years with several designed peptide agonists and antagonists with drug-like properties either in the development stage or already practiced in the clinic [1]. Peptides are highly potent in nature, and they have been the source of templates for lead design and inspiration for medicinal chemists to optimize the drug-like properties [2]. Traditionally, peptide-based drug discovery fundamentally relied on peptides extracted from natural sources such as peptide antibiotics, peptidic hormones, peptides extracted from snake venoms, etc.; however, with the advent of high-throughput techniques such as phage display, mRNA display, SICLOPPS and one-bead one-peptide and genome-encoded libraries, the repertoire of bioactive peptides has enormously increased [3]. Additionally, in the recent years, de novo-designed peptides are also aiding in filling up the bioactive space covered by peptides [4]. Like antibodies, peptides are highly specific towards their target protein of interest. Nevertheless,

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_2, © Springer Science+Business Media, LLC, part of Springer Nature 2019

17

18

Bhavesh Khatri et al.

although protein therapeutics and antibodies are selective towards their target, their application is mostly confined to extracellular biomolecular interactions, and both these therapeutic modalities are prone to aggregation with a probability of eliciting immune response [5]. However, unlike small molecule and protein therapeutics, peptides are intermediate in size, highly selective and less toxic [6]. Although functional and structural genomics have identified several drug targets, only few of them are druggable by small molecules, owing to the shallow binding clefts and grooves present on the protein surface. Furthermore, several protein-protein interactions (PPIs) are intricately linked to the progression of diseases like cancer, metabolic disorders and host-pathogen interactions [7]. An important feature of these PPI interface is their flat, solvent-exposed surface epitope that often consists of secondary structural elements such as α-helix, β-sheets and loops, which makes it a daunting task to target these inaccessible sites by small molecules. However, peptides derived from the epitope of PPI can be used to target these undruggable sites with high affinity owing to the chemical diversity of different amino acid side chains, which are involved in a variety of non-covalent interactions with the target. Nevertheless, in several cases, linear peptides derived from the PPI interface often tend to lose their bioactive conformation due to their inherent conformational flexibility, making them less selective towards the target. Furthermore, despite the high specificity, potency and low toxicity of peptide leads, most of them suffer from serious drawbacks like poor metabolic stability, membrane impermeability and concomitant lack of oral bioavailability [6]. The proteolytic enzymes present in the gastrointestinal tract and serum coupled with the metabolic enzymes present in the liver pose a serious threat to the half-life of peptides administered in vivo. Therefore, various peptidomimetic strategies have been developed to combat the poor metabolic stability of peptide therapeutics. Peptide cyclization is one such approach, which, in addition to reducing the conformational flexibility of the peptide chain and allowing for the static 3-D presentation of the pharmacophore, provides protection against the exopeptidases [8]. Such constrained macrocyclic peptides cannot only be obtained by cyclization (via end-to-end/side chain-to-end and side chain-to-side chain), they can also be designed by grafting onto small non-antibody-based scaffolds [9] and cyclotides [10]. Depending on their structural context, several cyclization strategies have been utilized to constrain a peptide motif into its bioactive conformation that additionally aids in improving its pharmacokinetic and pharmacodynamics properties, e.g. peptides derived from an α-helix of a protein can be constrained by the stapling approach [11]. However, it is important to note that cyclization of peptides alone does not necessarily confer resistance against the proteolytic enzymes, since

Metabolic Stability Strategies

19

Fig. 1 Various cyclization strategies

endopeptidases can easily access their cleavage site even in a cyclic backbone. To circumvent this problem, several other chemical modifications are introduced into the cyclic peptide backbone, such as D-amino acid substitution, incorporation of non-proteinogenic amino acids, substitution of an α-amino acid with β- or γ-amino acid, modification of the amide bond by N-methylation and thionation, etc. Thus, in this chapter we will shed light on the various cyclization strategies (Fig. 1) and their further modifications that have been utilized to enhance the metabolic stability of peptides.

2

Classical and High-Throughput Methods Backbone cyclization of linear peptides has been shown to enhance the metabolic stability by removing both the N-terminal and C-terminal recognition sites for amino- and carboxypeptidases. Cyclization also introduces constrainment (reduced conformational flexibility) that leads to shielding of the catalytic sites of endoproteases [12]. Metabolic stability and protease resistance of linear and cyclic peptides are typically assessed by serum stability and enzymatic assays. Several linear peptides act as signal messengers in the human body such as peptide hormones and neurotransmitters. Peptide hormones including somatostatin, vasopressin, oxytocin, melanocortins and gonadotrophins are important class of biomolecules, and their deregulated signal transduction is shown

20

Bhavesh Khatri et al.

to be involved in several metabolic disorders. Thus to prevent the disease progression associated with these peptide hormones, the simplest strategy is to administer these hormones produced by the synthetic or recombinant methods either via the intravenous or the oral route [13]. However, because of their biocompatibility, high flexibility and extended conformation, they are prone to fast degradation, which significantly limits their therapeutic utility. It was shown in several reports that cyclization of the minimal pharmacophore in these peptide hormones and neurotransmitters can successfully yield an analog with better pharmacokinetic properties with retained bioactivity. Upon cyclization, the stability of these cyclic peptides substantially increased over their linear counterpart. For example, somatostatin is a peptide hormone that exhibits inhibitory effects on both endocrine (e.g. growth hormone, insulin, glucagon, etc.) and exocrine (gastric acid, intestinal fluid and pancreatic enzymes) secretion by binding to its receptors on the cell surface known as G-protein-coupled receptors (GPCR) [14]. Somatostatin (Fig. 2a) is a disulphide-bridged peptide hormone, and despite its potent activity and strong binding affinity (in the nanomolar range) against the somatostatin receptor subtypes (sst1–5), it exhibits a very short plasma half-life (500-fold stability of the peptide by constraining the α-helical conformation (without compromising its affinity). Likewise, utilizing the hydrogen bond surrogate approach where the main chain i ! i + 4 hydrogen bond is replaced by a hydrocarbon tether, Arora et al. could improve the proteolytic stability of the BH3 peptide by 30–60-folds [67]. In certain PPIs, where the binding interface is large requiring lengthy peptides to inhibit such irregular interactions, the efficacy of such peptides in vivo could be compromised by rapid proteolysis resulting in lower bioavailability. For example, in anti-HIV-1 therapy, the disruption of the six-helix bundle viral fusion apparatus formed by the gp41 envelope glycoprotein becomes a challenge.

Metabolic Stability Strategies

31

Enfuvirtide, a FDA-approved peptide-based fusion inhibitor, is a 36-amino acid α-helical peptide drug, but it suffers from rapid clearance and lack of oral bioavailability. Walensky et al. used hydrocarbon stapling approach to improve its pharmacokinetic properties. [68]. Since this peptide was longer compared to earlier bioactive stapled peptides, they checked the impact of double stapling. Surprisingly, the double stapled peptide showed 82-fold stability enhancement compared to the unmodified peptide and tenfold compared to the singly stapled peptide. Since double stapling enhances the hydrophobicity of peptide resulting in aggregation, they ruled out the possibility of aggregation-mediated stability enhancement by demonstrating the existence of monomeric species of these peptides under the assay conditions. They eventually demonstrate that either direct shielding of the proteolytic sites by the staple and/or the reduced conformational flexibility that retains the local α-helical conformation of the peptide is responsible for the stark prevention of these peptides against proteolytic digestion. Since hydrocarbon-based peptide stapling increases the hydrophobicity of the peptide and results in reduced aqueous solubility, Fairlie et al. utilized the side-chain to side-chain cyclization approach and designed the shortest single-turn α-helical structures in water [69]. In this approach, they synthesized cyclic pentapeptide sequence Ac-[XARAX]-NH2 in which lysine, ornithine, aspartic acid and glutamic acid were systematically substituted at i and i + 4 positions. These short peptides achieved 100% helicity confirmed through CD and NMR analysis. Further they evaluated the metabolic stability via trypsin digestion and observed that the acyclic analogs were completely hydrolysed within a few minutes, whereas the cyclic peptides remained intact over 2.5 h. This approach could be an attractive option where the binding pocket on the protein target is short with access to a single helical turn. Lactam-based stapling approach has also been used to enhance the stability of Glucagon-like peptide-1 [GLP-1] that regulates the blood glucose level and is very important for the treatment of type 2 diabetes [70]. GLP-1 is rapidly degraded by the enzymes dipeptidyl peptidase-IV (DPP-IV) and the neutral endopeptidase [NEP 24.11] resulting in its half-life of 2 min in circulation. Since GLP-1 has helical segments in solution, thus to induce conformational restraint in GLP-1, multiple lactam bridges were introduced through lysine and glutamic acid side chains at i and i + 4 positions. These modifications enhanced the helical content of the peptide, which interacted stronger with the GLP-1 receptor than the native GLP-1. Interestingly, Ahn et al. observed that despite nonoverlapping enzyme recognition site and the lactam bridge site, the peptide showed a half-life of >24 h against DPP-IV, and they could achieve an exceptional stability (half-life of >96 h) against the NEP 24.11 enzyme.

32

Bhavesh Khatri et al.

Fig. 9 Different cyclization approaches to increase the proteolytic stability of β-hairpins

β-Hairpin/β-sheet is yet another important protein secondary structure that plays a significant role in PPI due to their large solvent-exposed surface. β-Hairpin is basically an antiparallel pair of β-sheets connected by a loop, the conformation of which in isolation has been stabilized by several macrocyclization strategies including disulphide bridge formation and backbone cyclization (Fig. 9). However, since the stability of disulphide bond in a reducing environment (e.g. inside cells) or in the presence of reducing agents is always questionable, disulphide bond surrogates have been developed. Meldal et al. incorporated a triazole via click chemistry between an azide and an alkyne moiety to replace disulphide bonds in the naturally occurring antimicrobial peptide tachyplesin-1 (TP-1) peptide. TP-1 has a β-hairpin structure stabilized by two disulphide bonds [71]. In this approach both the disulphide bonds were replaced by triazoles, and these modified peptides displayed similar or even better MIC values than the wildtype TP-1. Waters et al. also used this approach to cyclize model β-hairpin peptides that were significantly stable against exo- and endo-serine proteases and trypsin [72]. It was also observed that under the experimental conditions, only the termini were cleaved leaving the clicked cycle intact. Recently, Kennedy et al. utilized this triazole-based β-hairpin stabilization approach to inhibit the dimerization and subsequent allosteric activation of epidermal growth factor receptor (EGFR) that is overexpressed in multiple carcinomas. The authors mimic the β-hairpin dimerization arm of the EGFR in the form of a stable triazole-linked β-hairpin peptide that prevents the receptor dimerization and eventually reduces the cell viability [73]. They demonstrate that the triazole-linked peptides were completely stable against trypsin and chymotrypsin digestion over 4 h, whereas the linear peptide showed a half-life of 1 h. This was also reflected in the mouse serum stability assay,

Metabolic Stability Strategies

33

where the linear peptide showed complete degradation within 2 h as opposed to a triazole-linked peptide that showed marginal degradation even after 16 h. They further show that the conformation of the triazole-modified β-hairpin was independent of the pH of the medium. One of the most prominent class of naturally occurring macrocyclic β-hairpin peptides is the cationic antimicrobial peptides that are stabilized by either one or two disulphide bridges and possess serious membranolytic activity. Robinson et al. utilized the constrained D-Pro-L-Pro dipeptide motif that forms a stable βII’-type turn to develop conformationally rigid macrocyclic scaffold for the design novel antibiotics [74]. They synthesized several macrocyclic peptides of the toxic but potent antimicrobial peptide protegrin-I (PG-I) utilizing the D-Pro-L-Pro-based scaffold. This led to the development of a highly potent and selective compound against Pseudomonas spp. with much improved drug-like properties. This compound, POL7080, currently in clinical trial phase III against hospital-acquired and ventilator-associated bacterial pneumonia displayed a significant enhancement in plasma half-lives.

4

Naturally Occurring Macrocyclic Scaffolds Nature harbours a large repertoire of cyclic peptide scaffolds from plant and animal origin that has recently gained significant attention. These peptides are small (10–40 amino acids) and highly constrained due to the presence of several disulphide bonds. In this section, we present two prominent classes of molecules that have helped in unlocking the potential of peptides for biotechnological application and drug discovery. Conotoxins, isolated from the venoms of marine cone sails, have gained much attention in drug development due to their higher affinity and selectivity for the physiological targets like ion channels and transporters [75]. Several conotoxins are currently undergoing preclinical or clinical evaluation, and recently one of the conotoxins ziconotide was approved by FDA in 2004 for the treatment of severe chronic pain [76]. Since conotoxins are small and can be easily synthesized in the laboratory, several groups have started working on chemical modifications to overcome the associated disadvantages like poor absorption and short biological halflives. These efforts were mainly focused on disulphide engineering, sequence mutation, truncation of the flexible terminal residues and cyclization through backbone. Craik et al. reported the cyclization strategy to enhance the stability of such peptides [77]. By careful analysis of MII, a 16-residue α-conotoxin peptide isolated from ˚ distance between Conus magus, they observed a significant 11 A N and C termini. It is interesting to note that this peptide has an α-helical conformation stabilized by two disulphide bridges,

34

Bhavesh Khatri et al.

Fig. 10 Reinforcing conotoxin conformational stability by linker-mediated cyclization

despite that a further cyclization was attempted (Fig. 10). Simple backbone cyclization of this peptide can cause strain in the molecule which could disturb its structure and potency. Thus, to circumvent this problem, they used several linkers of varying length consisting of 5–7 Gly and Ala residues, where the cyclic conotoxins with 6and 7-residue linker retained the bioactivity. These two cyclic peptides demonstrated significant protection against the EndoGluC over the parent MII with 15–20% stability improvement in human blood plasma. Utilizing beautifully designed deuterium exchange NMR experiments, they could clearly demonstrate that the enhanced proteolytic stability is a direct consequence of reduced backbone flexibility, possibly via the increased protection of the intramolecular hydrogen bond network. A similar approach was utilized subsequently by the same group to convert the α-helical double disulphide bridged conotoxin Vc1.1 into an orally available analgesic for the treatment of neuropathic pain in rat models [78]. Here they demonstrate that the additional linkermediated cyclization of the conotoxin prevents it from degradation not only in serum but also in gastric and intestinal fluids. Furthermore, they show that a significant proportion of the parent conotoxin undergoes disulphide shuffling to an inactive isomer in serum and intestinal fluid (but not in gastric fluid), possibly due to the near neutral pH. However, the cyclized variant did not show this phenomenon and retained the bioactive disulphides throughout the duration of the experiment. Cyclotides, isolated from plants, are exceptionally stable due to their conserved disulphide bonds [79]. One such cyclotide, kalata B1, is stable at near boiling temperature, resistant to denaturation by chaotropes and degradation by trypsin, endoproteinase Glu-C, thermolysin and human serum. The combination of macrocyclic backbone and disulphide bonds confer them exceptional stability with favourable pharmacokinetic properties. Additionally, cyclotides display remarkable tolerance to amino acid mutations in

Metabolic Stability Strategies

35

Fig. 11 Grafting of bioactive linear peptide sequence on cyclotides

their loops that makes them ideal candidate for grafting of bioactive peptide sequence that otherwise degrades or denatures within minutes (Fig. 11). It has also been observed that some cyclotides are capable of penetrating the cell membrane and enter cells, which could be harnessed to target intracellular proteins. Several naturally occurring scaffolds like SFTI-1, MCoTI-II, θ-defensin and kalata B1 have been used for the grafting purpose. Craik et al. successfully grafted linear proangiogenic peptides derived from the extracellular matrix proteins laminin and osteopontin and vascular endothelial growth factor (VEGF) onto the 34-residue Momordica cochinchinensis trypsin inhibitor-II (MCoTI-II) and 14-residue sunflower trypsin inhibitor-1 (SFTI-1) to derive potent angiogenic peptides [80]. While all the linear peptides showed significant degradation within 3 h in human serum, the grafted peptides stayed 100% intact even after 24 h displaying angiogenic activity at nanomolar concentration. Additionally, two of those peptides even showed stability against thrombin and matrix metalloproteinase-9 (MMP-9). Camarero et al. utilized this grafting approach to target the intracellular p53-Hdm2/HdmX interaction [81]. In this study they used the cyclotide MCoTI-I that has cell-penetrating ability and grafted a loop with a p53-derived α-helical peptide. This engineered cyclotide induced cytotoxicity in a p53-dependent manner in cancer cells harbouring the wild-type p53 in vitro and in vivo. This was achievable purely due to the exceptional stability of the grafted cyclotide in human serum (half-life of 30 h as opposed to 0.7 h for the linearized peptide). This approach involved grafting

36

Bhavesh Khatri et al.

Table 1 Naturally occurring cyclotides and their usage in therapeutic intervention via the grafting approach

Scaffold SFTI-1

Origin of functional epitope VEGF Thrombospondin-1 (TSP-1) CD2 adhesion domain Bradykinin receptor antagonist

Kalata B1 VEGF-A Bradykinin antagonist peptide Myelin oligodendrocyte glycoprotein (MOG) MCoTI

VEGF p53 CVX15 Lyp1 (derived from phage display) Apolipoprotein E

Comparison of serum stability Target

Epitope

Angiogenesis t1/2 ~2 h Anti-angiogenesis 100% degradation within 24 h Autoimmune diseases Pain and t1/2 ~1 h inflammation VEGF-A t1/2 ~2 h antagonist Pain and t1/2 ~5 min inflammation Multiple sclerosis t1/2 < 1 h t1/2 ~2 h

Angiogenesis Cancer HIV infection Cancer

t1/2 ¼ 21 h t1/2 ¼ 4 h

Cancer

t1/2 < 2 h

Grafted scaffold

Refs.

24 h (100% intact) [80] [82] t1/2 ¼ 24 h t1/2 ¼ 24 h

[83]

6 h (>90% intact) [84] 8 h (>95% intact) [85] 6 h (>90% intact) [84] t1/2 > 8 h

[86]

24 h (100% intact) t1/2 ~ 30 h t1/2 ¼ 62 h 24 h (>85% intact) 24 h (>90% intact)

[80] [81] [87] [88] [89]

the functional epitope on a single loop; however, several studies have focused on engineering more than one loop. In the past few years, several groups have focused on this grafting approach to target various intracellular and extracellular PPIs and cell surface receptors that are listed in Table 1. References 1. Henninot A, Collins JC, Nuss JM (2018) The current state of peptide drug discovery: back to the future? J Med Chem 61:1382–1414. https://doi.org/10.1021/acs.jmedchem. 7b00318 2. Lau JL, Dunn MK (2018) Therapeutic peptides: historical perspectives, current development trends, and future directions. Bioorg Med Chem 26:2700–2707. https://doi.org/ 10.1016/j.bmc.2017.06.052 3. Davis AM, Plowright AT, Valeur E (2017) Directing evolution: the next revolution in drug discovery? Nat Rev Drug Discov 16:681–698. https://doi.org/10.1038/nrd. 2017.146

4. Hosseinzadeh P, Bhardwaj G, Mulligan VK et al (2017) Comprehensive computational design of ordered peptide macrocycles. Science 358:1461–1466. https://doi.org/10.1126/ science.aap7577 5. Roberts CJ (2014) Therapeutic protein aggregation: mechanisms, design, and control. Trends Biotechnol 32:372–380. https://doi. org/10.1016/j.tibtech.2014.05.005 6. Craik DJ, Fairlie DP, Liras S et al (2013) The future of peptide-based drugs. Chem Biol Drug Des 81:136–147. https://doi.org/10. 1111/cbdd.12055 7. Gao M, Cheng K, Yin H (2015) Targeting protein-protein interfaces using macrocyclic

Metabolic Stability Strategies peptides. Biopolymers 104:310–316. https:// doi.org/10.1002/bip.22625 8. Driggers EM, Hale SP, Lee J et al (2008) The exploration of macrocycles for drug discovery—an underexploited structural class. Nat Rev Drug Discov 7:608–624. https://doi. org/10.1038/nrd2590 9. Gilbreth RN, Koide S (2012) Structural insights for engineering binding proteins based on non-antibody scaffolds. Curr Opin Struct Biol 22:413–420. https://doi.org/10. 1016/j.sbi.2012.06.001 10. Gould A, Ji Y, Aboye TL et al (2011) Cyclotides, a novel ultrastable polypeptide scaffold for drug discovery. Curr Pharm Des 17:4294–4307 11. Verdine GL, Hilinski GJ (2012) Stapled peptides for intracellular drug targets. Methods Enzymol 503:3–33. https://doi.org/10. 1016/B978-0-12-396962-0.00001-X 12. Kessler H (1982) Peptide conformations.19. Conformation and biological-activity of cyclic-peptides. Angew Chem Int Ed 21:512–523. https://doi.org/10.1002/anie. 198205121 13. Hruby VJ (2002) Designing peptide receptor agonists and antagonists. Nat Rev Drug Discov 1:847–858. https://doi.org/10.1038/ nrd939 14. Weckbecker G, Lewis I, Albert R et al (2003) Opportunities in somatostatin research: biological, chemical and therapeutic aspects. Nat Rev Drug Discov 2:999–1017. https:// doi.org/10.1038/nrd1255 15. Zorzi A, Deyle K, Heinis C (2017) Cyclic peptide therapeutics: past, present and future. Curr Opin Chem Biol 38:24–29. https://doi.org/ 10.1016/j.cbpa.2017.02.006 16. Wikberg JES, Mutulis F (2008) Targeting melanocortin receptors: an approach to treat weight disorders and sexual dysfunction. Nat Rev Drug Discov 7:307–323. https://doi. org/10.1038/nrd2331 17. Alobeidi F, Castrucci AMD, Hadley ME et al (1989) Potent and prolonged acting cyclic lactam analogs of alpha-melanotropin—design based on molecular-dynamics. J Med Chem 32:2555–2561. https://doi.org/10.1021/ jm00132a010 18. Clayton AH, Lucas J, DeRogatis LR et al (2017) Phase I randomized placebocontrolled, double-blind study of the safety and tolerability of bremelanotide coadministered with ethanol in healthy male and female participants. Clin Ther 39:514–526. https:// doi.org/10.1016/j.clintiera.2017.01.018

37

19. Howell SM, Fiacco SV, Takahashi TT et al (2014) Serum stable natural peptides designed by mRNA display. Sci Rep 4:6008–6012. https://doi.org/10.1038/srep06008 20. Morioka T, Loik ND, Hipolito CJ et al (2015) Selection-based discovery of macrocyclic peptides for the next generation therapeutics. Curr Opin Chem Biol 26:34–41. https://doi.org/ 10.1016/j.cbpa.2015.01.023 21. Sako Y, Goto Y, Murakami H et al (2008) Ribosomal synthesis of peptidase-resistant peptides closed by a nonreducible inter-side-chain bond. ACS Chem Biol 3:241–249. https:// doi.org/10.1021/cb800010p 22. Assem N, Ferreira DJ, Wolan DW et al (2015) Acetone-linked peptides: a convergent approach for peptide macrocyclization and labeling. Angew Chem Int Ed 54:8665–8668. https://doi.org/10.1002/anie.201502607 23. de Araujo AD, Mobli M, Castro J et al (2014) Selenoether oxytocin analogues have analgesic properties in a mouse model of chronic abdominal pain. Nat Commun 5:3165–3176. https://doi.org/10.1038/ncomms4165 24. Nutt RF, Veber DF, Saperstein R (1980) Synthesis of non-reducible bicyclic analogs of somatostatin. J Am Chem Soc 102:6539–6545. https://doi.org/10.1021/ ja00541a025 25. Kourra CMBK, Cramer N (2016) Converting disulfide bridges in native peptides to stable methylene thioacetals. Chem Sci 7:7007–7012. https://doi.org/10.1039/ c6sc02285e 26. Chatterjee J, Gilon C, Hoffman A et al (2008) N-methylation of peptides: a new perspective in medicinal chemistry. Acc Chem Res 41:1331–1342. https://doi.org/10.1021/ ar8000603 27. Zhang JQ, Mulumba M, Ong H et al (2017) Diversity-oriented synthesis of cyclic azapeptides by A(3)-macrocyclization provides highaffinity CD36-modulating peptidomimetics. Angew Chem Int Ed 56:6284–6288. https:// doi.org/10.1002/anie.201611685 28. Shin SBY, Yoo B, Todaro LJ et al (2007) Cyclic peptoids. J Am Chem Soc 129:3218–3225. https://doi.org/10.1021/ja066960o 29. Sherman DB, Spatola AF (1990) Compatibility of thioamides with reverse turn features—synthesis and conformational-analysis of 2 model cyclic pseudopeptides containing thioamides as backbone modifications. J Am Chem Soc 112:433–441. https://doi.org/10.1021/ ja00157a064 30. Choudhary A, Raines RT (2011) An evaluation of peptide-bond isosteres. Chembiochem

38

Bhavesh Khatri et al.

12:1801–1807. https://doi.org/10.1002/ cbic.201100272 31. Chatterjee J, Rechenmacher F, Kessler H (2013) N-methylation of peptides and proteins: an important element for modulating biological functions. Angew Chem Int Ed 52:254–269. https://doi.org/10.1002/anie. 201205674 32. Cody WL, He JX, Reily MD et al (1997) Design of a potent combined pseudopeptide endothelin-A/endothelin-B receptor antagonist, Ac-DBhg(16)-Leu-Asp-Ile-[NMe]IleTrp(21) (PD 156252): examination of its pharmacokinetic and spectral properties. J Med Chem 40:2228–2240. https://doi.org/10. 1021/jm970161m 33. Biron E, Chatterjee J, Ovadia O et al (2008) Improving oral bioavailability of peptides by multiple N-methylation: somatostatin analogues. Angew Chem Int Ed 47:2595–2599. https://doi.org/10.1002/anie.200705797 34. Cort JH, Schuck O, Stribrna J et al (1975) Role of disulfide bridge and C-terminal tripeptide in antidiuretic action of vasopressin in man and rat. Kidney Int 8:292–302. https://doi. org/10.1038/ki.1975.116 35. Zaoral M, Kolc J, Sorm F (1966) Synthesis of D-Arg8- and D-Lys8- Vasopressin. Collect Czech Chem C 31:382–383. https://doi. org/10.1135/cccc19660382 36. Bauer W, Briner U, Doepfner W et al (1982) Sms 201-995—a very potent and selective octapeptide analog of somatostatin with prolonged action. Life Sci 31:1133–1140. https://doi.org/10.1016/0024-3205(82) 90087-X 37. Schumacher TNM, Mayr LM, Minor DL et al (1996) Identification of D-peptide ligands through mirror-image phage display. Science 271:1854–1857. https://doi.org/10.1126/ science.271.5257.1854 38. Eckert DM, Malashkevich VN, Hong LH et al (1999) Inhibiting HIV-1 entry: discovery of D-peptide inhibitors that target the gp41 coiled-coil pocket. Cell 99:103–115. https:// doi.org/10.1016/S0092-8674(00)80066-5 39. Goodman M, Chorev M (1979) Concept of linear modified retro-peptide structures. Acc Chem Res 12:1–7. https://doi.org/10.1021/ ar50133a001 40. Wermuth J, Goodman SL, Jonczyk A et al (1997) Stereoisomerism and biological activity of the selective and superactive alpha(v)beta (3) integrin inhibitor cyclo(-RGDfV-) and its retro-inverso peptide. J Am Chem Soc 119:1328–1335. https://doi.org/10.1021/ ja961908l

41. Jameson BA, Mcdonnell JM, Marini JC et al (1994) A rationally designed Cd4 analog inhibits experimental allergic encephalomyelitis. Nature 368:744–746. https://doi.org/10. 1038/368744a0 42. Li C, Pazgier M, Li J et al (2010) Limitations of peptide retro-inverso isomerization in molecular mimicry. J Biol Chem 285:19572–19581. https://doi.org/10. 1074/jbc.M110.116814 43. Batjargal S, Huang Y, Wang YXJ et al (2014) Synthesis of thioester peptides for the incorporation of thioamides into proteins by native chemical ligation. J Pept Sci 20:87–91. https://doi.org/10.1002/psc.2589 44. Mukherjee S, Verma H, Chatterjee J (2015) Efficient site-specific incorporation of thioamides into peptides on a solid support. Org Lett 17:3150–3153. https://doi.org/10. 1021/acs.orglett.5b01484 45. Chen X, Mietlicki-Baase EG, Barrett TM et al (2017) Thioamide substitution selectively modulates proteolysis and receptor activity of therapeutic peptide hormones. J Am Chem Soc 139:16688–16695. https://doi.org/10. 1021/jacs.7b08417 46. Verma H, Khatri B, Chakraborti S et al (2018) Increasing the bioactive space of peptide macrocycles by thioamide substitution. Chem Sci 9:2443–2451. https://doi.org/10.1039/ c7sc04671e 47. Rhodes CA, Pei DH (2017) Bicyclic peptides as next-generation therapeutics. Chem-Eur J 23:12690–12703. https://doi.org/10.1002/ chem.201702117 48. Heinis C, Rutherford T, Freund S et al (2009) Phage-encoded combinatorial chemical libraries based on bicyclic peptides. Nat Chem Biol 5:502–507. https://doi.org/10.1038/ nchembio.184 49. Hacker DE, Hoinka J, Iqbal ES et al (2017) Highly constrained bicyclic scaffolds for the discovery of protease-stable peptides via mRNA display. ACS Chem Biol 12:795–804. https://doi.org/10.1021/acschembio. 6b01006 50. Angelini A, Morales-Sanfrutos J, Diderich P et al (2012) Bicyclization and tethering to albumin yields long-acting peptide antagonists. J Med Chem 55:10187–10197. https://doi. org/10.1021/jm301276e 51. Bartoloni M, Jin X, Marcaida MJ et al (2015) Bridged bicyclic peptides as potential drug scaffolds: synthesis, structure, protein binding and stability. Chem Sci 6:5473–5490. https://doi. org/10.1039/c5sc01699a

Metabolic Stability Strategies 52. Kale SS, Villequey C, Kong XD et al (2018) Cyclization of peptides with two chemical bridges affords large scaffold diversities. Nat Chem 10:715–723. https://doi.org/10. 1038/s41557-018-0042-7 53. Wells JA, McClendon CL (2007) Reaching for high-hanging fruit in drug discovery at protein-protein interfaces. Nature 450:1001–1009. https://doi.org/10.1038/ nature06526 54. Bruzzoni-Giovanelli H, Alezra V, Wolff N et al (2018) Interfering peptides targeting proteinprotein interactions: the next generation of drugs? Drug Discov Today 23:272–285. https://doi.org/10.1016/j.drudis.2017.10. 016 55. Marqusee S, Baldwin RL (1987) Helix stabilization by glu- ... lys+ salt bridges in short peptides of denovo design. Proc Natl Acad Sci U S A 84:8898–8902. https://doi.org/10.1073/ pnas.84.24.8898 56. Walensky LD, Bird GH (2014) Hydrocarbonstapled peptides: principles, practice, and progress. J Med Chem 57:6275–6288. https://doi. org/10.1021/jm4011675 57. Wang YX, Chou DHC (2015) A thiol-ene coupling approach to native peptide stapling and macrocyclization. Angew Chem Int Ed 54:10931–10934. https://doi.org/10.1002/ anie.201503975 58. Spokoyny AM, Zou YK, Ling JJ et al (2013) A perfluoroaryl-cysteine SNAr chemistry approach to unprotected peptide stapling. J Am Chem Soc 135:5946–5949. https://doi. org/10.1021/ja400119t 59. Scrima M, Le Chevalier-Isaad A, Rovero P et al (2010) Cu-I-catalyzed azide-alkyne intramolecular i-to-(i+4) side-chain-to-side-chain cyclization promotes the formation of helix-like secondary structures. Eur J Org Chem 3:446–457. https://doi.org/10.1002/ejoc. 200901157 60. Mendive-Tapia L, Preciado S, Garcia J et al (2015) New peptide architectures through C–H activation stapling between tryptophanphenylalanine/tyrosine residues. Nat Commun 6:7160–7168. https://doi.org/10. 1038/ncomms8160 61. Patgiri A, Jochim AL, Arora PS (2008) A hydrogen bond surrogate approach for stabilization of short peptide sequences in alphahelical conformation. Acc Chem Res 41:1289–1300. https://doi.org/10.1021/ ar700264k 62. Bird GH, Mazzola E, Opoku-Nsiah K et al (2016) Biophysical determinants for cellular uptake of hydrocarbon-stapled peptide helices.

39

Nat Chem Biol 12:845–852. https://doi.org/ 10.1038/nchembio.2153 63. Walensky LD, Kung AL, Escher I et al (2004) Activation of apoptosis in vivo by a hydrocarbon-stapled BH3 helix. Science 305:1466–1470. https://doi.org/10.1126/ science.1099191 64. Bernal F, Tyler AF, Korsmeyer SJ et al (2007) Reactivation of the p53 tumor suppressor pathway by a stapled p53 peptide. J Am Chem Soc 129:2456–2457. https://doi.org/10.1021/ ja0693587 65. Checco JW, Lee EF, Evangelista M et al (2015) alpha/beta-peptide foldamers targeting intracellular protein-protein interactions with activity in living cells. J Am Chem Soc 137:11365–11375. https://doi.org/10. 1021/jacs.5b05896 66. Lee EF, Smith BJ, Horne WS et al (2011) Structural basis of Bcl-xL recognition by a BH3-mimetic alpha/beta-peptide generated by sequence-based design. Chembiochem 12:2025–2032. https://doi.org/10.1002/ cbic.201100314 67. Wang D, Liao W, Arora PS (2005) Enhanced metabolic stability and protein-binding properties of artificial alpha helices derived from a hydrogen-bond surrogate: application to Bcl-xL. Angew Chem Int Ed Eng 44:6525–6529. https://doi.org/10.1002/ anie.200501603 68. Bird GH, Madani N, Perry AF et al (2010) Hydrocarbon double-stapling remedies the proteolytic instability of a lengthy peptide therapeutic. Proc Natl Acad Sci U S A 107:14093–14098. https://doi.org/10. 1073/pnas.1002713107 69. Shepherd NE, Hoang HN, Abbenante G et al (2005) Single turn peptide alpha helices with exceptional stability in water. J Am Chem Soc 127:2974–2983. https://doi.org/10.1021/ ja0456003 70. Murage EN, Gao G, Bisello A et al (2010) Development of potent glucagon-like peptide-1 agonists with high enzyme stability via introduction of multiple lactam bridges. J Med Chem 53:6412–6420. https://doi.org/10. 1021/jm100602m 71. Holland-Nell K, Meldal M (2011) Maintaining biological activity by using triazoles as disulfide bond mimetics. Angew Chem Int Ed Eng 50:5204–5206. https://doi.org/10.1002/ anie.201005846 72. Park JH, Waters ML (2013) Positional effects of click cyclization on beta-hairpin structure, stability, and function. Org Biomol Chem

40

Bhavesh Khatri et al.

11:69–77. https://doi.org/10.1039/ c2ob26445e 73. Hanold LE, Oruganty K, Ton NT et al (2015) Inhibiting EGFR dimerization using triazolylbridged dimerization arm mimics. PLoS One 10:e0118796. https://doi.org/10.1371/jour nal.pone.0118796 74. Srinivas N, Jetter P, Ueberbacher BJ et al (2010) Peptidomimetic antibiotics target outer-membrane biogenesis in Pseudomonas aeruginosa. Science 327:1010–1013. https:// doi.org/10.1126/science.1182749 75. Brady RM, Baell JB, Norton RS (2013) Strategies for the development of conotoxins as new therapeutic leads. Mar Drugs 11:2293–2313. https://doi.org/10.3390/md11072293 76. Olivera BM, Cruz LJ, de Santos V et al (1987) Neuronal calcium channel antagonists. Discrimination between calcium channel subtypes using omega-conotoxin from Conus magus venom. Biochemistry 26:2086–2090. https:// doi.org/10.1021/bi00382a004 77. Clark RJ, Fischer H, Dempster L et al (2005) Engineering stable peptide toxins by means of backbone cyclization: stabilization of the alpha-conotoxin MII. Proc Natl Acad Sci U S A 102:13767–13772. https://doi.org/10. 1073/pnas.0504613102 78. Clark RJ, Jensen J, Nevin ST et al (2010) The engineering of an orally active conotoxin for the treatment of neuropathic pain. Angew Chem Int Ed Eng 49:6545–6548. https:// doi.org/10.1002/anie.201000620 79. de Veer SJ, Weidmann J, Craik DJ (2017) Cyclotides as tools in chemical biology. Acc Chem Res 50:1557–1565. https://doi.org/ 10.1021/acs.accounts.7b00157 80. Chan LY, Gunasekera S, Henriques ST et al (2011) Engineering pro-angiogenic peptides using stable, disulfide-rich cyclic scaffolds. Blood 118:6709–6717. https://doi.org/10. 1182/blood-2011-06-359141 81. Ji Y, Majumder S, Millard M et al (2013) In vivo activation of the p53 tumor suppressor pathway by an engineered cyclotide. J Am Chem Soc 135:11623–11633. https://doi. org/10.1021/ja405108p

82. Chan LY, Craik DJ, Daly NL (2015) Cyclic thrombospondin-1 mimetics: grafting of a thrombospondin sequence into circular disulfide-rich frameworks to inhibit endothelial cell migration. Biosci Rep 35:e00270–e00281. https://doi.org/10.1042/BSR20150210 83. Sable R, Durek T, Taneja V et al (2016) Constrained cyclic peptides as immunomodulatory inhibitors of the CD2:CD58 protein-protein interaction. ACS Chem Biol 11:2366–2374. https://doi.org/10.1021/acschembio. 6b00486 84. Qiu YB, Taichi M, Wei N et al (2017) An orally active bradykinin B-1 receptor antagonist engineered as a bifunctional chimera of sunflower trypsin inhibitor. J Med Chem 60:504–510. https://doi.org/10.1021/acs.jmedchem. 6b01011 85. Gunasekera S, Foley FM, Clark RJ et al (2008) Engineering stabilized vascular endothelial growth factor-a antagonists: synthesis, structural characterization, and bioactivity of grafted analogues of cyclotides. J Med Chem 51:7697–7704. https://doi.org/10.1021/ jm800704e 86. Wang CK, Gruber CW, Cemazar M et al (2014) Molecular grafting onto a stable framework yields novel cyclic peptides for the treatment of multiple sclerosis. ACS Chem Biol 9:156–163. https://doi.org/10.1021/ cb400548s 87. Aboye TL, Ha H, Majumder S et al (2012) Design of a novel cyclotide-based CXCR4 antagonist with anti-human immunodeficiency virus (HIV)-1 activity. J Med Chem 55:10729–10734. https://doi.org/10.1021/ jm301468k 88. Conibear AC, Chaousis S, Durek T et al (2016) Approaches to the stabilization of bioactive epitopes by grafting and peptide cyclization. Biopolymers 106:89–100. https://doi.org/ 10.1002/bip.22767 89. D’Souza C, Henriques ST, Wang CK et al (2016) Using the MCoTI-II cyclotide scaffold to design a stable cyclic peptide antagonist of SET, a protein overexpressed in human cancer. Biochemistry 55:396–405. https://doi.org/ 10.1021/acs.biochem.5b00529

Chapter 3 Designing Cell-Permeable Macrocyclic Peptides George Appiah Kubi, Patrick G. Dougherty, and Dehua Pei Abstract Peptides provide an attractive modality for targeting challenging drug targets such as intracellular proteinprotein interactions. Unfortunately, peptides are generally impermeable to the cell membrane and inherently susceptible to proteolytic degradation in vivo. Macrocyclization of peptides greatly increases their proteolytic stability and in some cases the cell-penetrating activity. Conjugation of peptidyl cargoes to cyclic cell-penetrating peptides has resulted in potent, cell-permeable, and metabolically stable macrocyclic peptides against intracellular protein targets. Proper conjugation/integration of a peptidyl cargo with a cyclic cell-penetrating peptide is critical to retain the activity of each component and generate a biologically active macrocyclic peptide. This chapter describes the different conjugation strategies that have been developed (including endocyclic, bicyclic, and reversible cyclization methods) and the detailed protocols for their preparation. Key words Bicyclic peptides, Cyclic peptides, Cyclic cell-penetrating peptides, Protein-protein interaction, Reversible cyclization

1

Introduction Traditional drug discovery has largely been dedicated to the discovery and optimization of small-molecule drugs that adhere to the rule of five (Ro5) guidelines [1, 2]. These small molecules have the ability to cross biological membranes by passive diffusion [3]. Unfortunately, molecules within the Ro5 boundary are generally ineffective against protein-protein interactions (PPIs), which represent the largest untapped class of therapeutic opportunities but usually do not contain well-defined, hydrophobic pockets required for small molecules to bind [4]. Biological drugs (e.g., monoclonal antibodies) have found considerable success in treating a wide range of diseases but cannot cross the cell membrane, limiting their usage to extracellular targets. An estimated ~80% of therapeutically relevant drug targets are currently undruggable by small molecules or biologics and most of them are intracellular [5]. Clearly, alternative drug modalities that effectively penetrate the cell membrane are needed.

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_3, © Springer Science+Business Media, LLC, part of Springer Nature 2019

41

42

George Appiah Kubi et al.

Peptides have emerged as a promising new drug modality because they have the ability to recapitulate the exceptional affinity and selectivity of proteins, while retaining some of the attributes of small molecules (e.g., synthetic accessibility, lower cost of production, and lower risk of immune response) [6]. However, traditional peptide therapeutics suffer from issues inherent to the class, such as rapid proteolytic degradation in vivo and lack of cell permeability [7]. To improve the metabolic stability of peptides, researchers have explored a variety of macrocyclization methods as well as incorporation of unnatural amino acids [7]. In general, cyclic peptides of small- and medium-sized rings (
10 aa) are relatively resistant to proteolytic degradation. Some of these cyclic peptides have already become successful therapeutic agents, clinical candidates, and/or valuable biological probes, including cyclic RGD peptides [8], disulfide-cyclized linaclotide [9], and hydrocarbon-stapled peptides against the MDM2-p53 interaction [10]. An ongoing and far more challenging effort is to improve the cell permeability of peptides (including cyclic peptides). Broadly speaking, researchers have been pursuing two different strategies to design cell-permeable cyclic peptides, corresponding to two different cellular entry mechanisms: passive diffusion vs. endocytosis. This chapter focuses on the latter approach, which involves conjugation or integration of a peptidyl cargo with a cell-penetrating peptide (CPP). 1.1 Cyclic CellPenetrating Peptides

The discovery of the Tat peptide (derived from HIV transactivator of transcription protein) as a CPP in the early 1990s led to the birth of a new field [11, 12]. Since then, a large number of CPPs have been discovered, including nonaarginine (R9) and penetratin (Antp), and have been used to deliver a wide range of cargoes in vitro and in vivo [13]. However, linear CPPs have encountered several challenges which have so far prevented their translation into the clinic [14], including rapid proteolytic degradation and poor pharmacokinetic properties [15], lack of biodistribution [16], and poor cytosolic delivery efficiencies [17, 18]. To overcome these challenges, Pei and colleagues explored cyclic peptides as CPPs and discovered cyclo(Phe-Nal-Arg-Arg-Arg-Arg-Gln) (cyclic CPP1, where Nal is L-2-naphthylalanine) as a significantly improved CPP [19] (Fig. 1). In addition to its excellent proteolytic stability, CPP1 showed 3- to 12-fold higher cytosolic delivery efficiency than Tat, R9, and Antp. Subsequent SAR and optimization led to the discovery of cyclic CPP9 [cyclo(phe-Nal-Arg-argArg-arg-Gln)] and CPP12 [cyclo(Phe-phe-Nal-Arg-arg-Arg-argGln)] as exceptionally active CPPs, having cytosolic delivery efficiencies of 62% and 120%, respectively (100% is defined as equal cargo concentration in the extracellular medium and cytosol) [20]. For comparison, Tat and R9 showed cytosolic delivery efficiencies of 2.0% and 4.4%, respectively, under the same assay condition. Cyclic CPPs bind directly to the plasma membrane

Cell-Permeable Cyclic Peptides

43

Fig. 1 Structure of cyclic CPP1

phospholipids and enter cells by endocytosis. While inside the early endosome, the cyclic CPPs bind to the endosomal membrane and induce membrane curvature and budding of small CPP-enriched vesicles into the cytosol. Subsequent collapse of the unstable vesicles resulted in the release of the vesicular contents (including the CPPs) into the cytosol. 1.2 Integration of Peptidyl Cargo with Cyclic CPP

Cyclic CPPs have no known biological activity on their own; their utility lies in their ability to transport biologically active but membrane-impermeable cargoes into the cytosol of mammalian cells. Their mechanism of action predicts that cyclic CPPs should be cargo agnostic, although the nature of the cargo can affect the cellular entry efficiency via several different mechanisms. For example, the cargo may bind to the plasma membrane and enhance the endocytic uptake of the CPP-cargo conjugate or to the endosomal membrane and enhance endosomal escape of the conjugate. The cargo may increase or decrease the binding of cyclic CPP to serum proteins during circulation, affecting the cellular uptake kinetics and/or efficiency. Finally, the cargo may physically interact with the CPP, either intramolecularly or intermolecularly, and decrease the activity of the cyclic CPP or the cargo. Therefore, proper conjugation of a given cargo with a cyclic CPP is critical to ensure that the activity of the CPP and/or the cargo is maintained or, in some cases, enhanced. So far, cyclic CPPs have been used to successfully deliver small molecules (e.g., fluorescent dyes), linear peptides, cyclic peptides, proteins, and nucleic acids [21]. Below, we describe the different strategies that have been developed to conjugate cyclic CPPs with peptidyl cargoes, including endocyclic, exocyclic, bicyclic, and reversible cyclization delivery methods (Fig. 2).

44

George Appiah Kubi et al.

a

b

Endocyclic conjugation

c

Exocyclic conjugation

d

Bicyclic conjugation

e

Plasma membrane

Extracellular

Intracellular

Reversible cyclization

Extracellular

Plasma membrane

Intracellular

Reversible bicyclization

Fig. 2 Schemes showing the different methods by which peptidyl cargoes may be conjugated with a cyclic CPP. (a) Endocyclic conjugation (b) exocyclic conjugation (c) bicyclic conjugation (d) reversible cyclization and (e) reversible bicyclization

1.2.1 Endocyclic Conjugation

Endocyclic conjugation involves direct insertion of a cargo sequence into a cyclic CPP ring, and the resulting conjugate is a monocyclic peptide [19]. This conjugation method is limited to relatively small cargoes (i.e., peptides of
5 aa). Longer peptide cargoes result in larger rings, which have progressively poorer cellular entry efficiencies, likely because increasing conformational flexibility decreases the entropic advantages offered by macrocyclization. Since the target-binding sequence and the CPP sequence are in the same ring, negatively charged cargo sequences may interact electrostatically with the cationic CPP and negatively impact the cellular uptake efficiency. On the other hand, the proximity of the cargo and CPP sequences makes it possible to design cyclic peptides in which the same residue(s) serves the dual function of cellular entry and target engagement. This can significantly reduce the size of macrocycles and make them more “drug-like.” By using this strategy, Bedewy et al. [22] designed a cycloheptapeptidyl inhibitor against peptidyl-prolyl cis-trans isomerase Pin1 (Fig. 3a). Despite the presence of a negatively charged phospho-Dthreonine as a key Pin1-binding motif, the macrocycle is cellpermeable and inhibited the Pin1 activity in HeLa cells. To our knowledge, this is the smallest macrocyclic peptide reported, which enters mammalian cells by an endocytic mechanism and elicits

Cell-Permeable Cyclic Peptides

a

45

b

Fig. 3 Structures of cell-permeable Pin1 inhibitors generated by endocyclic (a) and bicyclic conjugation methods (b)

biological activity. Upadhyaya et al. designed a combinatorial library of monocyclic peptides by integrating a CPP-like motif and randomized peptide sequences [23]. Screening of this library for binding to G12 V mutant K-Ras followed by optimization led to a potent, cell-permeable cycloundecapeptidyl K-Ras inhibitor, which selectively blocked Ras-GTP from binding to its downstream effector proteins such as Raf and PI3 kinases and induced apoptosis of mutant K-Ras-driven cancer cells. 1.2.2 Bicyclic Conjugation

When the biologically active cargo is also a cyclic peptide, one can fuse the cyclic peptide with a cyclic CPP to form a bicyclic peptide, in which one ring ensures efficient cellular entry, while the other ring binds to the target of interest (or bicyclic conjugation). A simple and yet general, highly effective method of bicyclization is to fuse the CPP and cargo sequences to form a linear peptide containing two side-chain amine-containing amino acids (e.g., lysine or 2,3-diaminopropionic acid (Dap)), one at the C-terminus and the other at the CPP-cargo junction. The resulting peptide is then converted into a bicyclic peptide by reacting the N-terminal amine and the two side-chain amines with trimesic acid (Fig. 2c) [24]. This rigid scaffold helps preorganize the peptide into productive binding conformations, increasing their binding affinity and specificity for protein targets. The planar scaffold also orients the CPP and cargo rings away from each other, minimizing their mutual interference. Most importantly, bicyclic conjugation can in principle accommodate cargoes of any size or sequence, since changes in the cargo ring does not affect the CPP ring or its cellular entry efficiency. By utilizing this strategy, Jiang and Pei designed a bicyclic peptide library aimed at finding a potent, cell-permeable non-phosphorylated bicyclic peptidyl inhibitor against Pin1

46

George Appiah Kubi et al.

[25]. In this library, one ring of the bicyclic peptides featured a fixed CPP motif, while the other ring consisted of a degenerate peptide sequence (Fig. 2c). Screening of the library against Pin1 led to the identification of a potent and highly selective Pin1 inhibitor (Fig. 3b) which bound to Pin1 with KD ¼ 0.12 μM and inhibited Pin1 activity in human cancer cells. The library approach was also applied to discover cell-permeable and biologically active inhibitors against protein-tyrosine phosphatases 1B [24] and TCPTP [26], as well as K-Ras [27]. 1.2.3 Reversible Cyclization

For some target proteins (e.g., PDZ domains), the peptide ligand must be in its linear, extended form to be biologically active. This type of linear peptidyl cargoes is not suitable for either the endocyclic or bicyclic conjugation method. Simply conjugating a linear peptide to the side chain of a cyclic CPP is often inadequate, because linear peptides, especially those consisting of primarily proteinogenic acids, are susceptible to rapid proteolytic degradation in vivo. To this end, Qian et al. introduced a reversible cyclization strategy which enhances both the proteolytic stability and the cell permeability of linear peptide cargoes [28]. In this strategy, the CPP sequence is fused with a peptidyl cargo sequence and cyclized through a disulfide bond. Cyclization improves both resistance to proteolysis and cellular uptake efficiency. Upon successful entry into the cytosol, the disulfide bond is reduced by intracellular thiols to liberate the biologically active, linear peptide for engagement of the desired intracellular target (Fig. 2d). Qian et al. have applied this strategy to generate cell-permeable peptide substrates for realtime detection of intracellular caspase activities [28]. The same investigators also developed a disulfide-cyclized peptidyl inhibitor against the CFTR-associated ligand CAL-PDZ domain as a potential treatment for cystic fibrosis [28] (Fig. 4a).

1.2.4 Reversible Bicyclization

Like endocyclic conjugation, disulfide-mediated cyclization of CPP-cargo fusion peptides is limited to relatively small ring sizes. Reversible bicyclization provides an alternative approach to introducing additional conformational rigidity into the macrocyclic peptide. Qian et al. first demonstrated this strategy by designing a cellpermeable inhibitor against NF-κB essential modulator (NEMO) [29]. They fused the NEMO-binding domain (NBD) of IκB kinase β (ALDWSWLQ) with a short CPP motif (RRRRΦF, where Φ is 2-naphthylalanine) and adding two cysteine residues, one at the C-terminus and the other in between the NBD and CPP sequences. The resulting peptide was cyclized into a bicycle by forming two pairs of disulfide bonds between the cysteines and an N-terminal 3,5-bis(mercaptomethyl)benzoyl moiety (Fig. 4b). Relative to a monocyclic peptide, bicyclization reduces the size of each ring to approximately half of the original size, greatly increasing the

Cell-Permeable Cyclic Peptides

47

a

CAL-PDZ inhibitor

b

NEMO inhibitor

Fig. 4 Structures of reversibly cyclized cell-permeable peptidyl inhibitors. (a) A monocyclic peptidyl inhibitor against the CFTR-CAL interaction and (b) a bicyclic peptidyl inhibitor against the NEMO-IKK interaction

conformational rigidity of the macrocycle. This in turn greatly improves the cellular uptake efficiency as well as resistance to proteolysis. The bicyclic NEMO inhibitor blocked the interaction between NEMO and IκB kinases and inhibited TNFα-induced NF-κB signaling in cell culture.

2

Materials

2.1 Endocyclic Conjugation

1. Rink amide resin LS (see Note 1). 2. Coupling reagents: 4 eq. 1-[bis(dimethylamino)methylene]1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxide hexafluorophosphate (HATU), 4 eq. N-hydroxybenzotriazole (HOBt), 8 eq. N,N-diisopropylethylamine (DIPEA) as solids or as prepared as stock solutions in DMF (see Note 2). 3. Deprotection reagents: 20% piperidine in DMF (v/v) (see Note 3). 4. Synthesis vessels: Pierce unpacked chromatography columns (Cat# 89898) (see Note 4). 5. Amino acids: 4 eq. of the desired Fmoc-protected amino acids dissolved in DMF, can be stored at 4  C (see Note 5).

48

George Appiah Kubi et al.

6. Synthesis solvents: DCM and DMF. 7. Cyclization reagents: (benzotriazol-1-yloxy)tripyrrolidinophosphonium hexafluorophosphate (PyBOP), HOBt, and DIPEA, dissolved in 1:1 DCM/DMF (v/v) (see Note 6). 8. Deallylation: 0.3 eq. tetrakis(triphenylphosphine)palladium (0) [Pd(PPh3)4] dissolved in dry DCM. Immediately before addition of this solution to the resin, add 10 eq. phenylsilane (see Note 7). 9. Side-chain deprotection and cleavage: freshly prepare a solution of 95:2.5:2.5 trifluoroacetic acid (TFA)/triisopropylsilane (TIPS)/water (see Note 8). 10. Inert gas (e.g., nitrogen or argon) and a gas manifold (see Note 9). 11. Diethyl ether, chilled to 20  C or below (see Note 10). 2.2 Bicyclic Peptide Library

1. TentaGel S NH2 resin (90 μm) (see Note 1). 2. Coupling reagents: 4 eq. HATU, 4 eq. HOBt, and 8 eq. DIPEA as solids or as prepared as stock solutions in DMF for standard amino acid couplings. 4 eq. diisopropylcarbodiimide (DIC) (see Note 2). 3. Deprotection reagents: 20% piperidine in DMF (v/v) (see Note 3). 4. Synthesis vessels: Pierce unpacked chromatography columns (Cat# 89898) (see Note 4). 5. Amino acids: 4 eq. of the desired Fmoc-protected amino acids dissolved in DMF, can be stored at 4  C (see Note 5). 6. Synthesis solvents: DCM and DMF. 7. 1,3,5-benzenetricarboxylic acid diallyl ester (see Note 11). 8. Cyclization reagents: PyBOP, HOBt, and DIPEA, dissolved in 1:1 DCM/DMF (v/v) (see Note 6). 9. Deallylation: 0.3 eq. Pd(PPh3)4 and 10 eq. triphenylphosphine (PPh3) dissolved in dry DCM. Immediately before addition of this solution to the resin, add 15 eq. of N-methylaniline (see Note 7). Resuspend the resin in a solution of 10% sodium dimethyldithiocarbamate (wt/v) in DMF for 10 min to scavenge residual palladium after deprotection. 10. Side-chain deprotection solution: freshly prepare a solution of 87.5:2.5:2.5:2.5:2.5:2.5 TFA/thioanisole/water/phenol/ 1,2-ethanedithiol (see Note 8). 11. Acylation: 10 eq. acetic anhydride and 20 eq. DIPEA combined in DCM immediately before addition to the resin.

Cell-Permeable Cyclic Peptides

2.3 Reversible Bicyclization

49

1. Rink amide resin HS (see Note 1). 2. Coupling reagents: 4 eq. HATU, 4 eq. HOBt, 8 eq. DIPEA as solids or as prepared as stock solutions in DMF (see Note 2). 3. Deprotection reagents: 20% piperidine in DMF (v/v), 2 M Hg (OAc)2 in DMF (see Note 3). 4. Synthesis vessels: Pierce unpacked chromatography columns (Cat# 89898) (see Note 4). 5. Amino acids: 4 eq. of the desired Fmoc-protected amino acids dissolved in DMF, can be stored at 4  C (see Note 5). 6. Synthesis solvents: DCM and DMF. 7. Cyclization reagents: PyBOP, HOBt, and DIPEA, dissolved in 1:1 DCM/DMF (v/v) (see Note 6). 8. Deallylation: 0.3 eq. Pd(PPh3)4 and 10 eq. PPh3 dissolved in dry DCM. Immediately before addition of this solution to the resin, add 15 eq. of N-methylaniline (see Note 7). 9. Side-chain deprotection and cleavage: freshly prepare a solution of 95:2.5:2.5 TFA/TIPS/water (see Note 8). 10. Inert gas (e.g., nitrogen or argon) and a gas manifold (see Note 9). 11. Diethyl ether, chilled to 20  C or below (see Note 10).

2.4 Reversible Cyclization

1. Rink amide resin HS (see Note 1). 2. Coupling reagents: 4 eq. HATU, 4 eq. HOBt, and 8 eq. DIPEA as solids or as prepared as stock solutions in DMF (see Note 2). 3. Deprotection reagents: 20% piperidine in DMF (v/v) (see Note 3). 4. Synthesis vessels: Pierce unpacked chromatography columns (Cat# 89898) (see Note 4). 5. Amino acids: 4 eq. of the desired Fmoc-protected amino acids dissolved in DMF, can be stored at 4  C (see Note 5). 6. Synthesis solvents: DCM and DMF. 7. Cyclization reagents: PyBOP, HOBt, and DIPEA, dissolved in 1:1 DCM/DMF (v/v) (see Note 6).

3

Methods

3.1 Endocyclic Conjugation

1. Swell 100 mg of Rink amide resin in DMF for 20 min (see Note 1). 2. Drain and add 20% (v/v) piperidine in DMF for 10 min (twice) to remove Fmoc group (see Note 3). 3. Drain and wash resin with DMF, DCM, and DMF (2–3 times each).

50

George Appiah Kubi et al.

4. Add 4 eq. Fmoc-Glu-OAll, 4 eq. HATU, 4 eq. HOBt, and 8 eq. DIPEA, and mix for 1 h in DMF (see Note 2). 5. Drain and wash exhaustively with DMF, DCM, and DMF. 6. Repeat steps 2 through 5 to couple the remaining amino acids in the linear sequence. 7. Add a solution of 0.3 eq. of Pd(PPh3)4 and 10 eq. of PhSiH3 in dry DCM to the resin and allow to mix for 15 min. Repeat the procedure three times (see Note 7). 8. Drain and wash exhaustively with DCM and DMF. 9. Incubate the resin in a solution of 10% sodium dimethyldithiocarbamate in DMF for 10 min (twice). 10. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 11. Drain and add 20% (v/v) piperidine in DMF for 10 min (twice) to remove Fmoc group (see Note 3). 12. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 13. Incubate resin with 1 M solution HOBt for 15 min (see Note 12). 14. Add a solution of 5 eq. of PyBOP, 5 eq. of HOBt, and 10 eq. of DIPEA in DMF and mix for 1.5 h (see Note 6). Repeat the procedure once. 15. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 16. Add a solution of 95:2.5:2.5 (v/v) TFA/H2O/TIPS for 3 h to release the cyclic peptide from the resin and effect side-chain deprotection (see Note 8). 17. Drain and concentrate the cleavage solution to a semisolid by gently blowing an inert gas over the solution inside a fume hood (see Note 9). 18. Add chilled diethyl ether to the concentrated cleavage solution to precipitate the peptide. Centrifuge at 7.5  103 rpm for 5 min, and remove supernatant with a pipette. Repeat the step three times. 19. Dissolve the triturated crude peptide in DMF, and purify it by reversed-phase HPLC (see Note 13). 3.2 Bicyclic Peptide Library

1. Swell 1 g of TentaGel S NH2 resin in DMF for 20 min (see Note 1). 2. Couple a linker sequence (β-Ala-β-Ala-Met) using standard Fmoc/HATU/DIPEA chemistry (see Note 2). 3. Soak beads in DMF for 30 min. 4. Drain and soak in 1:1 (v/v) degassed DMF/water mixture for 1 h (twice).

Cell-Permeable Cyclic Peptides

51

5. Drain and soak beads in 1:4 (v/v) degassed DMF/water for 1 h (twice). 6. Drain, wash with degassed water, and soak beads in degassed water overnight. 7. Carefully drain and quickly resuspend beads in 55:45 DCM/diethyl ether (v/v) containing 0.4 eq. of Fmoc-OSu and 1 eq. DIPEA, and allow to mix for 30 min. 8. Drain, wash with DMF, and incubate with a solution of 5 eq. of di-tert-butyl dicarbonate and 0.1 eq. of DMAP in DMF for 30 min. 9. Drain and wash with DMF. Add 20% piperidine in DMF solution to the resin, and incubate for 10 min to remove the Fmoc group from the outer layer. 10. Drain and wash resin with DMF, DCM, and DMF (2–3 times each) (see Note 14). 11. Couple 4-(hydroxymethyl)benzoic acid (HMB) to the outer layer using HATU/HOBt/DIPEA (see Note 2). 12. Couple β-Ala to the HMB linker with Fmoc-β-Ala-OH/DIC/ DMAP (5, 5.5, 0.1 eq., respectively) for 2 h. 13. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 14. Couple the sequence Dap(alloc)-β-Ala-β-Ala-Pra using standard Fmoc/HATU/HOBt/DIPEA chemistry. 15. Treat resin with 95/2.5/2.5 (v/v) mixture of TFA/H2O/ TIPS to remove Boc group from the inner layer. 16. Couple the sequence Dap(Mtt)-Phe-Nal-Arg-Arg-Arg-Arg using standard Fmoc/HATU/HOBt/DIPEA chemistry. 17. Divide the resin into desired number of equal portions corresponding to the number of amino acid building blocks to be incorporated into the sequence (see Note 15). 18. Couple a different amino acid building block to each aliquot of the resin by standard Fmoc chemistry (see Note 2). 19. Combine the resin together in the original synthesis vessel. 20. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 21. Treat resin with a solution of 2% (v/v) TFA in DCM for 5 min, and repeat until the solution is clear to remove the Mtt group on the internal Dap residue (see Note 16). 22. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 23. Add a solution of 3 eq. of Fmoc-OSu and 10 eq. DIPEA to the resin, and incubate with mixing for 1 h.

52

George Appiah Kubi et al.

24. Remove the allyl protecting group on the C-terminal Dap residue by treating the resin with a solution of tetrakis(triphenylphosphine)palladium/triphenylphosphine/N-methylaniline (0.3, 5, 15 eq., respectively) in dry DCM for 12 h (see Note 7). 25. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 26. Incubate the resin in a solution of 10% sodium dimethyldithiocarbamate in DMF for 10 min (twice). 27. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 28. Add a solution of 3 eq. diallyl trimesic acid, 3 eq. HATU, and 6 eq. DIPEA in DMF, and incubate for 1 h to acylate the exposed amine (see Note 11). 29. Remove the allyl protecting groups on trimesic acid by treating the resin with a solution of tetrakis(triphenylphosphine)palladium/triphenylphosphine/N-methylaniline (0.3, 5, 15 eq., respectively) in dry DCM for 12 h (see Note 7). 30. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 31. Incubate the resin in a solution of 10% (wt/v) sodium dimethyldithiocarbamate in DMF for 10 min (twice). 32. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 33. Remove the Fmoc groups at the N-terminus as well as on the internal Dap residue as described in steps 6 and 7. 34. Remove the Fmoc group at the N-terminus as well as on the internal Dap residue by adding 20% (v/v) piperidine in DMF for 10 min (twice) to remove Fmoc group (see Note 3). 35. Incubate with 1 M HOBt for 10 min (see Note 12). 36. To cyclize the library, add a solution of 5 eq. of PyBOP, 5 eq. of HOBt, and 10 eq. of DIPEA in DMF, and incubate with mixing for 1.5 h (twice) (see Note 6). 37. Drain and wash resin thoroughly with DMF followed by DCM. 38. Add a modified reagent K (TFA/thioanisole/water/phenol/ 1,2-ethanedithiol, 82.5:5:5:5:2.5 v/v) to the cyclized library, and allow to mix for 3 h (see Note 8). 39. Wash the resulting library thoroughly with DCM, DMF, 5% (v/v) DIPEA in DMF, 1:1 (v/v) DCM/diethyl ether, DMF, and DCM.

Cell-Permeable Cyclic Peptides

3.3 Reversible Bicyclization

53

1. Swell 100 mg of Rink amide resin LS in DMF for 20 min (see Note 1). 2. Drain and add 20% (v/v) piperidine in DMF for 10 min (twice) to remove Fmoc group (see Note 3). 3. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 4. To synthesize a disulfide-mediated bicyclic peptides, first synthesize the corresponding linear peptide containing two Acm-protected cysteine residues (one at the C-terminus and one at the CPP-cargo junction) using standard Fmoc/HATU chemistry (see Note 2). 5. To remove the Acm group, add 2 mL of 2 M mercury (II) acetate in DMF to the resin, and incubate with mixing overnight (see Fig. 5). 6. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 7. Incubate the resin with 2 mL of 20% β-mercaptoethanol in DMF for 2 h (twice) to release the free thiol. 8. Wash exhaustively with DMF to remove all reducing agents. 9. Incubate the resin overnight with a solution of 1 eq. 3,5-bis ((pyridin-2-yldisulfanyl)methyl)benzoic acid in methanol containing 1% acetic acid (v/v) (see Note 17). 10. Treat resin with 20% piperidine in DMF for 10 min (twice) to remove N-terminal Fmoc protecting group. 11. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 12. Add 1 eq. of HATU and 5 eq. of DIPEA in DMF and incubate for 2 h. 13. Drain and wash resin with DMF, DCM, and DMF (2–3 times each).

Fig. 5 Scheme showing the solid-phase synthesis of disulfide-mediated bicyclic peptides

54

George Appiah Kubi et al.

14. Deprotect and release peptide by treating the resin with a solution of 85:10:2.5:2.5 (v/v) TFA/DCM/water/TIPS for 2 h (see Note 8). 15. Drain and concentrate the cleavage solution as described above (see Note 9). 16. Triturate the crude peptide with chilled diethyl ether, and purify by reversed-phase HPLC as described above. 3.4 Reversible Cyclization

1. Swell 100 mg of Rink amide resin LS in DMF for 20 min (see Note 1). 2. Drain and add 20% (v/v) piperidine in DMF for 10 min. twice to remove Fmoc group (see Note 3). 3. Drain and wash resin with DMF, DCM, and DMF (2–3 times each). 4. To synthesize the disulfide-mediated cyclic peptide, first synthesize the corresponding linear peptide containing Cys(Trt) residue at the C-terminus using standard Fmoc/HATU chemistry. 5. Remove N-terminal Fmoc group using 20% (v/v) piperidine in DMF for 10 min (twice) (see Note 3). 6. Install N-terminal thiol by treating with 5 eq. of 3,30 -dithiodipropionic acid, 5 eq. N,N0 - DIC, and 0.1 eq. 4-(dimethylamino)pyridine (DMAP) in anhydrous DCM for 2 h (see Note 18). 7. Incubate resin with 2 mL of 20% β-mercaptoethanol in DMF for 2 h (twice) to expose the free thiol. 8. Wash exhaustively with DMF followed by DCM. 9. Add a solution of 90:2.5:2.5:2.5:2.5 (v/v) TFA/water/phenol/TIPS/1,2-ethanedithiol (EDT), and incubate for 2 h. 10. Drain and concentrate the cleavage solution as described above (see Note 9). 11. Triturate the crude peptide with chilled diethyl ether as described above. 12. Incubate the crude peptide with 5% (v/v) DMSO in PBS (pH 7.4), and mix gently overnight to effect intramolecular cyclization (see Note 19). 13. Purify the crude mixture on reversed-phase HPLC to isolate the intramolecularly cyclized peptide product.

4

Notes 1. There are a large number of different resins that have been developed to facilitate solid-phase peptide synthesis. In general, when an amidated C-terminus is desired, Rink amide resin

Cell-Permeable Cyclic Peptides

55

serves as a good all-purpose choice. In the case of library synthesis, a more hydrophilic support such as TentaGel is desirable and is recommended. For peptides with a free C-terminus, Wang resin and DIC coupling (see Note 2) are recommended. 2. The uronium-based (e.g., HATU, HBTU, and HCTU) coupling reagents enable cost- and time-effective syntheses with minimal racemization. Phosphonium- (e.g., PyBOP) or carbenium-based (e.g., COMU) reagents are also effective. Carbodiimide-based (e.g., DIC) reagents are not recommended for non-microwave-assisted syntheses, as they often cause unacceptable levels of racemization, especially when used in combination with DMAP. However, DIC/DMAP is required for effective coupling of amino acids to alcohols, such as those on Wang resin. 3. If Fmoc deprotection with 20% piperidine in DMF is incomplete (as revealed by MS or ninhydrin tests), a mixture of 2% 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) and 2% piperidine in DMF (v/v/v) is recommended. Twenty percent piperazine in DMF is also an acceptable alternative. 4. The choice of vessel for synthesis is highly specific to the equipment present in each laboratory. In our experience, unpacked chromatography columns provide an economical solution, but these vessels are generally non-reusable, and some researchers may prefer to use glass vessels designed specifically for SPPS. 5. Alternatively, N-methylpyrrolidinone (NMP) may be used in place of DMF to prepare Fmoc-protected AA solutions when solubility issues are encountered. 6. Uronium-based coupling reagents (e.g., HATU) should not be used for peptide cyclization, as they can react with the peptide N-terminus to form N-guanidinylation products. Phosphonium-based reagents do not have this problem and are recommended. 7. Anhydrous solvents should be used for these steps, as the palladium catalyst is water-sensitive. Facile deprotection can be accomplished by using phenylsilane (PhSiH3) with a typical reaction condition of 0.3 eq. Pd(PPh3)4 and 10 eq. PhSiH3 in dry DCM for 3  15 min. 8. The listed condition is for general side-chain deprotection and cleavage from the resin. If peptides contain redox-sensitive amino acids (e.g., Trp, Cys, or Met), addition of 2.5% 1,2-ethanedithiol or 2,20 -(ethylenedioxy)diethanethiol is advised to prevent side-chain oxidation during cleavage. 9. Concentration of peptide solution under inert atmosphere is only required for peptides containing redox-sensitive residues; for all other sequences, regular purified air is acceptable.

56

George Appiah Kubi et al.

10. Diethyl ether used for trituration should be chilled to the lowest, conveniently allowed temperatures (e.g., 20  C) to minimize unwanted product loss. Multiple trituration is recommended to remove as much of the scavengers as possible. More hydrophobic ethers, such as methyl tert-butyl ether (MTBE), are also acceptable solvents. 11. Diallyl trimesic acid is synthesized using a two-step protocol, first by refluxing trimesic acid with thionyl chloride in allyl alcohol, followed by hydrolysis with 1 eq. KOH in allyl alcohol. 12. Incubation and washing with 1 M HOBt ensures complete removal of piperidine from the preceding step. Piperidine can compete with peptide N-terminal amine during peptide cyclization and decrease the reaction yield. 13. Dissolving the conjugates for RP-HPLC can be difficult for highly hydrophobic sequences. The crude peptide may first be dissolved in DMSO or DMF (if containing redox-sensitive residues). Dilution of the solution is best accomplished by gradually adding the hydrophobic component of the mobile phase (e.g., acetonitrile) to the crude peptide solution and then adding water dropwise until the desired volume is reached. Additional organic solvent (DMSO or DMF) can be added if necessary. 14. Successful bead segregation is crucial to the quality and usefulness of the library and can be assessed by performing a chloranil test. Remove a small amount of the resin, deprotect the Fmoc group, add the chloranil reagents, and then examine the beads under a light microscope. A reddish ring on the exterior of the bead indicates successful segregation. If segregation is unsuccessful, remove the Boc group in the inner layer by incubation with 50% TFA/DCM solution for 1 h, and repeat steps 3 through 10. 15. Resin splitting is best accomplished volumetrically, by suspending the resin in a known volume of 1:1 DMF/DCM, withdrawing equal aliquots of the suspension, and transferring them into individual synthesis vessels. Washing and deprotection steps can be conveniently performed in the main synthesis vessel as opposed to in each of the smaller positional synthesis vessels by combining the resin after each coupling. 16. Removal of the Mtt protecting group produces a yellow color due to the resulting methyltrityl cation, providing a visual indication for the reaction progression. Two percent TIPS can be included in the deprotection solution to scavenge this cation and prevent reattachment. 17. The synthetic route to this compound can be found in ref. 29.

Cell-Permeable Cyclic Peptides

57

18. Other thiol-containing acids (e.g., cysteine) can also be used for this purpose. 19. Intramolecular cyclization can be challenging depending on the sequence. Cyclization can be performed directly with the crude linear peptide following trituration or with purified peptide under the same conditions. Reaction progress can be monitored by HPLC/MS. If necessary, additional DMSO can be added (up to 10% v/v), or higher pH is used (e.g., NaHCO3, pH 8.5) to facilitate the cyclization. Higher-order disulfide-mediated conjugates (e.g., dimers/trimers) may be separated by HPLC, reduced, and subjected to another round of intramolecular disulfide formation.

5

Conclusion and Future Directions Although not the focus of this article, several powerful combinatorial libraries technologies have been developed over the past decade or so and can be applied to generate macrocyclic peptidyl ligands with antibody-like affinity and specificity against essentially any protein target [30]. The recent discovery of cyclic CPPs as powerful intracellular delivery vehicles has now made it possible to deliver these macrocyclic peptidyl ligands into the cytosol and nucleus of mammalian cells. Furthermore, as demonstrated by the examples covered in this article, the two platform technologies (library screening and cyclic CPPs) can be integrated to rapidly discover potent, specific, cell-permeable, and metabolically stable macrocyclic peptides to modulate intracellular targets that are undruggable by current drug modalities (i.e., small molecules and biologics). These advancements have ushered in a new era for peptide therapeutics, which may well become the third major drug modality, occupying a vast and presently barren land bounded by molecular weights of 500–5000. These macrocyclic peptides will also provide powerful tool compounds for biological and biomedical research. To realize the potential of macrocyclic peptides as the third major drug modality, we believe that research in the following areas will be essential. First, we need to gain a better mechanistic understanding of how cyclic CPPs (and CPPs in general) achieve cytosolic entry. This knowledge will help predict how this new class of molecules behave in vivo as well as further improve the CPPs. Second, the pharmacokinetics, cytotoxicity, and immunogenic activity of this class of compounds need to be systemically evaluated, especially since poor PK properties were considered as the major limitations of peptidyl drugs in the past. Finally, the chemistry for synthesizing macrocyclic peptides will need significant improvements, at both medicinal and process chemistry stages. In particular, commercial availability of a large collection of affordable

58

George Appiah Kubi et al.

non-proteinogenic amino acids will greatly facilitate the lead optimization efforts.

Acknowledgments This work was supported by NIH grant GM122459 to D.P. References 1. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ (1997) Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 23:3–25 2. Yang NJ, Hinner MJ (2015) Getting across the cell membrane: an overview for small molecules, peptides, and proteins. Methods Mol Biol 1266:29–53 3. Singh SS (2006) Preclinical pharmacokinetics: an approach towards safer and efficacious drugs. Curr Drug Metab 7:165–182 4. Arkin MR, Tang Y, Wells JA (2014) Smallmolecule inhibitors of protein-protein interactions: progressing toward the reality. Chem Biol 21:1102–1114 5. Kumar V, Sanseau P, Simola DF, Hurle MR, Agarwal P (2016) Systematic analysis of drug targets confirms expression in disease-relevant tissues. Sci Rep 6:36205 6. Zorzi A, Deyle K, Heinis C (2017) Cyclic peptide therapeutics: past, present and future. Curr Opin Chem Biol 38:24–29 7. Gentilucci L, De Marco R, Cerisoli L (2010) Chemical modifications designed to improve peptide stability: incorporation of non-natural amino acids, pseudo-peptide bonds, and cyclization. Curr Pharm Des 16:3185–3203 8. Dechantsreiter MA, Planker E, Matha B, Lohof E, Holzemann G, Jonczyk A, Goodman SL, Kessler H (1999) N-Methylated cyclic RGD peptides as highly active and selective alpha(V)beta(3) integrin antagonists. J Med Chem 42:3033–3040 9. Corsetti M, Tack J (2013) Linaclotide: a new drug for the treatment of chronic constipation and irritable bowel syndrome with constipation. United European Gastroenterol J 1:7–20 10. Chang YS, Graves B, Guerlavais V, Tovar C, Packman K, To KH, Olson KA, Kesavan K, Gangurde P, Mukherjee A et al (2013) Stapled alpha-helical peptide drug development: a potent dual inhibitor of MDM2 and MDMX for p53-dependent cancer therapy. Proc Natl Acad Sci U S A 110:E3445–E3454

11. Frankel AD, Pabo CO (1988) Cellular uptake of the tat protein from human immunodeficiency virus. Cell 55:1189–1193 12. Green M, Loewenstein PM (1988) Autonomous functional domains of chemically synthesized human immunodeficiency virus tat transactivator protein. Cell 55:1179–1188 13. Bechara C, Sagan S (2013) Cell-penetrating peptides: 20 years later, where do we stand? FEBS Lett 587:1693–1702 14. Reissmann S (2014) Cell penetration: scope and limitations by the application of cellpenetrating peptides. J Pept Sci 20:760–784 15. Henninot A, Collins JC, Nuss JM (2018) The Current State of Peptide Drug Discovery: Back to the Future? J Med Chem 61:1382–1414 16. Dubikovskaya EA, Thorne SH, Pillow TH, Contag CH, Wender PA (2008) Overcoming multidrug resistance of small-molecule therapeutics through conjugation with releasable octaarginine transporters. Proc Natl Acad Sci U S A 105:12128–12133 17. El-Sayed A, Futaki S, Harashima H (2009) Delivery of macromolecules using argininerich cell-penetrating peptides: ways to overcome endosomal entrapment. AAPS J 11:13–22 18. Erazo-Oliveras A, Muthukrishnan N, Baker R, Wang TY, Pellois JP (2012) Improving the endosomal escape of cell-penetrating peptides and their cargos: strategies and challenges. Pharmaceuticals (Basel) 5:1177–1209 19. Qian Z, Liu T, Liu Y-Y, Briesewitz R, Barrios AM, Jhiang SM, Pei D (2013) Efficient delivery of cyclic peptides into mammalian cells with short sequence motifs. ACS Chem Biol 8:423–431 20. Qian Z, Martyna A, Hard RL, Wang J, AppiahKubi G, Coss C, Phelps MA, Rossman JS, Pei D (2016) Discovery and mechanism of highly efficient cyclic cell-penetrating peptides. Biochemistry 55:2601–2612 21. Wang F, Wang Y, Zhang X, Zhang W, Guo S, Jin F (2014) Recent progress of cellpenetrating peptides as new carriers for

Cell-Permeable Cyclic Peptides intracellular cargo delivery. J Control Release 174:126–136 22. Bedewy W, Liao H, Abou-Taleb NA, Hammad SF, Nasr T, Pei D (2017) Generation of a cellpermeable cycloheptapeptidyl inhibitor against the peptidyl-prolyl isomerase Pin1. Org Biomol Chem 15:4540–4543 23. Upadhyaya P, Qian Z, Selner NG, Clippinger SR, Wu Z, Briesewitz R, Pei D (2015) Inhibition of ras signaling by blocking ras-effector interactions with cyclic peptides. Angew Chem Int Ed 54:7602–7606 24. Lian W, Jiang B, Qian Z, Pei D (2014) Cellpermeable bicyclic peptide inhibitors against intracellular proteins. J Am Chem Soc 136:9830–9833 25. Jiang B, Pei D (2015) A selective, cellpermeable nonphosphorylated bicyclic peptidyl inhibitor against peptidyl–prolyl isomerase Pin1. J Med Chem 58:6306–6312 26. Liao H, Pei D (2017) Cell-permeable bicyclic peptidyl inhibitors against T-cell protein

59

tyrosine phosphatase from a combinatorial library. Org Biomol Chem 15:9595–9598 27. Trinh TB, Upadhyaya P, Qian Z, Pei D (2016) Discovery of a direct ras inhibitor by screening a combinatorial library of cell-permeable bicyclic peptides. ACS Comb Sci 18:75–85 28. Qian Z, Xu X, Amacher JF, Madden DR, Cormet-Boyaka E, Pei D (2015) Intracellular delivery of peptidyl ligands by reversible cyclization: discovery of a PDZ domain inhibitor that rescues CFTR activity. Angew Chem Int Ed 54:5874–5878 29. Qian Z, Rhodes CA, McCroskey LC, Wen J, Appiah-Kubi G, Wang DJ, Guttridge DC, Pei D (2017) Enhancing the cell permeability and metabolic stability of peptidyl drugs by reversible bicyclization. Angew Chem Int Ed 56:1525–1529 30. Dougherty PG, Qian Z, Pei D (2017) Macrocycles as protein-protein interaction inhibitors. Biochem J 474:1109–1125

Chapter 4 Computational Methods for Studying Conformational Behaviors of Cyclic Peptides Fan Jiang and Hao Geng Abstract Molecular dynamics (MD) simulations play more and more important roles in studying conformations of cyclic peptides in solution. Here we describe how to use replica-exchange molecular dynamics (REMD) implemented in Gromacs software package to simulate peptides with backbone cyclization and stapled peptides with side-chain linkages. Some of our methods for trajectory analyses and our residue-specific force fields are also described. Key words Molecular dynamics simulation, Cyclic peptides, Stapled peptides, Gromacs, Residuespecific force fields, Clustering analysis

1

Introduction Protein-protein interactions (PPIs) are often difficult targets for small-molecule drugs. As a surrogate of the functional region of a protein involved in PPI, peptides are promising candidates for targeting PPI with high potency and specificity. However, linear peptides are often conformationally flexible, lowering their binding affinity due to the entropic loss required to achieve their bound (functional) conformation. Also, they often have low bioavailability and poor in vivo stability. Cyclization of linear peptides, either by backbone or side-chain linkage, can achieve much lower conformational flexibility and better drug-like properties [1]. Recently, using computational simulations, we found that the conformational sampling of residues in small cyclic peptides (CPs) is much more similar to that in globular proteins, compared with that in linear peptides [2]. Indeed, computational methods play an important role in studying the conformational behaviors of CPs. Compared with linear peptides, CPs can have higher energy barriers for conformational transitions. For the structure prediction of CPs, conformational search algorithms have been developed previously to locate low-energy minima more efficiently

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_4, © Springer Science+Business Media, LLC, part of Springer Nature 2019

61

62

Fan Jiang and Hao Geng

[3–5]. However, each of these methods should be coupled with an effective energy function (either physics based or knowledge based) with implicit solvation to evaluate the free energy of a certain structure. This significantly limited the accuracies of these methods. For example, Rosetta cannot correctly predict the experimental structure of the α-conotoxin GI (a marine snail toxin, 13-residue peptide with two disulfide bonds) [6]. In recent years, with the rapid increase of computational power, molecular dynamics (MD) simulation of CPs in explicit solvent has become more and more feasible [7–11], and further facilitated by developments in both force fields and enhanced sampling methods. Here, we describe the computational methods for studying CPs used in our recent studies, based on widely used replicaexchange molecular dynamics (REMD) [12]. In a REMD simulation, multiple parallel MD simulations (replicas) are performed at different temperatures (Ts). After every short period of time (picoseconds), attempts of exchanging the structures between neighboring replicas (Ts) are tried such that the detailed balance condition can be fulfilled. During a REMD simulation, high-T replica can overcome energy barriers for conformational transition, and low-T replica can give correct conformational distributions near room temperature. The number of replicas needed to span a given temperature range (e.g., 300–500 K) is roughly in proportion to the square root of the system size (number of atoms). Thus, REMD is especially suitable for small systems such as free cyclic peptides, in contrast to larger proteins. There are also other enhanced sampling methods, such as bias-exchange metadynamics [9], which may achieve higher efficiency in conformational sampling compared with REMD. However, they are usually not as user friendly as REMD.

2

Materials and Methods

2.1 Head-to-Tail Cyclic Peptides

Here, the Gromacs software package (version
4.5) is used to perform all MD simulations. As far as we know, it is the fastest MD engine for small systems running on CPU, and it is also very flexible in terms of potential energy functions (force fields). Carry out the following procedure:

2.1.1 Build Initial 3D Structure (File in PDB Format)

Most MD software packages are designed to simulate linear polypeptide chain(s). A special procedure is needed. 1. Using software such as HyperChem or Chem3D to build corresponding linear peptide. 2. Make a single bond between the first N atom and the last C atom.

Computation of Cyclic Peptides Conformation

63

3. Do energy minimization, and save the optimized structure. Note: Other methods to generate initial PDB file are also OK, but make sure that the first backbone N atom has its amide H atom attached. 2.1.2 Modifications of Some Gromacs Library Files

An amide bond should form between the first and the last residues, such that a topology file with correct connectivity can be generated. 1. Add a new line in the $GMXLIB/specbond.dat file, like for the following example. Then change the total line number in the first line accordingly: ALA N 1 PRO C 1 0.134 ALA PRO Example when the first residue is Ala and the last one is Pro 2. Modify the $GMXLIB/amber99sb.ff/aminoacids.r2b file to avoid the first and last residues being recognized as terminals, by changing corresponding NXXX and CYYY to XXX and YYY. For example, change the name “NALA” to “ALA” and the name “CPRO” to “PRO.” 3. Modify the $GMXLIB/amber99sb.ff/aminoacids.n.tdb file to add new treatment for terminal residues. If the first residue is non-proline, add: [ NH ] [ replace ] N

N

14.0067

0.41570

H

H

1.008

0.27190

If the first residue is proline, please add: [ NH ] [ replace ]

2.1.3 Preparing Gromacs Simulation Files

N

N

14.0067

0.25480

H

H

1.008

0.27190

Generate the initial structure file (.gro) and topology file (.top) using the pdb2gmx command, with some precautions. 1. Look for the “Linking . . .” information after the “Special Atom Distance matrix” in the output, to make sure that the head and the tail have been correctly linked. 2. If being asked to choose terminal types: For “Start terminus,” please choose “NH” For “End terminus,” please choose “None” 3. To use our RSFF2 force field, use the g_mod_top_RSFF2_CycPep.py program instead of the g_mod_top_RSFF2.py program to modify the topology file from Amber99sb force field to RSFF2.

64

Fan Jiang and Hao Geng

2.1.4 Running, Preparing, and Producing Simulations

Solvate the CP with explicit water molecules (usually TIP3P model), using editconf and genbox commands. We have found that 500–1000 water molecules are enough for small CPs of 5–12 residues. After energy minimization and a 300 K NPT MD simulation, perform a 600 K NVT MD simulation of 30 ns to obtain the initial structures for REMD. For each REMD simulation, use temperature range from 300 to 600 K. The intermediate temperatures and number of replicas (>24 replicas depending on system size) were chosen following a recent study [13] to obtain uniform exchange rates of >20%. Exchanges were attempted between neighboring replicas every 1.0 ps. The structures were saved every 1.0 ps. We have found that trajectory length of 100–200 ns/replica is sufficient to achieve well-converged conformational sampling of CPs of 5–12 residues. The electrostatics were treated using the particle-mesh Ewald (PME) with a real-space cutoff of 0.9 nm and van der Waals interactions were also cut off at 0.9 nm with long-range dispersion correction for energy and pressure. A velocity rescaling thermostat with τT ¼ 0.2 ps and a Berendsen barostat with τP ¼ 0.5 ps were used to maintain constant temperature and constant pressure (for NPT simulations). All bonds involving hydrogen were constrained using LINCS, and a time step of 2 fs was used for cyclic peptide simulations. At the same time, the mass of water oxygen atom was reduced from 16 to 2 amu to increase the sampling efficiency without altering the thermodynamics equilibrium properties.

2.2 Side-Chain Stapled Peptides [14]

As in Subheading 2.1.1, starting structures can be constructed using the HyperChem software and saved in pdb file format.

2.2.1 Build Initial 3D Structure 2.2.2 Create Topologies and Force Field Parameters for New Residues

For a stapled peptide, there are two modified (nonnatural amino acid) residues at positions i and i þ 4 (or i þ 3). 1. For each modified residue, build a model compound incorporating the modified side-chain part (preferably using GaussView). Do geometry optimization using Gaussian quantum mechanics software, with keywords like “# HF/6-31G* opt.” 2. For each model compound, calculate electrostatic potential using the optimized structure, with keywords like “# HF/631G* SCF¼Tight Pop¼MK iop(6/33¼2, 6/41¼10, 6/42¼15).” Suppose the output file from this Gaussian calculation is model.log. 3. Derive new atomic charges using the restrained electrostatic potential (RESP) fitting method [15]. This can be done by running antechamber with commands like:

Computation of Cyclic Peptides Conformation

65

Antechamber -i model.log -fi gout -o model.mol2 -fo mol2 -c resp Charges can be found in the output file model.mol2 4. First create new entries for the two modified residues, in the aminoacids.rtp file. Atom-type(s) in AMBER protein force field (in upper-case letters) is preferred. However, if there are missing bond/bond-angle/dihedral-angle parameters, general amber force field (GAFF) atom-type(s) (in lower-case letters) can be used instead [16]. 5. Add the two newly defined residues in $GMXLIB/residuetypes. dat file and the new atomtypes (if any) in $GMXLIB/ amber99sb.ff/atomtypes.atp file. 2.2.3 Preparing Gromacs Simulation Files

1. Add information related to the bond linking the two stapled side chains in the peptide in $GMXLIB/specbond.dat file, similar to Subheading 2.1.2, item 1. 2. Generate the initial structure file (.gro) and topology file (.top) using the pdb2gmx command, using the pdb file created in Subheading 2.2.1. 3. To use our RSFF2 force field, use the g_mod_top_RSFF2.py program to modify the topology file from Amber99sb force field to RSFF2.

2.2.4 Running, Preparing, and Production Simulations

The procedure is very similar to the simulation of head-to-tail CPs described in Subheading 2.1.4.

2.3 Methods for Conformational Analysis

From a simulation trajectory at 300 K, with the first 20 ns discarded as pre-equilibrium.

2.3.1 How to Obtain the Ramachandran Plot

Ramachandran (ϕ, ψ) plot is crucial for visualizing the backbone conformational sampling of a single residue or the whole CP. There are two simple yet different approaches: 1. To draw the plot, the whole 360  360 ϕ, ψ space is divided into 300  300 small bins, and the relative probability of each bin was calculated from the counts. 2. Estimate the probability density distribution using a 2D Gaussian kernel:   X Δ2 ðϕ; ψ; ϕi ; ψ i Þ pðϕ; ψ Þ ¼ exp  2σ 2 i Here (ϕi, ψ i) are backbone conformations of certain residue in ith structure frame sampled in a MD simulation. Larger σ parameter results in smoother p(ϕ, ψ) distributions. Δ2 is the

66

Fan Jiang and Hao Geng

squared distance between (ϕ, ψ) and the ith observed conformation, considering the periodicity of the dihedral angles:  2  Δ2 ðϕ; ψ; ϕi ; ψ i Þ ¼ min jϕi  ϕj; 360  jϕi  ϕj  2  þ min jψ i  ψ j; 360  jψ i  ψ j Use a logarithmic scale to visualize the probability distribution, such that contours are drawn at a regular free energy difference. 2.3.2 Clustering Analysis

The simplest approach is to use the clustering tool (g_cluster) within the Gromacs package. For example, we have performed clustering analysis on 10,000 snapshots, using the “gromos” method with 1.0 A˚ cutoff based on RMSD of backbone and Cβ atoms. However, the results are sensitive to the choice of RMSD cutoff, and each cluster does not usually correspond to one conformation. Thus, we have developed a new clustering method, with the following procedure. 1. Obtain 30,000 structure frames evenly sampled in the trajectory. 2. Cluster the 30,000 backbone (ϕ, ψ) values (dots in Ramachandran plot) of each residue into several conformational states based on the Rodriguez-Laio algorithm [17]. Distance dij between two date points (ϕi, ψ i) and (ϕj, ψ j) is defined as r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi    d ij ¼

ϕi  ϕ j

2

þ ψi  ψ j

2

The local density ρi around point i is defined as X   χ d ij  d c ρi ¼ j

where χ(x) ¼ 1 if x < 0 and χ(x) ¼ 0 otherwise, and dc is a cutoff distance, which is set to 5 . Basically, ρi is equal to the number of points within a radius of dc around point i. Then, we compute the minimum distance between point i and any other point with higher density:   δi ¼ min d ij j :ρ j >ρi

The all cluster centers (density maxima) were defined as points with δi > 60 degrees. After the cluster centers have been found, each observed (ϕ, ψ) point is assigned to the same cluster as its nearest neighbor of higher density. (ϕ, ψ) points in the same (ϕ, ψ) cluster belong to the same single-residuelevel conformation (Fig. 1a).

Computation of Cyclic Peptides Conformation

67

Fig. 1 (a) Clustering analysis of the ϕ, ψ plot of each residue based on the Rodriguez-Laio algorithm. Points are colored according to their assigned clusters. Clusters are numbered in order of decreasing populations. (b) Representative structure of each of the five most populated conformations, including conformation code (sequence of residue conformational states) and relative free energy

3. Then, an entire CP structure can be encoded as a string of the conformational states of all residues along the sequence (Fig. 1b). Two structures with the same sequence of singleresidue conformations are considered to have the same conformation. For CPs with repeated sequence, proper treatment of symmetry should be taken into account. 4. Finally, the population, relative free energy, and representative structure of each CP conformation can be obtained (Fig. 1b). The relative free energy of each conformation was calculated based on ΔG ðConformationi Þ ¼ RT

pðConformationi Þ pðConformation0 Þ

Here, p(Conformationi) is the probability of a certain conformation and p(Conformation0) is the probability of the most populated conformation. The representative structure of each CP conformation was chosen as the structure with maximal ρsum (the summation of local densities ρ of all its residues), among all structures that belong to this conformation.

68

Fan Jiang and Hao Geng

2.3.3 Calculation of Backbone Conformational Entropies

The conformational entropy was calculated based on the Boltzmann-Shannon entropy: S ¼ R

N X

pi ln pi

i¼1

where pi is the probability distribution in a conformational space divided into N equal-sized bins (a bin size of 10 can be used). Here, we calculate the conformational entropy of a single residue j, S(1)j, based on the its two-dimensional (ϕj, ψ j) distribution. To include the influence of neighboring residue, we also calculated the conformational entropy of residue pairs: S(2)j-1,j and S(2)j,j+1 based on the four-dimensional distributions (ϕj, ψ j, ϕj, ψ j) and (ϕj, ψ j, ϕj, ψ j), respectively. Finally, the reported residue-specific conformational entropy for jth residue is [18] h i S j ¼ S ð1Þ j þ S ð2Þ j 1, j þ S ð2Þ j ,j þ1 =2:

3

Notes

3.1 Residue-Specific Force Fields (RSFFs)

The development of a force field involves the balance between accuracy, efficiency (speed), and transferability. For the simulation of biomolecules such as cyclic peptides, rather high accuracy in the description of the conformational energetics is needed. Commonly used force fields also have problems in describing the different intrinsic conformational propensities of different amino acid (AA) residues [19]. On the other hand, for simulation speed and ease of implementation and parameterization, it would be very beneficial to increase the force field accuracy without sacrificing its efficiency. Therefore, we choose to sacrifice the transferability by using different force field parameters for different AA residues, thanks to the very limited chemical space of these building blocks of peptides/proteins. Also, we are interested in the behaviors of peptides/proteins in aqueous solution, which are quite different from those in vacuum. The parameterization of RSFFs is based on the conformational free energy surfaces of short peptides in aqueous solution. Due to the lack of high-resolution free energy surfaces from direct experiments, we use the statistics of many protein crystal structures instead. Specifically, we constructed the so-called coil library from the residues outside the secondary structure regions (helices, sheets, turns), to approximate the intrinsic conformational preferences of AA residues without the influence of inter-residue backbone hydrogen bonding. All the bond stretching, bond-angle bending, van der Waals interactions (Lennard-Jones potential), and electrostatic interactions (Coulomb potential) in RSFF1 [20] and RSFF2 [21] are

Computation of Cyclic Peptides Conformation

69

adopted from the OPLS-AA/L and AMBER-99SB force fields, respectively. On the other hand, the dihedral-angle (torsion) potentials for all the rotatable bonds were reparameterized using the statistical free energy surfaces (potential of mean force, PMF) for protein coil library as reference, such that the backbone ϕ, ψ, and side-chain χ PMF obtained for dipeptide simulations agree well with the reference data. Unlike commonly used protein force fields, special parameters were used for some 1–5 and 1–6 van der Waals interactions to optimize the coupling between neighboring torsions. The RSFF1 force field has been parameterized with the TIP4PEw water model, and the RSFF2 has been parameterized with the TIP3P water model. Thus, they are usually used with these two water models, respectively. Simulations of all dipeptides (blocked amino acids, Ac-X-Nme) using RSFF1 and RSFF2 give 3JHNacoupling constants in excellent agreement with NMR measurements, much better than some recent force fields. Also, an independent benchmark study using 256 two-residue peptides (Ac-XY-NH2) has shown that RSFF2 can improve the modeling of sequence-specific conformational behavior of peptides [22]. 3.2 Cyclization of the Peptide

Here we use the pdb2gmx tool to automatically generate the cyclizing bond and associated angles and dihedral angles, with the information of the linkage added in the $GMXLIB/specbond.dat file. An alternative approach is to add all the bond, angles, and dihedral angles manually in the topology file. However, this is much more complicated and time consuming.

3.3 Treatment of D-Amino Acids

Unnatural D-amino acids, which are mirror images of natural (proteinogenic) L-amino acids, have often been used in designing cyclic peptides. However, in most popular protein force fields and our RSFFs, all dihedral-angle potential functions are even functions, and all distances and angles are not affected by enantiomeric inversion. Thus, RSFF1 and RSFF2 are expected to be equally applicable to both L- and D-enantiomers of an amino acid. Our own simulation studies strongly support this point.

3.4 Peptide Bond Cis–Trans Isomerization

In high-T (>500 K) MD simulation, cis–trans isomerization of peptide bond can occur, especially for that preceding proline [23]. This can also happen in a REMD simulation if T of the highest-T replica is too high. We have found that setting the highest T to 500 K can prevent cis–trans isomerization for OPLS-AA/L, AMBER-99sb, RSFF1, and RSFF2 force fields.

3.5 Treatment of Symmetry for CPs with Repeated Sequences

Some CPs have repeated sequences, such as cyclo-(aGPFaGPF) (“a” for D-alanine) and cyclo-(VPG)4. In the conformational clustering analysis of these CPs, equivalent conformational sequences are considered as identical CP conformation. An example: for

70

Fan Jiang and Hao Geng

cyclo-(aGPFaGPF), 11222222 and 22221122 belong to one conformation. Another example: for cyclo-(VPG)4, 111112113112 and 112113112111 belong to one conformation. For these cyclic peptides, the RMSD of a certain structure with respect to corresponding crystal structure was calculated as the lowest one in all possible structure alignments: for example, using two alignments for cyclo-(aGPFaGPF) and four alignments for cyclo-(VPG)4. References 1. Craik DJ, Fairlie DP, Liras S, Price D (2013) The future of peptide-based drugs. Chem Biol Drug Des 81:136–147 2. Geng H, Jiang F, Wu YD (2016) Accurate structure prediction and conformational analysis of cyclic peptides with residue-specific force fields. J Phys Chem Lett 7:1805–1810 3. Rayan A, Senderowitz H, Goldblum A (2004) Exploring the conformational space of cyclic peptides by a stochastic search method. J Mol Graphics Model 22:319–333 4. Beaufays J, Lins L, Thomas A, Brasseur R (2012) In silico predictions of 3D structures of linear and cyclic peptides with natural and non-proteinogenic residues. J Pept Sci 18:17–24 5. Goldtzvik Y, Goldstein M, BennyGerber R (2013) On the crystallographic accuracy of structure prediction by implicit water models, tests for cyclic peptides. Chem Phys 415:168–172 6. Das R (2011) Four small puzzles that Rosetta doesn’t solve. PLoS One 6:e20044 7. Damas JM, Filipe LCS, Campos SRR, Lousa D, Victor BL, Baptista AM, Soares CM (2013) Predicting the thermodynamics and kinetics of helix formation in a cyclic peptide model. J Chem Theory Comput 9:5148–5157 8. Razavi AM, Wuest WM, Voelz VA (2014) Computational screening and selection of cyclic peptide hairpin mimetics by molecular simulation and kinetic network models. J Chem Inf Model 54:1425–1432 9. Yu H, Lin YS (2015) Toward structure prediction of cyclic peptides. Phys Chem Chem Phys 17:4210–4219 10. Lama D, Quah ST, Verma C, Lakshminarayanan R, Beuerman RW, Lane DP, Brown CJ (2013) Rational optimization of conformational effects induced by hydrocarbon staples in peptides and their binding interfaces. Sci Rep 3:3451 11. Joseph TL, Lane DP, Verma C (2012) Stapled BH3 peptides against MCL-1: mechanism and

design using atomistic simulations. PLoS One 7(8):e43985 12. Sugita Y, Okamoto Y (1999) Replica-exchange molecular dynamics method for protein folding. Chem Phys Lett 314:141–151 13. Prakash MK, Barducci A, Parrinello M (2011) Replica temperatures for uniform exchange and efficient roundtrip times in explicit solvent parallel tempering simulations. J Chem Theory Comput 7:2025–2027 14. Hu K, Geng H, Zhang Q, Liu Q, Xie M, Sun C, Li W, Lin H, Jiang F, Wang T, Wu YD, Li Z (2016) An in-tether chiral center modulates the helicity, cell permeability, and target binding affinity of a peptide. Angew Chem Int Ed 55:8013–8017 15. Wang J, Cieplak P, Kollman PA (2000) How well does a restrained electrostatic potential (RESP) model perform in calculating conformational energies of organic and biological molecules? J Comput Chem 21:1049–1074 16. Wang J, Wolf RM, Caldwell JW, Kollman PA, Case DA (2004) Development and testing of a general amber force field. J Comput Chem 25:1157–1174 17. Rodriguez A, Laio A (2014) Clustering by fast search and find of density peaks. Science 344:1492–1496 18. Baxa MC, Haddadian EJ, Jha AK, Freed KF, Sosnick TR (2012) Context and force field dependence of the loss of protein backbone entropy upon folding using realistic denatured and native state ensembles. J Am Chem Soc 134:15929–15936 19. Meral D, Toal S, Schweitzer-Stenner R, Urbanc B (2015) Water-centered interpretation of intrinsic pPII propensities of amino acid residues: in vitro-driven molecular dynamics study. J Phys Chem B 119:13237–13251 20. Jiang F, Zhou CY, Wu YD (2014) Residuespecific force field based on the protein coil library. RSFF1: Modification of OPLS-AA/L. J Phys Chem B 118:6983–6998

Computation of Cyclic Peptides Conformation 21. Zhou CY, Jiang F, Wu YD (2015) Residuespecific force field based on protein coil library. RSFF2: Modification of AMBERff99SB. J Phys Chem B 119:1035–1047 22. Li S, Elcock AH (2015) Residue-specific force field (RSFF2) improves the modeling of

71

conformational behavior of peptides and proteins. J Phys Chem Lett 6:2127–2133 23. Neale C, Pomes R, Garcia AE (2016) Peptide bond isomerization in high-temperature simulations. J Chem Theory Comput 12:1989–1999

Chapter 5 Computational Opportunities and Challenges in Finding Cyclic Peptide Modulators of Protein–Protein Interactions Fergal Duffy, Nikunj Maheshwari, Nicolae-Viorel Buchete, and Denis Shields Abstract Peptide cyclization can improve stability, conformational constraint, and compactness. However, apart from beta-turn structures, which are well incorporated into cyclic peptides (CPs), many primary peptide structures and functions are markedly altered by cyclization. Accordingly, to mimic linear peptide interfaces with cyclic peptides, it can be beneficial to screen combinatorial cyclic peptide libraries. Computational methods have been developed to screen CPs, but face a number of challenges. Here, we review methods to develop in silico computational libraries, and the potential for screening naturally occurring libraries of CPs. The simplest and most rapid computational pharmacophore methods that estimate peptide threedimensional structures to be screened versus targets are relatively easy to implement, and while the constraint on structure imposed by cyclization makes them more effective than the same approaches with linear peptides, there are a large number of limiting assumptions. In contrast, full molecular dynamics simulations of cyclic peptide structures not only are costly to implement, but also require careful attention to interpretation, so that not only is the computation time rate limiting, but the interpretation time is also rate limiting due to the analysis of the typically complex underlying conformational space of CPs. A challenge for the field of computational cyclic peptide screening is to bridge this gap effectively. Natural compound libraries of short cyclic peptides, and short cyclized regions of proteins, encoded in the genomes of many organisms present a potential treasure trove of novel functionality which may be screened via combined computational and experimental screening approaches. Key words Peptide, Cyclic peptide, Cheminformatics, Bioinformatics, Molecular dynamics, Computational biology, Virtual libraries

1

Introduction Cyclic peptides are derivatives of linear peptides where the linear peptide has been pinned into one or more macrocycles by the addition of chemical bond(s). These bonds can be between amino acid side chains, or the peptide N- and C-termini, or a combination thereof. Cyclic peptides are of interest as a class of molecules for their ability to mimic specific, high-affinity binding of certain known linear peptides while potentially avoiding the drawbacks of

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_5, © Springer Science+Business Media, LLC, part of Springer Nature 2019

73

74

Fergal Duffy et al.

linear peptides, which include poor oral bioavailability, poor membrane permeability, vulnerability to proteolytic degradation, and lack of a rigid three-dimensional structure. Different chemical strategies can be used to create cyclized peptides. These include (1) disulfide bonding via the thiol side chain of two cysteines, common in natural proteins; (2) head-tail bonding, where the peptide N-terminus forms a peptide bond with its C-terminus, effectively removing the peptide termini; (3) N-terminus to acidic side chain; (4) C-terminus to basic side chain; (5) side chain to side chain; and (6) more complex modifications, such as the cysteine bridging to serine or threonine seen in lanthipeptides. While cheaper and easier to synthesize, disulfide bonds are unstable in the reducing intracellular environment. Typically cyclic peptides are at least four amino acids in length to be practically synthesizable, with the exception of the special structure of the diketopiperazines lactam ring formed from two amino acid head-tail-bonded peptides [1]. This review further develops aspects of computational cyclic peptide analysis we previously reviewed a number of years ago [2], and considers what are the key challenges and opportunities for this field into the future.

2

Cyclic Peptide Role in Drug Discovery Many drugs are small molecules that bind within a deep protein cavity. However, it is often more difficult to find small molecules that bind the large, relatively flat surfaces involved in protein interfaces [3]. In silico screening of compounds that can modulate protein interactions requires consideration of the features of the typical binding pockets [4]. Small-molecule drugs are typically planar molecules of low molecular weight with simple stereochemistry, and larger and more complex molecules may be required to effectively target protein–protein interactions [5]. The PPI (Protein–Protein Interaction Inhibition) library was designed with this aim in mind [6]. Against this background, cyclic peptides may offer useful features for certain target sites. Oligopeptides (with roughly 2–20 amino acids) are intermediate in size between small-molecule drugs and large therapeutic molecules such as antibodies [7] and proteins [8, 9]. They are a promising drug choice, given that their building blocks and degradation products are chemically identical to natural proteins. However, their short length means they often lack fixed secondary structural elements, negatively affecting peptide stability versus proteolysis, and also negatively affecting binding affinity [10], in part due to the entropic cost of fixing a flexible molecule into a defined shape on binding to a protein. Macrocyclic drugs [11, 12], which have ring structures of larger than 8–12 atoms, overcome some of these limitations. These include natural products [13], synthetic peptidomimetic macrocycles [14], stapled peptides [15], and CPs.

Computation for Cyclic Peptides PPI Modulators

3

75

Properties of Cyclic Peptides In principle, conformational constraint is beneficial not only in reducing the computational space that must be searched when finding optimal structural solutions that prioritize the most bioactive CP of interest from a library, but also in improving drug specificity compared to less constrained molecules which may adopt many conformations in vivo, which may contribute to adverse events [16]. Drug-target binding may be influenced by enthalpy, entropy, or both [17]. Oral bioavailability [18, 19] is useful to predict as well as target affinity. Cyclic peptides have specific features relating to membrane permeability that distinguish them from linear peptides [20]. CPs typically have reduced proteolytic sensitivity, extending half-life, and certain modifications can improve transmembrane permeability, even though cyclic peptides are larger than typically predicted for orally bioavailable drugs [21]. Rezai et al. [22] optimized internal hydrogen bonding in membrane-permeating cyclo [Leu-Leu-Leu-Leu-Pro-Tyr] CPs, with membrane diffusion rates corresponding to the degree of intramolecular hydrogen bonding. Selective N-methylation of backbone amide groups in cyclic peptides [23] can improve membrane permeability by “hiding” hydrophilic amides [24, 25]. The resulting alterations in structural conformation may, however, either decrease or increase specificity [26, 27], often altering the preferential location for insertion of a beta-turn that accommodates the cyclic constraint. Membrane permeability may also be improved by covalent attachment of a polyethylene glycol (PEG) group [28–30], or by considering the ability to target transport mechanisms such as PEPT1 and PEPT2 transporters [31] or endocytosis. Conformational constraint can prevent degradation by proteases [32, 33], and nonnatural amino acids, in particular D-enantiomer amino acids, can prevent protease recognition in both linear [34] and cyclic peptides [35]. Thus, conformational libraries of CPs may consider not only the amino acid sequences designed to confer specificity of interaction with the target site, but also chemical modifications designed to improve bioavailability.

4

Cyclic Peptides and Protein–Protein Interactions Protein–protein interactions (PPIs) pose difficulties for traditional small molecules, often due to the large, shallow interfaces [36], and the dynamic instability of transient pockets [37]. Nevertheless, small-molecule protein–protein interaction inhibitors have been identified [3, 38], and those with structural information available are curated in the 2P2I database [39]. A typical protein–protein

76

Fergal Duffy et al.

˚ 2, but may be much larger interaction surface is roughly 1600 A than a cyclic peptide can engage with [40], focusing attention on hot spots of binding energy [41]. Cyclic peptides can conformationally mimic key features of one binding surface, such as β-turns [42, 43], which are ligands for over 100 G protein-coupled receptors [44]. The PepX [45] database curates a set of protein-binding peptides. Cyclic peptides cannot always be used to mimic a linear peptide, as linear peptides often bind in an extended linear conformation [46] incompatible with the geometry of a cyclic peptide. Nevertheless, the set of protein-binding peptides contains promising lead compounds for cyclic peptides to mimic, such as protease inhibitors [47]. Recurring domain-motif interactions, curated on the Eukaryotic Linear Motif (ELM) database [48], typically involve globular protein interaction with 3–12 residue short linear motifs (SLiMs) in disordered protein contexts. They promote PPIs involved in protein targeting, modification, cleavage, and signaling. Protein–peptide interactions, of which SLiMs are a subset, comprise an estimated 15–40% of protein interactions [49]. SLiM-mediated interactions are smaller than most protein–protein interactions making them more tractable for drug targeting. As well as a number of antibiotics, cyclic peptide drugs include the angiogenesis inhibitor cilengitide [50], developed from the linear RGD motif, and peptides targeting the SH3 domain [51]; the somatostatin mimic octreotide [52]; and linaclotide agonist of guanylate cyclase 2C used for IBS abdominal pain [53]. Also of note is the linear amylin-mimetic davalintide, to reduce food intake [54] (Fig. 1). The observed dataset of cyclic drugs typically has molecular weights of 1–2 kDa. For computational peptide analysis, these databases of sequences, structures, and compounds relating to interactions, motifs, and peptides provide a potential reference base for screening of compounds, and for control structures and sequences when developing libraries and decoy libraries for different search strategies. They also provide guidance when deciding sensible constraints and limits to use when designing in silico compound libraries. Smaller bioactive cyclic peptides can range in size from 5 backbone amino acids in the case of octreotide and lypressin to 12 backbone amino acids in the case of cyclosporin [2]. Larger cyclic compounds take on some of the properties of small proteins, frequently with clear secondary and tertiary structural organization. While side chain functionality is important, the constrained backbone often plays a key role in binding interactions, as shown by the backbone hydrogen bonding interactions of cyclosporin with its target, human cyclophilin A [55].

Computation for Cyclic Peptides PPI Modulators

A

77

B H2N NH

HN

H NH 2

O H

H N

CH 3

H

O

H

O

H N

NH O

CH33 CH

O

H N

H N H

H

N H HO

CH 3 H N

HNH N H

O

O

S

O HN

H

H

HO

NH

S

N H

N H

H

O

O

H

O NH 2

H

O H3C

O

H

OH

OH

C

D

HO H

O

CH 3

H2N

HO OO H HN

O H

H N

H2N

H

O

H SH S N N

O O

H N

H

H

H NH

H

O O

H HN

N H

O

N H HO

H

H N

H CH 3 H

H3C

H CH 3

O O O

H N

H

H

O

CH 3 O

O

NH2

CH 3

NH

HO

H

H N

HN H N

N H

H

O H

HO

H

O H3C

NH H

O H

CH 3 O

O

CH 3 NH

HN

H

H H2N

NH

HN

N NH 2

NH

O

NH2

O

NH 2

H HN

N H O H

H HO

CH 3

H N

HN

H N

H N

O

O

HN

CHO3

NH

N H

N

H

O

H

H

H

O

O

O

OH

O HO

N O H CH 3 O

H3C

H N

OH

N H O

H

O

H2N

N H O

O H

H HN

H

H

H3C H N

HN

NH

O

O

NH

OH

OH

H3C O

N H NH 2

S H

H2N

H

O

O

CH 3

H O

H

NHH

H

CH 3

HO

CH 3

HN

HO

OH

O NH

S

O

S

H

S

H NH

NH

S

O H3C NH

N H

H

O

H S

O

H

H N

H

H2N

O

H2N

O

H

H

O

NH OH

O

H2N O OH

Fig. 1 Two-dimensional structures of peptide inhibitors of protein–protein interactions. (a) Cilengitide, (b) octreotide, (c) linaclotide, (d) davalintide

5

Experimental Screening for Cyclic Peptides Phage display enables the display of libraries of 1010 peptides simultaneously [56], with typical peptide sizes between 5 and 20 residues [57], and has been used to successfully identify highaffinity disulfide-bonded cyclic peptides [58], typically using combinatorial libraries, with or without partial sequence constraints. Suga has pioneered highly successful combinations of in vitro translation and mRNA display to screen libraries for larger cyclic peptides allowing nonnatural amino acids [59]. SICLOPPS [60] (split intein-mediated circular ligation of peptides and proteins) can recombinantly generate head-tail-bonded cyclic peptides inside a cell, allowing screening for a CP with desired bioactivity [61]. Natural sources contain a rich diversity of cyclic peptide, and cyclic

78

Fergal Duffy et al.

peptide-like structures. Natural cyclic peptides come from two principal sources—they can be synthesized, like proteins, from DNA, or they can be non-ribosomal peptide natural products, synthesized by specialized non-ribosomal peptide synthetases in microorganisms like bacteria and fungi which can incorporate a great variety of nonnatural amino acids and posttranslational modifications and possess a vast chemical diversity [13]. Non-ribosomal cyclic peptides are the principal source of cyclic peptide antibiotic structures such as tyrocidine [62] and daptomycin [63]. The cyanotoxins include CPs, such as microcystins [64], which inhibit protein phosphatases type 1 and 2A, and nodularins [65]. Non-ribosomal cyclic peptides also include anticancer drugs, such as the epothilones [66] (Fig. 2).

A

H2N

O

B

NH 2 O O H H N

H

O

H N

O

O

OH

H O

HO

HN

H

H

H

O

H N

H N

NH2 H

O

NH

O

NH

O

O

HN

H

HO

H HN

O O

NH

O

H

O

H

HN

N H

H

H

O

N H

NH 2

O

H O

H H2N

D H N

N

O

O

O

HO

O O

H

O

N H

N H

O

O

O

C

H

H N

N H

O H

H

H

N H

O

O

HN

HO

H

O

O

N H2N

NH

H

HO N

H O

OH

NH HN

H2N

NH2

H O

N

OH NH

O

H

H

O

HN

H

O

H N

O H

H O

NH

O

HN

H

H H

N H

H

H

O

H

O

O HO

O H

H

OH O

H O

H

O

NH

H N H

HN

NH

HO

N H2N

H

O

O

NH 2

E O H H

H

H

S N

OH

O

H O

H H

O OH

Fig. 2 Examples of cyclic peptides derived from natural products. (a) Tyrocidine, (b) daptomycin, (c) microcystin-LR, (d) nodularin-R, (e) epothilone A

Computation for Cyclic Peptides PPI Modulators

6

79

Virtual Screening Ligand-based screening identifies molecules similar to already known ligands, and includes fingerprinting methods, pharmacophore matching, and shape-based matching. Fingerprint methods represent the molecule with a bit string, with each bit representing either the presence or the absence of a chemical fingerprint (structure based [67]), or a numerical value to the atomic and bonding properties of linear substructures of the molecule that is passed through a hash function to create a bit string (hash based [68]). Molecule comparison is done by calculating the number of shared “on” bits in the fingerprints divided by the total “on” bits in both (the Tanimoto score). In pharmacophore matching, the molecule is broken into a set of points or volumes in space representing a chemical feature, such as a hydrogen bond donor/acceptor, +/ charge, lipophilic regions, and aromatic groups. A query pharmacophore model can be built from features of a known ligand known to be important for binding, and the matching score is based on a feature volume overlap score between the query pharmacophore and the candidate ligand features. Schemes may be simple points surrounded by spheres, or more complex, for example allowing the plane of aromatic groups to be identified [69–72]. In shape matching, the similarity of two molecules is compared based on their threedimensional shape, either by aligning 3D structures of the molecules and calculating the root mean squared deviation (RMSD) or by comparing statistical measures of molecular shape [73–76]. Ligand-based screening methods can be extremely quick, especially those that reduce a molecule to a bit string of 0s and 1s representing its chemical properties, as comparing bit strings is very fast computationally. Bit strings can be pre-calculated for large libraries of candidate compounds, allowing them to be easily rescreened against many true ligand structures. In the absence of structural information on the target protein, ligand-based methods have been shown to be just as accurate as structure-based methods [77], although this may be down to an imperfect understanding of how to design a docking scoring function that works well across diverse target types [78]. Structure-based screening is based on exploiting the known three-dimensional structure of the target and the topology of its ligand-binding surface to design or choose possible active molecules, often by attempting to “dock” the prospective ligand molecule via iterative pose generation and scoring, into a predefined binding site on the target. Scoring methods may be empirical, counting favorable interactions or change in solvent-accessible area [79, 80]; or a molecular mechanical force field, such as AMBER, to estimate binding affinities based on van der Waals,

80

Fergal Duffy et al.

hydrogen bond, and charged contacts [81]; or knowledge based, assessing docking score based on the statistical similarity of docked conformations to known protein ligand structures [82, 79], such as those in the Protein Data Bank [83]. Pose generation may systematically explore rotating bonds, often followed by sampling a diverse subset of generated structures [84]; or may incrementally construct poses from fragments of the input molecule [81, 85]; or may use genetic algorithms with the docking score as the fitness function, and representing features as a linear gene that are mutated, recombined, and rescored iteratively [86]; or may use Monte Carlo approaches to statistically sample random modifications to approach a solution [80]. Docking approaches may be rigid or flexible approaches—rigid docking is very fast, but less accurate. Rigid-body docking has been used by Mosca et al. [87] to accurately identify interacting proteins in the Saccharomyces cerevisiae interactome. Generally, when using virtual screening to find protein ligands, a flexible ligand—static protein or static protein backbone model—is used. Yuriev at al. [88] have reviewed different docking approaches and challenges in detail. However, there has been no systematic evaluation of which methods are best suited to cyclic peptide ligands, which share some of the properties of linear unconstrained peptides, and some of the properties of more rigid small-molecule ligands. Virtual screening software includes commercial, such as MOE [70], OEChem [89], the Schro¨dinger [90] suite of programs, Accelrys Discovery Studio [72], and open-source toolkits [91] such as RDKit [92], OpenBabel [93], and the Chemistry Development Kit [94] (CDK). Commercial virtual screening approaches generally provide a complete graphical environment. While the open-source equivalents usually require the user to construct their workflow using a computer scripting language such as Python, there have been efforts to make more user-friendly open-source screening tools, such as the Knime [95] workbench.

7

Combinatorial Library Generation Virtual screening requires the known structure of a macromolecule such as a protein, and a library of candidate compounds. Available small-molecule virtual libraries include those curated by the ChEMBL and ZINC [96] databases, and pharmaceutical companies generally curate their own compound collections. For cyclic peptide compounds, the most straightforward method of assembling a library is by combinatorial generation, assembling possible amino acid sequences in a cyclic peptide library. Burns et al. [97] docked cyclic peptide virtual libraries to find RNA-binding partners. Nonnatural amino acids and modifications that may be difficult or expensive to synthesize may be screened in virtual libraries,

Computation for Cyclic Peptides PPI Modulators

81

and it is then only necessary to synthesize compounds that are among the top hits. A large diversity of amino acid structures are commercially available, for example on the ZINC database. Virtual libraries allow basic validation of possible compounds before any complex chemistry takes place. We have developed CycloPs [98] software designed for the generation of virtual libraries of cyclic peptides, which can incorporate a variety of cyclic peptide constraint strategies, as well as user-defined amino acid structures (allowing, for example, the inclusion of amino acids including posttranslational modifications in the library), and the ability to filter out cyclic peptides that are likely to be difficult to synthesize.

8

Structural Optimization 2D virtual screening represents the molecule as a graph structure of atoms joined by bonds, and calculates molecular similarity based on substructures in these graphs, or by the various possible paths through the graph. 3D approaches, such as pharmacophore matching, use the predicted three-dimensional shape of the molecule to score hits. Due to the conformational restraint of cyclic peptides, accurate solvation structures may be predicted. The DEEPSAM structure prediction algorithm from the Tinker [99] molecular modeling package was used to accurately predict the solution structures of a set of five small cyclic peptides [100]. However, the potential conformational change upon binding of a cyclic peptide to its target means that it is useful to predict a range of likely conformations for a cyclic peptide to be used in any rigid-body virtual screening step, rather than attempting to predict one single shape. The constrained central ring structure of cyclic peptides reduces the number of poses to consider, not only speeding computation but also likely reducing error, in contrast to linear peptide screening. Most software packages have built-in routines to generate the set of rigid conformers of a molecule to screen versus a target. Some conformer generation software designed for small molecules [101] generate conformers by systematically rotating flexible bonds, and are not capable of varying macrocyclic rings, but others permit this; for example the loop prediction software in the Protein Local Optimization Program [102] was used to predict the solution structures of cyclic hexapeptides [22]. The RDKit [92] cheminformatics library may be used to combine stochastic conformer generation and subsequent optimization that can generate diverse, low-energy conformations, including varying ring structures [103].

82

9

Fergal Duffy et al.

Screening Cyclic peptides are often very structurally similar, with a large shared peptide backbone and a limited set of side chain groups, reducing the power of substructural searches to discriminate between structures. When screening against cyclic peptide libraries, fingerprint approaches are not suitable, failing to discriminate between cyclic peptides with a different sequence, but the same amino acid composition (such as CGVPRRC and CRVGPRC). Two-dimensional virtual screening methods have proved as effective as three-dimensional methods for small molecules [104], which are typically planar molecules without complex stereochemistry or structure [5]. However, cyclic peptides are larger, and more three dimensional than drug-like small molecules. This makes three-dimensional screening methods, such as docking, or threedimensional pharmacophore matching more appropriate. Pharmacophore matching is a ligand-based screen, with key pharmacophore points taken from a 3D structure of a known ligand interacting with the protein target. It is also possible to include structural information by the use of exclusion volumes, where the candidate molecule must match the key pharmacophore features of the known ligand, while avoiding steric clashes with the protein target. We used the known 3D structures of short peptide sequences, including 57 short linear motifs, 161 protein-binding peptides, and 154 turn structures at protein–protein interfaces to create a set of 3D pharmacophore templates for virtual screening [105]. These were used to rapidly computationally screen multiconformer libraries of ~100,000 random 5- and 6-residue disulfidebonded peptides. Top hits were enriched for matches to protein turn structures, including several of potential clinical utility. In particular, many hits were identified to loops located on the surface of thrombin, a serine protease that is a key player in the coagulation pathway. Synthesis and testing of the top 20 hits to this loop revealed a cyclic peptide that significantly slowed thrombininduced platelet activation in vitro, and showed moderate binding affinity via surface plasmon resonance [42].

10

Computational Road Blocks While ligand-based screening using fixed conformer libraries is rapid, even this can run into issues when very large libraries are screened, especially if decoy or alternative targets are screened simultaneously. A typical therapeutic cyclic peptide has 5–16 amino acids. To fully explore the set of head-tail 10-mer cyclic peptides combinatorially would require generating multiple conformers for over 1013 cyclic peptide structures—a prohibitively

Computation for Cyclic Peptides PPI Modulators

83

large amount. Knowledge of the target and ligands can help in reducing this complexity, by preselecting a restricted library of amino acids, or limiting the number of variable positions in a peptide. To make ligand-based screening methods more efficient, evolutionary algorithms (also known as genetic algorithms) used to explore ligand flexibility in docking [106–108] may also be applied to screen combinatorial libraries [109], to design de novo molecules from fragments [110, 111], or by pseudo-retrosynthesis, where a molecule is broken up into building blocks which can then be recombined [112]. This approach is equally well suited to peptide design [113–115]. Cyclic peptides may be represented as a sequence string, which can be converted into a chemical structure using a tool such as CycloPs [98]. Fitness functions for selection could be the standard pharmacophore screening [69–72], and shape matching [73–76] functions. Matching compounds to target structures faces substantial computational burdens. Probing their conformational space and the possible peptide-peptide interactions is extremely challenging with current computational resources and quantitative (though, not exhaustive) studies are necessarily relying on hybrid coarsegrained (i.e., such as docking) approaches for screening the complex underlying conformational space and atomistic (typically MD-based) approaches for validating and analyzing in more detail the results of the coarse-grained studies. One solution is to pre-screen with a pharmacophore-based screen, and only dock the best hits. Another approach is to limit the rotatable bonds on the cyclic peptide. Molecular dynamics (MD) models, which are more accurate and informative than docking or pharmacophore approaches, are extremely computationally intensive, and also require substantial interpretation and manual checking of results in order for the gain in accuracy to be typically realized, making them difficult to implement in high-throughput computational screens. In spite of their intrinsic limited accuracy, docking methods are emerging as useful techniques that can enable MD-based studies of complex molecular systems of biomedical interest for which all-atom modeling approaches are typically too expensive to apply on their own. For example, protein docking programs such as Zdock [116] may generate initial structures for further molecular dynamics refinement (e.g., for native and inhibitory peptides targeting interaction surfaces such as SARAH domain interactions [117]), or docking programs optimized for small-molecule binding such as Autodock Vina [80] may be used to dock peptides to enzyme-active sites or other pockets (such as screening for short peptide targeting [118]). While it might be anticipated that small highly constrained cyclic peptides with few rotational bonds may be better modeling with docking methods optimized for smallmolecule docking, this has yet to be proven, as there has been little evaluation of the efficacy of docking approaches for cyclic peptide

84

Fergal Duffy et al.

compounds, along the lines of benchmarking using small-molecule compounds [78, 119]. Establishment of benchmarking datasets would greatly enable the advancement of understanding of what tools are best suited to the docking of cyclic peptides, which are intermediate in size between small molecules and proteins. The computational road blocks at all three current levels of cyclic peptide screening (pharmacophore, docking, and MD simulation) are shared by more general compound docking, but may be more critical for cyclic peptides, where the relative homogeneity of the compound libraries (different sequences of 20 or more amino acids) may flatten the matching distance between the best fitting solutions and the various nearest alternatives in the search library, making it harder to resolve true binders at each level. For this reason, it may be more critical for virtual cyclic peptide library screening to develop hybrid computational approaches that bridge the gap between the rapid pharmacophore matching and docking, and that bridge the gap between the higher throughput docking and the very-low-throughput MD simulation. One way to alleviate the problems inherent to the high complexity of the conformational space of atomistic models of peptides is to study their dynamics using either coarse-grained or atomistic methods, but to map their dynamics on simplified kinetic models of transitions between well-defined metastable conformational states of the system of interest. These approaches are referred to as “dynamic coarse graining” methods and are more systematic (and thus more computationally demanding as well) than docking or spatial based (e.g., united residues) studies. They map the highdimensional (e.g., atomistic dynamics of molecular systems) on a coarse-grained network of conformational transitions between states that are assumed (and tested) to be Markovian in nature, allowing one to use coarse master equations or the associated Markov state models [120] to quantify exhaustively the system dynamics and a relevant level of its coarse-grained representation [121]. While the use of atomistic MD with explicit water molecules still limits dramatically the possibility to describe exhaustively the conformational space of even small peptides, the use of master equations and MSMs enables more systematic studies as shown recently in the application to dimeric and tetrameric peptides [122, 123]. We have recently been investigating similar approaches to analyze and characterize the conformational dynamics of cyclosporin A (CsA), a medium-sized (11 residues long) cyclic peptide widely used as an important immunosuppressant drug. Atomistic MD studies in explicit water molecules were performed and used to characterize and compare the differences in the conformational dynamics of N-methylated (CsA) and de-methylated (dCsA) versions of this cyclic peptide in order to understand the effects of N-methylation and its implications in CsA propensity to interact with lipids and to translocate through lipid membranes—an

Computation for Cyclic Peptides PPI Modulators

85

important factor in their pharmacological role [124]. Here, datadriven methods, such as principal component analysis (PCA), were used to cluster the conformational states of these cyclic peptides, providing an increasingly automatic method of state identification. Representative structures obtained in this way were analyzed further using the coarse master equation-based kinetic analysis method. This formalism—where kinetic analysis is implemented and rate matrices obtained using either propagators (i.e., transition probabilities for fixed time intervals) or lifetime-based methods— can be used for a systematic comparison of the conformational dynamics of cyclic peptides in MD simulations in explicit water, under various conditions, quantifying the mechanisms and effects of N-methylation and temperature [124]. One approach that has been effectively used to bridge the gap between docking and more computationally expensive peptide modeling is Rosetta FlexPepDock [125]. This has been shown to be able to retrieve near-native peptide conformations in a variety of docking experiments. It is, however, significantly more computationally expensive than other docking approaches. London et al. [126] have used this approach to test peptides binding to Bcl-2, and validated their results using peptide arrays. Mandal et al. have used docking to model the interaction of conformationally constrained phosphopeptides to the SH2 domain of the signal transducer and activator of transcription 3 (Stat3) protein—involved in aberrant growth in malignant tumor cells [127]. The Rosetta modeling framework has also been used to design peptides filling particular spaces [128], but this has been for larger peptides. It has also been used to systematically predict the conformations of short 7–14-mer cyclic and bicyclic peptides, followed by NMR validation of predicted structures [129]. This virtual structural library of cyclic peptide compounds provides an important resource for further analyses, for experimental peptide library development, for further virtual screening, and for scaffold development of novel peptidomimetics. The computational design of in silico libraries may be improved by limiting to cyclic peptides that are predicted to form a stable structure, and structural modeling [130] can thus define the subset of peptides that may be of more interest for further experimental analysis, as their structures are inherently more predictable, making computational matches to mimicked surfaces more realistic.

11

Computational Mining of Cyclic Peptides from Biological Sources An alternative approach to combinatorial virtual libraries is to take inspiration from nature [131], assembling cyclic peptide libraries from biological sources, thus exploiting the diversity of cyclic peptides that have developed through evolution in the genomes of

86

Fergal Duffy et al.

diverse organisms. Natural cyclic peptides provide a useful source of guidance for virtual screening. Focusing libraries on structures based on those found in nature is a method of stacking the deck of peptides toward bioactivity while retaining a manageable number of structures. Protein structures may be mined to find “self-inhibitory” peptides, where a peptide derived from a protein–protein interface inhibits the formation of that interface, and it has been observed that many protein–protein interaction surfaces are dominated by short segments of peptides [132]. More recently, this approach was used to identify peptides that inhibit viral membrane fusion [133]. While these studies principally examined short linear peptides, the same principles can be used to identify cyclic peptides, either by cyclizing bioactive linear peptides to improve bioavailability or by searching for bioactive peptides with natural cyclic shapes, derived from loop or turn regions of protein secondary structure. A well-known example of the use of cyclic peptides to mimic protein loops is the RGD-based cyclic peptides including cilengitide [27]. The RGD tripeptide motif is a cell attachment β-turn motif found in numerous proteins, and cyclic peptides containing this motif have been shown to inhibit integrin αVβ3 activity, which plays an important role in tumor metastasis. We have applied a structural mining approach to search the Protein Data Bank for surface-exposed short (4–11 residues) disulfide-bonded loops located at protein–protein interfaces [134]. These natural loop structures generally agreed closely with computational structure predictions, suggesting that natural peptide loops retain their native structure when removed from the context of their parent protein. They also showed increased evolutionary conservation compared with adjacent residues outside the loop. Among these loops was a cyclic peptide derived from a Bowman-Birk serine protease inhibitor previously shown to have independent protease inhibition activity [135], validating the potential of this approach to design enriched bioactive peptide libraries [134]. Structural modeling of loops may also facilitate the design of cyclic peptides mimicking them [136]. Cyclic peptides may also be computationally identified as the natural products of genes, including both ribosomal (where cyclization between residues may be inferred based on rules for disulfide or lanthionine bridging) and non-ribosomal synthesis. Linear azol(in)econtaining peptides (LAPs) were first identified in 1900s [137], and several have been identified by genome mining of bacterial genomes (mainly Firmicutes) since. LAPs contain several combinations of azoline and (methyl) oxazole rings by modification of cysteine, serine, and threonine residues [138]. The enzyme complex involved in LAP biosynthesis is functionally similar to other RiPP families such as lanthipeptides, cyanobactins, and

Computation for Cyclic Peptides PPI Modulators

87

thiopeptides [139]. Cyanobactins are N-to-C macrocyclic peptides, typically 6–20 amino acids long [140]. Over 200 cyanobactins have been identified, of which a third have arisen from genome mining [141]. They have enormous amount of sequence and functional group diversity, which may relate to their broad spectrum of activity such as anticancer, antiviral, and antibiotic [142]. Thiopeptides contain several thiazole rings and multiple dehydrated amino acid residues [143]. Over 100 thiopeptides have been identified, mainly from Actinobacteria. They possess antibacterial, antimalarial, and anticancer properties. Lastly, lasso peptides are a growing class of bacteriocins, consisting of a knotted structure called the lasso fold [144]. They contain 16–21 amino acids, of which only four residues are critical for posttranslational modification [145]. This low substrate specificity, along with possible insertions of novel functional entities in the lasso fold [146], is beneficial from a bioengineering point of view. The fact that these cyclic peptides are ribosomally synthesized and have several enzymes necessary for posttranslational modifications means that it is possible to mine for these genes in publicly available bacterial genomes. The BAGEL3 [147] pipeline and antiSMASH [148–150] pipelines and database are designed to identify the relevant gene clusters and potential gene products. Other computational predictions of peptide cyclization based on biologically occurring peptides generated by cyanobactin macrocyclases have been suggested to have some general predictive power for the propensity for peptides to be cyclized [151], and such predictions may be incorporated into virtual library generation. Libraries of potentially synthesizable novel non-ribosomal peptides [152] may be computationally derived by reassembling alternative cassettes of the peptide synthesis genes, since gene order is a major determinant of residue addition. These naturally occurring virtual peptide libraries offer a rich resource for further computational and experimental investigation. The Norine database provides a resource highlighting many known cyclic structures in non-ribosomal peptides [153], and its associated Smiles2monomer software identifies peptide monomer blocks from compound SMILES descriptors [154, 155]. Phage display technology is traditionally associated with the generation of combinatorial libraries of random cyclic compounds, presented on the surface of a larger phage protein. However, more recently, the lower cost of oligonucleotide synthesis has allowed phage display to be used to present not random short peptides, but instead large libraries of known peptides selected or completely mapped from a given proteome [156–158]. This approach has the potential to be applied to known or predicted short constrained peptides in various proteomes, such as those presented in short disulfide bond loops, to determine whether those short cyclic regions have specific functionalities, conferred by their structurally

88

Fergal Duffy et al.

constrained regions. While these approaches are experimental, the design of the phage display libraries themselves and their subsequent analysis both require extensive computation.

12

Conclusions Despite their promise for use in applications not well suited to traditional small molecules, virtual screening methods for cyclic peptides, and peptides in general, have not been subjected to systematic benchmarking tests of the alternative available or novel computational approaches. The systematic experimental screening of pre-chosen (as opposed to randomized) peptide sequences is likely to provide a set of useful large-scale benchmarks over the next few years that will in turn lead to better computational choice of peptides for screening, whether linear or cyclic. Cyclic peptides are computationally more tractable than linear peptides for many screens of targets of known structure [129]. While sophisticated experimental screening may eliminate some of the need for computational prediction of cyclic peptides [59], there are some practical limitations to such approaches, so that there is still a place for computational screening of cyclic peptide libraries, for example to create smaller targeted sets of sequences for further experimental analysis. Cyclic peptides occupy a niche between typical small molecules and larger peptides and antibodies, with some of the potential advantages and disadvantages of both. Virtual screening has not quite reached its potential, likely due to our incomplete knowledge of the fundamental nature of ligand binding [159], and must be used with an awareness of its fundamental limitations. While a variety of computational methodologies are now available to investigate cyclic peptides, clearer independent benchmarking datasets are needed to assess the relative performance of different generative or analytic approaches.

Acknowledgments This work was supported in part by Science Foundation Ireland (grant 08/IN.1/B1864 supporting FD). References 1. Martins MB, Carvalho I (2007) Diketopiperazines: biological activity and synthesis. Tetrahedron 63:9923–9932. https://doi.org/10. 1016/j.tet.2007.04.105 2. Duffy FJ, Devocelle M, Shields DC (2015) Computational approaches to developing short cyclic peptide modulators of

protein–protein interactions. Methods Mol Biol 1268:241–271. https://doi.org/10. 1007/978-1-4939-2285-7_11 3. Wells JA, McClendon CL (2007) Reaching for high-hanging fruit in drug discovery at protein-protein interfaces. Nature

Computation for Cyclic Peptides PPI Modulators 450:1001–1009. https://doi.org/10.1038/ nature06526 4. Kuenemann MA, Sperandio O, Labbe´ CM et al (2015) In silico design of low molecular weight protein-protein interaction inhibitors: overall concept and recent advances. Prog Biophys Mol Biol 119(1):20–32. https:// doi.org/10.1016/j.pbiomolbio.2015.02. 006 5. Huigens RW, Morrison KC, Hicklin RW et al (2013) A ring-distortion strategy to construct stereochemically complex and structurally diverse compounds from natural products. Nat Chem 5:195–202. https://doi.org/10. 1038/nchem.1549 6. Milhas S, Raux B, Betzi S et al (2016) Proteinprotein interaction inhibition (2P2I)-oriented chemical library accelerates hit discovery. ACS Chem Biol 11(8):2140–2148. https://doi. org/10.1021/acschembio.6b00286 7. Beck A, Wurch T, Bailly C, Corvaia N (2010) Strategies and challenges for the next generation of therapeutic antibodies. Nat Rev Immunol 10:345–352. https://doi.org/10. 1038/nri2747 8. Leader B, Baca QJ, Golan DE (2008) Protein therapeutics: a summary and pharmacological classification. Nat Rev Drug Discov 7:21–39. https://doi.org/10.1038/nrd2399 9. Chames P, Van Regenmortel M, Weiss E, Baty D (2009) Therapeutic antibodies: successes, limitations and hopes for the future. Br J Pharmacol 157:220–233. https://doi.org/ 10.1111/j.1476-5381.2009.00190.x ´ , Zheng G (2012) Flexible or fixed: a 10. Roxin A comparative review of linear and cyclic cancertargeting peptides. Future Med Chem 4:1601–1618. https://doi.org/10.4155/ fmc.12.75 11. Driggers EM, Hale SP, Lee J, Terrett NK (2008) The exploration of macrocycles for drug discovery—an underexploited structural class. Nat Rev Drug Discov 7:608–624. https://doi.org/10.1038/nrd2590 12. Kotz J (2012) Bringing macrocycles full circle. Sci Business eXchange 5:1176. https:// doi.org/10.1038/scibx.2012.1176 13. Schwarzer D, Finking R, Marahiel MA (2003) Nonribosomal peptides: from genes to products. Nat Prod Rep 20:275–287. https:// doi.org/10.1039/b111145k 14. Mullard A (2012) Protein–protein interaction inhibitors get into the groove. Nat Rev Drug Discov 11:173–175 15. Verdine GL, Hilinski GJ (2012) Stapled peptides for intracellular drug targets, 1st edn. Elsevier, Amsterdam

89

16. Arrowsmith J (2011) Trial watch: phase III and submission failures: 2007–2010. Nat Rev Drug Discov 10:87. https://doi.org/10. 1038/nrd3375 17. Freire E (2008) Do enthalpy and entropy distinguish first in class from best in class? Drug Discov Today 13:869–874. https:// doi.org/10.1016/j.drudis.2008.07.005 18. Hamman JH, Enslin GM, Kotze´ AF (2005) Oral delivery of peptide drugs: barriers and developments. BioDrugs 19:165–177 19. Ranade V (1991) Drug delivery systems 5A. Oral drug delivery. J Clin Pharmacol 31:2–16 20. Qian Z, Dougherty PG, Pei D (2017) Targeting intracellular protein–protein interactions with cell-permeable cyclic peptides. Curr Opin Chem Biol 38:80–86. https://doi. org/10.1016/j.cbpa.2017.03.011 21. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ (1997) Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 23:3–25 22. Rezai T, Yu B, Millhauser GL et al (2006) Testing the conformational hypothesis of passive membrane permeability using synthetic cyclic peptide diastereomers. J Am Chem Soc 128:2510–2511. https://doi.org/10.1021/ ja0563455 23. Biron E, Chatterjee J, Ovadia O et al (2008) Improving oral bioavailability of peptides by multiple N-methylation: somatostatin analogues. Angew Chem Int Ed Eng 47:2595–2599. https://doi.org/10.1002/ anie.200705797 24. Ovadia O, Greenberg S, Chatterjee J et al (2011) The effect of multiple N-methylation on intestinal permeability of cyclic hexapeptides. Mol Pharm 8:479–487 25. White TR, Renzelman CM, Rand AC et al (2011) On-resin N-methylation of cyclic peptides for discovery of orally bioavailable scaffolds. Nat Chem Biol 7:810–817. https:// doi.org/10.1038/nchembio.664 26. Doedens L, Opperer F, Cai M et al (2010) Multiple N-methylation of MT-II backbone amide bonds leads to melanocortin receptor subtype hMC1R selectivity: pharmacological and conformational studies. J Am Chem Soc 132:8115–8128. https://doi.org/10.1021/ ja101428m 27. Dechantsreiter MA, Planker E, Math€a B et al (1999) N-methylated cyclic RGD peptides as highly active and selective alpha(V)beta (3) integrin antagonists. J Med Chem 42:3033–3040. https://doi.org/10.1021/ jm970832g

90

Fergal Duffy et al.

28. Roberts MJ, Bentley MD, Harris JM (2012) Chemistry for peptide and protein PEGylation. Adv Drug Deliv Rev 64:116–127. https://doi.org/10.1016/j.addr.2012.09. 025 29. Cefalu WT (2004) Concept, strategies, and feasibility of noninvasive insulin delivery. Diabetes Care 27:239–246. https://doi.org/10. 2337/diacare.27.1.239 30. Chen X, Park R, Shahinian AH et al (2004) Pharmacokinetics and tumor retention of 125I-labeled RGD peptide are improved by PEGylation. Nucl Med Biol 31:11–19. https://doi.org/10.1016/j.nucmedbio. 2003.07.003 31. Rubio-Aliaga I, Daniel H (2002) Mammalian peptide transporters as targets for drug delivery. Trends Pharmacol Sci 23:434–440. https://doi.org/10.1016/S0165-6147(02) 02072-2 32. Rozek A, Powers J-PS, Friedrich CL, Hancock REW (2003) Structure-based design of an indolicidin peptide analogue with increased protease stability. Biochemistry 42:14130–14138. https://doi.org/10. 1021/bi035643g 33. Getz JA, Rice JJ, Daugherty PS (2011) Protease-resistant peptide ligands from a knottin scaffold library. ACS Chem Biol 6:837–844. https://doi.org/10.1021/ cb200039s 34. Guichard G, Benkirane N, Zeder-Lutz G et al (1994) Antigenic mimicry of natural L-peptides with retro-inverso-peptidomimetics. Proc Natl Acad Sci U S A 91:9765–9769 35. Fernandez-Lopez S, Kim HS, Choi EC et al (2001) Antibacterial agents based on the cyclic D,L-alpha-peptide architecture. Nature 412:452–455. https://doi.org/10.1038/ 35086601 36. Whitty A, Kumaravel G (2006) Between a rock and a hard place? Nat Chem Biol 2:112–118. https://doi.org/10.1038/ nchembio0306-112 37. Metz A, Pfleger C, Kopitz H et al (2012) Hot spots and transient pockets: predicting the determinants of small-molecule binding to a protein-protein interface. J Chem Inf Model 52:120–133. https://doi.org/10.1021/ ci200322s 38. Fasan R, Dias RLA, Moehle K et al (2004) Using aβ-hairpin to mimic anα-helix: cyclic peptidomimetic inhibitors of the p53–HDM2 protein–protein interaction. Angew Chem 116:2161–2164. https://doi. org/10.1002/ange.200353242

39. Betzi S, Restouin A, Opi S et al (2007) Protein-protein interaction inhibition (2P2I) combining high throughput and virtual screening: application to the HIV-1 Nef protein. Proc Natl Acad Sci U S A 104:19256–19261. https://doi.org/10. 1073/pnas.0707130104 40. Lo Conte L, Chothia C, Janin J (1999) The atomic structure of protein-protein recognition sites. J Mol Biol 285:2177–2198 41. London N, Movshovitz-Attias D, SchuelerFurman O (2010) The structural basis of peptide-protein binding strategies. Structure 18:188–199. https://doi.org/10.1016/j.str. 2009.11.012 42. Arbor S, Kao J, Wu Y, Marshall GR (2008) c [D-pro-Pro-D-pro-N-methyl-Ala] adopts a rigid conformation that serves as a scaffold to mimic reverse-turns. Biopolymers 90:384–393 43. Larregola M, Lequin O, Karoyan P et al (2011) beta-Amino acids containing peptides and click-cyclized peptide as beta-turn mimics: a comparative study with “conventional” lactam- and disulfide-bridged hexapeptides. J Pept Sci 17:632–643. https:// doi.org/10.1002/psc.1382 44. Tyndall JD, Pfeiffer B, Abbenante G, Fairlie DP (2005) Over one hundred peptideactivated G protein-coupled receptors recognize ligands with turn structure. Chem Rev 105:793–826. https://doi.org/10.1021/ cr040689g 45. Vanhee P, Reumers J, Stricher F et al (2010) PepX: a structural database of non-redundant protein-peptide complexes. Nucleic Acids Res 38:D545–D551. https://doi.org/10.1093/ nar/gkp893 46. Stanfield RL, Wilson IA (1995) Proteinpeptide interactions. Curr Opin Struct Biol 5:103–113. https://doi.org/10.1016/ 0959-440X(95)80015-S 47. Luckett S, Garcia RS, Barker JJ et al (1999) High-resolution structure of a potent, cyclic proteinase inhibitor from sunflower seeds. J Mol Biol 290:525–533. https://doi.org/10. 1006/jmbi.1999.2891 48. Gould CM, Diella F, Via A et al (2010) ELM: the status of the 2010 eukaryotic linear motif resource. Nucleic Acids Res 38:D167–D180. https://doi.org/10.1093/nar/gkp1016 49. Petsalaki E, Russell RB (2008) Peptidemediated interactions in biological systems: new discoveries and applications. Curr Opin Biotechnol 19:344–350. https://doi.org/10. 1016/j.copbio.2008.06.004

Computation for Cyclic Peptides PPI Modulators 50. Reardon DA, Fink KL, Mikkelsen T et al (2008) Randomized phase II study of cilengitide, an integrin-targeting arginine-glycineaspartic acid peptide, in recurrent glioblastoma multiforme. J Clin Oncol 26:5610–5617. https://doi.org/10.1200/ JCO.2008.16.7510 51. Gril B, Vidal M, Assayag F et al (2007) Grb2SH3 ligand inhibits the growth of HER2+ cancer cells and has antitumor effects in human cancer xenografts alone and in combination with docetaxel. Int J Cancer 121:407–415. https://doi.org/10.1002/ijc. 22674 52. Lamberts SW, van der Lely AJ, de Herder WW, Hofland LJ (1996) Octreotide. N Engl J Med 334:246–254. https://doi.org/10. 1056/NEJM199601253340408 53. Andresen V, Camilleri M, Busciglio IA et al (2007) Effect of 5 days linaclotide on transit and bowel function in females with constipation-predominant irritable bowel syndrome. Gastroenterology 133:761–768. https://doi.org/10.1053/j.gastro.2007.06. 067 54. Mack CM, Soares CJ, Wilson JK et al (2010) Davalintide (AC2307), a novel amylinmimetic peptide: enhanced pharmacological properties over native amylin to reduce food intake and body weight. Int J Obes 34:385–395. https://doi.org/10.1038/ijo. 2009.238 55. Kallen J, Mikol V, Taylor P, Walkinshaw MD (1998) X-ray structures and analysis of 11 cyclosporin derivatives complexed with cyclophilin A. J Mol Biol 283:435–449. https://doi.org/10.1006/jmbi.1998.2108 56. Hoogenboom HR, de Bruı¨ne AP, Hufton SE et al (1998) Antibody phage display technology and its applications. Immunotechnology 4:1–20 57. Willats WGT (2002) Phage display: practicalities and prospects. Plant Mol Biol 50 (6):837–854 58. McLafferty MA, Kent RB, Ladner RC, Markland W (1993) M13 bacteriophage displaying disulfide-constrained microproteins. Gene 128:29–36. https://doi.org/10.1016/ 0378-1119(93)90149-W 59. Bashiruddin NK, Suga H (2015) Construction and screening of vast libraries of natural product-like macrocyclic peptides using in vitro display technologies. Curr Opin Chem Biol 24:131–138. https://doi.org/ 10.1016/j.cbpa.2014.11.011

91

60. Horswill AR, Benkovic SJ (2005) Cyclic peptides, a chemical genetics tool for biologists. Cell Cycle 4:552–555 61. Kritzer JA, Hamamichi S, Mccaffery JM et al (2009) Rapid selection of cyclic peptides that reduce α-synuclein toxicity in yeast and animal models. Nat Chem Biol 5:655–663. https:// doi.org/10.1038/nchembio.193 62. Gale EF, Taylor ES (1946) Action of tyrocidine and detergents in liberating amino acids from bacterial cells. Nature 157:549 63. Arbeit RD, Maki D, Tally FP et al (2004) The safety and efficacy of daptomycin for the treatment of complicated skin and skin-structure infections. Clin Infect Dis 38:1673–1681. https://doi.org/10.1086/420818 64. Dawson R (1998) The toxicology of microcystins. Toxicon 36:953–962. https://doi. org/10.1016/S0041-0101(97)00102-5 65. Namikoshi M, Choi BW, Sakai R et al (1994) New nodularins: a general method for structure assignment. J Organomet Chem 59:2349–2357. https://doi.org/10.1021/ jo00088a014 66. Goodin S, Kane MP, Rubin EH (2004) Epothilones: mechanism of action and biologic activity. J Clin Oncol 22:2015–2025. https://doi.org/10.1200/JCO.2004.12. 001 67. Anderson S (1984) Graphical representation of molecules and substructure-search queries in MACCStm. J Mol Graph 2:83–90. https:// doi.org/10.1016/0263-7855(84)80060-0 68. Daylight Chemical Information Systems (2012) Daylight Toolkit, Laguna Niguel, http://www.daylight.com/products/toolkit. html 69. Taminau J, Thijs G, De Winter H (2008) Pharao: pharmacophore alignment and optimization. J Mol Graph Model 27:161–169. https://doi.org/10.1016/j.jmgm.2008.04. 003 70. Chemical Computing Group (2012) Molecule Operating Environment (MOE), Montreal, https://www.chemcomp.com/ Products.htm 71. Koes DR, Camacho CJ (2011) Pharmer: efficient and exact pharmacophore search. J Chem Inf Model 51:1307–1314. https:// doi.org/10.1021/ci200097m 72. Dassault Syste`mes BIOVIA (2017) Discovery Studio Modeling Environment, San Diego, https://www.3dsbiovia.com/products/col laborative-science/biovia-discovery-studio/

92

Fergal Duffy et al.

73. Ballester PJ, Richards WG (2007) Ultrafast shape recognition to search compound databases for similar molecular shapes. J Comput Chem 28:1711–1723. https://doi.org/10. 1002/jcc.20681 74. Schreyer AM, Blundell T (2012) USRCAT: real-time ultrafast shape recognition with pharmacophoric constraints. Aust J Chem 4:27. https://doi.org/10.1186/1758-29464-27 75. GRANT JA, GALLARDO MA, PICKUP BT (1996) A fast method of molecular shape comparison: a simple application of a Gaussian description of molecular shape. J Comput Chem 17:1653–1666. https://doi.org/10. 1002/(SICI)1096-987X(19961115) 17:143.0.CO;2-K 76. Sastry GM, Dixon SL, Sherman W (2011) Rapid shape-based ligand alignment and virtual screening method based on atom/feature-pair similarities and volume overlap scoring. J Chem Inf Model 51:2455–2466. https://doi.org/10.1021/ci2002704 77. Evers A, Hessler G, Matter H, Klabunde T (2005) Virtual screening of biogenic aminebinding G-protein coupled receptors: comparative evaluation of protein- and ligandbased virtual screening protocols. J Med Chem 48:5448–5465. https://doi.org/10. 1021/jm050090o 78. Warren GL, Andrews CW, Capelli A-M et al (2006) A critical assessment of docking programs and scoring functions. J Med Chem 49:5912–5931. https://doi.org/10.1021/ jm050362n 79. Certara (2013) Sybyl-X, Princeton, https:// omictools.com/sybyl-x-tool 80. Trott O, Olson AJ (2010) AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31:455–461. https://doi.org/10.1002/jcc. 21334 81. Lang PT, Brozell SR, Mukherjee S et al (2009) DOCK 6: combining techniques to model RNA-small molecule complexes. RNA 15:1219–1230. https://doi.org/10.1261/ rna.1563609 82. Zsoldos Z, Reid D, Simon A et al (2007) eHiTS: a new fast, exhaustive flexible ligand docking system. J Mol Graph Model 26:198–212. https://doi.org/10.1016/j. jmgm.2006.06.002 83. Berman HM, Westbrook J, Feng Z et al (2000) The protein data bank. Nucleic Acids Res 28:235–242

84. Viji SN, Prasad PA, Gautham N (2009) Protein-ligand docking using mutually orthogonal Latin squares (MOLSDOCK). J Chem Inf Model 49:2687–2694. https://doi. org/10.1021/ci900332a 85. Pearce BC, Langley DR, Kang J et al (2009) E-novo: an automated workflow for efficient structure-based lead optimization. J Chem Inf Model 49:1797–1809. https://doi.org/10. 1021/ci900073k 86. Morris GM, Huey R, Lindstrom W et al (2009) AutoDock4 and AutoDockTools4: automated docking with selective receptor flexibility. J Comput Chem 30:2785–2791. https://doi.org/10.1002/jcc.21256 87. Mosca R, Pons C, Ferna´ndez-Recio J, Aloy P (2009) Pushing structural information into the yeast interactome by high-throughput protein docking experiments. PLoS Comput Biol 5:e1000490. https://doi.org/10.1371/ journal.pcbi.1000490 88. Yuriev E, Agostino M, Ramsland PA (2011) Challenges and advances in computational docking: 2009 in review. J Mol Recognit 24:149–164. https://doi.org/10.1002/jmr. 1077 89. OpenEye Scientific Software (2010) OEChem Toolkit, Santa Fe, https://www. eyesopen.com/cheminformatics 90. Schrodinger LLC (2012) Schrodinger, New York City, https://www.schrodinger.com/ 91. Guha R, Howard MT, Hutchison GR et al (2006) The blue obelisk-interoperability in chemical informatics. J Chem Inf Model 46:991–998. https://doi.org/10.1021/ ci050400b 92. Landrum G (2018) RDKit: Open-source cheminformatics, Basel, https://www.rdkit.org/ 93. O’Boyle NM, Banck M, James CA et al (2011) Open babel: an open chemical toolbox. Aust J Chem 3:33. https://doi.org/10. 1186/1758-2946-3-33 94. Steinbeck C, Hoppe C, Kuhn S et al (2006) Recent developments of the chemistry development kit (CDK)—an open-source Java library for chemo- and bioinformatics. Curr Pharm Des 12:2111–2120 95. Mazanetz MP, Marmon RJ, Reisser CBT, Morao I (2012) Drug discovery applications for KNIME: an open source data mining platform. Curr Top Med Chem 12:1965–1979 96. Irwin JJ, Shoichet BK (2005) ZINC—a free database of commercially available compounds for virtual screening. J Chem Inf Model 45:177–182. https://doi.org/10. 1021/ci049714+

Computation for Cyclic Peptides PPI Modulators 97. Burns VA, Bobay BG, Basso A et al (2008) Targeting RNA with cysteine-constrained peptides. Bioorg Med Chem Lett 18:565–567. https://doi.org/10.1016/j. bmcl.2007.11.096 98. Duffy FJ, Verniere M, Devocelle M et al (2011) CycloPs: generating virtual libraries of cyclized and constrained peptides including nonnatural amino acids. J Chem Inf Model 51:829–836. https://doi.org/10.1021/ ci100431r 99. Ponder JW (2013) Tinker: software tools for molecular design, St. Louis, https://dasher. wustl.edu/tinker/ 100. Goldtzvik Y, Goldstein M, Benny Gerber R (2013) On the crystallographic accuracy of structure prediction by implicit water models: tests for cyclic peptides. Chem Phys 415:168–172. https://doi.org/10.1016/j. chemphys.2013.01.039 101. O’Boyle NM, Vandermeersch T, Flynn CJ et al (2011) Confab-systematic generation of diverse low-energy conformers. Aust J Chem 3:8. https://doi.org/10.1186/1758-29463-8 102. Jacobson MP, Pincus DL, Rapp CS et al (2004) A hierarchical approach to all-atom protein loop prediction. Proteins Struct Funct Genet 55:351–367 103. Ebejer JP, Morris GM, Deane CM (2012) Freely available conformer generation methods: how good are they? J Chem Inf Model 52:1146–1158. https://doi.org/10.1021/ ci2004658 104. Venkatraman V, Pe´rez-Nueno VI, Mavridis L, Ritchie DW (2010) Comprehensive comparison of ligand-based virtual screening tools against the DUD data set reveals limitations of current 3D methods. J Chem Inf Model 50:2079–2093. https://doi.org/10.1021/ ci100263p 105. Duffy FJ, O’Donovan D, Devocelle M et al (2015) Virtual screening using combinatorial cyclic peptide libraries reveals protein interfaces readily targetable by cyclic peptides. J Chem Inf Model 55:600–613. https://doi. org/10.1021/ci500431q 106. Oshiro CM, Kuntz ID, Dixon JS (1995) Flexible ligand docking using a genetic algorithm. J Comput Aided Mol Des 9:113–130. https://doi.org/10.1007/BF00124402 107. Jones G, Willett P, Glen RC et al (1997) Development and validation of a genetic algorithm for flexible docking. J Mol Biol 267:727–748. https://doi.org/10.1006/ jmbi.1996.0897

93

108. Morris GM, Goodsell DS, Halliday RS et al (1998) Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J Comput Chem 19:1639–1662. https://doi.org/10.1002/( SICI)1096-987X(19981115)19:143.0.CO;2-B 109. Sheridan RP, Kearsley SK (1995) Using a genetic algorithm to suggest combinatorial libraries. J Chem Inf Model 35:310–320. https://doi.org/10.1021/ci00024a021 110. Westhead DR, Clark DE, Frenkel D et al (1995) PRO-LIGAND: an approach to de novo molecular design. 3. A genetic algorithm for structure refinement. J Comput Aided Mol Des 9:139–148 111. Schneider G, Lee ML, Stahl M, Schneider P (2000) De novo design of molecular architectures by evolutionary assembly of drugderived building blocks. J Comput Aided Mol Des 14:487–494 112. Schneider G, Hartenfeller M, Reutlinger M et al (2009) Voyages to the (un)known: adaptive design of bioactive compounds. Trends Biotechnol 27:18–26. https://doi.org/10. 1016/j.tibtech.2008.09.005 113. Belda I, Madurga S, Llora` X et al (2005) ENPDA: an evolutionary structure-based de novo peptide design algorithm. J Comput Aided Mol Des 19:585–601. https://doi. org/10.1007/s10822-005-9015-1 114. Hohm T, Limbourg P, Hoffmann D (2006) A multiobjective evolutionary method for the design of peptidic mimotopes. J Comput Biol 13:113–125. https://doi.org/10.1089/ cmb.2006.13.113 115. Knapp B, Giczi V, Ribarics R, Schreiner W (2011) PeptX: using genetic algorithms to optimize peptides for MHC binding. BMC Bioinformatics 12:241. https://doi.org/10. 1186/1471-2105-12-241 116. Pierce BG, Wiehe K, Hwang H et al (2014) ZDOCK server: interactive docking prediction of protein-protein complexes and symmetric multimers. Bioinformatics 30:1771–1773. https://doi.org/10.1093/ bioinformatics/btu097 117. Sa´nchez-Sanz G, Tywoniuk B, Matallanas D et al (2016) SARAH domain-mediated MST2-RASSF dimeric interactions. PLoS Comput Biol 12(10):e1005051. https://doi. org/10.1371/journal.pcbi.1005051 118. Norris R, Casey F, FitzGerald RJ et al (2012) Predictive modelling of angiotensin converting enzyme inhibitory dipeptides. Food Chem 133:1349–1354. https://doi.org/10. 1016/j.foodchem.2012.02.023

94

Fergal Duffy et al.

119. Chen H, Lyne PD, Giordanetto F et al (2006) On evaluating molecular-docking methods for pose prediction and enrichment factors. J Chem Inf Model 46:401–415 120. Husic BE, Pande VS (2018) Markov state models: from an art to a science. J Am Chem Soc 140:2386–2396. https://doi.org/10. 1021/jacs.7b12191 121. Buchete N, Hummer G (2008) Coarse master equations for peptide folding dynamics. J Phys Chem B 112:6057–6069 122. Narayan B, Herbert C, Yuan Y et al (2018) Conformational analysis of replica exchange MD: temperature-dependent Markov networks for FF amyloid peptides. J Chem Phys 149:072323. https://doi.org/10.1063/1. 5027580 123. Leahy CT, Murphy RD, Hummer G et al (2016) Coarse master equations for binding kinetics of amyloid peptide dimers. J Phys Chem Lett 7:2676–2682. https://doi.org/ 10.1021/acs.jpclett.6b00518 124. McCartan S, Crowe A, Sa´nchez-Sanz G, et al (2019) Conformational dynamics of CsA: temperature and N-methylation effects, 2018, in prep.). Prep 125. Raveh B, London N, Zimmerman L, Schueler-Furman O (2011) Rosetta FlexPepDock ab-initio: simultaneous folding, docking and refinement of peptides onto their receptors. PLoS One 6:e18934. https://doi. org/10.1371/journal.pone.0018934 126. London N, Gulla´ S, Keating AE, SchuelerFurman O (2012) In silico and in vitro elucidation of BH3 binding specificity toward Bcl-2. Biochemistry 51:5841–5850. https:// doi.org/10.1021/bi3003567 127. Mandal PK, Limbrick D, Coleman DR et al (2009) Conformationally constrained peptidomimetic inhibitors of signal transducer and activator of transcription. 3: evaluation and molecular modeling. J Med Chem 52:2429–2442. https://doi.org/10.1021/ jm801491w 128. Bhardwaj G, Mulligan VK, Bahl CD et al (2016) Accurate de novo design of hyperstable constrained peptides. Nature 538:329–335. https://doi.org/10.1038/ nature19791 129. Hosseinzadeh P, Bhardwaj G, Mulligan VK et al (2017) Comprehensive computational design of ordered peptide macrocycles. Science 358:1461–1466. https://doi.org/10. 1126/science.aap7577 130. Slough DP, McHugh SM, Cummings AE et al (2018) Designing well-structured cyclic pentapeptides based on sequence–structure

relationships. J Phys Chem B 122:3908–3919. https://doi.org/10.1021/ acs.jpcb.8b01747 131. Claro B, Bastos M, Garcia-Fandino R (2018) 4—Design and applications of cyclic peptides. In: Peptide Applications in Biomedicine, Biotechnology and Bioengineering, 87–129 132. London N, Raveh B, Movshovitz-Attias D, Schueler-Furman O (2010) Can selfinhibitory peptides be derived from the interfaces of globular protein-protein interactions? Proteins 78:3140–3149. https://doi.org/10. 1002/prot.22785 133. Xu Y, Rahman NA, Othman R et al (2012) Computational identification of selfinhibitory peptides from envelope proteins. Proteins 80:2154–2168. https://doi.org/ 10.1002/prot.24105 134. Duffy FJ, Devocelle M, Croucher DR, Shields DC (2014) Computational survey of peptides derived from disulphide-bonded protein loops that may serve as mediators of proteinprotein interactions. BMC Bioinformatics 15:305. https://doi.org/10.1186/14712105-15-305 135. Domingo GJ, Leatherbarrow RJ, Freeman N et al (1995) Synthesis of a mixture of cyclic peptides based on the Bowman-Birk reactive site loop to screen for serine protease inhibitors. Int J Pept Protein Res 46:79–87. https://doi.org/10.1111/j.1399-3011. 1995.tb00585.x 136. Karami Y, Guyon F, De Vries S, Tuffe´ry P (2018) DaReUS-loop: accurate loop modeling using fragments from remote or unrelated proteins. Sci Rep 8:13673. https://doi.org/ 10.1038/s41598-018-32079-w 137. Besredka A (1921) de l’Inst. Pasteur 35:291 138. Li Y-M, Milne JC, Madison LL et al (1996) From peptide precursors to oxazole and thiazole-containing peptide antibiotics: microcin B17 synthase. Science 274:1188–1193 139. Walsh CT, Malcolmson SJ, Young TS (2012) Three ring posttranslational circuses: insertion of oxazoles, thiazoles, and pyridines into protein-derived frameworks. ACS Chem Biol 7:429–442 140. Ireland C, Scheuer PJ (1980) Ulicyclamide and ulithiacyclamide, two new small peptides from a marine tunicate. J Am Chem Soc 102:5688–5691 141. Sivonen K, Leikoski N, Fewer DP, Jokela J (2010) Cyanobactins-ribosomal cyclic peptides produced by cyanobacteria. Appl Microbiol Biotechnol 86:1213–1225

Computation for Cyclic Peptides PPI Modulators 142. Houssen WE, Jaspars M (2010) Azole-based cyclic peptides from the sea squirt lissoclinum patella: old scaffolds, new avenues. Chembiochem 11:1803–1815 143. Bagley MC, Dale JW, Merritt EA, Xiong X (2005) Thiopeptide antibiotics. Chem Rev 105:685–714 144. Maksimov MO, Pan SJ, Link AJ (2012) Lasso peptides: structure, function, biosynthesis, and engineering. Nat Prod Rep 29:996–1006 145. Knappe TA, Linne U, Robbel L, Marahiel MA (2009) Insights into the biosynthesis and stability of the lasso peptide capistruin. Chem Biol 16:1290–1298 146. Knappe TA, Manzenrieder F, Mas-Moruno C et al (2011) Introducing lasso peptides as molecular scaffolds for drug design: engineering of an integrin antagonist. Angew Chem 123:8873–8876 147. van Heel AJ, de Jong A, Montalba´n-Lo´pez M et al (2013) BAGEL3: automated identification of genes encoding bacteriocins and (non-)bactericidal posttranslationally modified peptides. Nucleic Acids Res 41: W448–W453. https://doi.org/10.1093/ nar/gkt391 148. Blin K, Wolf T, Chevrette MG et al (2017) AntiSMASH 4.0—improvements in chemistry prediction and gene cluster boundary identification. Nucleic Acids Res 45: W36–W41. https://doi.org/10.1093/nar/ gkx319 149. Blin K, Medema MH, Kottmann R et al (2017) The antiSMASH database, a comprehensive database of microbial secondary metabolite biosynthetic gene clusters. Nucleic Acids Res 45:D555–D559. https://doi.org/ 10.1093/nar/gkw960 150. Blin K, Kazempour D, Wohlleben W, Weber T (2014) Improved lanthipeptide detection and prediction for antiSMASH. PLoS One 9 (2):e89420. https://doi.org/10.1371/jour nal.pone.0089420 151. Booth J, Alexandru-Crivac CN, Rickaby KA et al (2017) A blind test of computational

95

technique for predicting the likelihood of peptide sequences to cyclize. J Phys Chem Lett 8:2310–2315. https://doi.org/10. 1021/acs.jpclett.7b00848 152. Stevens BW, Joska TM, Anderson AC (2005) Progress toward re-engineering non-ribosomal peptide synthetase proteins: a potential new source of pharmacological agents. Drug Dev Res 66:9–18. https://doi.org/10.1002/ ddr.20041 153. Flissi A, Dufresne Y, Michalik J et al (2016) Norine, the knowledgebase dedicated to non-ribosomal peptides, is now open to crowdsourcing. Nucleic Acids Res 44(D1): D1113–D1118. https://doi.org/10.1093/ nar/gkv1143 154. Lecle`re V, Weber T, Jacques P, Pupin M (2016) Bioinformatics tools for the discovery of new nonribosomal peptides. Methods Mol Biol 1401:209–232 155. Dufresne Y, Noe´ L, Lecle`re V, Pupin M (2015) Smiles2Monomers: a link between chemical and biological structures for polymers. Aust J Chem 7:62. https://doi.org/ 10.1186/s13321-015-0111-5 156. Nim S, Jeon J, Corbi-Verge C et al (2016) Pooled screening for antiproliferative inhibitors of protein-protein interactions. Nat Chem Biol 12(4):275–281. https://doi. org/10.1038/nchembio.2026 157. Davey NE, Seo MH, Yadav VK et al (2017) Discovery of short linear motif-mediated interactions through phage display of intrinsically disordered regions of the human proteome. FEBS J 284(3):485–498. https://doi. org/10.1111/febs.13995 158. Sundell GN, Ivarsson Y (2014) Interaction analysis through proteomic phage display. Biomed Res Int 2014:176172. https://doi. org/10.1155/2014/176172 159. Schneider G (2010) Virtual screening: an endless staircase? Nat Rev Drug Discov 9:273–276. https://doi.org/10.1038/ nrd3139

Chapter 6 Design of Cyclic Peptides as Protein Recognition Motifs Ye Che Abstract Protein–protein interactions are ubiquitous, essential to almost all known biological processes, and offer attractive opportunities for therapeutic intervention. Linear peptide drugs, however, can be applied therapeutically as protein recognition motifs only to a limited extent because of their poor permeability, decreased receptor selectivity, and proteolytic stability. A major strategy in peptide chemistry is directed toward chemical modification and macrocyclization in order to limit a peptide’s conformational possibilities, to increase its chemical and enzymatic stability, to prolong the time of action, and to increase activity and selectivity toward the receptor. Key words Protein–peptide interaction, Protein surface mimetics, Conformationally constrained peptides, Macrocyclization, α-Helix, β-Strand, Reverse turn

1

Introduction Protein–protein interactions dominate molecular recognition in biological systems. One major challenge for drug discovery arises from the very large surfaces that are characteristic of many protein–protein interactions [1]. Recent studies of protein interactions involved in cell regulation and signaling have identified a large number in which one component involves a flexible or unstructured region of the polypeptide chain under physiological condition that folds into ordered structures only on binding to their cellular targets. Coupled folding and binding often gives to a protein complex high specificity and relative low affinity, which is appropriate for signal transduction proteins that must not only associate specifically to initiate the signaling but must also be capable of dissociation when signaling is complete [2]. Nature optimizes rates and system dynamics rather that affinities per se. In addition, conformational flexibility allows a protein to bind to both its physiologic target and modifying enzymes to facilitate posttranslational modifications. The thermodynamic consequence is that there is an entropic cost associated with the disorder-to-

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_6, © Springer Science+Business Media, LLC, part of Springer Nature 2019

97

98

Ye Che

order transition that accompanies the binding of an intrinsically unstructured protein to its target. It is estimated that elimination of a single rotational degree of freedom of a peptide by preorganization to stabilize the receptor-bound conformation enhances affinity by about 1.2–1.6 kcal/mole assuming complete loss of rotational degrees of freedom [3]. Therefore, it has been proposed that protein–protein interactions involving one intrinsically disordered partner are more druggable sites of interactions. In addition, studies indicated that there is high abundance of intrinsic disorder in proteins associated with cancer, neurodegenerative, and cardiovascular diseases [2]. The recognition of peptide hormones by their receptors can be viewed as a special case of protein–protein interactions involving one unstructured partner. Peptides can be directly applied as pharmacologically active compounds to only a very limited extent. The major disadvantages of the application of a linear peptide in a biological system—for example, rapid degradation by proteases, hepatic clearance, undesired side effects by interaction of conformationally flexible linear peptides with different receptors, and low membrane permeability due to their hydrophilic character—are in most cases detrimental to oral application. On the other hand, cyclic peptides have long tantalized drug designers with their potential ability to combine the best attributes of antibodies and small molecules [4]. A cyclic peptide drug with only limited conformational flexibility would combine the specificity and protein–protein interaction-blocking capabilities of an antibody with the bioavailability and cellular permeability of a small-molecule drug. Although the number of possible conformations for a small, linear peptide sequence can be limited by certain steric hindrances and allowable torsional angles, the overall number often is still large because of the flexibility of the peptide backbone chain. To further restrict the number of possible conformations into a handful of allowed backbone structures, cyclic constraints are often incorporated. There are six main ways to form cyclic peptides [5]. These cyclizations, as illustrated in Fig. 1, involve bridging between (1) N-terminal amino and C-terminal carboxylate groups; (2) two side chains; (3) two internal backbone NH groups; (4) an internal backbone NH group and a side chain; (5) a side chain to N terminus; and (6) a side chain to C terminus. Cyclization of a peptide chain can dramatically change the overall conformation of a ligand favoring certain secondary structures. Greatly reducing the degrees of freedom in a newly formed cyclic structure can stabilize the bioactive conformation of the peptide ligand. This chapter highlights recent advances in the design of cyclic peptides that stabilize common protein recognition motifs, such as α-helix, β-strand, and reverse turns, to modulate protein–protein interactions.

Cyclic Peptides as Protein Recognition Motifs

99

Fig. 1 Types of peptide cyclization: (1) head-to-tail, (2) side chain-to-side chain, (3) backbone-to-backbone, (4) backbone-to-side chain, (5) head-to-side chain, and (6) side chain-to-tail

2

α-Helix Cyclic Peptides α-Helices on protein surfaces often function as recognition motifs for protein–protein, protein–DNA, and protein–RNA interactions. Consequently, these helical recognition motifs represent attractive targets for potential therapeutics in a broad spectrum of diseases. For example, transcriptional activators (e.g., p53, NF-kBp65, VP16c), apoptosis regulators (e.g., Bak), and RNA-transporter proteins (e.g., Rev) all contain a short α-helical sequence that mediates function by direct interaction with a receptor. A short synthetic peptide corresponding to a helical recognition motif does not typically fold stably in isolation and is usually flexible and conformationally disordered in solution. Such flexible peptides present side chains in a plethora of relative orientations increasing undesirable interactions at multiple recognition sites. This inherent flexibility also limits binding affinity when these peptides bind to their receptors in a unique conformation, due to a more significant loss of entropy. Marshall and Bosshard [6] predicted in 1972 that α,α-dialkyl amino acids (Fig. 2a), such as α-aminoisobutyric acid (Aib) would severely restrict the backbone Φ- and ψ-torsional angles of that residue to those associated with right- or left-handed helices. Subsequent experimental validation of that prediction is abundant. An example where α,α-dialkyl amino acids were used to induce an α-helical peptide in water that enhanced binding involves the p53/MDM2 helix recognition: IC50 of 5 nM for an Aib-containing peptide and 8.7 μM for the native α-helical peptide [7]. Alternatively, the α-helical structure can be stabilized through the incorporation of covalent cyclization between side chains of two

100

Ye Che

Fig. 2 Helix stabilization methods: (a) α,α-disubstituted unnatural amino acids, (b) side chain cross-linked stapled peptides, (c) intrachain H-bond covalent surrogates, (d) end-capping templates, and (e) ALRN-6924, currently in phase II clinical trials, an example of stapled α-helical cyclic peptides

residues separated in sequence, but spatially close in a helix, such as residues i and i + 4 or i and i + 7 of an α-helix. Examples of cyclic linkages shown to enhance helical propensity include disulfide bonds, lactam bridges, hydrocarbon staplings, diaminoalkanes, and acetylenes (Fig. 2b). These side chain-to-side chain cyclic peptides have been demonstrated to yield a marketed enhancement of peptide helicity, stability, and in vitro and in vivo biological activity. For example, the interaction between the proapoptotic protein BID and the antiapoptotic protein Bcl-xL was disrupted by a hydrocarbon-stapled helix combined with α-methyl substituents on the two linked amino acids [8]. This conformationally constrained cyclic peptide, derived from the helical BH3 domain of BID, was found to be protease resistant, cell permeable, and bound to Bcl-xL with a sixfold higher affinity than the unconstrained helix peptide. Cellular uptake was observed and apoptosis was activated within cells after treatment with the cyclic peptide. In addition, the stapled cyclic peptide effectively inhibited the growth of human leukemia xenografts in vivo. α-Helical peptides are stabilized by extensive but weak intrachain H-bonds. Design of backbone-to-backbone covalent linkage of intrachain H-bonds reinforces the helical structure (Fig. 2c). Such cyclic peptides are attractive scaffolds for molecular recognition because the backbone H-bond surrogate neither blocks solvent-exposed recognition surface nor removes important side chain functionalities. For example, one cyclic peptide analog of a human papillomavirus peptide segment was conformationally restricted to an α-helical structure using a hydrazine linkage and was shown to have a very strong reaction with sera from women having cervical carcinoma [9]. Though the main body of an α-helical peptide is stabilized by intrachain H-bonds, free amino

Cyclic Peptides as Protein Recognition Motifs

101

groups at the N terminus and carboxyl groups at the C terminus of the peptide do not participate in such internal peptide H-bonding. Thus, preorganized helix-nucleating templates, via side chain-toside chain cyclic constraints, have been developed [10] in which the orientation of the first 4 amino groups or the last 4 carboxyl groups were fixed in a rigid cyclic structure to template α-helical formation and prevent fraying of either end (Fig. 4d). An example of α-helical cyclic peptides currently in human clinical trials is a potent and selective dual inhibitor of MDM2 and MDMX, ALRN-6924 (Fig. 2e), which effectively activates the p53 pathway in tumors in vitro and in vivo [11]. Specifically, ALRN-6924 consists an all-hydrocarbon cyclic linkage, which was prepared via two α-methyl-substituted amino acids having terminal olefin side chains to enable ring-closing metathesis using Grubbs ruthenium catalyst. ALRN-6924 binds both MDM2 and MDMX with nanomolar affinities, shows submicromolar cellular activities in cancer cell lines in the presence of serum, and demonstrates highly specific, on-target mechanism of action. Phase I trial of agent at 0.16–4.4 mg/kg/week for 3 weeks i.v. and 0.32–2.7 mg/kg b.i.w. for 2 weeks i.v., respectively, in 41 and 30 patients with solid tumors displayed evidence of antitumor activity of drug with few adverse effects. The candidate is currently in phase II trials to treat peripheral T-cell lymphoma.

3

β-Strand Cyclic Peptides A β-strand peptide, which has an extended backbone conformation, often recognizes other protein surfaces via not only side chains but also backbone H-bonding interactions. Strands are now recognized in their own right as important recognition motifs in protein–protein interactions [12, 13]. More than 1500 crystal and solution structures of proteases with their corresponding peptide substrates or inhibitors show that local regions of the peptide bound to the active sites adopt an extended strand conformation, in which side chains are readily accessible and the backbone amide bonds play significant roles in recognition. In immune defense, the major histocompatibility complex (MHC) selectively bind the extended β-strand conformation of peptides derived from intracellular processing of viral, bacterial, and endogenous proteins and present them at the cell surface for recognition and immunological destruction. β-Sheet formation between proteins occurs widely in normal biological functions as exemplified by the dimerization interface of HIV protease, the binding of the Ras oncoproteins to their kinase receptors, the formation of cell-cell junctions through the interactions between PDZ or PTB domains with peptide segments from other proteins, etc. β-Sheet formation is also critical in protein

102

Ye Che

aggregation which occurs in a variety of diseases and is a common feature of many neurodegenerative diseases. A key structural feature of the β-strand is that alternating side chains (e.g., i, i + 1) pointing in opposite directions and thus first and third residues (e.g., i, i + 2) in a peptide sequence have side chains on the same face of a β-strand, bringing them into proximity for potential ring closure to cyclic peptides. β-Strand cyclic peptides have been extensively used in the development of protease inhibitors used in clinics. For example, ring-closing metathesis has been widely used for the formation of such macrocyclic linkages and has given rise to hepatitis C virus (HCV) drugs, such as vaniprevir and danoprevir, as shown in Fig. 3. Available X-ray crystallographic structures of the NS3 protease domain bound with linear peptide inhibitors revealed a relatively shallow and highly solvent-exposed active site occupied by linear peptide inhibitors in extended β-strand conformations. Aided by computational modeling, vaniprevir [14] was prepared with a side chain-to-side chain linkage between P2 and P4 to form a 20-membered macrocyclic HCV inhibitor, with improved enzyme potency, cellular activity, good plasma exposure, and excellent liver exposure in multiple species

Fig. 3 Examples of clinical HCV NS3/4A protease inhibitors, vaniprevir and danoprevir, with P2–P4 and P1–P3 macrocyclic constraints stabilizing a β-strand bound conformation, respectively

Cyclic Peptides as Protein Recognition Motifs

103

following oral dosing. Similarly, a 15-membered macrocyclic HCV NS3/4A protease inhibitor, danoprevir [15], has a side chain-toside chain linkage between P1 and P3. It displays unique preclinical characteristics, e.g., a slow off rate, and elicits potent antiviral activity in the clinic.

4

Reverse-Turn Cyclic Peptides A reverse turn is a structural motif that invariably lies on the surface of proteins that often participate in protein–protein interactions (Fig. 4a). Receptor recognition, substrate specificity, and catalytic function generally reside in these loop regions, which often connect residues of adjacent α-helices and β-strands, contributing to the structural stability of proteins. Reverse turns comprise a diverse group of structures with a well-defined three-dimensional orientation of amino acid side chains. β-Turns constitute the most important subgroup and are formed by four consecutive amino acids. Examples of turns as recognition motifs can be readily found in peptide antigen-antibody complexes. Structure-activity relationship studies of many peptide hormones interacting with G protein-coupled receptors (GPCRs) have indicated that the hormones are probably in reverse-turn conformations when bound to their receptors [16]. Conformational and topographical restrictions are particularly suited as manipulation for reverse-turn stabilization toward an increase of receptor selectivity, metabolic stability, and the development of highly potent protein recognition motifs for modulating protein interactions [17]. One straightforward approach for peptide modification is to introduce a covalent linkage between residues i and i + 3, such as head-to-tail cyclization (Fig. 4b), which retains the reverse-turn conformation. Cyclic peptides form a large class of naturally occurring compounds with a variety of biologic activities, such as hormones, antibiotics, ion-transport regulators, and toxins. They have been reported to bind multiple, unrelated classes of receptors with high affinity. Thus, cyclic peptides are considered to be privileged structures capable of providing useful ligands for more than one receptor, due to their high content of reverse-turn motifs. Another approach is to incorporate heterochiral dipeptides at residues i + 1 and i + 2 (Fig. 4c). Nearly all biological polymers are homochiral: all amino acids coded and incorporated by protein synthesis are left-handed, whereas all sugars in DNA/RNA and in metabolic pathways are right-handed. It is the homochirality of naturally occurring amino acids that allows proteins to adopt regular conformations, such as the α-helix and the β-strand. The incorporation of heterochiral (D, L-alternating) dipeptides into a peptide chain abruptly changes the direction of the peptide. For example, Marshall and

104

Ye Che

Fig. 4 Reverse-turn stabilization methods: (a) a β-turn motif with an intramolecular bond between residues i and i + 3, (b) end-to-end cyclization, (c) heterochiral dipeptide, (d) dipeptide lactam, and (e) PMX53, an example of reverse-turn cyclic peptides, with a heterochiral dipeptide and a side chain-to-tail covalent linkage

co-workers [18] suggested that D-Pro-L-Pro, L-Pro-D-Pro, D-ProL-Pip, L-Pro-D-Pip, D-Pro-L-NMe-AA, and L-Pro-D-NMe-AA (where AA, amino acid other than Gly; Pip, pipecolic amino acid; NMe, N-methylation) offer relatively rigid scaffolds on which to orient side chains for interactions with receptors that recognize reverse-turn structures. Similarly, Gellman and colleagues [19] described that the β-amino acid heterochiral dinipecotic acid segments, R-Nip-S-Nip and S-Nip-R-Nip (where Nip, nipecotic acid), could also promote reverse-turn formation. Very similarly, conformationally constrained dipeptide mimics can be introduced to replace residues i + 1 and i + 2 to enhance a reverse-turn structure. The dipeptide lactam (Fig. 4d), the bicyclic dipeptide BTD, and spirolactam bicyclic and tricyclic systems based on Pro are all examples which partially constrain the four backbone torsional angles of residues i + 1 and i + 2 and enhance reverse-turn propensity. PMX53 [20], a potent orthosteric antagonist of the C5a receptor, currently in phase II clinical trials, is a reverse-turn cyclic peptide with both a heterochiral dipeptide and a side chain-to-tail covalent linkage (Fig. 4e). It was designed to mimic the C terminus of a 74 residue pro-inflammatory human protein, complement C5a. It features a pentapeptide macrocycle formed through a side chain to C terminus lactam bridge and a D-Cha-L-Trp heterochiral dipeptide turn-inducing constraint. PMX53 has a 106-fold higher affinity for the C5a receptor than the hexapeptide C terminus of C5a, showing potent full antagonism with high selectivity and long residence time. Recently, crystal structures of human C5a receptor

Cyclic Peptides as Protein Recognition Motifs

105

in complexes with PMX53 have been reported [21], and it confirms that PMX53 adopts a reverse-turn-bound conformation centered on the D-Cha-L-Trp heterochiral dipeptide.

Acknowledgments I am grateful to Prof. Garland R. Marshall for his mentorship at Washington University on peptide chemistry and to many Pfizer colleagues, e.g., Thomas Maggie, Peter Jones, Matthew Hayward, and Adam Gilbert, for their insights and fruitful discussions on the design of cyclic peptides for human diseases. References 1. Jin L, Wang W, Fang G (2014) Targeting protein–protein interaction by small molecules. Annu Rev Pharmacol Toxicol 54:435–456 2. Wright PE, Dyson HJ (2015) Intrinsically disordered proteins in cellular signalling and regulation. Nat Rev Mol Cell Biol 16:18–29 3. Mammen M, Shakhnovich EI, Whitesides GM (1998) Using a convenient, quantitative model for torsional entropy to establish qualitative trends for molecular processes that restrict conformational freedom. J Org Chem 63:3168–3175 4. Zorzi A, Deyle K, Heinis C (2017) Cyclic peptide therapeutics: past, present and future. Curr Opin Chem Biol 38:24–29 5. White CJ, Yudin AK (2011) Contemporary strategies for peptide macrocyclization. Nat Chem 3:509–524 6. Marshall GR, Bosshard HE (1972) Angiotensin II. Studies on the biologically active conformation. Circ Res 31:143–150 7. Garcia-Echeverria C, Chene P, Blommers MJ, Furet P (2000) Discovery of potent antagonists of the interaction between human double minute 2 and tumor suppressor p53. J Med Chem 43:3205–3208 8. Walensky LD, Kung AL, Escher I, Malia TJ, Barbuto S, Wright RD, Wagner G, Verdine GL, Korsmeyer SJ (2004) Activation of apoptosis in vivo by a hydrocarbon-stapled BH3 helix. Science 305(5689):1466–1470 9. Calvo JC, Choconta KC, Diaz D, Orozco O, Bravo MM, Espejo F, Salazar LM, Guzman F, Patarroyo ME (2003) An alpha helix conformationally restricted peptide is recognized by cervical carcinoma patients’ sera. J Med Chem 46:5389–5394 10. Kemp DS, Boyd JG, Muendel CC (1991) The helical s constant for alanine in water derived

from template-nucleated helices. Nature 352:451–454 11. Chang YS, Graves B, Guerlavais V, Tovar C, Packman K, To KH, Olson KA, Kesavan K, Gangurde P, Mukherjee A, Baker T, Darlak K, Elkin C, Filipovic Z, Qureshi FZ, Cai H, Berry P, Feyfant E, Shi XE, Horstick J, Annis DA, Manning AM, Fotouhi N, Nash H, Vassilev LT, Sawyer TK (2013) Stapled alpha-helical peptide drug development: a potent dual inhibitor of MDM2 and MDMX for p53-dependent cancer therapy. Proc Natl Acad Sci U S A 110:E3445–E3454 12. Madala PK, Tyndall JD, Nall T, Fairlie DP (2010) Update 1 of: proteases universally recognize beta strands in their active sites. Chem Rev 110:PR1–P31 13. Loughlin WA, Tyndall JD, Glenn MP, Hill TA, Fairlie DP (2010) Update 1 of: beta-strand mimetics. Chem Rev 110:PR32–PR69 14. McCauley JA, McIntyre CJ, Rudd MT, Nguyen KT, Romano JJ, Butcher JW, Gilbert KF, Bush KJ, Holloway MK, Swestock J, Wan BL, Carroll SS, DiMuzio JM, Graham DJ, Ludmerer SW, Mao SS, Stahlhut MW, Fandozzi CM, Trainor N, Olsen DB, Vacca JP, Liverton NJ (2010) Discovery of vaniprevir (MK-7009), a macrocyclic hepatitis C virus NS3/4a protease inhibitor. J Med Chem 53:2443–2463 15. Jiang Y, Andrews SW, Condroski KR, Buckman B, Serebryany V, Wenglowsky S, Kennedy AL, Madduru MR, Wang B, Lyon M, Doherty GA, Woodard BT, Lemieux C, Geck Do M, Zhang H, Ballard J, Vigers G, Brandhuber BJ, Stengel P, Josey JA, Beigelman L, Blatt L, Seiwert SD (2014) Discovery of danoprevir (ITMN-191/R7227), a highly selective and potent inhibitor of

106

Ye Che

hepatitis C virus (HCV) NS3/4A protease. J Med Chem 57:1753–1769 16. Ruiz-Gomez G, Tyndall JD, Pfeiffer B, Abbenante G, Fairlie DP (2010) Update 1 of: over one hundred peptide-activated G proteincoupled receptors recognize ligands with turn structure. Chem Rev 110:PR1–P41 17. Che Y, Marshall GR (2008) Privileged scaffolds targeting reverse-turn and helix recognition. Expert Opin Ther Targets 12:101–114 18. Chalmers DK, Marshall GR (1995) Pro-DNMe-amino acid and D-Pro-NMe-amino acid: simple, efficient reverse-turn constraints. J Am Chem Soc 117:5927–5937

19. Chung YJ, Huck BR, Christianson LA, Stanger HE, Krauth€auser S, Powell DR, Gellman SH (2000) Stereochemical control of hairpin formation in β-peptides containing dinipecotic acid reverse turn segments. J Am Chem Soc 122:3995–4004 20. Finch AM, Wong AK, Paczkowski NJ, Wadi SK, Craik DJ, Fairlie DP, Taylor SM (1999) Low-molecular-weight peptidic and cyclic antagonists of the receptor for the complement factor C5a. J Med Chem 42:1965–1974 21. Liu H, Kim HR, Deepak R, Wang L, Chung KY, Fan H, Wei Z, Zhang C (2018) Orthosteric and allosteric action of the C5a receptor antagonists. Nat Struct Mol Biol 25:472–481

Chapter 7 Design and Synthetic Strategies for Helical Peptides Licheng Tu, Dongyuan Wang, and Zigang Li Abstract Abnormal protein–protein interactions (PPIs) are the basis of multiple diseases, and the large and shallow PPI interfaces make the target “undruggable” for traditional small molecules. Peptides, emerging as a new therapeutic modality, can efficiently mimic PPIs with their large scaffolds. Natural peptides are flexible and usually have poor serum stability and cell permeability, features that limit their further biological applications. To satisfy the clinical application of peptide inhibitors, many strategies have been developed to constrain peptides in their bioactive conformation. In this report, we describe several classic methods used to constrain peptides into a fixed secondary structure which could significantly improve their biophysical properties. Key words Stapled peptides, Hydrogen bond surrogates (HBSs), Cross-linking, Protein–protein interactions, Helicity, Secondary conformation

1

Introduction Protein–protein interactions (PPIs) play pivotal roles in mediating intracellular biological processes. The targeting of dysfunctional PPIs is broadly utilized for therapeutics development [1]. However, small molecule ligands are generally incapable of interrupting PPIs with large, shallow, or discontinued surfaces, making many PPIs “undruggable” [2]. Compared with small molecule ligands, peptides have larger interaction areas and are broadly utilized as PPI modulating probes [3]. Based on a statistical analysis of known PPIs, over 50% of them involve α-helices interactions [4]. Numerous peptide-based drugs are on the market, including insulin, glucagon-like peptide 1 (GLP-1) analogues, and hormones. However, in general, peptides suffer from poor stability and cell permeability and are believed to be less druggable than small molecules [3]. Over the past two decades, peptide stapling technology has been well developed to stabilize peptides into a

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_7, © Springer Science+Business Media, LLC, part of Springer Nature 2019

107

108

Licheng Tu et al.

fixed secondary structure with significantly enhanced cellular uptake and stability [5]. Successful attempts at applying these chemically stabilized peptides are reported with various intracellular PPIs, and stapled peptide Ph-substituted and R configuration (PhR) targeting the p53/Mdm2 interaction has recently reached a phase III clinical trial [6]. In general, peptide stabilization methods can be divided into two categories: side chain–side chain cross-links or N-terminal capping [7]. The preorganization of peptides into their functional conformation shows improvement in target binding affinity and stability [8]. This chapter will briefly introduce the preparation of commonly used stabilized peptides.

2

Materials and General Method

2.1 Reagents and Method

DCM

Dichloromethane

DIC

N,N0 -Diisopropylcarbodiimide

DIPEA N,N-Diisopropylethylamine DMF

N,N-Dimethylformamide

EDTA

Ethylenediaminetetraacetic acid

FITC

Fluorescein isothiocyanate

Fmoc

9-Fluorenylmethyloxycarbonyl

HCTU 2-(1H-6-chlorobenzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate HoAt

1-Hydroxy-7-azabenzotriazole

HOBt

1-Hydroxybenzotriazole

MAP

4-methoxyacetophenone

MMP

2-hydroxy-4’-(2-hydroxyethoxy)-2-methylpropiophenone

NMM

N-Methyl morpholine

PyBOP Benzotriazol-1-yl-oxytripyrrolidinophosphonium hexafluorophosphate TFA

Trifluoroacetic acid

TIS

Triisopropylsilane

All solvents and reagents used for solid phase peptide synthesis were purchased from commercial suppliers including GL Biochem (Shanghai) Ltd., Shanghai Hanhong Chemical Co., J&K Co. Ltd., Shenzhen Tenglong Logistics Co., or Energy Chemical Co., and were used without further purification unless otherwise stated.

Design and Synthetic Strategies for Helical Peptides

2.2 General Method for Solid-Phase Peptide Synthesis

109

Currently, most laboratory-scale peptide synthesis is carried out with the Fmoc strategy, mainly for safety reasons. In general, peptide synthesis is performed on resins (Rink Amide MBHA resin and Wang resin are commonly utilized) by standard Fmoc-based solidphase peptide synthesis (Fig. 1). Using Rink Amide MBHA resin as an example: 1. Rink amide MBHA resin is swelled with DCM for 10 min. 2. Fmoc deprotection is performed with morpholine (50% in DMF) for 30 min (repeated if necessary). Then the resin is washed with DMF (five times), DCM (five times), and DMF (five times). 3. Fmoc-protected amino acids (6.0 eq. according to initial loading of the resin) and HCTU (5.9 eq.) are dissolved in DMF, then DIPEA is added (12.0 eq.). The mixture is pre-activated for 1 min and added to the resin for 1–2 h, and then the resin is washed with DMF (five times), DCM (five times), and DMF (five times). 4. Upon completion of peptide assembly, if necessary, peptides can be N-terminally modified.

2.2.1 Acetylation

1. Swell the Fmoc-deprotected resin in DMF for 10 min and drain the solvent. 2. Add a solution of acetic anhydride and DIPEA in DMF (1:1:8 by volume) to the resin. Gently agitate the mixture with bubbling under nitrogen for 1 h and then drain. 3. Wash the resin and reaction vessel thoroughly from the top with DCM while applying vacuum to wash off residual reaction solution.

Fig. 1 Schematic presentation of standard 9-fluorenylmethyloxycarbonyl (Fmoc)based solid-phase peptide synthesis

110

Licheng Tu et al.

4. For cleavage or storage (cleavage cocktail: TFA/TIS/H2O 95/2.5/2.5, 1 mL per 100 mg resin), gently agitate the reaction on an orbital shaker in a fume hood at room temperature for 2 h. Remove most of the TFA by evaporation under a stream of nitrogen, then add diethyl ether to precipitate the products. Finally, air-dry the residue and add 50% (vol/vol) aqueous acetonitrile to combine the filtrates for highperformance liquid chromatography (HPLC) purification. For storage, wash the resin with DCM three times and then wash with methanol to shrink the resin. Dry the resin under a stream of nitrogen. 2.2.2 FITC Labeling

For FITC-labeled peptides, a β-Ala residue needs to be added to the N-terminus of the peptide before FITC attachment, to avoid Edman degradation [9]. 1. Swell the Fmoc-deprotected resin in DMF for 10 min and drain the solvent. 2. Couple Fmoc-β-Ala-OH. 3. Wash the resin and deprotect the Fmoc group using the method described above. 4. Wash the resin and reaction vessel thoroughly from the top with DCM while applying vacuum to wash off residual reaction solution. 5. Prepare a solution of FITC (7 eq.) and DIPEA (14 eq.) in DMF and add to the resin. 6. Cap the top of the vessel and gently agitate on a shaker with protection from light at 20–25  C for 12–24 h. 7. Then, wash the resin and reaction vessel thoroughly from the top with DMF while applying vacuum to wash off residual reaction solution. 8. For cleavage or storage, use the same procedure as for acetylation.

2.3

Notes

1. It should be noted that N,N-dimethylformamide (DMF), DCM, and HCTU are all toxic and harmful on inhalation, ingestion, or by skin contact. In addition, all organic solvents are flammable, and diethyl ether is highly flammable [10]. Trifluoroacetic acid (TFA) and acetic anhydride are corrosive. Therefore, all organic solvents and chemicals should be handled with appropriate personal protective equipment (nitrile gloves, lab coat, protective eye glasses) and handled inside a chemical fume hood.

Design and Synthetic Strategies for Helical Peptides

111

2. A single coupling reaction with HCTU and a 45-min reaction time is used for most standard amino acids. However, a longer coupling time (e.g., 90 min) is recommended if coupling after α-methyl α-alkenyl amino acids [10]. For the coupling of sterically hindered amino acid derivatives immediately following an α-methyl, α-alkenyl amino acid, the use of double coupling with HCTU is advised. For the α-methyl, α-alkenyl amino acids, a single 2-h coupling reaction is used with HCTU. 3. HCTU is a light-sensitive reagent and therefore needs to be appropriately protected, by covering the container with aluminum foil or by using an amber-colored glass container. Since the present protocol uses the same volumes of the stock solutions of amino acid and coupling reagent for activation, it is convenient to prepare a slightly lower concentration for HCTU solution than that for amino acid (0.38 M versus 0.4 M) [10]. 4. For cleavage, the resin must be thoroughly washed and completely dried, as reagent or solvent carry-overs may affect cleavage reactions; for example, DMF will restrain the TFA-mediated cleavage of the peptides and residual methanol will cause undesired esterification of the carboxyl group of aspartate and glutamate residues. 5. TFA is highly corrosive; wear protective clothing and work in an efficient fume hood.

3 3.1

All-Hydrocarbon-Linked Peptides Stapled Peptides

3.1.1 Introduction

An all-hydrocarbon cross-linked peptide constructing strategy, named “stapled peptides,” was reported by Schafmeister et al. (Verdine’s group) in 2000, using olefin metathesis [11]. The two main features of this strategy include the methylation of the α carbon atom of the non-natural amino acids at the bridgeheads and the utilization of olefin metathesis to achieve clean conversion. For example, staples between the positions of i and i þ 4, two unnatural amino acids, should be S and S absolute configuration at the α-carbon, and the optimal tether length should be 8 atoms. While for the i, i þ 7 staple, the unnatural amino acids should be R and S absolute configuration, and the optimal tether length should be 11 atoms. Peptides stapled with this method showed good helix enhancement, improved anti-proteolytic degradation ability, and increased cellular uptake. The group of Verdine et al. also developed i, i þ 3 stapled peptides, which also enhanced cellular uptake, but to a lesser extent than their i, i þ 4 analogues [12–14]. In 2014, Verdine et al.

112

Licheng Tu et al.

S5

B5 2

2

2

S8 5

S5

S5 2

i, i+4, i+11 stitch

n

n

R5

2

i, i+4

S5

R8

S5

2

5

2

i, i+7

i, i+3 S5: n=2 S8: n=5 Fmoc

n N H

COOH

R5: n=2 R8: n=5 Fmoc

n N H

COOH

B5 Fmoc

N H

COOH

Fig. 2 Schematic representation of stapled peptides

further expanded their concept of “stapled peptides” and developed the second generation of the all-carbon side chain system, the so-called stitched peptide [15]. The construction of stitched peptides beautifully utilized the selectivity of olefin metathesis and precisely arranged unnatural amino acids to construct two consecutive all-hydrocarbon side chains. The constructed stitched peptides exhibited better stability and membrane penetration than linear peptides [15]. This type of polypeptide resembles a mini-protein with a compact secondary structure, large interacting surface area, and enhanced cellular uptake. This method greatly expands the chemical space of peptides and has been broadly utilized in various biological applications (Fig. 2) [16]. 3.1.2 Synthetic Methods

1. For natural amino acids, the coupling is performed according to the ‘General Method’ above. 2. For unnatural amino acids, in a vial, mix 3 eq. of a 0.4 M amino acid stock solution for a single 30-μM reaction (0.03 mM per coupling  3 eq.  1 coupling  1000 μL/mL  0.4 mM/ mL) with the same volume of 0.4 M HCTU stock solution. 3. Add 6 eq. of DIPEA (0.03 mM per coupling  6 (eq.)  1 coupling  129.25 mg/mM  0.742 mg/μL) and thoroughly mix the resulting solution. 4. Add the resulting activated amino acid solution to the Fmocdeprotected peptide resin, gently agitate with nitrogen bubbling for 2 h, and then drain the solution by vacuum filtration. 5. Repeat steps 1–3 if double coupling is necessary.

Design and Synthetic Strategies for Helical Peptides

113

Fig. 3 Synthetic route lines of stapled peptide

6. For the ring-closing metathesis (RCM) reaction, the RCM is carried out on resin-bound, fully protected peptides in 1,2-dichloroethane (DCE) at room temperature (20–25  C), using Grubbs’ first-generation catalyst. Treat the resin (30 μM) with 6 mM solution of Grubbs’ first-generation catalyst in DCE (4.94 mg/mL; 20 mol% with regard to the resin substitution). 7. Further modification, if necessary, and peptide cleavage and purification are conducted as indicated in ‘General Method’ above (Fig. 3). 3.1.3 Notes

1. In most cases, the RCM provides yields of greater than 90% after the first round of the 2-h metathesis and is completed after the second round of catalyst treatment [10]. We recommend confirming completion of the RCM before carrying out further modifications. The progress of the metathesis can be monitored by performing analytical cleavage of the peptide products. 2. Nitrogen bubbling is recommended for the metathesis reaction, as it removes the ethylene by-product from the reaction and drives the RCM to high conversion. To minimize solvent evaporation, set the nitrogen pressure as low as possible for bubbling.

114

Licheng Tu et al.

Fig. 4 N-terminal hydrogen bond surrogate (HBS) system

3.2 Hydrogen Bond Surrogates (HBSs) for N-Terminal Helix Nucleation 3.2.1 Introduction

3.2.2 Synthetic Methods

Another commonly utilized method of constructing fixed secondary structure peptides is N-terminal nucleation [17]. The α-helix is usually characterized by a 13-membered intramolecular hydrogen bond between the i C¼O and the i þ 4 NH. Thus, chemically installing a suitable covalent linker to replace the intramolecular hydrogen bond may constrain peptides into fixed conformations. The replacement of putative hydrogen bonds in an α-helix with a covalent bond was first envisioned and developed by Cabezas and Satterthwait in 1999 [18]. In 2004, Arora’s group developed an N-terminal hydrogen bond surrogate (HBS) strategy to replace an intra-helix hydrogen bond with a covalent bond constructed by an olefin metathesis reaction, as shown in Fig. 4 [16, 19–21]. The i, i þ 4 hydrogen bond in an N-terminal helix was replaced with a covalent carbon-carbon bond, which significantly improved nucleation and allowed the peptide to maintain a helical conformation in solution. This method can be applied to construct peptide modulators of various intracellular PPIs, including HIF-1α, Ras, and p53/Mdm2 [22–26]. With the HBS, the side chain amino acid residues are protected from loss during cyclization, without affecting the molecular recognition of the target. This is especially important when the α-helix participates in multiple face recognition or is completely embedded inside the target [26]. Therefore, N-terminal nucleation and the side chain coupling strategy complement each other (Fig. 4). 1. For natural amino acids, the coupling is performed according to the ‘General Method’ above. 2. For the RCM reaction, this reaction is commonly performed in refluxing DCM with Hoveyda-Grubbs catalyst. 3. Coupling of Fmoc-amino acids to N-allylglycine peptide: (a) Preparation of solution: dissolve the desired Fmoc amino acid (2.0 mM), DIC (310 μL; 2 mmol), and HoAt

Design and Synthetic Strategies for Helical Peptides

115

Fig. 5 Synthesis of an internally constrained α-helix

(136 mg; 1 mM) in DMF (3–5 mL) and allow and stir the solution for 15 min at room temperature. (b) Next, transfer the resin bearing the N-allylglycine monomer to a microwave vessel equipped with a cap and magnetic stir bar. Treat the resin with the prepared solution and irradiate the microwave vessel. (c) Synthesis of HBS helix: Weigh 12.5 mg of the Hoveyda-Grubbs II catalyst (0.20 moles of catalyst for each mole of bis-olefin) and add to the microwave vessel containing the bis-olefin peptide Allow the nitrogen gas to flow for 30 min. Under nitrogen, add 2 mL anhydrous 1,2-DCEper 0.10 mol of resin and stir for 15 min. Irradiate the microwave vessel containing the reaction mixture. Next, wash the resin and carry out the necessary posttreatment according to your own needs (Fig. 5). 3.2.3 Notes

1. Resin needs to be extensively dried for this reaction. 2. Use freshly distilled 1,2-DCE [17].

116

4

Licheng Tu et al.

Thiol-Ether

4.1 Nucleophilic Reaction 4.1.1 Introduction

4.1.2 Synthetic Methods

Cysteine, an amino acid with high nucleophilicity, undergoes facile modifications. Cysteine-based peptide stabilization strategy has undergone considerable development in recent years. The nucleophilicity of the thiol group of cysteine can electively react with an electrophile, thus providing a plurality of molecules containing a biphilic functional group for side chain coupling. Well-known examples of this strategy include the reaction of cysteine at different positions with dibromo-m-xylene [27, 28]. In 2013, Spokoyny et al. reported perfluoroaryl-cysteine based nucleophilic aromatic substitution reaction (SNAr) nucleophilic reactions to construct stabilized peptides at the i, i þ 4 and i, i þ 7 positions [29]. Compared with the linear peptide, the perfluoroaryl-stapled peptide shows significantly enhanced binding, cell and blood-brain barrier permeability, and proteolytic stability properties. In 2015, Vinogradova et al. developed a palladium-catalyzed cysteine-based bioconjugation reaction (Fig. 6) [30]. 1. The target linear peptide was synthesized by solid-phase peptide synthesis. 2. To a solid sample of peptide (7.5 μM) in a plastic Eppendorf tube was added 1.9 mL of 100 μM solution (~25 eq.) of hexafluorobenzene in DMF and 1.5 mL of 50 mM solution

Fig. 6 Conformational stabilization of α-helical and cyclic peptides by the stapling together of two cysteines with a suitable cross-linker

Design and Synthetic Strategies for Helical Peptides

F

F

F F

F

F F

F

F F

F F

F

F

F

F F F

HS

F

SH F

F

117

F

F

F

F

F F

F

Fig. 7 Perfluoroarylation of cysteine residues in an unprotected peptide

of TRIS base in DMF. The contents of the tube were vigorously mixed on a shaker for 30 s and left at room temperature for 4.5 h. 3. The reaction mixture with peptides was characterized by liquid chromatography-mass spectrometry (LC-MS). Finally, the peptides were separated and purified by HPLC. 4. For the rest of the linker (shown in Fig. 6), the cross-linking reactions were carried out by incubating the purified dicysteine-containing peptides with 1.5 eq. of linker in a mixed solvent of acetonitrile/30 mM NH4HCO3 buffer (1:4–2:3 depending on solubility), pH 8.5, to obtain a final peptide concentration of 1 mM. The mixture was stirred at room temperature for 1.5–2 h. 5. Afterwards, the solvents were evaporated and the excess amount of cross-linker was removed by washing the residue with diethyl ether. Then the residue was purified by preparative reverse-phase (RP) HPLC to give the purified peptide (Fig. 7). 4.1.3 Notes

1. The nucleophilic substitution reaction mentioned here is carried out under weak basic conditions. 2. The linker should be in excess of the polypeptide, around 1.5 eq [29].

4.2 Light-Induced Intermolecular Thiol-ene or Thiol-yne Reaction 4.2.1 Introduction

The thiol-ene reaction has gained prominence in recent years for its feasibility and excellent functional group tolerance [31–33], which allows a broad range of applications in the syntheses of polymers and other materials [34–37]. Because of its bio-orthogonal nature, the light-induced thiol-ene reaction has also been widely used in protein and peptide modification [38]. With the reactive nucleophilic sulfhydryl group, cysteine is considered as a useful active site for biocompatible cross-linking. The bioconjugation of peptide side chains can be accomplished between sulfhydryl and olefin groups under ultraviolet (UV) light irradiation. This mild reaction is also called “thiol-ene” coupling, which is triggered by a radical initiator from the sulfhydryl group. In 2015, Wang and Chou reported the discovery of a peptide

118

Licheng Tu et al.

SH

hn,365 nm photo-induced

SH +

S

S

on-resin intra-molecular thiol reaction

SH n

365 nm Photo-induced intramolecular thiol-yne coupling m

S n

m

n=0,1 m=3

Fig. 8 Preparation of cross-linked peptides via intramolecular thiol-ene and thiol-yne reaction

stapling and macrocyclization methodology using thiol-ene reactions between two cysteine residues and a diene, with high yields [39]. This new approach was able to selectively modify cysteine residues in native, unprotected peptides with a variety of stapling modifications for helix stabilization or general macrocyclization. By introducing the alkyne group in unnatural amino acids and cysteine at the i, i þ 4 site, 365-nm UV irradiation could initiate high efficiency ring closure [40]. This strategy can stabilize the peptide into a regular α-helical conformation, which is stable at high temperatures and high pH. In addition, the thiol-ether cross-linker increases the serum stability and cell penetration of the stabilized peptide (Fig. 8). 4.2.2 Synthetic Methods

1. The solid-phase peptide synthesis (SPPS) was performed by the above protocol. The following Cys(Trt) deprotection was carried out with 3% TFA/5% TIS in DCM (4  10 min) until the drained solution was no longer yellow [38]. 2. After being dried with MeOH under vacuum, the resin was used in the intra-molecular thiol-ene reaction. The resin was suspended in DMF, and the photoinitiator 3,2-hydroxy-1[4-(2-hydroxyethoxy)phenyl]-2 -methyl-1-propanone (1 eq.) and MAP. (1 eq.) were added. 3. The solution was degassed with N2 and subsequently irradiated with UV light at room temperature for 1 h.

Design and Synthetic Strategies for Helical Peptides

119

1. Thiol-ene and thiol-yne are most efficient at the 365-nm wavelength [38].

4.2.3 Notes

2. The Cys (trt) protection group has to be cleaned. Recently, the effects of in-tether modifications have been intensively studied in established methodologies [5]. Li et al. discovered that the introduction of a chiral center in the tether of a thioetherstabilized peptide could increase the rigidity of the peptide and help to stabilize a peptide to the α-helix. The protocol for synthesizing chiral-center peptides is similar to the thiol-yne method and the two epimers can be easily separated by HPLC. Although R/S epimers share the same chemical sequence, their biophysical properties and helical content are very different. Compared with the S epimer, which is a largely random coil, the R epimer exhibits a helical conformation, as confirmed by X-ray and nuclear magnetic resonance (NMR) analysis. In addition, the R epimer shows significantly increased serum stability, cell permeability, and target binding affinity. The improvement of biophysical properties is attributed to the introduction of the new chiral center, which gives this strategy its name: chiral-center-induced helicity (CIH) strategy (Fig. 9). Subsequently, the concept of a chiral-induced helix has been further extended to sulfoxide [5, 41], sulfilimine [42, 43], and sulfonium [44] systems. Similar patterns were also observed in an i, i þ 7 system [45]. Notably, the additional chiral center can serve as a handle for the secondary modification of the peptides for further diversity of modification [46]. The CIH strategy was utilized to construct peptide modulators for the KLA(K: Lysine, L: Leucine, A: Alanine) apoptotic peptide and the p53/Mdm2 interaction [47].

4.3 4.3 ChiralCenter-Induced Helicity (CIH) Strategy 4.3.1 Introduction

NHTs S i+4

R H

S i

i

i+4

i

S

S isomer:random coil

i+4

S isomer:random coil

.. H i

S

random coil

R

i

S

i+4

R isomer:a-helix R isomer:a-helix In-tether chirality induced conformation change

Fig. 9 Chiral center-induced peptide helicity

NHTs i+4

120

Licheng Tu et al.

4.3.2 Synthetic Methods

1. The CIH peptides were synthesized on Rink Amide MBHA resin (loading capacity: 0.54 mmol/g) with a standard solidphase peptide synthesis protocol. 2. The thiol-ene cyclization step was conducted under 365-nm UV light with photo initiators MAP/MMP (1:1) for 2 h. After cleavage and precipitation, the peptide was isolated by centrifugation. 3. The resulting residue was dried in vacuo and dissolved in degassed DMF to reach a concentration of 0.5 mM (based on resin substitution). Photo initiator (0.5 eq.) was added and the solution was degassed with N2 and subsequently irradiated with UV light at room temperature for 0.5–1 h without agitation. 4. The crude peptide solution was purified by RP-HPLC, as mentioned above, with a linear gradient of 5–45% acetonitrile in 50 min. Fractions containing the desired peptide were combined and lyophilized.

4.3.3 Notes

1. DMF in step 3 was then evaporated and the crude residue was dissolved in water, followed by the addition of ether to dissolve the organic by-product [47]. 2. Precipitate with diethyl ether to remove some impurities that are soluble in ether.

5

Click Chemistry

5.1 Cu(I)-Catalyzed Azide-Alkyne Cycloaddition (CuAAC) 5.1.1 Introduction

The Cu(I)-catalyzed azide–alkyne cycloaddition (CuAAC) reaction, or “click reaction,” is a regioselective reaction between alkyne and azide functional groups, which gives rise to 1,4-disubstituted 1,2,3-triazoles under mild conditions: pH=7, room temperature. Triazole can be used as an amide bond surrogate because of its similarity to the molecular dimensions of amide bonds in terms of distance and planarity [48]. Huisgen first reported the 1,3-dipolar cycloaddition between an azide and a terminal or internal alkyne to generate 1,2,3-triazole [49]. Kolb and Sharpless and Torna et al. “re-invented” this reaction under mild conditions and termed this cycloaddition a “click reaction” [50, 51]. In the past two decades or so, the alkyne and azide cycloaddition reactions have been broadly utilized in various areas, including peptide stabilization [52]. In 2011, Wang et al. reported the use of copper-catalyzed terminal alkyne and azide cycloadditions to construct a series of α-helical peptides that interacted with β-catenin/ Bcl-9 [53]. Lau et al., in 2014, reported a two-component terminal alkyne/azide cycloaddition reaction. With this method, it is

Design and Synthetic Strategies for Helical Peptides

121

convenient to introduce the side-chain-regulating peptide with different modifications to further improve the biological effect of the cell-penetrating ability (Fig. 10) [54–56]. 5.1.2 Synthetic Methods

1. A solution of diazide peptide (1 eq.) and dialkynyl linker (1.1 eq.) in 1:1 tert-butanol/water (1 mL/mg peptide) was degassed with nitrogen for 15 min, followed by the addition of copper(II) sulfate pentahydrate (1 eq.), tris(3-hydroxypropyltriazolylmethyl)amine (1 eq.), and sodium ascorbate (3 eq.). 2. After stirring under nitrogen at room temperature for 16 h, the reaction mixture was lyophilized and purified by HPLC to give the final stapled peptide (Fig. 11). 1. The monovalent copper ion achieves the purpose of catalysis by forming a coordination compound with an alkyne group, and it is necessary to add barium sulfate to the copper sulfate [48].

5.1.3 Notes

linker

N3

N3

N

N N

N N N

linker

Cu(I)

Fig. 10 Diazidopeptide is combined with a dialkynyl linker under Cu(I) catalysis to give a bis-triazole stapled peptide

X

N N3

N3

X

O

(Arg)3 NH2

X= H CuSO4.5H2O Sodium ascorbate THPTA t-BuOH/H2O(1:1)

Fig. 11 Double-click stapling of peptides with tunable linkers

N N

N N

N

122

6 6.1

Licheng Tu et al.

Amide Bonds Proteins and peptides are composed of amide bonds, the formation of which is the most important reaction in peptide synthesis. Thanks to various orthogonal protection/deprotection strategies, amide bond formation is commonly utilized to construct cyclic lactam-bridged peptides. In 1990, Osapay and Taylor reported a series of systematic studies on different pairs in stabilizing helical peptides, via both single and multiple bridges [57]. In 2005, Shepherd et al. reported work that determined a Lys-Asp linkage as a robust amide bond cross-linking to stabilize short helices [58]. They also reported short peptides with five amino acid residues (one helix unit) to exclude the influence of non-helical residues [58]. These are the smallest alpha helical peptides in water. In 2016, Zhao et al. developed a helix nucleation template based on cross-linked aspartic acid (terminal aspartic acid; TD) [59]. The TD strategy primarily increases permeability through conformational constraints rather than lipophilic adjustments [15]. Therefore, the TD strategy provides a good platform to study how mutations in short helices affect their cell penetration and translocation into the nucleus. Recently, Tian et al. found a proline-derived N cap as a helical nucleation template in a variety of bio-related peptide sequences by macrocyclic lactamization on the resin. The method utilizes the N-capping nature of proline and the synergistic stabilization of the main chain hydrogen bond and covalent bond substitution (Figs. 12, 13, and 14) [60].

Introduction

1. Solid-phase peptide synthesis (SPPS) was performed by the standard protocol outlined in the ‘General Method’ section.

6.2 Synthetic Methods

2. Upon completion of peptide assembly, allyl and alloc protecting groups were deprotected by Pd(PPh3)4 (0.05 eq.) and N-methyl-D-aspartate (NMDA; 4.0 eq.) dissolved in DCM. 3. Peptide cyclization was performed with PyBOP (2.4 eq.), HOBt (2.4 eq.), and NMM (4.0 eq.) in DMF. Peptides were cleaved by TFA/TIS/H2O (95:2.5:2.5) and precipitated in diethyl ether. NHAlloc NH2

COOAllyl 3 Pd(PPh3)4 i

O

NH

COOH 3

3

PyBOP,HOBt

i+4 i

i+4

Fig. 12 Shepherd et al.’s [58] TD stabilization of peptide

i

i+4

Design and Synthetic Strategies for Helical Peptides

123

O

O

O

O R

R

R

N H

NH O

R

O R N H

R

O

O

NH O

Dap

O R

N H

O

R

N H

O

R

N H NH

N H OO

NH diacid

N-cap template

R crosslinking and subsitution tuning

Fig. 13 Cross-linked diacids as N-cap nucleating templates

O

R O

AllocHN

HN O

O N H

FmocHN

COOH

SPPS

R

COOH

N H

N HN

O

R

1.Deprotection 2.Macrolactamization 3.Acylation

R

R

AllocHN

– HN HN O O

O

N H

R

N HN

O R

HN COOAllyl

O

AllylOOC HN

R

O

HN O

HN O NHAc

Fig. 14 Proline-derived transannular N-cap peptides synthesized efficiently via macrocyclic lactamization on resin

4. The resulting residue was then dried in vacuo and dissolved in 25% (v/v) CH3CN/H2O. Crude peptides were purified on RP-HPLC, confirmed by LC-MS (Fig. 15). 6.3

Notes

1. The two protection groups allyl and alloc require Pd(PPh3)4 as a catalyst to be removed under anhydrous and anaerobic conditions. Since Pd(PPh3)4 is in contact with water, it fails to catalyze the reaction [60]. 2. After deprotection, the residual palladium is washed away with a saturated solution of sodium diethyldithiocarbamate (trihydrate) in DMF. Wash several times for about 30 min until the color no longer fades.

124

Licheng Tu et al.

FmocHN

O

H N

O

SPPS

N H

R

Fmoc/tBu strategyFmocHN

R

O

H N

R N H

O AllocHN

O

H N O

R

Rink Amide MBHA Resin OAllyl

O

2 times

O

N H

R

FmocHN

R

O

H N

O Pd(PPh3)4

H N

O

R N H

O H2N

O

H N O

R

OH

O

O

H N

PyBOP,HOBt R FmocHN

R N H

H N O

O

N H

R N H

O

H N O

R

O R=residue

Fig. 15 Synthesis method of amide bond ring

7

Characterization of Helical Peptides Commonly utilized techniques to determine peptides’ secondary structures include circular dichroism (CD) spectroscopy, X-ray crystallography, and NMR spectroscopy. Among common secondary structures, the helix is the best studied structure, in part because of its highly compact and well-structured features. The molecular structures of many protein complexes have been unlocked by the technique of X-ray crystallography [61]. Later, NMR was also applied to reveal the molecular structure of protein complexes and to facilitate the identification of weak PPIs [62].

7.1 Circular Dichroism

Circular dichroism (CD) is a simple and convenient method for analyzing the secondary structure of proteins and peptides. Only chiral molecules can exhibit characteristic absorption bands in their CD spectra, while randomly oriented systems have no CD intensity. Proteins and peptides consisting of chiral amino acid residues exhibit characteristic CD uptake, according to their different secondary structures. Owing to the sensitivity and convenience of CD measurements, CD has become a routine tool for determining protein and peptide structure [63]. The characteristic structural features of the alpha helix include the i/i þ 4 hydrogen bond mode. The 3.6 amino acid residues per turn, the 100 turn angle, ˚ translational features between each residue allow the and the 1.5-A α-helix to be distinguished from other structural motifs in the protein and peptide. The α-helix shows two separate negative maxima signals, at 222 nm (n ! π* transition) and 208 nm (part of the π ! π* transition), and another positive signal (part of π) at 195 nm; π* transition). Other common structural unit β-sheets

Design and Synthetic Strategies for Helical Peptides

a

125

b 4

q (deg cm2 dmol–1)

4×10

S 2×10

Me

1a 1b

4

1a(S) Ac– –C– –A– –A– –A– –S5– –

0

1b(R)

–2×104 –4×104

S5(2-Me):

Me

NH2

200 220 240 wavelength (nm)

COOH

c Entry

Peptide

[θ]222

[θ]205

[θ]190

Helicity

1b

Cyclo-CAAAS5(2-Me)

–11792

–27010

17974

0.87

Fig. 16 Helicity enhancements with an in-tether chiral center. (a) Circular dichroism (CD) spectra of cyclic pentapeptides 1a/1b in phosphate-buffered saline (PBS; pH ¼ 7.0) at 20  C. (b) Constrained peptide preparation. (c) Molar ellipticities and percentage of helicity of peptide 1b in PBS (pH ¼ 7.0) at 20  C

show a negative band at about 220 nm compared with the characteristic absorption of an α-helix, showing another positive band at about 200 nm in the corresponding CD spectrum. It is worth noting that CD-based tools predict that α-helices in proteins and peptides are very sensitive to the environment, and the CD spectrum itself is not often used as conclusive evidence of α-helices (Fig. 16). 7.2 X-ray Crystallography

X-ray crystallography is a technique for determining the precise atomic and molecular structure of a crystal, in which a crystalline atom causes an incident X-ray beam to be diffracted into many specific directions. This method has been used to reveal the structure and function of many biomolecules, including drugs, proteins, and nucleic acids [64]. Although X-ray crystallography requires more specific sample preparation work and expertise than CD spectroscopy, it can provide a more detailed and more accurate presentation of the secondary structure of the peptide. Single-crystal X-ray crystallography technology usually requires three basic steps: 1. Peptide crystals are obtained for size requirements (usually greater than 0.1 mm for all sizes), regularity, and purity, and the crystals should show no significant internal defects such as cracks or twins. This is usually the most difficult step, and twins are the main obstacle to using this technique for peptides. It may be helpful to use a mixture of peptide enantiomers;

126

Licheng Tu et al.

Fig. 17 Thermal ellipsoid and backbone H-bonds in a crystal structure of pentapeptide 10b: cyclo-Ac-CAAIS5(2-Me)-NH2

2. The peptide is irradiated with a finely focused monochromatic X-ray beam to achieve reasonable reflections; 3. Two-dimensional (2D) images taken in different directions are converted into 3D models. The transformation is based on the electron density within the crystal, determined by the Fourier transform mathematical method. These data are then combined with the calculations for determining refinement (Fig. 17). 7.3 Nuclear Magnetic Resonance (NMR)

X-ray crystallography requires suitable crystals that are typically prepared under specific and sometimes extreme conditions, whereas NMR can reflect the real-time structure of the peptide in solution. In 1D and 2D NMR spectroscopy, nuclear overhauser effects (NOEs), temperature dependence, and the coupling constants of NMR spectroscopy are three important indicators for determining the secondary structure of a peptide [65]. The measurement of the chemical shift index and the exchange rate with the nucleus are also two conventional methods for the conformational analysis of peptides. There is a very close correlation between the NOEs and the distance between protons. Typically, the NOE signal between two protons can occur when the spatial distance is less than 5 A˚, indicating strong evidence of secondary structure [66]. In the rotating frame nuclear overhauser effect spectroscopy (ROESY) spectrum, some representative non-sequential medium cross-peaks, such as dalphaN(i, i þ 4), dalphaN(i, i þ 3), and dalphabeta(i, i þ 3) support a helical conformation [55]. Among these different types of coupling, the homonuclear 3 JHNCalphaH is easy to measure and is therefore particularly important. However, according to the Karplus curve, although 3JHNCalphaH depends on the angle φ, there is no one-to-one

Design and Synthetic Strategies for Helical Peptides

127

correspondence between the measured coupling constant and the bond angle. In fact, in most cases, a coupling constant can respond to four different bond angles φ. Therefore, a convincing conclusion from the 3JHNCalphaH measurement is ambiguous and difficult to implement, so it can be considered as a complementary method to analyze the conformation of the peptide based on the bond angle φ [67]. In addition, observations of 1H NMR chemical shift characteristics are very helpful in determining the position and comparing the content of different secondary structure motifs in proteins and peptides. Compared with the traditional NOE-based secondary structure determination method, with this method one needs to measure only the α-CH 1H resonance distribution, and the method has proven to be accurate and useful by a large number of examples [68]. In general, amide protons that participate in intramolecular hydrogen bonding are expected to exhibit lower exchange rates in a deuterated solvent, while those protons that are not present in the hydrogen bond will be exchanged more quickly. Finally, the exchange rate with the addition of deuterium from D2O is a convenient method for measuring the molecular dynamics of the peptide (Fig. 18) [69].

Fig. 18 Nuclear overhauser effect (NOE) summary diagram of pentapeptide 10b: cyclo-Ac-CAAIS5(2-Me)-NH2 (measured in 10% D2O in H2O, 25  C). Bar thickness shows the intensity of the NOE signals

128

8

Licheng Tu et al.

Summary The discovery of bioactive molecules to efficiently target PPIs is a challenging endeavor. The rational design of peptide inhibitors to mimic the PPI epitope could be an alternative approach for drug discovery. Unmodified short peptides usually have a flexible conformation with poor binding affinity and limited stability and cellular uptake. In recent decades, many chemical methodologies have been developed to stabilize peptide secondary structures. One of the most common constraints is the side chain-to side chain lactam bridge, which can change linear peptides into a helical structure through the cyclization of a side chain carboxylic group to an amino group. This method has provided the well-defined helical structure but has limited cell permeability. Another classic method is the stapled peptide, developed by Verdine et al., who constrained the peptide’s helical structure through the orthogonal reaction of olefin metathesis [10]. This method can significantly improve peptide cellular uptake and serum stability and has been applied to target important PPIs for cancer therapy [6]. In addition to stapled peptides, another all-hydrocarbon cross-linking HBS, is the most widely used nitro-terminal helix-nucleating template. To improve a given peptide inhibitor’s drug-like properties, different methodologies are being developed, focusing on tuning the chemical structure of the side chains. Meanwhile, the structure of the constrained peptides can be identified and studied by many techniques, such as (CD, X-ray crystallography, and NMR), which help us to better understand the secondary structure of peptides and pave the way for the development of peptide drugs.

References 1. Pelay-Gimeno M, Glas A, Koch O, Grossmann TN (2015) Structure-based design of inhibitors of protein–protein interactions: mimicking peptide binding epitopes. Angew Chem Int Ed Engl 54(31):8896–8927 2. Yin H, Hamilton AD (2005) Strategies for targeting protein–protein interactions with synthetic agents. Angew Chem Int Ed Engl 44(27):4130–4163 3. Tsomaia N (2015) Peptide therapeutics: targeting the undruggable space. Eur J Med Chem 94:459–470 4. Fujiwara D, Fujii I (2013) Phage selection of peptide “microantibodies”. Curr Protoc Chem Biol 5(3):171–194 5. Hu K, Geng H, Zhang Q, Liu Q, Xie M, Sun C et al (2016) An in-tether chiral center modulates the helicity, cell permeability, and target

binding affinity of a peptide. Angew Chem Int Ed Engl 55(28):8013–8017 6. Walensky LD, Bird GH (2014) Hydrocarbonstapled peptides: principles, practice, and progress. J Med Chem 57(15):6275–6288 7. Zhao H, Jiang Y, Tian Y, Yang D, Qin X, Li Z (2017) Improving cell penetration of helical peptides stabilized by N-terminal crosslinked aspartic acids. Org Biomol Chem 15 (2):459–464 8. Lau YH, de Andrade P, Wu Y, Spring DR (2015) Peptide stapling techniques based on different macrocyclisation chemistries. Chem Soc Rev 44(1):91–102 9. Kashina AS (2015) Protein arginylation: methods and protocols. Humana Press, New York, NY 10. Kim YW, Grossmann TN, Verdine GL (2011) Synthesis of all-hydrocarbon stapled α-helical

Design and Synthetic Strategies for Helical Peptides peptides by ring-closing olefin metathesis. Nature Protocols 6(6):761–771 11. Schafmeister CE, Julia Po A, Verdine GL (2000) An all-hydrocarbon cross-linking system for enhancing the helicity and metabolic stability of peptides. J Am Chem Soc 122 (24):5891–5892 12. Kim YW, Kutchukian PS, Verdine GL (2010) Introduction of all-hydrocarbon i, i þ 3 staples into alpha-helices via ring-closing olefin metathesis. Org Lett 12(13):3046–3049 13. Kim YW, Verdine GL (2009) Stereochemical effects of all-hydrocarbon tethers in i, i þ 4 stapled peptides. Bioorg Med Chem Lett 19 (9):2533–2536 14. Shim SY, Kim YW, Verdine GL (2013) A new i, i þ 3 peptide stapling system for alpha-helix stabilization. Chem Biol Drug Des 82 (6):635–642 15. Hilinski GJ, Kim YW, Hong J, Kutchukian PS, Crenshaw CM, Verdine GL (2014) Stitched alpha-helical peptides via bis ring-closing metathesis. J Am Chem Soc 136 (35):12314–12322 16. Chapman RN, Gianluca Dimartino A, Arora PS (2004) A highly stable short α-helix constrained by a main-chain hydrogen-bond surrogate. J Am Chem Soc 126 (39):12252–12253 17. Liu J, Wang D, Zheng Q, Lu M, Arora PS (2008) Atomic structure of a short alpha-helix stabilized by a main chain hydrogen-bond surrogate. J Am Chem Soc 130(13):4334–4337 18. Cabezas E, Satterthwait AC (1999) The hydrogen bond mimic approach: solid-phase synthesis of a peptide stabilized as an α-helix with a hydrazone link. J Am Chem Soc 121 (16):3862–3875 19. Gianluca D, Deyun W, Ross NC, Arora PS (2005) Solid-phase synthesis of hydrogenbond surrogate-derived α-helices. Org Lett 7 (12):2389–2392 20. Miller SE, Kallenbach NR, Arora PS (2012) Reversible α-helix formation controlled by a hydrogen bond surrogate. Tetrahedron 68 (23):4434–4437 21. Mahon AB, Arora PS (2012) Design, synthesis and protein-targeting properties of thioetherlinked hydrogen bond surrogate helices. Chem Commun (Camb) 48(10):1416–1418 22. Henchey LK, Porter JR, Ghosh I, Arora PS (2010) High specificity in protein recognition by hydrogen-bond-surrogate alpha-helices: selective inhibition of the p53/MDM2 complex. Chembiochem 11(15):2104–2107

129

23. Wang D, Liao W, Arora PS (2005) Enhanced metabolic stability and protein-binding properties of artificial alpha helices derived from a hydrogen-bond surrogate: application to Bcl-xL. Angew Chem Int Ed Engl 44 (40):6525–6529 24. Wang D, Lu M, Arora PS (2008) Inhibition of HIV-1 fusion by hydrogen-bond-surrogatebased alpha helices. Angew Chem Int Ed Engl 47(10):1879–1882 25. Henchey LK, Kushal S, Dubey R, Chapman RN, Olenyuk BZ, Arora PS (2010) Correction to inhibition of hypoxia inducible factor 1-transcription coactivator interaction by a hydrogen bond surrogate α-helix. J Am Chem Soc 132(3):941–943 26. Patgiri A, Yadav KK, Arora PS, Bar-Sagi D (2011) An orthosteric inhibitor of the Ras–Sos interaction. Nat Chem Biol 7(9):585–587 27. Miller SE, Watkins AM, Kallenbach NR, Arora PS (2014) Effects of side chains in helix nucleation differ from helix propagation. Proc Natl Acad Sci U S A 111(18):6636 28. Muppidi A, Wang Z, Li X, Chen J, Lin Q (2011) Achieving cell penetration with distance-matching cysteine cross-linkers: a facile route to cell-permeable peptide dual inhibitors of Mdm2/Mdmx. Chem Commun (Camb) 47(33):9396–9398 29. Spokoyny AM, Zou Y, Ling JJ, Yu H, Lin YS, Pentelute BL (2013) A perfluoroaryl-cysteine S (N)Ar chemistry approach to unprotected peptide stapling. J Am Chem Soc 135 (16):5946–5949 30. Vinogradova EV, Zhang C, Spokoyny AM, Pentelute BL, Buchwald SL (2015) Organometallic palladium reagents for cysteine bioconjugation. Nature 526(7575):687–691 31. Diderich P, Bertoldo D, Dessen P, Khan MM, Pizzitola I, Held W et al (2016) Phage selection of chemically stabilized alpha-helical peptide ligands. ACS Chem Biol 11 (5):1422–1427 32. Dondoni A, Marra A (2012) Recent applications of thiol-ene coupling as a click process for glycoconjugation. Chem Soc Rev 41 (2):573–586 33. Hoyle CE, Bowman CN (2010) Thiol-ene click chemistry. Angew Chem Int Ed Engl 49 (9):1540–1573 34. Dondoni A (2008) The emergence of thiol-ene coupling as a click process for materials and bioorganic chemistry. Angew Chem Int Ed Engl 47(47):8995–8997 35. Hagberg EC, Malkoch M, Ling Y, Hawker CJ, Carter KR (2007) Effects of modulus and

130

Licheng Tu et al.

surface chemistry of thiol-ene photopolymers in nanoimprinting. Nano Lett 7(2):233–237 36. Kato LK, Campos LM, Hawker CJ (2008) Robust, efficient, and orthogonal synthesis of dendrimers via thiol-ene “click” chemistry. J Am Chem Soc 130(15):5062–5064 37. Natarajan LV, Shepherd CK, Brandelik DM, Sutherland RL, Chandra S, Tondiglia, VP (2003) Switchable holographic polymerdispersed liquid crystal reflection gratings based on thiol-ene photopolymerization. Chem Mat 15(12):2477–2484 38. Campos LM, Meinel I, Guino RG, Schierhorn M, Gupta N, Stucky GD et al (2010) Highly versatile and robust materials for soft imprint lithography based on thiolene click chemistry. Adv Mater 20 (19):3728–3733 39. Wang Y, Chou DH (2015) A thiol-ene coupling approach to native peptide stapling and macrocyclization. Angew Chem 127 (37):11081–11084 40. Tian Y, Li J, Zhao H, Zeng X, Wang D, Liu Q et al (2016) Stapling of unprotected helical peptides via photo-induced intramolecular thiol-yne hydrothiolation. Chem Sci 7 (5):3325–3330 41. Zhang Q, Jiang F, Zhao B, Lin H, Tian Y, Xie M et al (2016) Chiral sulfoxide-induced single turn peptide alpha-helicity. Sci Rep 6:38573 42. Li J, Tian Y, Wang D, Wu Y, Ye X, Li Z (2017) An in-tether sulfoxide chiral center influences the biophysical properties of the N-capped peptides. Bioorg Med Chem 25 (6):1756–1761 43. Lin H, Jiang Y, Zhang Q, Hu K, Li Z (2016) An in-tether sulfilimine chiral center induces helicity in short peptides. Chem Commun (Camb) 52(68):10389–10391 44. Shi XD, Zhao RT, Jiang YX, Zhao H, Tian Y, Jiang YH et al (2018) Reversible stapling of unprotected peptides via chemoselective methionine bis-alkylation/dealkylation. Chem Sci 9(12):3227–3232 45. Hu K, Sun C, Yang D, Wu Y, Shi C, Chen L et al (2017) A precisely positioned chiral center in an i, i þ 7 tether modulates the helicity of the backbone peptide. Chem Comm 53 (50):6728–6731 46. Hu K, Sun C, Yu M, Li W, Lin H, Guo J et al (2017) Dual in-tether chiral centers modulate peptide helicity. Bioconjug Chem 28 (5):1537–1543 47. Hu K, Yin F, Yu M, Sun C, Li J, Liang Y et al (2017) In-tether chiral center induced helical peptide modulators target p53-MDM2/

MDMX and inhibit tumor growth in stemlike cancer cell. Theranostics 7(18):4566–4576 48. Kolb HC, Finn MG, Sharpless KB (2010) Click chemistry: diverse chemical function from a few good reactions. Angew Chem Int Ed Engl 32(35):2004–2021 49. Huisgen R (1961) Proceedings of the Chemical Society. Proc Chem Soc:357–396 50. Kolb HC, Sharpless KB (2003) The growing impact of click chemistry on drug discovery. Drug Discov Today 8(24):1128–1137 51. Torna˜ ECW, Christensen C, Meldal M (2002) Peptidotriazoles on solid phase: [1,2,3]triazoles by regiospecific copper(i)-catalyzed 1,3-dipolar cycloadditions of terminal alkynes to azides. J Org Chem 67(9):3057–3064 52. Castro V, Rodriguez H, Albericio F (2016) CuAAC: an efficient click chemistry reaction on solid phase. ACS Comb Sci 18(1):1–14 53. Madden MM, Muppidi A, Li Z, Li X, Chen J, Lin Q (2011) Synthesis of cell-permeable stapled peptide dual inhibitors of the p53-Mdm2/Mdmx interactions via photoinduced cycloaddition. Bioorg Med Chem Lett 21(5):1472–1475 54. Lau YH, Andrade PD, Quah ST, Rossmann M, Laraia L, Sko¨ld N et al (2014) Functionalised staple linkages for modulating the cellular activity of stapled peptides. Chem Sci 5 (5):1804–1809 55. Lau YH, de Andrade P, Skold N, McKenzie GJ, Venkitaraman AR, Verma C, Spring DR et al (2014) Investigating peptide sequence variations for ‘double-click’ stapled p53 peptides. Org Biomol Chem 12(24):4074–4077 56. Lau YH, de Andrade P, McKenzie GJ, Venkitaraman AR, Spring DR (2014) Linear aliphatic dialkynes as alternative linkers for double-click stapling of p53-derived peptides. Chembiochem 15(18):2680–2683 57. Osapay G, Taylor JW (1990) Multicyclic polypeptide model compounds. 1. Synthesis of a tricyclic amphiphilic alpha-helical peptide using an oxime resin, segment-condensation approach. J Am Chem Soc 112 (16):6046–6051 58. Shepherd NE, Hoang HN, Abbenante G, Fairlie DP (2005) Single turn peptide alpha helices with exceptional stability in water. J Am Chem Soc 127(9):2974–2983 59. Zhao H, Liu QS, Geng H, Tian Y, Cheng M, Jiang YH et al (2016) Crosslinked aspartic acids as helix-nucleating templates. Angew Chem Int Ed Engl 55(39):12088–12093 60. Tian Y, Wang D, Li J, Shi C, Niu X, Li Z et al (2016) A proline-derived transannular

Design and Synthetic Strategies for Helical Peptides N-cap for nucleation of short alpha-helical peptides. Chem Commun (Camb) 52 (59):9275–9278 61. Zhu G (2012) NMR of proteins and small biomolecules. Springer, Berlin, Heidelberg 62. Lo Conte L, Chothia C, Janin J (1999) The atomic structure of protein-protein recognition sites. J Mol Biol 285:2177–2198 63. Vinogradova O, Qin J (2012) NMR as a unique tool in assessment and complex determination of weak protein–protein interactions. Top Curr Chem 326(326):35–45 64. Souers AJ, Leverson JD, Boghaert ER, Ackler SL, Catron ND, Chen J et al (2013) ABT-199, a potent and selective BCL-2 inhibitor, achieves antitumor activity while sparing platelets. Nat Med 19(2):202–208 65. Schulze-Gahmen U, Rini JM, Arevalo J, Stura EA, Kenten JH, Wilson IA (1988) Preliminary crystallographic data, primary sequence, and

131

binding data for an anti-peptide fab and its complex with a synthetic peptide from influenza virus hemagglutinin. J Biol Chem 263 (32):17100–17105 66. Brandts JF, Halvorson HR, Brennan M (1975) Consideration of the possibility that the slow step in protein denaturation reactions is due to cis-trans isomerism of proline residues. Biochemist 14(22):4953–4963 67. Bystrov VF, Arseniev AS, Gavrilov YD (1978) NMR spectroscopy of large peptides and small proteins. J Magn Reson 30(2):151–184 68. Wishart DS, Sykes BD, Richards FM (1992) The chemical shift index: a fast and simple method for the assignment of protein secondary structure through NMR spectroscopy. Biochemist 31(6):1647–1651 69. Kessler H (2010) Conformation and biological activity of cyclic peptides. Angew Chem Int Ed 21(7):512–523

Chapter 8 Click Chemistry for Cyclic Peptide Drug Design Adel Ahmed Rashad Abstract Click chemistry is a powerful tool in constraining peptides into their active conformations. This chapter presents recent advancements involving the use of copper-catalyzed [3 + 2] azide–alkyne cycloaddition (CuAAC), better known as “click reaction” in the design and synthesis of cyclic peptide and cyclic peptidomimetic compounds. The usage of “click chemistry” reactions includes various topics: (a) mimicking peptide bonds; (b) synthesis of ordered structures; (c) ligation of peptidomimetic scaffolds; and most importantly in this chapter (d) cyclization of peptidomimetic scaffolds using the triazole ring as constraint of conformation. Key words Click chemistry, Cyclic peptide, Triazole, Peptide cyclization

1

Introduction Since the discovery of the regioselective copper-catalyzed click reaction in 2002, the drug discovery arena has witnessed massive developments in both synthesis and impressive biological activities of the new triazole-containing scaffolds [1, 2]. The stability of the alkynes and azides in various reaction conditions and the facile chemical synthesis, in addition to the ease of forming the triazole ring in different solvents, have made this technique widely popular for organic and medicinal chemists (Fig. 1). One compelling advantage of this ease of chemical synthesis and the orthogonal nature of the Cu-mediated alkyne–azide couplings is that they can be used to prepare cyclic peptides, compared to the chemically labile disulfide-containing macrocyclic peptides [3]. The triazole ring, especially the 1,4-disubstituted-1,2,3-triazole, has some interesting properties. The molecular dimensions are somewhat similar to amide bonds (distance and planarity). Moreover, the triazole linkage geometry and isosterism are also compatible with the overall structural features of the -Pro-Glysequences in some β-turns (Fig. 2). β-turns are short peptide sequences incorporating -Gly-Pro- residues, and they serve to

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_8, © Springer Science+Business Media, LLC, part of Springer Nature 2019

133

134

Adel Ahmed Rashad

Fig. 1 Click reaction conditions and the corresponding outcome in absence/presence of catalysts

Fig. 2 Peptide bond mimicry by the 1,2,3-triazole ring and the ability to conserve a similar β-turn

retain the direction of peptide backbones. The β-turn unit is commonly found in a variety of protein–protein interfaces, making it a desirable synthetic target for therapeutics [4]. Another application of the β-turn mimicry abilities of the triazole ring is the synthesis of cyclic turns and short-length cyclic peptide to mimic protein loops, hairpins, and β-turn structures found in full-length proteins. Besides β-turns, α-helices are common secondary structure motifs, which play important functions in many proteins. This fact has attracted medicinal chemists to develop molecules, which can mimic α-helices that are found at the binding interface between two proteins. Such mimetic molecules can act as competitive inhibitors of protein–protein interactions (PPIs) [5]. Many of these

Click Chemistry for Cyclic Peptide Drug Design

135

Fig. 3 Diagram showing α-helical peptide highlighting the residues (i, i + 4, 7, or 11) that can be stapled together

PPIs were previously thought to be “undruggable” by small molecules; however, recent drug discovery studies revealed the potential use of cyclic and stapled peptides to efficiently target PPIs [6]. The widespread popularity of the CuAAC click reaction has been used as biocompatible ligation technique for peptide stapling. It was shown that replacement of lactam cross-links by [1,2,3]triazole rings provides peptides with similar α-helical content. Chorev and co-workers reported the first instance of generating helical structures using the CuAAC reaction between i and i + 4 residues of a model peptide based on parathyroid hormonerelated peptide [7] (Fig. 3). In this chapter, we describe some of the applications of the “click reaction” to make cyclic peptides, β-turns, and α-helices mimetics.

2

Triazole Linkage as a Turn Inducer for Macrocyclization The specific three-dimensional protein structures determine their functions, and these specific structures arise through highly ordered folding of their polypeptide backbones [8]. Peptide and protein natural macrocyclization frequently occurs as a constraining element in turn structures, for example through thioether bridges or disulfide bonds [9, 10]. In synthetic chemistry and drug discovery, peptide cyclization is mostly achieved through either disulfide bond formation between two cysteine residues or isopeptide bond formation (lactam formation through either side chain to side chain, main chain to main chain, or side chain to main chain) as shown in Fig. 4. This latter process includes selective incorporation and deprotection of the COOH and NH2 groups of the side chains and/or main chain during the synthesis process. Moreover, peptide cyclization through lactam ring formation might not be enough for structuring the macrocycle to populate a certain active conformation [11]. Using the triazole linkage for macrocyclization (Fig. 4) offers two major advantages: (a) the chemoselectivity of the

136

Adel Ahmed Rashad

Fig. 4 Different synthetic strategies to cyclize linear peptide sequences

Fig. 5 CuAAC substrate cyclization minimizes the production of the dimeric product

azide–alkyne reaction in the presence of other functional groups, and (b) the structuring effect induced by the triazole ring which is not fully achieved by other cyclization methods. Angell and Burgess, in 2005, synthesized peptide-based macrocyclic β-turn mimics using the CuAAC reaction. In this study, a propargyl amine is ligated to an activated C-terminal amino acid. The amino acid chain is then extended in the opposite direction to include an azido-functionalized benzamide group. The authors demonstrate that these azide- and alkyne-functionalized linear peptide substrates can be cyclized via the CuAAC reaction with high efficiency (>70% average yield) as shown in Fig. 5 [12]. The monocyclic compound was the expected major product from a click-mediated macrocyclization reaction. However, a surprising number of reports have described macrocyclic dimers as prevalent products in similar processes [13, 14]. The mechanism for this dimerization was proposed by Finn and co-workers (Fig. 6). It was proposed that the reaction proceeds via two alkynes bound to a dicopper intermediate where the azide units interact with the Cu atom that is not attached to the alkyne of the same substrate [14]. Kinetic studies supported this by showing that the immediate precursor contains two copper atoms and possibly two alkynes [15]. Intermediates A and B are possible candidates.

Click Chemistry for Cyclic Peptide Drug Design

137

Fig. 6 CuAAC mechanisms of forming monomeric and dimeric products

Fig. 7 Triazole-containing peptide 2 as functional analog to peptide 1

To form a monomeric cyclization product, the regioselectivity of the copper-catalyzed cycloaddition for 1,4-disubstituted triazoles requires the endo-like conformation. This endo-like conformation would be less populated than exo-like ones that are intrinsically less ordered. Angell and Burgess [12] suggested that the preference for the exo-like orientation B prevails over the conformational biases of the peptide part, or secondary copper coordination effects, that might sometimes favor reactions via endo-like conformations. Angell and Burgess also suggested that cyclodimerization seems to be favored by low catalyst concentrations, where using more than 0.5 equivalents of copper could reduce the yields of the cyclodimer compounds [16]. Synthesis of natural cyclic peptide mimetics has then witnessed fast developments using optimized copper-catalyzed cycloaddition click reaction, to increase the yield of the monocyclic peptide. In 2006, Maarseveen et al. replaced the amide ring of naturally occurring cyclic peptide 1, a potent tyrosinase inhibitor, by a triazole rings to synthesize the triazole analog 2 (Fig. 7). Cyclization through isopeptide bond formation fails to provide the desired macrocycle, but the triazole tetrapeptide was successfully produced using CuI-catalyzed alkyne–azide coupling at 110  C in 70% yield, using CuBr and DBU in toluene [17]. Copper-catalyzed cycloaddition click reaction was also successful in the synthesis of cyclic depsipeptides (Fig. 8). In 2014, Kaushik and co-workers incorporated the triazole ring formation

138

Adel Ahmed Rashad

Fig. 8 CuAAC formation of cyclic depsipeptide

Fig. 9 Microwave-assisted CuAAC peptide cyclization via triazole linkage

in the synthesis of the ester-containing cyclic scaffold. Interestingly, the CuBr/DBU/toluene click conditions were very successful in this case. In this study, the structure of these cyclic depsipeptides was confirmed by X-ray crystallography [18]. Microwave-assisted click reaction was also used in cyclizing a peptide derived from Alzheimer Aβ(16–22) sequence which has a strong tendency to form antiparallel β-sheets (Fig. 9). The cyclization was achieved using copper acetate in DMSO for 30 min at 100  C [19]. The use of the copper(I)-assisted azide–alkyne cycloaddition (CuAAC, or “click” reaction) as a method of β-hairpin stabilization was demonstrated by several reports. Waters and co-workers investigated the effect of the triazole ring at several different positions to determine the impact on hairpin structure and function. The CuAAC reaction was found to be suitable for locking the β-hairpin conformation in some peptides. In addition, the triazole-containing β-hairpin showed improved thermal stability and resistance to proteolysis compared to their linear peptides. The potency of the triazole β-hairpin was not affected by cyclization. For the click reaction in this study, 1 mM of peptide solution phosphate buffer was added to a mixture of tris-tri(methylazolyl)amine ligand (2.7 eq), 1.8 eq of [Cu(CH3CN)4][PF6], and 2.1 eq sodium ascorbate in acetonitrile (Fig. 10). The reaction mixture was stirred in the dark [20].

Click Chemistry for Cyclic Peptide Drug Design

139

Fig. 10 Locking the β-hairpin conformation in some peptides via head-to-tail triazole linkage

3

Peptide Stapling Using the Triazole Linkage Stapled helical peptides are cyclic peptides constrained in their helical forms to maintain the bioactive conformation and at the same time increase protease resistance and improve cell permeability. Peptide stapling is one of the most established methods for making helical peptides. The stapling techniques, including hydrocarbon, thiol, and lactam based, have been well described in literature [11, 21]. Replacing of any of these “staples” by one triazole ring generates peptide with similar α-helical content. This technique is called a one-component stapling technique. Two-component triazole stapling involves a bifunctional linker (azide or alkyne), which forms a staple by reacting with two complementary nonnative amino acids in the peptide. In 2010, the first triazole-linked helical peptide derived from parathyroid hormone-related peptide was prepared (Fig. 11). The triazole formation stapling was carried out in solution on fully cleaved and deprotected peptide, in positions i and i + 4. NMR and circular dichroism confirmed conformational similarities to an analogous lactam-stapled peptide in the literature. It was also found that five or six methylene units in the staple were able to stabilize α-helicity. The reaction conditions were tenfold molar excess of CuSO4·5H2O plus ascorbic acid in a mixture of tBuOH/H2O (1:2, v/v). No oligomeric products resulting from intermolecular click reactions were observed [22]. In 2012, Wang and co-workers reported another successful application in triazole stapling. In this work, a peptide segment from the B-cell CLL/lymphoma 9 (BCL9) was stapled by i and i + 4 trizole linkage (Fig. 12). The stapled helical peptide bound in the β-catenin-binding groove more efficiently than the linear precursor and inhibited BCL9 oncogenic interaction with β-catenin. The click reaction between the alkyne and the azide moieties was

140

Adel Ahmed Rashad

Fig. 11 Triazole stapling (i, i + 4) of helical peptide derived from parathyroid hormone-related peptide

Fig. 12 In solution (i, i + 4) triazole peptide stapling technique used for a peptide segment from the B-cell CLL/lymphoma 9

performed in solution after cleavage from resin, using CuSO4·5H2O (4.4 equiv) and sodium L-ascorbate (4.4 equiv) in a mixture of tBuOH/H2O, and stirred at room temperature for 30–90 min. The stereochemistry of the used nonnative amino acids as well as the position of the triazole ring relative to the stapling loop were found to affect the percentage helicity and activity [23]. In 2015, the Wang group also reported first-in-class triazolestapled peptides that block the protein protein interaction between repressor activator protein 1 (RAP1) and telomeric repeat-binding factor 2 (TRF2) with a high affinity (Ki value of 7 nM) which is >100 times more potent than the corresponding native linear TRF2 peptide (Fig. 13). The authors used the same click reaction conditions (4.4 equivalent of CuSO4·5H2O, 4.4 equivalent of sodium L-ascorbate in a mixture of tBuOH/H2O). The optimal number of the linker atoms was found to be 8 [24]. The triazole-stapling technique has also been used for stabilizing the α-helical structure of natural antimicrobial peptide PolybiaMPI (MPI) isolated from the venom of the social wasp Polybia paulista. Figure 14 shows that two triazole-stapled peptides were prepared (i, i + 4) and (i, i + 6). The (i, i + 4) triazole-stapled peptide showed similar activity to the linear peptide, adopted α-helical structure in aqueous solution, and had increased α-helical content in 30 mM sodium dodecyl sulfate with 50% trifluoroethyl alcohol. The (i, i + 6) triazole-stapled peptide did

Click Chemistry for Cyclic Peptide Drug Design

141

Fig. 13 Chemical structure of a triazole-stapled peptide that inhibits the protein protein interaction between repressor activator protein 1 (RAP1) and telomeric repeat-binding factor 2

Fig. 14 Stabilizing α-helical conformation through (i, i + 4, top) whereas the (i, i + 6, bottom) triazole-stapled peptide did not adopt α-helical conformation

not adopt α-helical conformation as judged by circular dichroism measurements. The click reaction conditions used in this study was CuSO4·5H2O and sodium L-ascorbate in a mixture of tBuOH/ H2O at room temperature [25]. The click reaction conditions (CuSO4·5H2O, sodium L-ascorbate in tBuOH/H2O mixture) can be replaced by CuSO4/ascorbic acid to generate the CuI catalyst in situ. Colombo and co-workers used this method on purified linear precursor peptide, in solution, to make a stabilized version of the native α-helical fold of the Pal3 peptide epitope from the antigen protein PalBp (BPSL2765) from Burkholderia pseudomallei, the etiological agent of melioidosis (Fig. 15). Pal3 shows no secondary structuring outside its native protein context; however, the triazole-stapled peptide (Pal3H) formed a stable α-helix [26]. The “click reaction” has also been used in two-component stapling systems using functionalized double-click linkers (Fig. 16). In 2008, Bong and co-workers used the bifunctional linker 1,5-hexadiyne, either on-resin or in-solution phase, to prepare i and i + 4 bis-triazole-stapled peptides, in a peptide derived from the GCN4 leucine zipper. For the in-solution click reaction, the 1,5-hexadiyne and diazopeptide were dissolved in Tris-buffered saline (TBS, 10 mm Tris, 110 mm NaCl, pH 8.5). A mixture of CuSO4 (4 equiv), sodium ascorbate (50 equiv), and bathophenanthroline disulfonate ligand (7.2 equiv) was added to the peptide solution and the mixture was stirred under argon for 2.5–12 h. For the on-resin click reaction, resin was swelled with DMF and treated with hexa-1,5-diyne (50 equiv), and CuI and DIEA (3 equivalents

142

Adel Ahmed Rashad

Fig. 15 Developing the triazole-stapled peptide (Pal3H) with a stable α-helical structure

Fig. 16 Two-component-stapling systems using functionalized double-click linkers

each in CH3CN) were then added. The resin was shaken overnight under argon and treated again with CuI/DIEA for 4–6 h after DMF wash, without additional 1,5-hexadiyne linker [27]. The Spring group at Cambridge University then used these bifunctional dialkynyl linkers as carriers to introduce new functional motifs or molecules to the stapled α-helical peptides, starting from a single unprotected diazido peptide, in one step. Their first application was targeting the p53/MDM2 interaction, a promising therapeutic target for cancer therapy. After preparing a p53-based peptide with two azidoornithine residues, the solution-phase double-click reaction was carried out with 3,5-diethynylbenzene, giving full reaction conversion to the desired stapled peptide without any protecting groups (Fig. 17). This ditriazolo-stapled peptide was found to bind with nanomolar affinity to MDM2 (IC50 ~88 nM and kd ~6 nM), compared to the corresponding wildtype and unstapled linear peptides (IC50 ~4 μM and kd ~480 nM). However, the new ditriazolo-stapled peptide showed no cellular uptake and was inactive in a cellular p53 reporter assay. When modified dialkynyl linkers bearing cationic arginine groups were stapled onto the same p53 peptide, the net positive charge increase via the introduced staple linker, cellular uptake, and p53 activation was achieved. The stapling double-click reaction conditions involved treating the unprotected peptide with an excess of dialkynyl linker, CuSO4, and sodium ascorbate in mixture (1:1) of acetonitrile and 20 mM sodium phosphate buffer at pH 7.6 [28].

Click Chemistry for Cyclic Peptide Drug Design

143

Fig. 17 Design of potent stapled peptide to target the p53/MDM2 interaction, using the two-component stapling technique

H 2N

H N O

HN NH2 STAPLE

HN

O

N DLWRLL H

O

N H O

STAPLE = TAMRA Ahx ETF

O HN O

N N N

N N N

O

NH

HN

NH2 NH NH NH2

N EN NH2 H

Fig. 18 Design of a platelet-penetrating stapled peptide via incorporating a membrane-penetrating linker

The Spring group also managed to utilize the benefits of the two-component CuAAC stapling technique to prepare a series of novel functionalized stapled Bim BH3 mimetics, from one synthesized linear sequence, to investigate phosphatidyl-serine (PS) exposure in platelets (Fig. 18). The new stapled peptides were able to enter the platelet cytosol and modulate activity. In this study, the authors used the following reaction conditions: 1.1 equivalent of the dialkynyl linker was added to the diazido-peptide in 1:1 mixture of tBuOH/H2O under nitrogen, 1 equivalent of CuSO4·5H2O, 1 equivalent of tris(3- hydroxypropyltriazolylmethyl)amine, and 3 equivalents of sodium ascorbate, and the reaction mixture was stirred for 15 min–2 h at room temperature [29].

144

4

Adel Ahmed Rashad

Conclusions The introduction of the copper-catalyzed [3 + 2] azide–alkyne cycloaddition (CuAAC) in peptide synthesis has greatly advanced the peptide science in general. In addition to keeping the peptide conformation similar to the parent unmodified peptide, the planner triazole ring, as a peptide/amide bond mimetic, can provide additional stability for the peptide scaffold. Using the triazole linkage cyclization method has greatly advanced the area of biologically active macrocyclic compounds due to the ease of synthesis, compared to the other peptide cyclization techniques, and to the great stability it can provide. An additional advantage of the triazole cyclization techniques is the ability to introduce new chemical entities that are capable of controlling the physiochemical properties of the molecules, without interference with their biological activities.

References 1. Rostovtsev VV, Green LG, Fokin VV, Sharpless KB (2002) A stepwise huisgen cycloaddition process: copper(I)-catalyzed regioselective “ligation” of azides and terminal alkynes. Angew Chem Int Ed Engl 41:2596–2599 2. Tornoe CW, Christensen C, Meldal M (2002) Peptidotriazoles on solid phase: [1,2,3]triazoles by regiospecific copper(i)-catalyzed 1,3-dipolar cycloadditions of terminal alkynes to azides. J Org Chem 67:3057–3064 3. Roice M, Johannsen I, Meldal M (2004) High Capacity Poly(ethylene glycol) Based Amino Polymers for Peptide and Organic Synthesis. QSAR Combinat Sci 23:662–673 4. Oh K, Guan Z (2006) A convergent synthesis of new β-turn mimics by click chemistry. Chem Commun:3069–3071 5. Bullock BN, Jochim AL, Arora PS (2011) Assessing helical protein interfaces for inhibitor design. J Am Chem Soc 133:14220–14223 6. Villar EA, Beglov D, Chennamadhavuni S, Porco JA Jr, Kozakov D, Vajda S, Whitty A (2014) How proteins bind macrocycles. Nat Chem Biol 10:723 7. Cantel S, Le Chevalier Isaad A, Scrima M, Levy JJ, DiMarchi RD, Rovero P, Halperin JA, D’Ursi AM, Papini AM, Chorev M (2008) Synthesis and conformational analysis of a cyclic peptide obtained via i to i+4 intramolecular side-chain to side-chain azide alkyne 1,3-dipolar cycloaddition. J Org Chem 73:5663–5674

8. Arkin MR, Wells JA (2004) Small-molecule inhibitors of protein-protein interactions: progressing towards the dream. Nat Rev Drug Discov 3:301–317 9. Musiol HJ, Siedler F, Quarzago D, Moroder L (1994) Redox-active bis-cysteinyl peptides. I. Synthesis of cyclic cystinyl peptides by conventional methods in solution and on solid supports. Biopolymers 34:1553–1562 10. Siedler F, Quarzago D, Rudolph-Bohner S, Moroder L (1994) Redox-active bis-cysteinyl peptides. II. Comparative study on the sequence-dependent tendency for disulfide loop formation. Biopolymers 34:1563–1572 11. Hill TA, Shepherd NE, Diness F, Fairlie DP (2014) Constraining cyclic peptides to mimic protein structure motifs. Angew Chem Int Ed Engl 53:13020–13041 12. Angell Y, Burgess K (2005) Ring closure to beta-turn mimics via copper-catalyzed azide/ alkyne cycloadditions. J Org Chem 70:9595–9598 13. Billing JF, Nilsson UJ (2005) C2-symmetric macrocyclic carbohydrate/amino acid hybrids through copper(I)-catalyzed formation of 1,2,3-triazoles. J Org Chem 70:4847–4850 14. Punna S, Kuzelka J, Wang Q, Finn MG (2005) Head-to-tail peptide cyclodimerization by copper-catalyzed azide-alkyne cycloaddition. Angew Chem Int Ed Engl 44:2215–2220 15. Rodionov VO, Fokin VV, Finn MG (2005) Mechanism of the ligand-free cui-catalyzed

Click Chemistry for Cyclic Peptide Drug Design azide-alkyne cycloaddition reaction. Angew Chem Int Ed Engl 44:2210–2215 16. Angell YL, Burgess K (2007) Peptidomimetics via copper-catalyzed azide-alkyne cycloadditions. Chem Soc Rev 36:1674–1689 17. Bock VD, Perciaccante R, Jansen TP, Hiemstra H, van Maarseveen JH (2006) Click chemistry as a route to cyclic tetrapeptide analogues: synthesis of cyclo-[Pro-Val-psi(triazole)-Pro-Tyr]. Org Lett 8:919–922 18. Agrawal SK, Panini P, Sathe M, Chopra D, Kaushik MP (2014) Design and synthesis of cyclic depsipeptides containing triazole (CDPT) rings. RSC Adv 4:10728–10730 19. Elgersma RC, van Dijk M, Dechesne AC, van Nostrum CF, Hennink WE, Rijkers DT, Liskamp RM (2009) Microwave-assisted click polymerization for the synthesis of Abeta (16-22) cyclic oligomers and their selfassembly into polymorphous aggregates. Org Biomol Chem 7:4517–4525 20. Park JH, Waters ML (2013) Positional effects of click cyclization on beta-hairpin structure, stability, and function. Org Biomol Chem 11:69–77 21. Pelay-Gimeno M, Glas A, Koch O, Grossmann TN (2015) Structure-based design of inhibitors of protein-protein interactions: mimicking peptide binding epitopes. Angew Chem Int Ed Engl 54:8896–8927 22. Scrima M, Le Chevalier-Isaad A, Rovero P, Papini AM, Chorev M, D’Ursi AM (2010) CuI-catalyzed azide–alkyne intramolecular i-to-(i+4) side-chain-to-side-chain cyclization promotes the formation of helix-like secondary structures. Eur J Org Chem 2010:446–457 23. Kawamoto SA, Coleska A, Ran X, Yi H, Yang CY, Wang S (2012) Design of triazole-stapled BCL9 alpha-helical peptides to target the beta-catenin/ B-cell CLL/lymphoma 9 (BCL9) proteinprotein interaction. J Med Chem 55:1137–1146

145

24. Ran X, Liu L, Yang CY, Lu J, Chen Y, Lei M, Wang S (2016) Design of high-affinity stapled peptides to target the repressor activator protein 1 (RAP1)/telomeric repeat-binding factor 2 (TRF2) protein-protein interaction in the shelterin complex. J Med Chem 59:328–334 25. Liu B, Zhang W, Gou S, Huang H, Yao J, Yang Z, Liu H, Zhong C, Liu B, Ni J, Wang R (2017) Intramolecular cyclization of the antimicrobial peptide Polybia-MPI with triazole stapling: influence on stability and bioactivity. J Pept Sci 23:824–832 26. Gori A, Peri C, Quilici G, Nithichanon A, Gaudesi D, Longhi R, Gourlay L, Bolognesi M, Lertmemongkolchai G, Musco G, Colombo G (2016) Flexible vs. rigid epitope conformations for diagnostic- and vaccine-oriented applications: novel insights from the Burkholderia pseudomallei BPSL2765 Pal3 epitope. ACS Infect Dis 2:221–230 27. Torres O, Yuksel D, Bernardina M, Kumar K, Bong D (2008) Peptide tertiary structure nucleation by side-chain cross-linking with metal complexation and double “click” cycloaddition. Chembiochem 9:1701–1705 28. Lau YH, de Andrade P, Quah S-T, Rossmann M, Laraia L, Sko¨ld N, Sum TJ, Rowling PJE, Joseph TL, Verma C, Hyvo¨nen M, Itzhaki LS, Venkitaraman AR, Brown CJ, Lane DP, Spring DR (2014) Functionalised staple linkages for modulating the cellular activity of stapled peptides. Chem Sci 5:1804–1809 29. Iegre J, Ahmed NS, Gaynord JS, Wu Y, Herlihy KM, Tan YS, Lopes-Pires ME, Jha R, Lau YH, Sore HF, Verma C, DH OD, Pugh N, Spring DR (2018) Stapled peptides as a new technology to investigate protein-protein interactions in human platelets. Chem Sci 9:4638–4643

Chapter 9 Frontier Between Cyclic Peptides and Macrocycles Philipp Ermert, Anatol Luther, Peter Zbinden, and Daniel Obrecht Abstract This review describes a selection of macrocyclic natural products and structurally modified analogs containing peptidic and non-peptidic elements as structural features that potentially modulate cellular permeability. Examples range from exclusively peptidic structures like cyclosporin A or phepropeptins to compounds with mostly non-peptidic character, such as telomestatin or largazole. Furthermore, semisynthetic approaches and synthesis platforms to generate general and focused libraries of compounds at the interface of cyclic peptides and non-peptidic macrocycles are discussed. Key words Macrocycle, Cellular permeability, Library, Modular synthesis, Semi-synthesis

1

Introduction: Privileged Properties of Macrocycles Medium-sized macrocyclic natural products in the MW range of 500–4000 Da have traditionally played an important role in drug discovery [1, 2], especially in the anti-infective [3, 4] and anticancer [5, 6] areas. Highly diverse compounds such as caspofungin, daptomycin, avermectin, halichondrin and its synthetic analog eribulin, rapamycin and everolimus, and cyclosporin A are well-known examples (Fig. 1). In addition, such macrocyclic natural products have been a rich source of inspiration for medicinal chemists to generate semi- and fully synthetic analogs, or de novo-designed molecules based on structural data available from X-ray or NMR structures [7]. While only ~3% of all natural products are macrocyclic in nature they constitute roughly 30% of the approved natural product-derived drugs [2, 8] underscoring the fact that macrocycles are privileged scaffolds for drug discovery. The field of medium-sized macrocycles has also had a recent revival by the fact that increasingly novel therapeutically relevant targets require new chemotypes, since both established small molecules and large biopharmaceuticals show limitations. In a recent review authors estimated a total of roughly 3000 disease-relevant targets. Currently 1600 FDA-approved drugs (all chemotypes)

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_9, © Springer Science+Business Media, LLC, part of Springer Nature 2019

147

Fig. 1 Medium-sized macrocyclic natural products

148 Philipp Ermert et al.

Frontier Between Cyclic Peptides and Macrocycles

149

target roughly 670 human-genome-derived proteins associated with a human disease [9]. Many of the remaining disease-relevant targets are predicted to involve extra- and intracellular proteinprotein interactions (PPIs) with large and flat binding interfaces typically less amenable to small-molecule drugs. Hence, mediumsized macrocycles potentially fill an important gap in the arsenal of drugs between small molecules and large biopharmaceuticals. This review focuses on macrocyclic approaches that use the strengths of natural product cyclopeptides to develop into non-peptide scaffolds with excellent drug-like properties. Macrocycles as a class have two important features: (a) they can ˚ 2 [10] as interact with PPI “hot spots” of typically 800–1200 A summarized recently [7] and (b) their semirigid nature allows to form tight complexes with dynamic protein surfaces by induced fit [7, 11]. The conformational flexibility of macrocycles can also translate into favorable physicochemical properties well known for cyclosporin A (CspA). CspA adopts a hairpin-like conformation in apolar medium (e.g., lipid membranes) reducing the number of solvent-exposed H-bond donors and acceptors which results in good membrane-permeability. In polar (cellular) medium, however, CspA undergoes a significant conformational change which exposes the H-bond donors required for binding to its target cyclophilin [12] (see Subheading 2.1). These “chameleonic” properties seem to be unique to macrocycles and translate into physicochemical properties beyond the rule of 5 [13]. Systematic analysis of drug-like properties of 100 macrocyclic drugs and drug candidates suggested that oral bioavailability with molecular weights of ˚ 2 can be up to 1000 Da and total polar surface areas
250 A achieved [13]. In a recent analysis efflux-inhibited (passive) Caco2 cell permeability correlated strongly with the compounds’ minimum solvent-accessible 3D polar surface areas (PSA). Interestingly, the PSA difference between minimum and maximum observed PSA ˚ 2, which is a direct result of its chameleonic for CspA was 79 A conformational properties [14]. Macrocycles in general consist of a macrocyclic core to which substituents (side chains) are appended. While the macrocyclic core scaffold has a big impact on the conformational properties of a macrocycle, the size and nature of the side chains are also of critical importance for the overall physicochemical properties. In CspA the side chain is mainly lipophilic and devoid of H-bond-donating functional groups which is important for its cell permeability and oral bioavailability. For macrocycles containing polar (neutral and charged) side chains with a significant number of H-bond donors the privileged chameleonic properties of the core scaffold are rather dominated by the physicochemical properties of the side chains. Bioactive natural product macrocycles encompass many different classes. Macrocyclic peptides/depsipeptides as well as macrolides seem to be most abundant. However, many macrocyclic

150

Philipp Ermert et al.

hybrids containing mixed amino acid, hydroxy acid, polyene, polyacetate, terpenoid, and various other building blocks do exist. In addition, an increasing number of fully synthetic macrocycles exhibiting an almost unlimited diversity have emerged as viable scaffolds for new drugs. Such fully synthetic scaffolds may have advantages over traditional cyclopeptides to obtain excellent drug-like properties. Cyclopeptides are particularly privileged as bioactive drugs because they can form densely packed complexes with target proteins mediated by numerous specific interactions of H-bond donor-acceptors from the backbone and side chains as well as strong hydrophobic interactions mainly driven by side-chain groups [15]. Systematic analysis of these peptide-protein complexes shows a high prevalence of involvement of pharmacophores forming privileged secondary structural motifs such as β-turns, β-hairpins, hot loops, and α-helices. Therefore, peptidic macrocycles mimicking such epitopes have proven valuable for drug discovery [16]. Although macrocyclic peptides, depsipeptides, and lipopeptides have provided important drug candidates and drugs, most of them are administered parenterally and show limited oral bioavailability. However, the following examples will highlight approaches that use fully synthetic macrocycles which were to some extent inspired by nature’s macrocycles.

2

Transition from Cyclopeptides to Non-peptidic Macrocycles As mentioned in Subheading 1, cyclic peptides do generally exhibit poor cell permeability, due to the polar nature of the backbone and the presence of polar and/or charged side chains. Molecular weight, total polar surface area, and number of H-bond donor/ acceptors put them outside classical drug-like properties. However, exceptions may comprise cyclic peptides with mostly lipophilic side chains [17–19] or combination of lipophilic side chains and the capability of conformational adaptation to the environment, which may result in significant penetration of eukaryotic cells by passive diffusion. Cyclosporin A [20], sanguinamide A [21], and phepropeptin C [17] have been reported to exhibit such “chameleonic” behavior [11], which confers not only permeability but also high aqueous solubility to these molecules. In addition, cyclosporin A and sanguinamide A are also orally bioavailable (Fig. 2, Table 1). Many natural products which act on intracellular targets are modified peptides (Fig. 2, Table 1). Examples comprise peptidepolyketide hybrid macrocycles like jasplakinolide (34) [38], apratoxin A (22) [34], sanglifehrin A (48) [47], and rapamycin. Azolecontaining peptides like patellamide C (7) [24] and compounds like virginiamycin M1 (83) [42] or largazole (60) [48] which contain several structural elements (peptidic, azole, polyketidic) as well as molecules showing the presence of concatenated azoles such

Fig. 2 Natural products at the interface between peptidic and non-peptidic macrocycles. Lipophilic side chains, removal of secondary amides by N-alkylation, heterocyclization or inclusion of a depsipeptide linkage, and integration of lipophilic elements like polyketide structures reduce their peptidic character

Frontier Between Cyclic Peptides and Macrocycles 151

763 Antiproliferative; modulation of multidrug resistance; target: unknown

709 Antiproliferative Target unknown

Patellamide C [24–26]

IB-01211/mechercharmycin A [27, 28]

Telomestatin [30–32]

582 Antiproliferative Target: telomerase, (c-Myb)

Permeability data

204 A˚2

185 A˚2 PAMPA Pe 17.8  106 cm/s

195 A˚2

Low!

600 μM 180 A˚2 PAMPA Pe 7  106 cm/s MDCK-LE Papp 7.7  106 cm/s

a

1.4%

21% (rat)

Structures cf. Fig. 2. Evidence for the passive membrane/cellular permeability of the compounds was obtained by direct measurements of PAMPA- Caco2- or MDCK (MadinDarby canine kidney low-efflux cells) permeability or could be demonstrated for example using derivatives with a fluorescence label. Cellular activity of a compound acting on an intracellular target allowed the assumption of cellular permeability a F almost 0% for the closely related virginiamycin M2 (94; cf. Scheme 6) [43]. TPSA values were calculated using Molecular Operating Environment (MOE), 2018.01; Chemical Computing Group ULC, 1010 Sherbrooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2018

526 Antimicrobial Target: ribosome

Virginiamycin M1 [42–44]

1090 Antiviral; target: cyclophilin A

840 Antiproliferative Target: secretion pathway Sec61 complex

Apratoxin A [34–37]

Sanglifehrin A [40, 41]

793 n.a.

Model cpd 21 [33]

Frontier Between Cyclic Peptides and Macrocycles 153

154

Philipp Ermert et al.

as IB-01211 (11) [27], GE2270 A [29], and telomestatin (17) [30] are just a few prominent examples demonstrating the large diversity observed in chimeric macrocyclic natural products. While compounds exhibiting potent biological activity serve frequently as a starting point for drug discovery programs, the isolation of the metabolite mycocyclosin (diketopiperazine with macrocyclic bridge) [49, 50], which is produced by Mycobacterium tuberculosis, led to the identification of a potential target to discover compounds inhibiting the growth of this bacterium. Structural motifs that promote the ability of the macrocycles to cross cell membranes comprise l

Lipophilic side chains

l

Removal of H-bond donors by formation of intramolecular H-bonds

l

Removal of H-bond donors by alkylation of backbone NH groups (methylation), with the additional effect of increased conformational flexibility

l

Removal of H-bond donors by alkylation of side-chain OH groups (methylation, prenylation)

l

Shielding of polar contacts by inclusion of amino acids with β-branched side chains (valine, isoleucine, tert-butyl glycine), tert. alcohols (3-hydroxy-valine), or ortho, ortho0 -disubstituted phenols (o,o0 -dichloro tyrosine)

l

Shielding of polar contacts by orientation into the center of bowl-shaped structures

l

Removal of polar groups through heterocyclization

l

2.1 Conformationally Flexible Peptides

Inclusion of β-hydroxy-γ-amino acids and vinylogous (α,β-unsaturated) amino acids

l

Depsipeptide linkage, resulting in removal of H-bond donor and increased conformational flexibility

l

Integration of lipophilic structural elements, e.g., polyketide motifs

l

Presence of structural elements (molecular hinges like tertiary amide bonds) conferring balanced conformational flexibility

As mentioned in Subheading 1, the undecapeptide cyclosporin A (CspA; Fig. 3) is the archetype of a conformationally flexible macrocyclic peptide, capable of conformational adaption to the environment: in apolar solvent CspA adopts a conformation with internalized polar contacts [20, 51]; in polar environment the hydrogen bond donors are exposed [12]. The conformational flexibility is related to the presence of tertiary amides and upon conformational change the 9,10-peptide bond isomerizes from the cis to trans configuration.

Frontier Between Cyclic Peptides and Macrocycles

155

Fig. 3 Apolar (left) [20, 51] and polar [12] conformation of cyclosporin A. Dashed lines indicate intramolecular hydrogen bonds. Apolar conformation: X-ray structure of CspA in free unbound state (crystallized from aprotic solvent; DEKSAN; CCDC code). Polar conformation: X-ray structure of CspA bound to cyclophilin and calcineurin (1MF8; PDB code). Min 3D SA PSA ¼ 208 A˚2, max. 3D SA PSA ¼ 287 A˚2 [14]. Tertiary amides confer conformational flexibility to the macrocyclic backbone (3D SA PSA: three-dimensional solvent-accessible polar surface area)

Phepropeptin C (1; Fig. 4) is a hexapeptide consisting of five lipophilic amino acid residues and a proline. In CDCl3 it was found to adopt a conformation with two transannular and one extraannular intramolecular hydrogen bond whereas in DMSO-d6 (more polar, but aprotic), however surprisingly, a conformation with four intramolecular hydrogen bonds (two transannular, two extra-annular). Lokey and coworkers [17] explained the high permeability (MDCK assay Papp ¼ 40  106 cm/s) by its ability of conformational equilibration between polar and less polar conformations. The D-proline epimer (2), however, exhibited a lower permeability (MDCK assay Papp ¼ 15  106 cm/s) than the natural stereoisomer although the aqueous solubility of the two isomers was found to be comparable at pH 7.4 (1: 0.06 g/L; 2: 0.07 g/L). In the cyclic heptapeptide sanguinamide A (3; Fig. 4) tertiary amide bonds and a thiazole ring are present. Based on NMR spectroscopic characterization in DMSO-d6 solution Fairlie and colleagues [21] reported the configuration at the Phe(3)-Pro(4) amide bond and at the Ile(5)-Pro(6) amide bond to be cis and trans. In solution the molecule forms an antiparallel β-sheet with a turn at each end and two amide NH groups participating in intramolecular hydrogen bonds. The molecule is orally bioavailable (F ¼ 7% in rat, dose 10 mg/kg) [21, 22]. Lokey and coworkers [23] found the effect of N-methylation to be position dependent; N-methylation of the exposed Phe(3)NH (providing 4) resulted in improved permeability, and rather surprisingly also solubility. However, reduced permeability was found upon N-methylation of amide bonds which participate in intramolecular hydrogen bonds. In addition, NMR spectra of these derivatives showed the presence

156

Philipp Ermert et al.

Fig. 4 Phepropeptin C [17] and sanguinamide A [21–23] show environment-dependent conformational flexibility

of multiple conformers. Replacement of Ala(2) of 4 by Leu gave 5, which was reported [23] to exhibit excellent solubility and permeability; NMR data (temperature coefficient TCNH) suggested that different conformations are adopted in polar DMSO-d6 and in less polar CDCl3, conferring passive permeability and solubility. 2.2 Azol(in)eContaining Macrocycles 2.2.1 Patellamides

Patellamides represent a family of macrocyclic octapeptide derivatives with alternating arrangement of linear and azol(in)eheterocyclized amino acid residues. Conformational flexibility of these molecules does not rely on the presence of tertiary amides and cis–trans amide isomerization, but rather is the consequence of the heterocyclization of every second amino acid (Fig. 5). The heterocycles reduce the number of hydrogen-bond donors and confer flexibility, allowing the population of backbone conformations which are not accessible for a classical peptide scaffold resulting in unusual β-turns in the twisted (or closed) conformer [53, 54]. Azole-containing macrocycles are often derived from ribosomally synthesized and post-translationally modified peptides

Frontier Between Cyclic Peptides and Macrocycles

157

Fig. 5 Structure of patellamides A–D and ascidiacyclamide [24, 52]

[55]. Serine (Ser) or threonine (Thr) and cysteine (Cys) residues are the precursors of oxazolines and thiazolines, respectively, which can be further oxidized to the hetero-aromatic oxazoles and thiazoles. Consistent with the ribosomal origin, these macrocycles rarely contain N-methyl groups [56]. The macrocyclization substrate is a peptide; the ring closure goes along with the cleavage of peptide bond (rather than with attack on the activated thioester used in the non-ribosomal peptide synthesis, see below, Subheading 2.3). Patellamides (Fig. 5) have been isolated from the ascidian Lissoclinum patella. They originate from its cyanobacterial symbiont Prochloron didemni [55, 57]. Patellamides A (9) and C (7) (Fig. 5) are derived from two eight-residue core peptides, which are integrated in the ribosomal 71-residue PatE protein. The biosynthesis has been recently reviewed [53, 55, 58]. The core peptides are flanked by an N-terminal protease signature and a C-terminal macrocyclization signature. The posttranslational steps comprise amino acid heterocyclization, and heterocycle oxidation, peptide cleavage, and peptide macrocyclization. During biosynthesis, the two centers adjacent to the thiazole rings are epimerized. Enzyme activities have been identified for all steps, except for the epimerization [53]. The heterocyclic rings result from cyclodehydration of the Ser/Thr OH group or the Cys SH group and the backbone carbonyl group of the N-terminally adjacent amino acid. The ring closure (Scheme 1) is catalyzed by the macrocyclase domain of PatG (PatGmac, residues 492–851). PatGmac (overexpressed in E. coli, crystal structure determination at 2.19 A˚ resolution [58]) is a subtilisin like Ser protease (catalytic triad Asp548, His618, Ser783), and unlike other members it contains an insertion located above the active site. It binds substrates in an unusual

158

Philipp Ermert et al.

Scheme 1 PatGmac-catalyzed macrocyclization of patellamides [53, 55, 58]

conformation, recognizing the thiazol(ine) residue in the P1 position and the Ala-Tyr-Asp macrocyclization signature in the positions P10 to P30 , and shields the acyl enzyme intermediate from hydrolysis. This results in cleavage of the C-terminal signature sequence and N–C macrocyclization [58] (Scheme 1). McIntosh et al. [59] reported that PatGmac is converting 29 known precursor peptides. The broad substrate tolerance is in line with the structural findings. As the enzyme is converting nonactivated peptide substrates, PatGmac is interesting and valuable for applications in organic synthesis (cf. Subheading 4). The backbone conformation of pseudo-C2 symmetric patellamide A (9) and the nonsymmetric patellamides B (6), C (7), and D (8) has been studied applying X-ray crystallography, NMR methods, and CD spectroscopy. Patellamide A (9) [52] was reported to adopt a saddle- or bowlshaped, nearly square type II conformation (Fig. 6a/b) in solid state (crystallized from aqueous methanol), which was also observed in solution and in solid state for the very closely related C2 symmetric ascidiacyclamide (10) [63]. For the two pairs of NH protons of 10, temperature coefficients and slow exchange rate (CD3OD addition to a CDCl3 solution of 10, t1/2 ca 2 days) indicate that the NH protons are sterically shielded from the solvent, as pointed out by Ishida et al. [63]. For patellamide D (8), however, a twisted solid-state conformation was found [62], which is stabilized by four transannular hydrogen bonds, two C¼O to H–N and two oxazoline O to H–N hydrogen bonds. Patellamides B (6) and C (7) were reported to populate twisted conformers in solution in protic and aprotic environment [54, 60, 61]. Depending on solvent and temperature, patellamide A (9) populates in solution either the square or the twisted conformation. CD spectra recorded by Freeman et al. [60] in methanol reveal that patellamide A (9) adopts at room temperature (26  C) a square type II conformation, whereas at low temperature

Frontier Between Cyclic Peptides and Macrocycles

159

Fig. 6 (a) Conformations of patellamides [52, 54, 60, 61]. (b) Patellamide A in the open (type II) conformation; X-ray structure [52] (YAYHUB; CCDC code) and patellamide D in the closed (type III) conformation; X-ray structure [62] (SAVBOG; CCDC code)

(74  C), however, a twisted type III conformation was observed. Based on NOE restraints, Milne et al. [54] found that 9 adopts in CDCl3 the twisted conformation and proposed on the basis of molecular dynamics simulation the interconversion of square and twisted conformation to occur through an intermediate type IV conformation resulting from sequential rotation of the oxazoline rings, different from the originally proposed [52, 61] path via type I intermediate (Fig. 6a).

160

Philipp Ermert et al.

In agreement with the internalization of H-bond donors, patellamide C (7) exhibited a high passive permeability of Pe ¼ 19.05  106 cm/s (log Pe ¼ 4.72) [25] quantified in a PAMPA assay. A comparable permeability was determined for cyclosporin A (Pe ¼ 9.8  106 cm/s (log Pe ¼ 5.01) [25]). Lissoclinum patella was reported to concentrate copper (ca 20,000-fold as compared to seawater), and it was speculated that patellamides may act as chelating agents [64]. Upon binding of Cu2+, patellamide C (7) undergoes a conformational change from the closed to the open conformation, creating a second Cu2+binding site [65]. Patellamides A, B, and C were found to exhibit equal cytotoxicity against L1210 murine leukemia cells (IC50 2–3.9 μg/mL). In addition, patellamide A (9) also inhibited the human ALL (T-cell acute leukemia) cell line CEM (IC50 ¼ 0.028 μg/mL) [24]. Patellamides B (6) and C (7) were also shown to reduce drug resistance in vitro in CEM/VBL100 cells ca tenfold (IC50 against vinblastine ¼ 90 nM in the absence of 6 or 7; IC50 ¼ 12 nM in the presence of 6 or 7) [26]. 2.2.2 IB-01211 and Telomestatin

A variety of natural products contain concatenated azoles/azolines together with amino acid residues; examples are IB-01211 (mechercharmycin; 11; Fig. 7) and the antibiotic GE 2270A (cf. Subheading 3.1). Whereas Wahyudi and McAlpine [68] recently reviewed structure-activity studies on peptide-based macrocycles, the synthesis of oxazole-containing natural products

Fig. 7 Structures of IB-01211/mechercharmycin A and related compounds [28, 66, 67]

Frontier Between Cyclic Peptides and Macrocycles

161

Table 2 Cytotoxic activities of IB-01211/mechercharmycin and derivatives [28]

Cpd

Lung carcinoma A-549 cells GI50 (μM)

Colon carcinoma HT-29 cells GI50 (μM)

Breast adenocarcinoma MDA-MB-231 cells GI50 (μM)

11

0.03

0.04

0.09

14

0.17

0.12

0.10

15

0.12

0.13

0.13

12

Inactive

Inactive

Inactive

originating from marine organisms was summarized by Tilvi and Singh [69]. IB-01211 (11; Fig. 7) was isolated from a marine microorganism strain ES7-008 [28]. The same structure was described for mechercharmycin A, isolated from Thermoactinomyces sp. [27]. The structure elucidation was achieved by X-ray crystallographic analysis [27], showing that the 24-membered macrocycle consists of a phenyl penta-azole pentazole is a cycle consisting of 5 N-atoms; penta-azole system means a sequence of 5 oxazole and thiazole rings, a D-allo-Ile-Val dipeptide, and an exocyclic methylidene group. A total synthesis [66] and syntheses of analogs [28, 67] were reported. IB-01211/mechercharmycin A (11) exhibits cytotoxic effects against human cancer cells (Table 2) [27, 28]. The linear congener, mechercharmycin B (13), however, was found to be inactive [27], which Wahyudi and McAlpine [68] attributed to the importance of the macrocyclic structure. Introduction of methyl substituents at the oxazole rings of 11 (intended to improve the solubility) reduced the potency by ca 10–60-fold [28], potentially due to changes in the conformation. IB-01211 (11), demethylidene analog 14, and derivative 15 were found to alter the cell cycle progression and to induce apoptosis (detection of caspase-3/-7 in colorimetric test). The presence of the exocyclic methylene group was not required for activity (cf. 14), whereas the presence of the thiazole ring was important: the oxazole analog 12 [67] was inactive. The biological target of 11 is not known; however, Hernandez et al. [28] speculated that it might act on telomerase based on the structural similarity with telomestatin (17; Fig. 8). YM-216391 (16) [73] is a constitutional isomer of 14 and was isolated from Streptomyces nobilis. It exhibits growth inhibition of human cervical cancer cells HeLa S3 (IC50 ¼ 14 nM) and cytotoxic activity in a panel of 39 human cancer cell lines. The mean GI50 was

162

Philipp Ermert et al.

Fig. 8 Telomestatin (17) and derivatives recognize G-quadruplex structures [32, 70–72]; telomestatin is a very potent telomerase inhibitor, which induces apoptosis in various cancer cells [31]

37 nM; compound 16 caused cell death with a mean LC50 of 550 nM. Telomestatin (17; isolated from Streptomyces anulatus) is a 24-membered macrocyclic structure consisting of eight concatenated azole or azoline rings (Fig. 8) [30]. The [R]-configuration at the thiazoline stereocenter was established by total synthesis [74]. Telomestatin (17) is derived from a ribosomally synthesized and post-translationally modified precursor peptide TlsC containing the -Cys-Thr-Thr-Ser-Ser-Ser-Ser-Ser- sequence [70]. Telomestatin (17) induces apoptosis in various cancer cells. Effects of telomestatin on 39 human cancer cell lines were reported [31] and cells derived from tumors of the CNS were found to be more sensitive. Telomestatin (17) was previously reported to inhibit telomerase activity (IC50 ca 2 μM) in BCR-ABL-positive leukemic cell lines OM9;22 and K562 [75]. Cell cultures in the presence of 2 μM of 17 exhibited after 15 days (OM9;22 cells) and 30 days (K562 cells) almost complete proliferation inhibition [75]. Telomestatin (R-17) is a very potent telomerase inhibitor (IC50 ¼ 5 nM [30], 20 nM [32]) due to its ability to stabilize the

Frontier Between Cyclic Peptides and Macrocycles

163

G-quadruplex structures [32, 70, 71]. The S-enantiomer of 17 (S-17) was reported to be fourfold more potent with an IC50 ¼ 5 nM [32]. The presence of either the natural product R-17 or the enantiomer S-17 with a randomly structured singlestranded d[TTAGGG]4 oligonucleotide induced an antiparallel G-quadruplex structure, regardless of the configuration of 17, as demonstrated by Doi et al. [32] using CD analysis. The melting temperatures Tm of the complexes with S-17 and R-17 were reported to be 62.3 and 56.3 , respectively [32], suggesting that the S-enantiomer is the stronger G-quadruplex binder. In a previous study, Rezler et al. [72] found that telomestatin (17) binds to the antiparallel basket-type G-quadruplex structure with a 2:1 stoichiometry to d[TTAGGG]4. Cancer cells are characterized by aberrant cell division and have an increased ability to extend their telomeres. The enzyme telomerase (a DNA polymerase with endogenous RNA template [75]) has an important role in the process of telomere elongation. Telomerase is active in a majority of cancer cells but is inactive in most normal somatic cells. For optimal telomerase activity the non-folded single-stranded telomere DNA overhang is required. The formation of G-quadruplex structures masks the singlestranded d[TTAGGG]n primer molecules required for telomerase activity [76]. Telomestatin inhibits telomerase activity through a primer sequestration mechanism [72]. The effect of telomestatin (17) on telomeric DNA is not the only mechanism by which the compound acts on cancer cells. Comparing the sensitivity to telomestatin treatment, Miyazaki et al. [31] found that the growth of patient-derived glioma stem cells (GSC) was strongly impaired while normal neuronal precursors derived from fetal brains were less sensitive. Searching for target genes of telomestatin, these authors identified the protooncogene c-Myb, which has a telomestatin-binding sequence in its promoter region and showed reduction of its expression. They concluded that telomestatin eradicates GSCs through telomere disruption and c-Myb inhibition. Whereas physicochemical or pharmacokinetic data of telomestatin (R-17) is hard to find in the public domain, its very hydrophobic nature and poor water solubility have been described [77, 78]. Difficulties in isolation and handling of R-17 [74] or its enantiomer S-17 [32] have been attributed to π-π stacking of the oxazole rings [32]. L1BOD-7OTD (19; Fig. 8) was developed as fluorescent ligand for G-quadruplexes. The compound can be used to visualize G-quadruplexes by green fluorescence. The compound was reported to permeate HeLa I.2.11 cells and nuclear membranes and to be enriched in the nucleus of the cells [79].

164

Philipp Ermert et al.

Maleki et al. [76] used the Cy5 fluorophore bearing oxazole telomestatin analog 20 (also derived from 18) [80] to study binding kinetics and dwell time in interaction with different G-quadruplex structures by employing single-molecule Fo¨rster resonance energy transfer (smFRET). They proposed that 20 might be able to bind in two different orientations (with different dwell times/stabilities) to G-quadruplex structures. 2.3 Peptide Polyketide Hybrid Macrocycles

In microorganisms, the peptide part of peptide-polyketide hybrid macrocycles is synthesized by non-ribosomal peptide synthetases (NRPS) and its structural diversity goes beyond the regular 20 proteinogenic amino acids. The polyketide part typically consists of a carbon chain with multiple stereocenters and double bonds and is frequently substituted with methyl and hydroxyl groups. Polyketide synthetases (PKS) are key for the biosynthesis [81, 82]. Accessory domains in NRPS and PKS control the modification of the growing peptide or polyketide chain or incorporated amino acids. Such modification can include heterocyclization, Nmethylation, epimerization, or macrocyclization. NRPS and PKS are multifunctional synthetases with modular organization where the order and number of the modules dictate the structure of the natural products. Both types of synthetases use carrier proteins with 40 -phosphopantetheine as prosthetic group to tether and activate the growing intermediate via thioester group to the enzyme. Elongation of the peptide chain by C–N-bond formation as well as growth of the polyketide fragment by C–C-bond formation are achieved through nucleophilic attack at the thioester function in biosynthesis [81, 82]. The polyketide elements are thought, as pointed out by Bockus et al. [56], to confer balanced conformational flexibility, allowing peptide-polyketide macrocycles to adopt conformations, which are not accessible for only peptide-based scaffolds. Substituents of the polyketide part (i.e., 1,3-arranged methyl groups) were seen by Maier and coworkers [83, 84] as conformational control elements, providing soft constraints, which allow an induced fit between ligand and receptor. The hydrophobic nature of the polyketide moieties is able to compensate for the polarity of the peptidic secondary amide bonds. N-methyl amino acids and additional structural elements derived from polyketide synthethases like β-hydroxylated- and α,β-unsaturated γ-amino acids (statins and vinylogous amino acids) promote the ability of the hybrid structures to cross cell membranes. Lokey and coworkers [33] studied the influence of statins and vinylogous amino acids on ADME properties of cyclic peptides. Two sets of stereoisomeric hexapeptides were investigated, consisting of three N-methyl leucines, one proline, and either two statins (general structure A, 10 compounds, Fig. 9) or one statin and one vinylogous amino acid (general structure B, 4 compounds, Fig. 9).

Frontier Between Cyclic Peptides and Macrocycles

165

Fig. 9 The influence of statins and vinylogous amino acids on ADME properties of cyclic peptides was investigated by Lokey and coworkers [33]

The compounds were found to cover a broad range of permeability values (passive membrane permeability of 1–20  106 cm/s in the PAMPA assay and cellular permeability of 0.8–11.2  106 cm/s in the MDCK assay. Compound 21, selected based on good cell permeability and low liver microsomal clearance, was further characterized in an oral PK experiment in rat. At a dose of 5 mg/kg an oral bioavailability of ca 21% was found [33]. These authors determined NMR-based solution conformations of these hybrid model compounds in CDCl3. CDCl3 (dielectric constant 4.8 [85]) is used to mimic the center of a lipid bilayer (dielectric constant 4.0 [86, 87]). They concluded that the permeability correlated with the capability of the NH groups to participate in intramolecular hydrogen bonding, as observed also for some classical cyclic hexapeptides [88]. The OH groups were reported to participate in hydrogen bonds in most of the structures investigated with no correlation with permeability. 2.3.1 Apratoxin A

In 2001, Luesch et al. [34] described apratoxin A (22, Fig. 10), which was isolated from the marine cyanobacterium Lyngbya majuscula. The all-S configured 25-membered macrolactone comprises three either Nα- or O-methylated amino acids, proline as regular amino acid, and an α,β-unsaturated modified cysteine, which forms a thiazoline ring with the adjacent polyketide part, the dihydroxylated fatty acid Dtena. Grindberg et al. [90] described the isolation of the biosynthetic gene cluster responsible for apratoxin A synthesis in the cyanobacterium Lyngbya bouillonii and proposed a biosynthesis with release of the growing chain by macrolactone formation. Doi and colleagues [91] investigated the solution structure of apratoxin A (22) and apratoxin C (24, Fig. 11) applying molecular modeling with constraints from NMR data recorded in CD3CN. These authors concluded that the two apratoxins adopt similar conformations and that the methyl group at C(37) and the isopropyl/tert-butyl group at C(39) play a role in maintaining the conformation while the methyl group at C(34) does not. They also

166

Philipp Ermert et al.

Fig. 10 Structure of apratoxin A (22) and in vitro cytotoxicity against human cancer cell lines. In vivo the compound was found to be not well tolerated [34]. Apratoxin B (23) exists as mixture of conformers in CDCl3 [89]. moCys: modified cysteine; Dtena: 3,7-dihydroxy-2,5,8,8-tetramethyl-nonanoic acid

Fig. 11 Natural apratoxins and synthetic hybrids of apratoxin A/E [92–94]

Frontier Between Cyclic Peptides and Macrocycles

167

reported the CD3CN structure model of 22 to be very similar to the CDCl3 model proposed by Luesch et al. [34]. No intramolecular H-bond was observed. Apratoxin B (23; Fig. 10) containing isoleucine instead of Nmethyl isoleucine was reported by Luesch et al. [89] to exist in CDCl3 as 3:1 mixture of two conformers, which differ in the configuration around the Tyr(Me)-MeAla amide bond. Close proximity of the Ile NH and the carbonyl CO of the moCys residue in the predominating cis form suggest the presence of an intramolecular hydrogen bond. Apratoxin A (22) was reported to exert in vitro a potent antiproliferative effect on human cancer cell lines and to induce G1 cell cycle arrest and apoptosis [34, 95, 96]. In vivo, the compound was found to be not well tolerated and only marginally active against a colon tumor implanted subcutaneously into mice [34]. Dosedependent anticancer activities were reported for other human cancer xenograft models (BxPC3 T1 pancreatic cancer model and A549 lung cancer model) to go along with body weight loss [35]. The mode of action of 22 was studied applying genomics [95] and chemical biological approaches [35–37]. Apratoxin A (22) was found to reversibly inhibit the secretory pathway for several cancerassociated receptors by preventing cotranslational translocation [36] and ultimately the Sec61 complex was identified as the molecular target of 22 [35, 37]. The oxazoline analog (26, Fig. 11) was reported to bind to Hsc70/Hsp70 and proposed to stabilize the interaction of Hsp90 client proteins with Hsc70/Hsp70, thus preventing their binding to Hsp90 [97]. Membrane proteins and secreted proteins contain an N-terminal signal sequence which is required for targeting the secretory pathway. After initiation of the protein synthesis from such m-RNA template by free ribosomes, the signal recognition particle (SRP) binds to the signal sequence and mediates the binding to the SRP receptor located in the ER. In mammalian cells, continued protein synthesis is concomitant with translocation to the ER lumen. This process is called cotranslational translocation and requires besides SRP and SRP receptor the Sec61 complex, which is the protein translocation channel localized in the ER membrane. The process is facilitated by accessory factors and finally signal peptides are then cleaved by cellular signal peptidases [36, 37]. Apratoxin A (22) inhibits protein production at the stage of cotranslational translocation; synthesis of the proteins that entered into the secretory pathway occurs but is diverted to the cytoplasm, where these proteins are readily degraded through a proteasomemediated pathway [36]. Apratoxin A prevents protein translocation by directly targeting Sec61α, the central subunit of the protein translocation channel, and blocks the biogenesis of Sec61 clients

168

Philipp Ermert et al.

Table 3 Tissue distribution of apratoxin A [35] AUC(0–48) (ng h/mL)

Half-life (h)

Vss (L/kg)

182

11.57

19.3

Liver

8677

25.15

Lung

3404

17.3

100,368

27.6

47,955

23.6

Plasma

Pancreas Salivary gland Tumor

3709

a

Dose: 1 mg/kg, i.v., mice with subcutaneously implanted A549 tumor a Long drug half-life or no drug concentration reduction after 48 h

[37], which results in downregulation of secretory and membrane proteins among them several cancer-associated growth factors and receptors [36]. HUN-7293, a macrocyclic fungal metabolite with anti-inflammatory activity, is also a Sec61 inhibitor; however, it has a different binding site from apratoxin A [35]. The main toxicity observed in mice was severe body weight loss [34, 35]. Huang et al. [35] found the pancreas to be the main target organ of apratoxin A (22). Pancreatic atrophy was observed in mice and a histopathological evaluation revealed degeneration/ single-cell apoptosis of pancreatic tissue. Other major organs were reported to show no detectable changes. High drug exposure was reported for all tissues tested (Table 3) [35]; however, highest values of 22 were found for pancreatic tissue. Huang et al. [35] concluded that high exposure might only partially explain the effects on pancreas and proposed 22 to inhibit the cotranslational translocation of target proteins, which are essential for the survival of pancreatic cells. Due to the 27-fold higher exposure in the pancreas than in the tumor, these authors considered it to be unlikely that the same target proteins are responsible for the effects on pancreatic cells and for the anticancer activity. Several research groups reported syntheses of apratoxins and analogs and established a structure activity correlation [91–93, 98, 99]. These efforts allowed to separate toxicity and anticancer activity [92] and most recently provided improved apratoxin analogs [98]. The synthesis and biological activity of apratoxin derivatives were reviewed by Rastelli and Coltart [99]. In line with the abovementioned conformational studies, apratoxin A (22) and its 34-epi analog were found to be equipotent [93]. Removal of all substituents of Dtena abolished the activity [93]. Removal of the methyl group at C(37) or inversion of the

Frontier Between Cyclic Peptides and Macrocycles

169

Table 4 Activity of apratoxin analogs [92, 98]

Compound

HCT116 Cell viability IC50 (nM)

HCT116 VEGF-A secretion IC50 (nM)

28

1.43

0.32

29

1.25

0.30

30

1.99

0.47

31

220

32

120

33

1.1

Apratoxin A

2.8

configuration at C(37) was reported to abolish antiproliferative activity of the oxazoline analog 26 of apratoxin A (22) [96]. The macrolactam analog 25 [94] was more than 1000-fold less active than 22. Conformation analysis (on the basis of NMR data recorded in CD3OD and molecular modeling) revealed deviations in the orientation of the O-methyl tyrosine side chain, the isoleucine side chain, and the C(35)-OH group. The Michael acceptor present in apratoxin A (22) was demonstrated to undergo additions of cellular nucleophiles like glutathione, cysteine, or N-acetyl cysteine [92]. Apratoxin E (27), a natural analog with saturated C(27)–C(29) carbonyl system, shows reduced antiproliferative activity due to the dehydration at C (35) [92]. Luesch and coworkers [92] reported the synthesis of the apratoxin A/E hybrid 28 (Fig. 11), containing the C(35)-OH group and lacking the Michael acceptor. This compound again exhibits low-nanomolar antiproliferative activity. Similar activity was observed for the C(34) epimer (34-epi-28) and for a simplified non-methylated analog 29 and the gem-dimethyl analog 30 (Table 4) [92]. The antiproliferative activity of these compounds was found by Luesch and colleagues to be paralleled by potent inhibition of the VEGF-A secretion in HTC116 cells (Table 4) [92]. The analog 30, which cannot be deactivated by dehydration, was further tested in a HCT116 xenograft mouse model and reported to show a dose-dependent tumor growth inhibition. 30 was administered at a dose of 0.25 mg/kg or 0.1 mg/kg once daily i.p. for 16 days. No body weight loss was observed. Luesch and coworkers concluded that the toxicity may be an off-target effect, related to the presence of the Michael acceptor [92].

170

Philipp Ermert et al.

Fig. 12 Replacement of the moCys and the Tyr(Me) moiety by amino acids [98]

Doi and coworkers [98] recently described analogs of apratoxin A with the moCys moiety replaced by seven amino acid linkers, conferring variable flexibility and possibilities to form hydrogen bonds. The analogs 31 and 32 with the more constraint motives, containing either 3-(N-methylamino)benzoic acid or piperidine-4carboxylic acid, respectively, exhibited moderate activity against HCT-116 cells with IC50 values in the range of 100–200 nM (Fig. 12, Table 4). Substitutions of amino acid residues in the Tyr(Me)-MeAlaMeIle moiety in 32 led to the identification of 33, which was again equipotent to apratoxin A. In 33, the O-methyl tyrosine was replaced with biphenyl alanine [98]. In a panel of human cancer cell lines, 33 and apratoxin A (22) exhibited comparable GI50 values (4–40 nM) against eight out of ten cell lines. 2.3.2 Jasplakinolide

Jasplakinolide (34, also named “jaspamide”; Fig. 13) is a 19-membered macrocyclic depsipeptide isolated from the marine sponge Jaspis splendens. It is freely cell permeable and binds to F-actin, stabilizing filaments in vivo [39]. The melting temperature Tm (denaturation) of Ca-F-actin filaments was reported to be at 67.3  C; in the presence of jasplakinolide Tm shifted to 87.7  C [104]. Because of its membrane permeability, jasplakinolide (34) is a commonly used tool in cell biology [105]. Jasplakinolide (34) was reported to also exhibit fungicidal, insecticidal, anthelmintic, herbicidal, ichthyotoxic (toxic to fish), and antimalarial properties; cf. [39] for leading reference. Jasplakinolide exhibits potent cytotoxicity to various cancer cell lines [39]. The biological evaluation of the natural analogs 35–37 and 39 [39, 100–102] and the simplified synthetic analogs 38 and 40–43 [39] contributed to

Frontier Between Cyclic Peptides and Macrocycles

171

Fig. 13 Jasplakinolide/jaspamide (34) and analogs; natural analogs: 35 (jaspamide J), 36 (jaspamide M), 37 (jaspamide F), and 39 (jaspamide H) [39, 100–102]. Cell-permeable probe 45 allows to fluorescence label F-actin in living cells [103]. Jasplakinolide analog 46 with rigidified non-peptide part [83]

the understanding of the structure-activity correlation and allowed the development of probes for selective chemical imaging of static F-actin in living cells (Fig. 13, Table 5) [103].

172

Philipp Ermert et al.

Table 5 Cytotoxicity of jasplakinolide (34) and analogs toward MCF-7 and HT-29 cells

Cpd

MCF-7 IC50 (μM)

HT-29 IC50 (μM)

Refs.

34

0.025

0.021

[39]

34

0.02

0.04

[100–102]

35

5.0

–

[100–102]

36

0.1

0.18

[100–102]

37

30

–

[100–102]

39

30

–

[100–102]

Truncation

38

0.023

0.011

[39]

Indole bromination

40

0.04

0.026

[39]

Indole bromination, C(6)–CH3

41

0.085

0.168

[39]

Indole bromination, C(6)–CH3, C(8)–CH3

42

>15

>15

[39]

Indole bromination, β-Tyr, C(6)–CH3

43

>15

>15

[39]

Indole bromination, β-Tyr, C(6)–CH3, C(8)–CH3

0.01

[103]

44

0.039

MCF-7: human breast adenocarcinoma cells; HT-29: colon carcinoma cells

The debromo analog 38 of jasplakinolide was reported to be equipotent [39], and the analog 36 with the D-tryptophan Nmethyl group lacking was found to be fivefold less cytotoxic [101]. Removal of the methyl group in position 2 (! 35) was described to go along with a two-order-of-magnitude loss in activity, while removal of the methyl group in position 4 (! 37) and in position 6 (! 39) led to inactive compounds [39, 100–102]. In contrast to this, the removal of the methyl group in position 6 of the debromo analog 38—leading to 40—surprisingly was tolerated, whereas compound 41, demethylated in positions 6 and 8, was ca tenfold less potent. The β-tyrosine moiety was found to play a key role as its replacement with β-alanine (! 42, 43) abolished cytotoxic activity [39]. In summary the 2-bromo substituent of the indole moiety of 34 is not required and the aliphatic methyl group in position 6 can be omitted to obtain the comparable efficacy (! 40). Maier and coworkers [83] prepared analogs of jasplakinolide (34) with a more rigid non-peptide part by replacing the 1,3-arranged allylic and homoallylic methyl groups in positions 4 and 6 of the ω-hydroxy acid by an aromatic ring and the ω-hydroxy group by an amino group. The resulting macrolactam 46 exhibited weak cytotoxic activity (L929 mouse fibroblasts, IC50 ¼ 25 μg/mL, and SKOV-3 ovary cancer cells,

Frontier Between Cyclic Peptides and Macrocycles

173

IC50 ¼ 20 μg/mL). The diastereomers of macrolactam 47 were found to be 75-fold and 300-fold less active in a cell growth inhibition assay (CA46 Burkitt lymphoma human cells) than the parent jasplakinolide (IC50 ¼ 10 nM) [106]. Ghosh et al. [106] concluded that the macrolactone motif seems to be important for biological activity. Derivatives without substituent in 2-position also exhibit reduced activity (cf. 35), and thus the conclusion is debatable. Investigations of Gala et al. [102] suggested that the alanine methyl residue of 34 can be modified. Analogs with alanine replaced by 2-aminobutyric acid (jaspamide D) or serine (jaspamide E) retain cytotoxic activity. Based on these findings the cell-permeable probe 45 (Fig. 13) was developed, which allows targeting of F-actin structures in living cells. Starting from 44, the lysine homolog of 40, the conjugate 45 was obtained by attachment of β-4,4-difluoro-5,7-dimethyl-4bora-3a,4a-diaza-s-indacenyl propionate via 6-aminohaxanoic acid linker to the lysine-ε-amino group [103]. The conjugate 45 is well cell permeable and was described to selectively label static longlived actin filaments while only weakly affecting dynamic actin structures with more rapid turnover. In contrast to 45, jasplakinolide (34) and the intermediate 44 were found to be not selective. Milroy et al. [103] commented that the fluorescence label reduces the compound’s actin disturbing activity and that further clarifying is required to explain why 45 is not toxic. 2.3.3 Sanglifehrin A

The bacterial 22-membered macrolactone natural product sanglifehrin A (48, Fig. 14) [40, 47, 107] consists of a tripeptide portion (valine-meta-tyrosine-piperazic acid) and an extended polyketide structure. It binds with nanomolar affinity to cyclophilin (Cyp) A, B, and C. The tripeptide moiety is the recognition motif [47, 107]. Sanglifehrin A (48) also exhibits immunosuppressive properties. Sanglifehrin A (48) is marginally soluble in PBS (45 μM; ¼ 0.049 g/L) and exhibits low microsomal stability (human and mouse t1/2 < 8 min); in human hepatocytes however, 48 was found to be more stable (t1/2 ¼ 145 min) [41]. The oral bioavailability is low (F ¼ 1.42%, dosed to mouse at 10 mg/kg); the long half-life (t1/2 ¼ 8 h) was attributed to the low clearance (0.054 L/ h/kg; corresponding to 1% of hepatic blood flow in mice) [41]. The liver penetration of sanglifehrin A (48) was confirmed by Gregory et al. [41], who found the concentrations of 48 to be 8–21 times higher in the liver than in whole blood. CypA is necessary for hepatitis C virus (HCV) replication and binds to the viral protein NS5A. To treat HCV infection, non-immunosuppressive CypA inhibitors would be of interest [41]. The aldehyde 49, obtained by oxidative cleavage of the C (26)–C(27) side-chain double bond, maintained affinity for the

174

Philipp Ermert et al.

Fig. 14 Sanglifehrin A (48) and truncated analogs [40, 47, 107, 108]

cyclophilins but showed no immune-suppressive activity [107]. A further simplified analog 52, which lacks the side chain at C (23) and the substituents in positions 14–17, exhibited still micromolar affinity for cyclophilins [47, 107].

Frontier Between Cyclic Peptides and Macrocycles

175

Table 6 Biochemical and cellular potency of cyclophilin A inhibitors 50–55

Compound

TR-FRET CypA binding KD (nM)

PPIase Functional assay Ki (nM)

HCV genotype 1b replicon cellular assay EC50 (nM)

50

25

16

600

51

65

65

240

52

2600

795

4500

53

64

14

320

54

11

4

36

55

24

7

87

CspA

17

CspA ¼ cyclosporin A; data taken from Steadman et al. [40]

Table 7 Thermodynamic data of inhibitors 50–52 binding to cyclophilin A [40] Compound

ΔG (kcal/mol)

ΔH (kcal/mol)

TΔS (kcal/mol)

50

10.3

9.97

0.32

51

9.5

8.6

0.93

52

8.2

7.0

1.2

Recently Steadman et al. [40] described the analogs 50 and 51 (with substituents present in positions 14–17), which were found to be more potent than 52 (Table 6). X-ray crystal structures of complexes of 50 or 52 with Cyp A revealed a well-conserved binding mode of the tripeptide portion (which is responsible for all direct hydrogen bonding interactions to CypA), while the conformation of the non-peptidic moiety was found to be different. Like sanglifehrin A and unlike 52, inhibitor 50 exhibits an intramolecular hydrogen bond between the amide NH of Val and the oxygen atom of the substituent at C(15). The affinity difference of the two compounds could be attributed to a difference in binding enthalpy (Table 7) [40]. The replacement of the diene by a meta-substituted styryl group further rigidified the non-peptidic portion, as found by Steadman et al. [40]. Compound 53 was reported to bind in a new mode. The phenolic side chain of m-Tyr is no longer hydrogen bonding to His(126) and the styrene moiety is undergoing a π-stacking interaction with the guanidinium side chain of Arg (55). The good activity of compound 55 (Table 6) confirmed

176

Philipp Ermert et al.

Fig. 15 Simplified analogs of largazole (60) and largazole thiol (61) [45, 48, 109, 110]

that the m-Tyr side chain is dispensable in the presence of the styryl linker [40]. Compounds 51 and 55 were found not to be immunosuppressive. A patent application, describing analogs of 55, was approved in 2018 [108]. Examples exhibiting in the cyclophilin TR-FRET competitive binding assay IC50 values
100 nM and anti-HCV replicon activity of 100% at 1 μM comprise, among others, compounds 56–59. These analogs feature a 21-membered macrolactone ring with an even further simplified and more rigid non-peptide part. 2.3.4 Largazole

Largazole (60; Fig. 15) is a natural 16-membered macrocyclic depsipeptide, combining proteinogenic and azol(ine)-heterocyclized amino acid residues with a polyketide moiety and a fatty acid S-acyl substituent, the latter acting as transient protection of the sulfhydryl group. Largazole is a potent class I selective histone deacetylase (HDAC) inhibitor [48] and was isolated from a marine cyanobacterium. HDACs are zinc metalloenzymes which catalyze the cleavage of an acetyl group from lysine residues in histones and other substrates [111, 112]. Cleavage of the octanoyl group releases the largazole thiol 61 which is thought to coordinate Zn2 + . The depsipeptide ring system shields the reactive thiol and acts as surface recognition group [48, 111, 112]. A multigram-scale synthesis of largazole was developed by Luesch and coworkers [113]. Several research groups described

Frontier Between Cyclic Peptides and Macrocycles

177

Table 8 IC50 values of largazole (60) and 69 against HDACs 1, 2, and 3 of class I and HDAC6 of class II [48] Compound

HDAC1 IC50 (nM)

HDAC2 IC50 (nM)

HDAC3 IC50 (nM)

HDAC6 IC50 (nM)

Largazole (60)

61

64

50

1400

69

21

28

27

13,000

largazole analogs [112, 114–117]. Poli et al. [118] recently reviewed their structure-activity relationships. Reddy et al. [109] synthesized simplified largazole analogs, where the Zn-binding groups were varied, the thiazole-thiazoline part was replaced by a proline dipeptide, and the β-hydroxy acid was substituted by aspartate, providing a cyclic peptide with a 13-atom macrocyclic core. The C-terminal aspartate amide bond served as an equivalent of the C-C double bond. Compounds 62 and 63 with a hydroxamic acid Zn-binding group were most active. Substitution of the thiazole-thiazoline largazole fragment with alternative groups, while maintaining the ring size and warhead, was described by Almaliti et al. [48]. The analogs 68–70 were biologically evaluated; the bipyridyl analog 69 (the best compound of the series described) showed potency against HDACs comparable to that of largazole. For 69, selectivity for class I HDACs over class II HDACs (Table 8) [48] was improved. All three compounds were tested on a NCI 60 cell line panel for antiproliferative activity. Inhibitor 69 was reported to be the most active compound with a profile similar to largazole. In a HCT116 cell growth inhibition assay 69 had a GI50 of 75 nM (largazole: GI50 ¼ 65 nM) [48]. Analogs with bi(hetero)aryl motif were previously obtained by Li et al. [110], who replaced the 4-methylthiazoline-4-carboxylic acid moiety by triazolo- and tetrazoloacetic acid, thus increasing the size of the macrocycle by one atom. In the enzymatic assay against HDAC1 the triazole (64; Fig. 15) as well as the tetrazole analog (66) were described to be 50- to 120-fold less potent than largazole (60). Largazole analogs with the thiazole ring substituted by a pyridine ring, maintaining the ring size at 16 atoms, and lactam isosteres have been studied in in vitro and in vivo experiments [45, 46]. Compared to largazole, pyridine analogs 73 and 75 (Fig. 16) exhibited reduced potency in the enzymatic assay; their thiol derivatives 74 and 76, however, were found to be as potent as largazole thiol (61). In the cellular assays the largazole analogs 73 and 75 (prodrugs) were more potent than the corresponding thiols with 75 even more potent than largazole (Table 9).

178

Philipp Ermert et al.

Fig. 16 Lactam isosteres of largazole and analogs [45, 48, 109, 110] Table 9 Enzymatic and cellular potency of largazole analogs and lactam isosteres reported by Clausen et al. [45] Compound

HDAC1 IC50 (nM)

HDAC2 IC50 (nM)

HDAC3 IC50 (nM)

HDAC6 IC50 (nM)

797 cell line IC50 (nM)

10,326 cell line IC50 (nM)

Largazole (60)

10.09

18.65

9.09

165.6

24

25

61

2.51

4.19

2.78

28.11

100

80

71

544.1

825.2

1151

–

110

170

72

1.95

3.38

2.59

102

1740

11,940

75

340

655

319

–

10

10

76

2.2

4.42

2.31

35.16

90

120

77

817

1240

846

–

70

130

78

13.2

20.8

14.59

2849

4250

8640

In the enzymatic assay, largazole thiol 61 and the lactam isostere 72 exhibit comparable potency against HDACs 1–3 [45, 46], whereas the thioester lactam isostere 71 is 44–126-fold less potent than parent largazole (60) (Table 9) [45]. In the cellular assays the thioester prodrugs were found to be more potent than the free thiols (60 vs. 61; 71 vs. 72; Table 9). Reduced potency (5- to 70-fold) was observed for the lactam

Frontier Between Cyclic Peptides and Macrocycles

179

Table 10 Enzymatic potency of largazole thiol (61), the pyridine analog 76, and the corresponding lactam isosteres 72 and 78 reported by Pilon et al. [46] Compound

HDAC1 IC50 (nM)

HDAC2 IC50 (nM)

HDAC3 IC50 (nM)

HDAC6 IC50 (nM)

61

3.4

10.5

10.4

203

72

5.8

10.0

12.0

2045

76

3.9

16.5

17.9

655

78

38.7

76.4

82.2

33,710

Table 11 In vitro antiproliferative activity of largazole, the pyridine analog 75, and the corresponding lactam isosteres 71 and 77 [46] Compound

HTC-116 IC50 (nM)

HT-29 IC50 (nM)

SW620 IC50 (nM)

A549 IC50 (nM)

MiaPaCa IC50 (nM)

Largazole (60)

3.5

16.2

26.5

3.8

206.4

71

103.3

81.0

224.8

281.2

842.7

75

0.7

0.3

8.2

0.18

91.8

77

82.9

31.5

81.0

542.7

3536

HCT-116, HT-29, and SW620 are colon cancer, A549 non-small cell lung cancer, and MiaPaCa pancreatic cancer cell lines

Table 12 PK properties of largazole (60) and largazole lactam isostere 71; mice, 5 mg/kg, i.v. [46] Compound

AUC (min μg/mL)

CL (mL/min/kg)

MRT (min)

Vss (L/kg)

Largazole (60)

31.34

160

84

13.45

71

44.69

111

108

12.10

isostere 71 compared to largazole 60 (Tables 9 and 11). The thiol 72 was found to be 17- and 150-fold less active than largazole thiol (61) in the cellular assay, while being equipotent in the enzymatic assay (Tables 9 and 10). This reduced cellular potency may illustrate the contribution of the lactone ring to cell permeability. Largazole (60) and the lactam isostere 71 exhibit comparable pharmacokinetic parameters (Table 12), the latter showing an increased drug exposure (AUC, MRT) as a result of slower elimination or better stability as argued by Pilon et al. [46]. In an A549 non-small cell lung carcinoma xenograft model (mice, i.p. administration of 5 mg/kg largazole or largazole analogs

180

Philipp Ermert et al.

every second day), largazole (60) and the lactam isostere 71 exhibited 32% and 66% tumor growth inhibition, respectively [46]. For comparison, largazole thiol 61 exhibited 42% tumor growth inhibition, demonstrating that in vivo the protective moiety is dispensable for antitumor effects, although it may reduce toxicity [46]. In addition, pyridine analog lactam isostere 78 was reported [45] to be 50- to 70-fold less potent than the free thiol 76 in the cellular assay (Table 9), while being almost equipotent (ca 5- to 10-fold difference) in the enzymatic assay (Tables 9 and 10). Potency differences for the thioester compounds 75 and 77 were found to vary a lot for different cell lines (10- to 3000-fold with the lactam being less potent; Tables 9 and 11). Despite good in vitro potency, the pyridyl analogs were surprisingly found to show no tumor growth inhibition under the above-described dosing regimen [46].

3

Semisynthetic Approaches Toward Analogs of Macrocyclic Natural Products A plethora of molecules at the interface of cyclic peptides and non-peptidic macrocycles can be found in nature as described in the previous sections. Many natural product-derived molecules were made using semisynthetic approaches, i.e., preparation of molecules by modification of a natural product starting material. Although total synthesis of natural product analogs was described, many drug discovery programs use semisynthetic approaches. Semi-synthesis entails biological production of a natural product usually by fermentation, followed by chemical derivatization. Such an approach allows a fast, economical, and time-efficient optimization versus otherwise tedious multistep total synthesis approaches. Due to the structural complexity of natural products, semisynthesis may be the only economical scalable route to manufacture drug candidates, provided that the natural product precursor can be fermented in sufficient quantities. There is a vast body of literature describing examples of semisynthetic alterations of natural peptidic macrocycles, in particular in the antibacterial (e.g., macrolides [119], glycopeptides like vancomycin analogs [120] or the mannopeptimycins [121], depsipeptides like daptomycin derivatives [122]), antifungal (e.g., echinocandin derivatives [123]), immunological (e.g., cyclosporine derivatives [124]), and oncological fields (e.g., temsirolimus [125], kahalalide F [126]), just to name a few of them. A few more recent examples without any intention of being comprehensive are described below. Interesting new members of thiazolyl actinomycete metabolite family [55, 127] have been described recently: GE2270 A, thiomuracin A, and baringolin [68]. These highly modified,

Frontier Between Cyclic Peptides and Macrocycles

181

Scheme 2 Semi-synthesis of LFF571 from GE2270 A 3.1 Thiazolyl Actinomycete Metabolites

sulfur-containing macrocyclic peptide structures possess a tri- or tetra-substituted nitrogen-containing heterocyclic core. They all show antimicrobial activity against Gram-positive bacteria inhibiting bacterial cell growth by targeting protein synthesis elongation factor Tu (Ef-Tu, located in the cytoplasm) which is involved in protein synthesis. These three thiopeptide-based natural products contain a central pyridine ring which is decorated by three thiazole or by two thiazole and one oxazole structural element. GE2270 A has been isolated from fermentation broth of Planospora rosea and the structure has been published first in 1991 [29]. This 29-membered macrocyclic peptide possesses a tri-substituted pyridine core and contains five thiazole units in the macrocycle and one exocyclic thiazole and oxazoline moiety (Scheme 2). GE2270 A exhibits good in vitro antimicrobial activity (MICs 25 steps) [134, 135]. In particular molecule 82 (Fig. 17) shows promising antibacterial activity on a selection of Gram-positive bacterial strains [135]. However, an economical scalable route to manufacture drug candidates based on the fully synthetic approach might be challenging. 3.2 Streptogramin Antibiotics

The emergence of multidrug-resistant bacteria initiated renewed interest in older drugs. Streptogramins, discovered in the 1950s, target the ribosome and inhibit protein synthesis [42, 136]. They consist of two structurally different subgroups. Group A compounds are 23-membered peptide/polyketide hybrid polyunsaturated macrolactones, incorporating an oxazole ring; group B molecules are 19-membered macrocyclic depsipeptides (Fig. 18). Streptogramins A and B (also called virginiamycins, pristinamycins) are used in combination. Whereas each individual component exhibits a bacteriostatic effect, their combination, however, is bactericidal [136, 137]. The renewed interest is reflected in a recent development of a fully modular synthesis of group A streptogramins in 2017 [138] (cf. Subheading 4), which could serve as a basis for the development of improved analogs. Furthermore, efforts to understand the structural basis of the synergistic effect of group A and B molecules resulted in the determination of X-ray structures of the clinically relevant streptogramin combinations (Synercid and NXL103)

184

Philipp Ermert et al.

Scheme 4 Semi-synthesis of dalfopristin (84) [42, 44]

bound to the 70S ribosome of E. coli [136] or to the 50S ribosomal subunit from Deinococcus radiodurans [137]. In addition, a structure of madumycin II (86) bound to the 70S ribosome of Thermus thermophilus was described [139]. Synercid is an injectable formulation of a 70:30 (w/w) mixture of dalfopristin (84) and quinupristin (88). The product has been approved in the USA in 1999 for the treatment of Gram-positive infections in hospital [43, 136], ca 40 years after the discovery of the compound class. The natural streptogramins show poor water solubility (virginiamycin M1: 50 mg/L at pH 7.4) [43]; the group A molecules furthermore exhibit limited chemical stability at acidic and basic pH due to the presence of the sensitive β-hydroxy ketone [43]. A semisynthesis program led to the identification of positions allowing the introduction of solubilizing groups without loss of biological activity [42, 44]. Dalfopristin (84) was derived from virginiamycin M1 (83) at low temperature by stereoselective 1,4-addition of diethylaminoethanethiol to the α,β-unsaturated lactone affording intermediate 90 (Scheme 4). The sulfide 90 was then oxidized to the sulfone dalfopristin (84) [42, 44]. In vivo, dalfopristin rapidly metabolizes to virginiamycin M1 [44, 140–142]. Quinupristin (88) was derived from virginiamycin S1 (87), which was converted into the α,β-unsaturated ketone 93 (Scheme

Frontier Between Cyclic Peptides and Macrocycles

185

Scheme 5 Semi-synthesis of quinupristin (88); the Mannich reaction was used to functionalize the α-position of the ketone 87 [42, 44]

5). The regioselectivity was reported to be a consequence of the conformation of the molecule with the neighboring phenyl substituent shielding the β-methylene group [42] of the 4-oxopipecolic acid moiety. The subsequent 1,4-addition of (3S)-quinuclidine thiol proceeded stereoselectively in acetone at low temperature to afford 88. Further efforts (Scheme 6) to modify group A streptogramins led to the semi-synthesis of flopristin (85) [43]. Reduction of the keto group of virginiamycin M2 (94), silyl protection of the hydroxyl group in position 14, and DAST fluorination led to the 16-deoxo-16-fluoro derivative 97. Subsequent desilylation afforded flopristin (85). Flopristin exhibited improved MICs against Gram-positive bacteria as a consequence of increased lipophilicity but also reduced aqueous solubility as compared with the parent compound 94. Flopristin (85) was found to be stable at pH 4–9 but still unstable at pH 2 and 12; it showed an oral bioavailability of 1.8% (while the oral bioavailability of 94 was almost 0%) [43]. Flopristin was

186

Philipp Ermert et al.

Scheme 6 Semi-synthesis of flopristin (85). Solubility at pH 7.4: Virginiamycin M2 (94) 3500 mg/L; flopristin (85): 220 mg/L. Chemical stability: Virginiamycin M2 (94) unstable at pH 2–4 and 7–12; flopristin (85): stable between pH 4 and 9; unstable at pH 2 and 12 [43]

clinically evaluated in combination with linopristin (89) as orally available streptogramin combination NXL 103 [43, 136] for the treatment of bacterial respiratory tract and skin infections. The trials were discontinued in Phase 2.

4

Library Approaches: Macrocycle Libraries by Modular Synthesis A variety of approaches toward arrays of compounds at the interface between cyclic peptides and macrocycles with reduced number of peptidic bonds have been developed over the last decade making a broad diversity of scaffolds accessible. They are mostly based on solid-phase organic synthesis, with the ring closure performed in solution, in the presence of a coupling agent, a synthetic catalyst, or even a macrocyclase enzyme. In order to capitalize on the modular architecture of macrocycle libraries, it is important to synthesize significant numbers of compounds with a limited effort and to benefit from a facilitated, rapid hit amplification and optimization. For the synthesis of larger number of final products, suitable building blocks must be accessible in multi-gram quantity. Usually, some

Frontier Between Cyclic Peptides and Macrocycles

187

Scheme 7 Synthesis of natural product-inspired cyclo-depsipeptides developed by Arndt et al. [143]. Peptide synthesis through DIC coupling of Fmoc amino acids; diketopiperazine formation during N-terminal deprotection of the second amino acid residue was controlled by minimal exposure of the substrate; conditions used: piperidine/DBU/DMF 2:2:96; 20 s. Release of the protected macrocyclic product 99 as 1:1.2 E/Z mixture by relay ring-closing metathesis (RRCM)

of the building blocks have to be prepared to adapt to the needs defined by the synthetic strategy (e.g., protective group pattern). The ring closure is often the most critical step. Robust cyclization conditions are of key importance to avoid loss of material in this late step. The modular synthesis of macrocyclic depsipeptides with peptide/polyketide structure developed by Arndt et al. [143] (Scheme 7) was inspired by natural products like jasplakinolide (34, Fig. 13). The approach is suitable for the synthesis of diverse compound arrays, since both the peptide and polyketide parts can be varied. This was illustrated by the synthesis of 11 prototype compounds with variation of the amino acid composition, length of the peptide sequence, and substitution pattern (number of methyl groups, configuration of chiral centers) of the polyketide part. The acyclic precursor chain was assembled on solid support, attached via an allylic 1,6-diene handle. Release from solid support was achieved by relay ring-closing metathesis (RRCM, via initial ruthenium carbene formation going along with cleavage from the solid support,

188

Philipp Ermert et al.

Fig. 19 MacroEvoLution is AnalytiCon Discovery’s platform for the generation of scaffold-diverse libraries of macrocyclic compounds [144]. Natural product-like motifs, e.g., hydroxy-methyl-substituted γ-amino acid or azole rings, are included. Boc-amino group, tert-butyl- and benzyl ester, and alkyne serve as functional groups for lateral derivatization

followed by closure of the macrocyclic ring in a second metathesis). The strategy was validated with the synthesis of seragamide A. A modular approach toward macrocyclic peptide-based scaffolds (Fig. 19) was recently presented by scientists of AnalytiCon Discovery [144]. Cyclic tripeptide scaffolds, consisting of natural and synthetic amino acids (β- and γ-amino acids, heterocyclic derivatives of amino acids, to reduce the peptidic character of the products), some of them with side-chain functional groups for later derivatization of each scaffold (by introduction of lateral diversity), were obtained in a two-step process. It comprised a cyclization screen and an upscale. Assembly of arrays of 8  8  8 building blocks (6 μmol scale) by solid-phase peptide synthesis (Fmoc strategy) was reported to afford up to 512 linear peptides; the combination of the selected building blocks would in principle allow to access 9- to 21-membered macrocyclic scaffolds. Subsequently, the ring closure was performed in solution under highdilution conditions (cfinal ¼ 1 mM) by PyBOP-mediated lactam formation, in 96 parallel synthesis format, using a microtiter plate. Analysis by LC-MS revealed that 100 macrocyclic tripeptides were obtained (success rate ca 20%). All ring sizes with the exception of 10-membered rings were represented whereby closure of larger rings was reported to be more often successful, 18-membered rings being however an exception. The cyclized compounds were synthesized on larger scale, followed by deprotection, side-chain derivatization, and preparative HPLC purification of the final products. The molecular weight of the compounds was generally kept below 1000 Da, in order to be compliant with the bRo5 criteria for permeability and oral bioavailability [13, 14]. To further reduce the peptidic character of the products, Saupe et al. [144] proposed to replace amide bonds by aryl ethers or

Frontier Between Cyclic Peptides and Macrocycles

189

Fig. 20 Ghrelin agonists (104, 105) were identified and developed applying Tranzyme’s MATCH™ technology [147–149]. The structurally closely related compounds 106 and 107 patented by Roche [150] are representatives of a series of analogs exhibiting antibiotic activity

carbon–carbon bonds. Jefferson et al. [145, 146] at Ibis Therapeutics described quinolone-macrocycle conjugates. The macrocyclic scaffold derived from two amino acids, an amino alcohol and a fluoronitro benzoic acid, was assembled and cyclized by SNAr aryl ether bond formation. Tranzym developed the MATCH™ approach [147–149] and independently Polyphor, the MacroFinder® platform. Both used variable modules (building blocks) to provide access to different scaffolds. Both technologies allow to obtain all stereoisomers of a given scaffold. Late-stage derivatization of lateral functional groups provided further amplification of the structural diversity of the products. The application of Tranzyme’s approach (MATCH™) allowed the discovery and development of potent and selective ghrelin agonists (104; Fig. 20) [147–149], which are based on compounds of the general formula 103 [148]. Macrocycles generated by assembly of a tripeptide sequence and a molecular tether were

190

Philipp Ermert et al.

Fig. 21 MacroFinder® platform allows modular assembly of chiral and achiral building blocks in high throughput; prototype compound 108. The orally available MacroFinder® compound 109 exhibits “chameleonic” behavior

intended to mimic the pharmacophore motif of peptidergic GPCR ligands. Starting from ghrelin agonist 104, hit to lead optimization led to 105 (TZP-101, Ulimorelin), which was evaluated in Phase III clinical trials for the treatment of postoperative ileus and acute gastroparesis. The clinical evaluation was discontinued after failing to meet efficacy endpoints [151]. More recently, structurally similar compounds 106 and 107 (Fig. 20) were patented [150] by Roche scientists. The compounds were described as potent antibiotics specifically against Acinetobacter baumannii. IC50 values 1000-fold) for the N-type calcium channel versus other calcium channels (like P/Q types). Ziconotide acts as a potent analgesic by reversibly blocking the N-type calcium channels localized on primary nociceptive afferent nerves of the dorsal horn of the spinal cord. Importantly and in

Natural Product Inspired Cyclopeptides and Macrocycles

225

Fig. 17 Structure of thiocillin I

contrast to opiates, ziconotide does not produce tolerance, a major hurdle in pain treatments. Ziconotide has only limited ability to cross the blood–brain barrier; thus, in order to achieve optimal analgesic efficacy with reduced risk for significant side effects, it must be administered intrathecally to patients. This spinal route of administration allows ziconotide to reach its maximum local concentration in a short time, leading to a rapid onset of analgesia [66]. The large macrocycle is delivered intrathecally via an implantable pump or by an external micro-infusion device. Ziconotide was approved by the FDA in 2004 and is currently indicated for the treatment of severe chronic pain [67] in patients with cancer or AIDS. The potent activity of the conotoxins has inspired multiple efforts to identify novel natural and synthetic analogs that can be orally bioavailable [5, 68]. 2.13

Thiocillin I

Thiocillin I (Fig. 17) is a natural thiopeptide macrocycle [69] with antibiotic activity isolated from Bacillus cereus. Thiopeptide antibiotics inhibit the growth of Gram-positive bacteria including MRSA and VRE at nanomolar concentrations [70]. Like many other thiopeptides, thiocillin targets the interface between ribosomal protein L11 and 23S rRNA. This compound led to renewed interest on biosynthetic research that has enabled the discovery of novel analogs through the genetic manipulation of producing organisms. This approach could lead to the discovery of new antiinfective compounds with improved therapeutic properties. Interestingly, thiocillin was selected as a macrocycle prototype to study the impact of ring entropy on binding and activity in a model natural product system, by combining systematic mutational analysis with computational modeling of ring entropy [71]. Macrocycles are conformationally constrained by cyclization, which has been hypothesized to reduce their molecular volume. In addition, this molecular pre-organization positions the compound into a low entropy state, facilitating permeation and target binding. However,

226

Maria-Jesus Blanco

Fig. 18 Structures of enigmazole A, 15-OMe-enigmazole A, and 13-OH-15-OMe-enigmazole A

macrocyclic rings of the same size can vary markedly in their conformational flexibility, due to the presence or absence of rigidifying elements such as double bonds or backbone rings. The outcome of the research was that a rigid macrocycle is a requisite for binding in the thiocillin system. Moreover, it was concluded that very small molecular modifications can generate significant increases in entropy calculated through computational modeling. The macrocyclic molecular arrangement greatly reduces backbone entropy; however, macrocyclization alone appears insufficient for pre-organization of the compound. In other macrocycle systems that are not heavily modified, entropy reduction can be achieved by intramolecular hydrogen bonding. Entropy is often an overlooked parameter in macrocycle design and the principles learned here can be applied to other natural and synthetic macrocycle systems. 2.14

Enigmazole A

Enigmazole A and the related congeners 15-OMe-enigmazole A and 13-OH-15-OMe-enigmazole A represent the first and to date only family of phosphate-containing marine macrolides [72] (Fig. 18). The natural macrocycles were isolated from a Papua New Guinea collection of the marine sponge Cinachyrella enigmatica. The structure of enigmazole A, including the absolute stereochemistry of the eight chiral centers, was determined by a combination of spectroscopic analyses and microscale chemical derivatization studies. Enigmazole A is an 18-membered phosphor-macrolide that contains an embedded exomethylenesubstituted tetrahydropyran ring and an acyclic portion that includes an oxazole moiety. Two closely related analogs, 15-Omethylenigmazole A and 13-hydroxy-15-O-methylenigmazole A, were also isolated and assigned. ()-Enigmazole A exhibited potent cytotoxic activity against the NCI 60 human tumor cell line, with a GI50 mean activity of 1.7 μM. The promising biological activity, in combination with the complex architecture, has attracted considerable interest in the synthetic community

Natural Product Inspired Cyclopeptides and Macrocycles

227

Fig. 19 Structure of natural macrocycle halichondrin B and synthetic analog eribulin

[73]. Those synthetic efforts could pave the way to additional medicinal chemistry efforts toward enigmazole A analogs in the near future. 2.15

Halichondrin B

Among many remarkable nature-inspired macrocycles that have led to clinical compounds and approved drugs, the story of halichondrin B and eribulin stands out as an excellent example in drug discovery. Halichondrin B [74] is a potent cytotoxic macrolide isolated from the sea sponge H. okadai in 1986 (Fig. 19). It is a complex marine natural product with 32 stereocenters. Its biological activity against microtubule assembly was evaluated across different cytotoxicity panels in human tumor cell lines. In 1992, Kishi [75] and his team completed the total synthesis of halichondrin B. Due to its potent anti-tumoral activity, additional SAR studies were pursued in collaboration with Eisai [76] toward a more potent, structurally simpler, and more stable ketone analog. Instability of the macrolactone acyl moiety was considered as the most likely reason for the lack of in vivo efficacy associated with the truncated macrolactone series. As a result of this effort, eribulin (E7389) was discovered, a simplified halichondrin B analog with 19 stereocenters, which preserves the promising biological properties of the natural product with improved pharmaceutical attributes

228

Maria-Jesus Blanco

including water solubility and chemical stability. Eribulin mesylate, marketed as Halaven (Fig. 19), was granted FDA approval in 2010 for metastatic breast cancer [77]. This is an extraordinary illustration of the power of synthetic organic chemistry yielding the final product in 62-step synthesis, offering this nature-inspired macrocyclic drug to the patients. 2.16

Cyclotides

2.17 Macrocyclic Peptides

One of the largest families of cyclopeptides found in nature is the cyclotides. They have been found in more than 20 species from the Violaceae, Rubiaceae, Cucurbitaceae, Fabaceae, and Solanaceae plant families. Cyclotides are 28–37 amino acids with a unique head-to-tail cyclized backbone, stabilized by three disulfide bonds forming a cystine knot. This unique topology makes them exceptionally stable to chemical, thermal, and biological degradation compared to other peptides of similar size [78]. One of the best known examples is Kalata-B1, an amphipathic peptide [79] containing 29 amino acid residues. Two of the largest cyclotides, MCoTI-I and MCoTI-II, are also trypsin inhibitors [12], and are found in the seeds of a bitter melon from Vietnam. Some primates express cyclopeptides called θ-defensins [80] as a part of their immune system, and to date they are the only known ribosomally synthesized cyclopeptides in mammals. Rhesus θ-defensin 1 (RTD-1) is a unique tri-disulfide, cyclic antimicrobial peptide formed by the ligation of two 9-residue sequences derived from heterodimeric splicing of similar 76-amino acid, α-defensin-related precursors, termed RTD1a and RTD1b. The characteristic structural motif of the θ-defensins is the cyclic cystine ladder, including a cyclopeptide backbone and three parallel disulfide bonds. Some examples are that RTD-1 was isolated from macaque leukocytes and some recent naturally occurring isoforms were isolated from baboon leukocytes. Naturally occurring macrocyclic peptides play a wide variety of roles, such as hormones (somatostatin [81], vasopressin) and neurotransmitters (oxytocin). Oxytocin and vasopressin [82] are cyclic peptide hormones released by primate posterior pituitary gland. They are nonapeptides (9 amino acids) with a single disulfide bridge, differing only by two substitutions in the amino acid sequence. Both are examples of cyclopeptides already available as drug therapeutics, being originally discovered as hormones found in mammals and later used as the starting point [83] in drug design in the pharmaceutical industry. For example, since oxytocin is highly susceptible to metabolic degradation, many medicinal chemistry approaches were directed toward increasing its half-life, to avoid the substantial loss of activity of the naturally occurring macrocycle. However common strategies to overcome peptide degradation, like the use of unnatural and D-amino acids, terminal capping, and chemical modification, are often unsuccessful for

Natural Product Inspired Cyclopeptides and Macrocycles

229

oxytocin since its biological activity is highly sensitive to structural change. In spite of the challenges, the medicinal chemistry community continues to expand the toolbox to increase the drug-like properties [84] of the naturally occurring macrocycles.

3

Conclusion Natural products play an essential role in modern drug discovery, and they continue to provide innovative compounds further advancing into clinical trials. The growing interest for peptidebased drugs has boosted R&D efforts toward nature-derived peptides for therapeutic applications. The rise of protein–protein interactions (PPIs) as promising targets for therapeutic intervention has greatly influenced the approaches on high-molecular-weight and complex drug leads, with medium-sized peptides and peptidomimetics emerging as key molecular mimics of the protein surfaces and peptide epitopes involved in protein–protein binding. Bioactive peptides exist in all organisms, and physiologically they can function as peptide hormones for cellular signaling, secretory peptides for interspecies communication, predatory peptide toxins, or antimicrobial host-defense peptides. These molecules have evolved over millions of years into a structurally sophisticated collection of compounds to modulate a diverse set of target proteins. More specifically, cyclopeptides have several structural features that modulate the hydrophobicity of native linear counterparts providing conformational stability and enhanced affinity to their target. Cyclization of linear peptides also increases the resistance to cleavage by proteolytic enzymes and membrane permeability, leading to superior overall pharmacokinetic profile. Several natural cyclopeptides have recently emerged as templates for drug design due to their resistance to chemical or enzymatic hydrolysis and high selectivity to receptors. With the exception of several naturally occurring cyclic peptides already approved as therapeutics, biologically active cyclic peptides were developed in recent years with genetic and synthetic approaches to enable various important applications, such as therapeutics, or diagnostics. The development of medicinal chemistry approaches that mimic the power of Nature’s strategies remains paramount for the advancement of novel macrocycles. A deeper knowledge of the action of macrocycles and cyclopeptides in Nature might lead to their utilization in different applications. However, in many instances, the three-dimensional structures of naturally occurring cyclopeptides and macrocycles are still unknown, and efforts to determine their structures can provide many important new lessons for synthetic chemists. Furthermore, without the ability to understand dynamic structural changes, the design and creation of new cyclopeptides will rely mostly on empirical rule-based approaches.

230

Maria-Jesus Blanco

More importantly, combining the knowledge of structure–activity relationships (SAR) of the naturally occurring macrocycles and cyclopeptides, with the factors that modulate the drug-like properties and with the current advances in rational drug design, synthetic methodology, and structure determination will result in the development of novel and more potent compounds. Overall the expectation is that nature-inspired macrocycles and cyclopeptides will lead the way to novel and breakthrough therapeutics for the treatment of human diseases, such as cancer, infection, neurodegeneration, and autoimmunity. References 1. Hill TA, Shepherd NE, Diness F, Fairlie DP (2014) Constraining cyclic peptides to mimic protein structure motifs. Angew Chem Int Ed Engl 53:13020–13041 2. Wessjohann LA, Ruijter E, Garcia-Rivera D, Brandt W (2005) What can a chemist learn from nature’s macrocycles? A brief, conceptual view. Mol Divers 9:171–186 3. Naylor MR, Bockus AT, Blanco MJ, Lokey RS (2017) Cyclic peptide natural products chart the frontier of oral bioavailability in the pursuit of undruggable targets. Curr Opin Chem Biol 38:141–147 4. Wetzler M, Hamilton P (2018) Peptides as therapeutics. In: Koutsopoulos S (ed) Peptide applications in biomedicine, biotechnology and bioengineering. Woodhead Publishing Elsevier Ltd. 5. Qvit N, Rubin SJ, Urban TJ, Mochly-Rosen D, Gross ER (2017) Peptidomimetic therapeutics: scientific approaches and opportunities. Drug Discov Today 22:454–462 6. Pomilio AB, Battista ME, Vitale AA (2006) Naturally-occurring cyclopeptides: structures and bioactivity. Curr Org Chem 10:2075–2121 7. Bockus AT, McEwen CM, Lokey RS (2013) Form and function in cyclic peptide natural products: a pharmacokinetic perspective. Curr Top Med Chem 13:821–836 8. Ghadiri MR, Granja JR, Milligan RA, McRee DE, Khazanovich N (1993) Self-assembling organic nanotubes based on a cyclic peptide architecture. Nature 366:324–327 9. Rosenthal-Aizman K, Svensson G, Unde´n A (2004) Self-assembling peptide nanotubes from enantiomeric pairs of cyclic peptides with alternating D and L amino acid residues. J Am Chem Soc 126:3372–3373 10. Qian Z, Dougherty PG, Pei D (2017) Targeting intracellular protein–protein interactions

with cell-permeable cyclic peptides. Curr Opin Chem Biol 38:80–86 11. Senthilkumar B, Rajasekaran R (2017) Analysis of the structural stability among cyclotide members through cystine knot fold that underpins its potential use as a drug scaffold. Inter J Peptide Res Therap 23(1):1 12. Molesini B, Treggiari D, Dalbeni A, Minuz P, Pandolfini T (2017) Plant cystine-knot peptides: pharmacological perspectives. Br J Clin Pharmacol 83:63–70 13. Dobson CM (2004) Chemical space and biology. Nature 432:824–828 14. Borel J.F. (1982) History of cyclosporin A and its significance in immunology. In: Cyclosporin A, pp 5–17 15. Borel JA, Feurer C, Gubler HU, St€ahelin H (1976) Biological effects of cyclosporin A: a new antilymphocytic agent. Agents Actions 6:468–475 16. Wenger RM (1984) Synthesis of cyclosporine. Total syntheses of ‘cyclosporin A’ and ‘cyclosporin H’, two fungal metabolites isolated from the species Tolypocladium inflatum GAMS. Helv Chim Acta 67:502–525 17. Sweeney ZK, Fu J, Wiedmann B (2014) From chemical tools to clinical medicines: nonimmunosuppressive cyclophilin inhibitors derived from the cyclosporin and sanglifehrin scaffolds. J Med Chem 57:7145–7159 18. Bai Y, King C, Francis C, Gooch J (2017) Cyclosporin A alters expression of renal MicroRNAs: new insights into calcineurin inhibitor nephrotoxicity. FASEB J 31:757–713 19. Naicker S, Yatscoff RW, Foster RT (2009) Deuterated cyclosporine analogs and methods of making the same. US Patent 7(521):421 20. Ahlbach CL, Lexa KW, Bockus AT, Chen V, Crews P, Jacobson MP, Lokey RS (2015) Beyond cyclosporine A: conformationdependent passive membrane permeabilities of

Natural Product Inspired Cyclopeptides and Macrocycles cyclic peptide natural products. Future Med Chem 7:2121–2130 21. Wang CK, Swedberg JE, Harvey PJ, Kaas Q, Craik DJ (2018) Conformational flexibility is a determinant of permeability for cyclosporin. J Phys Chem B 122:2261–2276 22. Rossi Sebastiano M, Doak BC, Backlund M, Poongavanam V, Over B, Ermondi G, Caron G, Matsson P, Kihlberg J (2018) Impact of dynamically exposed polarity on permeability and solubility of chameleonic drugs beyond the rule of 5. J Med Chem 61:4189–4202 23. R€ader AF, Reichart F, Weinmu¨ller M, Kessler H (2018) Improving oral bioavailability of cyclic peptides by N-methylation. Bioorg Med Chem 26:2766–2773 24. Consden R, Gordon AH, Martin AJ, Synge RL (1947) Gramicidin S: the sequence of the amino-acid residues. Biochem J 41:596 25. Kondejewski LH, Farmer SW, Wishart DS, Hancock RE, Hodges RS (1996) Gramicidin S is active against both gram-positive and gram-negative bacteria. Chem Biol Drug Des 47:460–466 26. Abraham T, Prenner EJ, Lewis RN, Mant CT, Keller S, Hodges RS, McElhaney RN (2014) Structure–activity relationships of the antimicrobial peptide gramicidin S and its analogs: aqueous solubility, self-association, conformation, antimicrobial activity and interaction with model lipid membranes. Biochim Biophys Acta 1838:1420–1429 27. Meanwell NA (2011) Synopsis of some recent tactical application of bioisosteres in drug design. J Med Chem 54:2529–2591 28. Xiao J, Weisblum B, Wipf P (2005) Electrostatic versus steric effects in peptidomimicry: synthesis and secondary structure analysis of gramicidin S analogues with (E)-alkene peptide isosteres. J Am Chem Soc 127:5742–5743 29. Davies J, Davies D (2010) Origins and evolution of antibiotic resistance. Microbiol Mol Biol Rev 74:417–433 30. Carney DW, Schmitz KR, Truong JV, Sauer RT, Sello JK (2014) Restriction of the conformational dynamics of the cyclic acyldepsipeptide antibiotics improves their antibacterial activity. J Am Chem Soc 136:1922–1929 31. Goodreid JD, Wong K, Leung E, McCaw SE, Gray-Owen SD, Lough A, Houry WA, Batey RA (2014) Total synthesis and antibacterial testing of the A54556 cyclic acyldepsipeptides isolated from Streptomyces hawaiiensis. J Nat Prod 77:2170–2181 32. Socha AM, Tan NY, LaPlante KL, Sello JK (2010) Diversity-oriented synthesis of cyclic acyldepsipeptides leads to the discovery of a

231

potent antibacterial agent. Bioorganic Med Chem 18:7193–7202 33. Goodreid JD, Janetzko J, Santa Maria Jr JP, Wong KS, Leung E, Eger BT, Bryson S, Pai EF, Gray-Owen SD, Walker S, Houry WA. (2016) Development and characterization of potent cyclic acyldepsipeptide analogues with increased antimicrobial activity. J Med Chem 59:624–646 34. Naylor M, Ly A, Schwochert J, Desai P, Gonzalez Valcarcel IC, Barrett J, Sawada G, Blanco MJ, Lokey S (2016) Amide-to-ester substitutions modify the permeability and ADME properties of natural and synthetic cyclic peptides. From abstracts of papers, 252nd ACS National Meeting & Exposition, Philadelphia, PA, United States, August 21–25, MEDI-344 35. Lukat P, Katsuyama Y, Wenzel S, Binz T, Ko¨nig C, Blankenfeldt W, Bro¨nstrup M, Mu¨ller R (2017) Biosynthesis of methyl-proline containing griselimycins, natural products with anti-tuberculosis activity. Chem Sci 8:7521–7527 36. Kling A, Lukat P, Almeida DV, Bauer A, Fontaine E, Sordello S, Zaburannyi N, Herrmann J, Wenzel SC, Ko¨nig C, Ammerman NC (2015) Targeting DnaN for tuberculosis therapy using novel griselimycins. Science 348:1106–1112 37. Dong M, Pfeiffer B, Altmann KH (2017) Recent developments in natural productbased drug discovery for tuberculosis. Drug Discov Today 22:585–591 38. Salvador-Reyes LA, Luesch H (2015) Biological targets and mechanisms of action of natural products from marine cyanobacteria. Nat Prod Rep 32:478–503 39. Taori K, Paul VJ, Luesch H (2008) Structure and activity of largazole, a potent antiproliferative agent from the Floridian marine cyanobacterium Symploca sp. J Am Chem Soc 130:1806–1807 40. Poli G, Di Fabio R, Ferrante L, Summa V, Botta M (2017) Largazole analogues as histone deacetylase inhibitors and anticancer agents: an overview of structure–activity relationships. ChemMedChem 12:1917–1926 41. Chen QY, Chaturvedi PR, Luesch H (2018) Process development and scale-up total synthesis of Largazole, a potent Class I histone deacetylase inhibitor. Org Process Res Dev 22:190–199 42. Cole KE, Dowling DP, Boone MA, Phillips AJ, Christianson DW (2011) Structural basis of the antiproliferative activity of largazole, a depsipeptide inhibitor of the histone deacetylases. J Am Chem Soc 133:12474–12477

232

Maria-Jesus Blanco

43. Almaliti J, Al-Hamashi AA, Negmeldin AT, Hanigan CL, Perera L, Pflum MK, Casero Jr RA, Tillekeratne LV. (2016) Largazole analogues embodying radical changes in the depsipeptide ring: development of a more selective and highly potent analogue. J Med Chem 59:10642–10660 44. Steenbergen JN, Alder J, Thorne GM, Tally FP (2005) Daptomycin: a lipopeptide antibiotic for the treatment of serious Gram-positive infections. J Antimicrob Chemother 55:283–288 45. Bionda N, Pitteloud JP, Cudic P (2013) Cyclic lipodepsipeptides: a new class of antibacterial agents in the battle against resistant bacteria. Future Med Chem 5:1311–1330 46. Daley P, Louie T, Lutz JE, Khanna S, Stoutenburgh U, Jin M, Adedoyin A, Chesnel L, Guris D, Larson KB, Murata Y (2017) Surotomycin versus vancomycin in adults with Clostridium difficile infection: primary clinical outcomes from the second pivotal, randomized, double-blind, Phase 3 trial. J Antimicrob Chemother 72:3462–3470 47. Yin N, Li J, He Y, Herradura P, Pearson A, Mesleh MF, Mascio CT, Howland K, Steenbergen J, Thorne GM, Citron D (2015) Structure–activity relationship studies of a series of semisynthetic lipopeptides leading to the discovery of Surotomycin, a novel cyclic lipopeptide being developed for the treatment of Clostridium difficile-associated diarrhea. J Med Chem 58:5137–5142 48. Lee CH, Patino H, Stevens C, Rege S, Chesnel L, Louie T, Mullane KM (2016) Surotomycin versus vancomycin for Clostridium difficile infection: Phase 2, randomized, controlled, double-blind, non-inferiority, multicentre trial. J Antimicrob Chemother 71:2964–2971 49. Borders DB, Leese RA, Jarolmen H, Francis ND, Fantini AA, Falla T, Fiddes JC, Aumelas A (2007) Laspartomycin, an acidic lipopeptide antibiotic with a unique peptide core. J Nat Prod 70:443–446 50. Kleijn LH, Oppedijk SF, ‘t Hart P, Van Harten RM, Martin-Visscher LA, Kemmink J, Breukink E, Martin NI. (2016) Total synthesis of laspartomycin C and characterization of its antibacterial mechanism of action. J Med Chem 59:3569–3574 51. Mi Y, Zhang J, He S, Yan X (2017) New peptides isolated from marine cyanobacteria, an overview over the past decade. Mar Drugs 15:132 52. Reese MT, Gulavita NK, Nakao Y, Hamann MT, Yoshida WY, Coval SJ, Scheuer PJ (1996) Kulolide: a cytotoxic depsipeptide

from a cephalaspidean mollusk, Philinopsis speciosa. J Am Chem Soc 118:11081–11084 53. Boudreau PD, Byrum T, Liu WT, Dorrestein PC, Gerwick WH (2012) Viequeamide A, a cytotoxic member of the kulolide superfamily of cyclic depsipeptides from a marine button cyanobacterium. J Nat Prod 75:1560–1570 54. Wang D, Song S, Tian Y, Xu Y, Miao Z, Zhang A (2013) Total synthesis of the marine cyclic depsipeptide viequeamide A. J Nat Prod 76:974–978 55. Almaliti J, Malloy KL, Glukhov E, Spadafora C, Gutie´rrez M, Gerwick WH (2017) Dudawalamides A–D, Antiparasitic Cyclic Depsipeptides from the Marine Cyanobacterium Moorea producens. J Nat Prod 80:1827–1836 ´ lvarez M (2014) 56. Just-Baringo X, Albericio F, A Chiral thiazoline and thiazole building blocks for the synthesis of peptide-derived natural products. Curr Top Med Chem 14:1244–1256 57. Nielsen DS, Hoang HN, Lohman RJ, Diness F, Fairlie DP (2012) Total synthesis, structure, and oral absorption of a thiazole cyclic peptide, sanguinamide A. Org Lett 14:5720–5723 58. Bockus AT, Schwochert JA, Pye CR, Townsend CE, Sok V, Bednarek MA, Lokey RS (2015) Going out on a limb: delineating the effects of β-branching, N-methylation, and side chain size on the passive permeability, solubility, and flexibility of Sanguinamide A analogues. J Med Chem 58:7409–7418 59. Desai PV, Raub TJ, Blanco MJ (2012) How hydrogen bonds impact P-glycoprotein transport and permeability. Bioorg Med Chem Lett 22:6540–6548 60. Zipperer A, Konnerth MC, Laux C, Berscheid A, Janek D, Weidenmaier C, Burian M, Schilling NA, Slavetinsky C, Marschal M, Willmann M (2016) Human commensals producing a novel antibiotic impair pathogen colonization. Nature 535:511–516 61. Krismer B, Peschel A, Grond S, Zipperer A, Konnerth MC, Janek D (2016) Patent PCT Int Appl WO 2016151005 62. Mousa WK, Athar B, Merwin NJ, Magarvey NA (2017) Antibiotics and specialized metabolites from the human microbiota. Nat Prod Rep 4:1302–1331 63. Sekizawa R, Momose I, Kinoshita N, Naganawa H, Hamada M, Muraoka Y, Iinuma H, Takeuchi T (2001) Isolation and structural determination of phepropeptins A, B, C, and D, new proteasome inhibitors, produced by Streptomyces sp. J Antibiot 54:874–881

Natural Product Inspired Cyclopeptides and Macrocycles 64. Schwochert J, Lao Y, Pye CR, Naylor MR, Desai PV, Gonzalez Valcarcel IC, Barrett JA, Sawada G, Blanco MJ, Lokey RS (2016) Stereochemistry balances cell permeability and solubility in the naturally derived phepropeptin cyclic peptides. ACS Med Chem Lett 7:757–761 65. Safavi-Hemami H, Brogan SE, Olivera BM (2018) Pain therapeutics from cone snail venoms: from Ziconotide to novel non-opioid pathways. J Proteomics 190:12–20 66. Newcomb R, Abbruscato TJ, Singh T, Nadasdi L, Davis TP, Miljanich G (2000) Bioavailability of Ziconotide in brain: influx from blood, stability, and diffusion. Peptides 21:491–501 67. Schmidtko A, Lo¨tsch J, Freynhagen R, Geisslinger G (2010) Ziconotide for treatment of severe chronic pain. Lancet 375:1569–1577 68. Thell K, Hellinger R, Sahin E, Michenthaler P, Gold-Binder M, Haider T, Kuttke M, Liutkevicˇiu¯te˙ Z, Go¨ransson U, Gru¨ndemann C, Schabbauer G (2016) Oral activity of a nature-derived cyclic peptide for the treatment of multiple sclerosis. Proc Natl Acad Sci U S A 113:3960–3965 69. Aulakh VS, Ciufolini MA (2011) Total synthesis and complete structural assignment of thiocillin I. J Am Chem Soc 133:5900–5904 ´ lvarez M (2014) 70. Just-Baringo X, Albericio F, A Thiopeptide antibiotics: retrospective and recent advances. Mar Drugs 12:317–351 71. Tran HL, Lexa KW, Julien O, Young TS, Walsh CT, Jacobson MP, Wells JA (2017) Structure–activity relationship and molecular mechanics reveal the importance of ring entropy in the biosynthesis and activity of a natural product. J Am Chem Soc 139:2541–2544 72. Oku N, Takada K, Fuller RW, Wilson JA, Peach ML, Pannell LK, McMahon JB, Gustafson KR (2010) Isolation, structural elucidation, and absolute stereochemistry of enigmazole A, a cytotoxic phosphomacrolide from the Papua New Guinea marine sponge Cinachyrella enigmatica. J Am Chem Soc 132:10278–10285 73. Ai Y, Kozytska MV, Zou Y, Khartulyari AS, Maio WA, Smith AB III (2018) Total synthesis of the marine phosphomacrolide, ()Enigmazole A, exploiting multicomponent Type I Anion Relay Chemistry (ARC) in conjunction with a late-stage Petasis–Ferrier union/rearrangement. J Org Chem 83:6110–6126

233

74. Towle MJ, Salvato KA, Budrow J, Wels BF, Kuznetsov G, Aalfs KK, Welsh S, Zheng W, Seletsky BM, Palme MH, Habgood GJ (2001) In vitro and in vivo anticancer activities of synthetic macrocyclic ketone analogues of halichondrin B. Cancer Res 61:1013–1021 75. Aicher TD, Buszek KR, Fang FG, Forsyth CJ, Jung SH, Kishi Y, Matelich MC, Scola PM, Spero DM, Yoon SK (1992) Total synthesis of halichondrin B and norhalichondrin B. J Am Chem Soc 114:3162–3164 76. Zheng W, Seletsky BM, Palme MH, Lydon PJ, Singer LA, Chase CE, Lemelin CA, Shen Y, Davis H, Tremblay L, Towle MJ (2004) Macrocyclic ketone analogues of halichondrin B. Bioorg Med Chem Lett 14:5551–5554 77. Cortes J, Vahdat L, Blum JL, Twelves C, Campone M, Roche´ H, Bachelot T, Awada A, Paridaens R, Goncalves A, Shuster DE (2010) Phase II study of the halichondrin B analog eribulin mesylate in patients with locally advanced or metastatic breast cancer previously treated with an anthracycline, a taxane, and capecitabine. J Clin Oncol 28:3922–3928 78. Camarero JA (2017) Cyclotides, a versatile ultrastable micro-protein scaffold for biotechnological applications. Bioorg Med Chem Lett 27:5089–5099 79. Eliasen R, Daly NL, Wulff BS, Andresen TL, Conde-Frieboes KW, Craik DJ (2012) Design, synthesis, structural and functional characterization of novel melanocortin agonists based on the cyclotide kalata B1. J Biol Chem 287:40493–40501 80. Tran D, Tran PA, Tang YQ, Yuan J, Cole T, Selsted ME (2002) Homodimeric thetadefensins from rhesus macaque leukocytes: isolation, synthesis, antimicrobial activities, and bacterial binding properties of the cyclic peptides. J Biol Chem 277:3079–3084 81. Reichlin S (1983) Somatostatin. N Engl J Med 309:1495–1501 82. Donaldson ZR, Young LJ (2008) Oxytocin, vasopressin, and the neurogenetics of sociality. Science 322:900–904 83. Beard R, Stucki A, Schmitt M, Py G, Grundschober C, Gee AD, Tate EW (2018) Building bridges for highly selective, potent and stable oxytocin and vasopressin analogs. Bioorg Med Chem 26:3039–3045 84. Poongavanam V, Doak BC, Kihlberg J (2018) Opportunities and guidelines for discovery of orally absorbed drugs in beyond rule of 5 space. Curr Opin Chem Biol 44:23–29

Chapter 11 Design of Oxytocin Analogs Kazimierz Wis´niewski Abstract The neurohypophyseal hormone oxytocin (OT) and related modulators of the oxytocin receptor (OTR) have been the subject of intensive research for nearly seven decades. Despite having rather poor drug-like properties, OT is used as a treatment for labor induction, postpartum hemorrhage, and lactation support. The potential use of OT in the treatment of central nervous system (CNS)-related diseases has recently renewed interest in the pharmacology of OT. Oxytocin is one of the most extensively studied cyclic peptides and since the elucidation of its structure in 1953 thousands of peptidic OT analogs with antagonistic and agonistic properties have been synthesized and biologically evaluated. Among them are atosiban, a mixed oxytocin receptor (OTR)/vasopressin 1a receptor (V1aR) antagonist used as a tocolytic agent approved (in certain countries), and carbetocin, a longer acting OTR agonist on the market for the treatment of postpartum hemorrhage. Many other OT analogs with improved pharmacological properties (e.g., barusiban, Antag III) have been identified. These peptides have been tested in clinical trials and/or used as pharmacological tools. In this chapter, the modifications of the OT molecule that led to the discovery of these compounds are reviewed. Key words Oxytocin (OT), Oxytocin receptor (OTR), OTR antagonists, OTR agonists, Selectivity, Duration of action, Clinical use

1

Introduction The neurohypophyseal hormone oxytocin (OT) displays a variety of biological actions mediated by its molecular target, the seventransmembrane-domain G protein-coupled oxytocin receptor (OTR) [1, 2]. OTR is expressed in numerous tissues both peripherally and centrally [1–4]. OT is centrally produced in the hypothalamus [5] and released to the peripheral circulation through the posterior pituitary gland. OT is also produced peripherally in the reproductive system, GI tract [6], heart [7], and bone [8]. The interactions of OT with peripherally located OTR mediate milk ejection and labor induction [9]. Other peripheral functions of oxytocin include inhibition of food intake [10, 11], particularly the intake of sweet-tasting carbohydrates [12], and bone [13, 14] and fat mass control [15]. OT increases energy expenditure and lipolysis

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_11, © Springer Science+Business Media, LLC, part of Springer Nature 2019

235

236

Kazimierz Wis´niewski

[16, 17], improves glucose homeostasis [10], and shows cardioprotective effects [18, 19]. Centrally mediated actions include induction of maternal behavior [20], pair and mother-child bonding [21], social behavior and cognition [21], and male and female sexual responses (prosocial and reproductive effects) [21, 22]. Oxytocin plays a neuromodulatory role in the brain [23] and has profound anxiolytic and stress-reducing effects [24]. Endogenous OT appears to have analgesic effect at the spinal and peripheral level [25] and plays a modulatory role in pain perception [26]. Peripheral actions of oxytocin are exploited clinically as the peptide is widely used for labor induction and augmentation [9, 27–29], prevention of postpartum hemorrhage [30], and, in a few countries, lactation support [31, 32]. Substantial scientific and medical interest exists in the therapeutic use of OT for the treatment of CNS disorders. Intranasal OT spray gained attention regarding CNS-mediated effects after it was reported that OT treatment enhanced trust [33]. Potential use in indications such as anxiety [24, 34], autism [35], alcohol and drug addiction [36–38], pain management [39, 40], posttraumatic stress disorder (PTSD) [41], schizophrenia [42, 43], Prader-Willi syndrome [44], postpartum depression (PPD) [45, 46], and dementia [47] has been investigated. Despite the initial positive reported results, additional clinical studies with intranasal OT trials have been conducted and were unable to replicate the effects on patients with Prader-Willi syndrome [44] and schizophrenia [48], or in youths with autism spectrum disorders [49]. As of today, OT has not been approved for any CNS-related indications. The structure of oxytocin (Fig. 1) was elucidated by du Vigneaud and his research team in the early 1950s [50] and the

Fig. 1 Structure of oxytocin (OT). The sequence positions are numbered at α-carbons

Design of Oxytocin Analogs

237

hormone was the first biologically active peptide prepared synthetically [51]. OT is a nonapeptide of the following sequence H-c(Cys-TyrIle-Gln-Asn-Cys)-Pro-Leu-Gly-NH2 that contains as a key structural feature a 20-membered, six-amino acid ring closed by a disulfide bridge between the cysteine residues in positions one and six and a C-terminal tripeptide amide. The cyclic part of the OT molecule is believed to interact with the second extracellular loop and upper portions of transmembrane domains 3, 4, and 6 of the OTR while the C-terminal part of OT binds to the first extracellular loop and the N-terminal domain of the receptor [1, 2, 52, 53]. The presence of divalent cations and cholesterol positively affects the affinity of OT and its analogs for the OTR [52]. The 20-membered ring of OT is essential for its functional activity and the OTR binding. It is dramatically exemplified by the total lack of ex vivo activity of linearized analogs [Ala1,6]OT and [Ser1,6]OT on rat mammary gland strips and/or isolated uterus [54]. No acyclic OTR binders have been found, although linear ligands for the related V1a and V2 receptors have been identified [55, 56]. Oxytocin is not selective for the OTR, also interacting with the related vasopressin V1a, V1b, and V2 receptors [57–59] (Table 2). As a moderately potent agonist of the hV2 receptor in the kidney, it can produce antidiuresis, inappropriate water conservation, and hyponatremia in patients receiving OT [60]. Clinically, guidelines for OT use have been developed to minimize the risk of hyponatremia during labor induction [28, 61] and several cases of severe hyponatremia by use of the nasal spray have been reported [62–64]. Body temperature and cardiac changes (bradycardia) after peripheral administration of OT have been postulated to be V1a receptor mediated [65]. The pharmacokinetic (PK) properties of oxytocin in humans and other species have been broadly studied. The hormone has rather short half-life (3–20 min) and its clearance (CL) in nonpregnant women is about 20 mL/min/kg [66–68]. A similar CL value (21
2.2 mL/min/kg) for OT was reported in rats [69]. The PK parameters change dramatically in pregnancy [66, 68] due to proteolysis of the hormone by oxytocinase [70, 71]. Considerable intranasal bioavailability of OT has been demonstrated in humans [67, 72] but no significant increase in OT plasma concentration was observed after buccal administration [67]. Low oral absorption of oxytocin is typical for similarly sized peptides [73, 74] and is exacerbated by rapid proteolysis in the gut by chymotrypsin and other enzymes [75, 76]. The presence of the disulfide bridge adds to low metabolic stability of OT as it can participate in thiol-disulfide exchange with the reduced form of glutathione [77]. Multiple enzymes in the protein thiol oxidoreductase family can catalyze the disulfide exchange [78]. No meaningful crossing of the blood-brain barrier (BBB) from the

238

Kazimierz Wis´niewski

circulation has been observed in rats where only 0.002% of the subcutaneously injected dose reached the CNS [79]. The direct penetration of the CNS via the intranasal route remains controversial [80]. Striepens et al. showed some increase in CNS OT concentration but only 75 min post-intranasal dosing in humans [81]. In macaque studies only up to 0.003% of the administered dose was accounted for in cerebrospinal fluid (CSF) [80, 82, 83]. Oxytocin is marketed under two brand names: Pitocin (10 IU/ mL, for i.v. administration) and Syntocinon (40 IU/mL, a nasal spray), both aqueous formulations. The stability of OT in common infusion solutions has been extensively studied [84–87], and addition of divalent metal ions stabilizes OT in solution [88, 89]. Mechanism of OT degradation has been investigated and monomeric (OT tri- and tetrasulfides) and dimeric degradation products have been identified [90]. In our laboratory we clarified the degradation mechanism, elucidated the structures of the dimeric degradants, proposed chemical modifications to stabilize the OT molecule, and demonstrated a positive correlation of the degradation rate with pH [91]. As outlined above, OT possesses poor drug-like properties and therefore is not an ideal agent for clinical use. The extensive search for agonist-antagonist switch residues and for new analogs with improved pharmacological properties (better potency and selectivity, improved chemical and metabolic stability, enhanced bioavailability, and lower clearance) started shortly after the OT structure was elucidated in 1953 [50]. Thousands of research papers on oxytocin have been published to date, so by necessity some hard choices had to be made to fit this vast knowledge into the format of the current review. Prior to cloning of the OT receptors, the potency, receptor selectivity, and duration of action data were generated mainly in rat in vivo and ex vivo assays and translatability to human effects was not predictable. Following cloning of the receptors, the mechanism for the substantial species differences (e.g., between rats and humans) was recognized [92, 93]. Knowing the sequence of the hOT receptor [94, 95] and the related vasopressin receptors hV1aR [96], hV1bR [97], and hV2R [98, 99] allowed for better intraspecies predictions of potency and selectivity and generation of more clinically relevant OT analogs. The remainder of the review is divided into two sections on the discovery and design of peptidic OTR antagonists and agonists, respectively. Small-molecule ligands for the OTR, either with antagonistic or agonistic properties, have been discovered [100–103] but are beyond the scope of this chapter and will not be discussed here. The abbreviated sequences of the compounds mentioned in the text are given as OT analogs. The structural changes versus the OT molecule are highlighted in blue.

Design of Oxytocin Analogs

2

239

Design of OTR Antagonists Since the discovery of oxytocin, the focus of extensive research has been on identifying clinically relevant OTR antagonists. OT and its receptor play a vital role in the initiation and maintenance of labor [104–106]. This is supported by evidence showing that OT-induced contractility increases as pregnancy approaches term [107] and that an increase in the concentration of uterine OT occurs during term and preterm labor in humans [108]. Therefore, antagonism of the OTR has been considered a viable strategy to treat pathophysiological conditions such as preterm labor [109], the major cause of perinatal death and morbidity [110]. The search for novel OTR antagonists with improved pharmacological properties has resulted in a massive body of literature. Due to space constraints, only certain aspects of the available research data are discussed below and the reader is urged to consult numerous reviews published on the subject [92, 111–118].

2.1 Analogs Modified in Positions 1, 2, and 8

A minor modification of the oxytocin molecule led to the first synthetic weak OTR antagonist - O-methyloxytocin [119]. The replacement of the Cys1 residue with deaminopenicillamine [120] resulted in another weak antagonist of oxytocin responses. Increasing the size and hydrophobicity [55] of the substituents at the β,β-position of residue 1 by replacing dimethyl-alkyl groups with diethyl- [121] or pentamethylene groups [122] further improved the potency of the resulting compounds. These changes combined with the Tyr(Me)2 and/or Thr4 resulted in promising lead OT antagonists [dPen1,Tyr(Me)2]OT, [dPen1,Tyr(Me)2,Thr4]OT, and [Pmp1,Thr4]OT [123]. These early studies demonstrated the importance of the N-terminal part and position 4 for switching agonist to antagonist activity and improving potency and selectivity for the OTR. Subsequent work in the Manning group focused on vasotocin analogs combining the N-terminal modifications with the basic residue in position 8. Sawyer et al. [124] described analogs of [dPen1]OT with Orn in position 8 showing enhanced potency in rat uterotonic and milk ejection assays. This effort produced two very potent analogs, namely [Pmp1, Tyr(Me)2,Orn8]OT and [Mpa(3,3-Et2)1,Tyr(Me)2,Orn8]OT [125], Fig. 2. Both compounds exhibited very good selectivity versus the rV2 receptor but were equipotent as antivasopressors in rat. The inversion of configuration of the position 2 residue was first proposed in the 1970s by Kaurov et al. [126, 127]. The modification was successfully employed by other research teams [128–130] and in some cases was sufficient for agonist-to-antagonist conversion [128].

240

Kazimierz Wis´niewski

Fig. 2 Structures of [Pmp1,Tyr(Me)2,Orn8]OT (left) and [Mpa(3,3-Et2)1,Tyr(Me)2,Orn8]OT (right)

Fig. 3 Structures of [Pmp1,D-Trp2,Arg8]OT (left) and Antag III (TT-235), [(S)Pmp1,D-Trp2,Pen6,Arg8]OT (right)

Flouret and his research team utilized the modifications discussed above to design some of the most potent peptidic OT antagonists to date. Starting with analog [Pmp1,D-Trp2,Arg8]OT (Fig. 3) previously discovered in their laboratory, a potent antagonist of the uterotonic effect of oxytocin in the rat and the baboon [131], the team explored replacements of the 4-methylene group

Design of Oxytocin Analogs

241

of the cyclohexyl ring of the Pmp1 residue with O, S, NH, or carbonyl [132]. The more hydrophilic substitutions (NH and carbonyl) resulted in inactive analogs but acylations of the NH group led to partial (For, Ac) or full (head-to-tail cyclization) recovery of the antagonistic activity in rats. The modification of the bicyclic analog with Pen led to compound cyclo(1–9)[(HN)Pmp1,D-Trp2,Pen6, Arg8]OT which was more potent and selective than the parent antagonist. The replacement of Pmp with sulfur-containing β,β-(3-thiapentamethylene)-β-mercaptopropionic acid ((S)Pmp) combined with the Pen6 modification resulted in the most potent and selective oxytocin antagonist [(S)Pmp1,D-Trp2,Pen6,Arg8]OT (Fig. 3) designed in Flouret’s lab. The compound, also known as Antag III or TT-235, showed prolonged duration of action in the in vivo model of suppressing uterus spontaneous contractions in rat [133] and in primates [134] and was stable in plasma from pregnant women [135]. The analog was previously in clinical trials for labor inhibition but no evidence of ongoing development progress could be found in clinical trial databases (e.g., ClinicalTrials.gov). Subsequent efforts to improve the pharmacological profile of Antag III by substituting the Arg8 residue [136] or introducing the ureido function in position 4 or 5 [137] resulted in analogs with better selectivity versus the rV1aR (e.g., [(S)Pmp1,D-Trp2,Pen6, Dap8]OT). 2.2 Monocarba Modifications

The carba modifications (i.e., the replacement of one disulfide bridge sulfur atoms with the methylene group) introduced by Lebl et al. in combination with the residue of the D-configuration in pos. 2 led to incremental improvements in OT antagonist activity [129]. The carba-6 compound [Mpa1,D-Phe(4-Et)2,Abu6]OT was particularly potent in the described set. As expected, its sulfoxide analog showed less activity in the rat uterotonic assay but was surprisingly more potent than the parent analog as an inhibitor of galactogogic response. The carba-6 alteration worked well in combination with the deaminopenicillamine (dPen1) and Tyr(Me)2 modifications as demonstrated by higher potency of peptide [dPen1,Tyr(Me)2,Abu6]OT in the rat uterotonic assays as compared to its disulfide counterpart [dPen1,Tyr(Me)2]OT [138]. In another study, the same group demonstrated that the carba-6 modification yields slightly more potent OT antagonists than the carba-1 replacement [139].

2.3 Position 9 Replacements

The substitutions of the Gly9 residue in vasotocin analogs were well tolerated by the OTR and led to some gains in selectivity versus V1aR [140, 141]. The replacement of the Gly9 residue with a variety of natural or unnatural amino acids of L or D configuration in deaminooxytocin led to potent and selective OTR antagonists in rat assays when combined with D-Aaa2 (e.g., [Mpa1,D-Tyr(Et)2,

242

Kazimierz Wis´niewski 9 D-Thi ]OT

and [Mpa1,D-Tyr(Et)2,D-Tic9]OT) [142]. The species differences were evident in this study as binding affinities to the hOTR did not entirely match the affinities at the rat OT receptor. The position 9 was utilized to design labeled OTR antagonists [143–145]. For example, widely used pharmacological tool compound [Pmp1,Tyr(Me)2,Thr4,Orn8,Tyr9]OT was radiolabeled with 125I in the aromatic ring of Tyr9 [144]. 2.4 Discovery of Atosiban

The OT antagonist program at Ferring started with the goal to rapidly identify analogs with potential clinical utility. Leaving aside the heavily exploited position 1 β,β-dialkylation discussed above, early efforts focused on the position 2 O-alkylated compounds with or without a positively charged residue in position 8 [146]. [Mpa1, Tyr(Et)2]OT identified in the study was shown to be an effective antiuterotonic agent in rats and humans both in vitro and in vivo [146–148]. Inversion of configuration at position 2 in this analog combined with modifications in positions 4 and 8 resulted in [Mpa1,D-Tyr(Et)2,Thr4,Orn8]OT (atosiban, Fig. 4) [149]. The compound was shown to be considerably more potent in rat antiuterotonic (pA2 increase from 7.2 to 8.29) and human myometrium assays in vitro than the parent analog [Mpa1,Tyr(Et)2] OT [150]. Atosiban was selected as a clinical candidate and was approved in Europe for the treatment of preterm labor [109, 151–154] under the trade name of Tractocile™. Atosiban is not very selective versus rV1aR [155] (Table 1) and it has been argued that the

Fig. 4 Structure of atosiban, [Mpa1,D-Tyr(Et)2,Thr4,Orn8]OT

7.91

[Pmp1,D-2Thi2,Thr4, Orn8,Tyr9]OT

Atosiban

7.72

10.6 (human)d

8.29 7.71

9.5 (human)d

8.86

4.55

5.95 52 76.4

7.56

[Pmp1,D-Trp2,Arg8]OT 7.77 7.54

Antag III (TT-235)

8.02

8.91

[Mpa(3,3-Et2)1,Tyr (Me)2,Orn8]OT

~5

0.02 (IU/μmol)e 6.14

7.47

7.96

8.52

[Pmp1,Tyr(Me)2,Orn8] OT

7.32

5.86

In vivo rat In vitro hOTR antivasopressor (Ki, nM) pA2

7.64

1.75

In vivo rat ED (nmol/kg)b

V1aR

[dPen ,Tyr(Me) ,Thr ] OT

4

[Pmp1,Thr4]OT

2

Compound

1

Rat antiuterotonic,a no Mg2+, pA2

OTR

Receptor

Table 1 Pharmacological properties of selected OTR antagonists discussed in the chapter

5.1

In vitro hV1aR (Ki, nM)

0.04 (IU/μmol)

(continued)

[159]

[150] [155] [162] [140] [170]

[132] [136] [170]

27,000 (EC50)d

0.5 (EC50)d

[213] 170 (EC50)b [59] 5680 (Ki) [162]

>10,000 (Ki)

[245]

[195]

[196]

[192] 3.5 (EC50)b [59] 981 (Ki) [228]

[196] 7.3 (EC50)b [59] 310 (Ki) [57, 162]

109 (EC50)e [228] 910 (Ki) [228]

0.9

0.002

5.3

2.1

19

5

Ref.

132 (IC50)d 56.8 (Ki)

41c (EC50)b

305 (Ki)

21c (EC50)b 3.8 (Ki)

10 (EC50)b 210 (Ki)

In vitro hV2R (nM)

[Suc1,Dap6,Sar7,Chg8]OT

0.70 (EC50)b 7.1 (Ki)

13 (Ki)

3 copies in the target selection and 0 copies in the NTC is kept for analysis. 3. The warhead counts from the cyclized library were then compared to those from the linear library. Any warhead with copies in the cyclized library that was not present in the linear library was reported as a cyclic specific binder.

3.4 Identification of Potential Peptide Hits

The selection of potential cyclic peptide hits is done firstly by generating minimum energy conformation of cyclic peptides followed by calculating the solvation energy of the minimum energy conformations and evaluation of the cyclic peptides for passive permeability using a ChromlogD/CMR model [7]. The following steps outline the procedure for generation of 3D conformations of cyclic peptides: 1. All the compounds in the dataset are protonated (hydrogen atoms added to the compound) at pH 7.4 using Protonate3D module in MOE 2015. The atomic charges of each compound are calculated using the MMFF94x force field in MOE. This is followed by minimizing each compound using MMFF94x force field. AM1 charges are derived for the minimized conformations of the molecules. 2. Multiple conformations of each peptide are generated using LowModeMD (CCG MOE software) at 300 K using Berendsen thermostat. The conformational search is done in vacuum and all the conformations within 20 kcal/mol of the lowest energy conformer are retained. The energy minimization ˚ . LowModeMD search threshold is set to 0.001 kcal/mol/A is configured to terminate, if after 500 attempts the method fails to generate any novel conformation up to a maximum of 10,000 iterations. A root mean square deviation (rmsd) threshold of 0.5 A˚ is set so that the conformations are not duplicated. Using this conformational analysis approximately 10,000 conformations are generated for each compound. The conformations of a peptide are sorted according to the intramolecular energy of that peptide.

DNA-Encoded Macrocyclic Peptide Library

279

3. The conformations of a cyclic peptide are clustered based on the backbone dihedral angles (φ, ψ, and ω) of the peptide. Conformations of a peptide with dihedral angles within a tol
erance limit of 30 are clustered together. The seed of each cluster is the conformation with the lowest conformational energy. 4. Solvation energy is calculated for the seed of each cluster using Poisson–Boltzmann calculations which are available from SVL exchange in MOE [8]. These calculations are used to calculate the free energy change of transferring a solute molecule from a low dielectric medium (to mimic a nonpolar environment) to a high dielectric medium (like water). The conformations of each peptide with lowest (strain) energy and solvation energy are used for comparison with similar 3D conformations of other compounds. 5. ChromlogD is calculated for each compound using a SVMchromlogD model [7]. To have a balance between permeability and solubility the following criteria should be followed to rank the cyclic peptides belonging to a series: 1. SVMchromlogD ranges between 4 and 6. 2. Solvation energy less than 40 kcal/mol was chosen to maintain a balance between water solubility and lipophilicity (for this series of peptides). 3. Compounds fall in the permeable region of the SVMchromlogD/CMR plot. 4. Chemical diversity is also taken into account to evaluate the compounds. 3.5 Synthesis of OffDNA Linear and Macrocyclic Peptides

Linear peptides are synthesized on the CEM Liberty 1 microwave peptide synthesizer using the general method below. 1. Linear peptide synthesis: Using an analytical balance, weigh Rink Amide AM resin (200400 mesh, Merck Millipore, 0.2 mmol scale) directly into the 30 mL Liberty microwave reaction vessel or transfer using dichloromethane. 2. Prepare solutions of the required Fmoc-protected amino acids (FmocAA-OH, 0 .2M in N-methyl pyrrolidinone) in separate Liberty 1 reagent bottles. Load reagent bottles onto the machine, taking note of the locations of each bottle to program the sequence. 3. Prepare solutions of 2-(6-chloro-1-H-benzotriazole-1-yl)1,1,3,3-tetramethylaminium hexafluorophosphate (HCTU, 0.5 M in N,N-dimethylformamide) and N,N-

280

Zhengrong Zhu et al.

diisopropylethylamine (DIPEA, 1 .0M in N,N-dimethylformamide) in Liberty 1 reagent bottles and load bottles onto the machine. 4. Follow the Liberty 1 software to load the required peptide sequence and initiate peptide synthesis according to the following conditions (single coupling): Fmoc-AA-OH (2.5 mL, 5 equiv., 0 .2M in N-methyl pyrrolidinone), HCTU (1.0 mL, 5 equiv., 0 .5M in N,N-dimethylformamide), DIPEA (1 mL, 1 .0M in N-methyl pyrrolidinone), 20 W (power), and 75
C for 10 min. N,N-dimethylformamide washing steps occurred after each step in the coupling. Double (two successive) couplings are completed for amino acids which directly follow sterically hindered amino acids. After each coupling, Fmoc cleavage proceeded with 20% piperidine/N,N-dimethylformamide solution followed by successive washes with N,Ndimethylformamide. 5. Mini-cleavage: To test for success of the linear peptide synthesis, perform a mini-cleavage. Transfer ~2 mg of the resin to a 5 mL Eppendorf tube and treat the resin with a solution of trifluoroacetic acid/triisopropylsilane/water (0.5 mL, 95:2.5:2.5). Seal and shake the Eppendorf tube on a platebased shaker for 1 h. 6. Filter off the resin using a filter tube, collecting the eluent in a 15 mL Falcon tube. Rinse the resin with dichloromethane (2 mL), collecting the washings into the same Falcon tube. The contents of the Falcon tube are concentrated by blowing under a stream of nitrogen until only the dry residue remained. 7. Dissolve the residue in methanol (2 mL) and transfer into a filter LC-MS vial. Analyze by LC-MS Method 1 to confirm the desired mass. If the desired linear peptide is synthesized proceed to the cyclization protocol as specified below. 8. Peptide cyclization: Transfer the remaining resin into a 5 mL Biotage glass microwave vial with stir bar. Add water (1.125 mL) and acetonitrile (1.125 mL) and stir the resin for 2 min to allow it to swell. 9. In a 5 mL Eppendorf tube, dissolve copper(II) sulfate pentahydrate (135 mg, 0.541 mmol) in 0.54 mL water. Using another 5 mL Eppendorf tube, dissolve sodium L-ascorbate (108 mg, 0.545 mmol) in 0.27 mL water. Add the two solutions to the resin mixture and seal the microwave vial using a crimp cap. 10. Heat and stir the microwave vial in the Biotage microwave according to the following conditions: temperature 60
C, maximum pressure 250 psi, ramp time 2 min, hold time 30 min, and power 2 W.

DNA-Encoded Macrocyclic Peptide Library

281

11. After the reaction is done, filter off the resin using a filter tube with open stopcock, collecting the eluent in a 50 mL centrifuge tube. Close the stopcock and treat the resin with the following solutions, opening the stopcock and draining the eluent into the centrifuge tube after each treatment: 5  2.5 mL water washings; 5  2.5 mL N,N-dimethylformamide; and 5  2.5 mL dichloromethane. After the final wash, the resin is left to dry for 20 min under a stream of air (see Note 3). 12. Mini-cleavage: Follow steps 5–7 to confirm cyclization of the peptide. If cyclization is confirmed then proceed to the full resin cleavage; if not repeat steps 8–11. 13. Full cleavage of cyclic peptide: Transfer the resin to a 15 mL Falcon tube and treat with a solution of trifluoroacetic acid/ triisopropylsilane/water (4 mL, 95:2.5:2.5). Seal and shake the Falcon tube on a plate-based shaker for 1 h at room temperature. 14. Filter off the resin using a filter tube, collecting the eluent in a 50 mL centrifuge tube. Rinse the resin with dichloromethane (8 mL), collecting the washings into the same centrifuge tube. The contents of the centrifuge tube are evaporated under a stream of nitrogen to remove the majority of the trifluoroacetic acid. 15. Dissolve a small amount of the residue in methanol (2 mL) and transfer into a filter LC-MS vial. Analyze by LC-MS using Method 1 (see Note 4) to confirm the desired mass. 16. Preparative HPLC purification: Purify the peptide by MDAPHPLC according to Method 2 (see Note 5). Combine the purified fractions and dilute to three times the initial volume with water. Pass through a pre-wetted (MeCN, then water) Oasis-HLB cartridge to remove the TFA salts. Elute the retained peptide out of the Oasis cartridge with dioxane, and then freeze-dry overnight. 17. Analyze the freeze-dried lyophilizate by LCMS for correct identity and purity (Method 3, Note 6). Report final compounds with overall yields based on the resin (0.2 mmol).

4

Notes 1. Either selection method (affinity resin tip or magnetic beads) can be used. Which one gives better result may vary from target to target, and thus needs to be tested. 2. Since protein molecules are much more than specific binding DNA-encoded molecules in a library during ELT selection, binding between protein and DNA-encoded molecules is driven by protein concentration. What protein concentration

282

Zhengrong Zhu et al.

to use in selection depends on what range of binding affinity is desired. Generally, 1 μM is a good starting point. If higher protein concentration is preferred, amount of magnetic beads may need to be increased accordingly. 3. When synthesizing off-DNA peptides, peptide-functionalized resin can be stored in the freezer (20
C) over several days. When ready to use, remove the resin from the freezer and allow to thaw at room temperature. 4. Method 1: The UPLC analysis was conducted on an Acquity UPLC CSH C18 column (50 mm  2.1 mm i.d. 1.7 μm packing diameter) at 40
C. The solvents employed are: A ¼ 0.1% v/v solution of trifluoroacetic acid in water B ¼ 0.1% v/v solution of trifluoroacetic acid in acetonitrile The flow rate is 1 (mL/min). The gradient employed is as follows: Method 1 Time (min)

%A

%B

0.0

98

2

0.2

98

2

2.5

3

97

2.9

3

97

3.0

98

2

The UV detection is a summed signal from wavelength of 210–350 nm. Injection volume: 0.3 μL MS: Waters ZQ Ionization mode: Alternate-scan positive and negative electrospray Scan range: 100–1000 AMU Scan time: 0.27 s Inter-scan delay: 0.10 s 5. Method 2: The MDAP-HPLC is conducted on an Xselect CSH C18 column (150 mm  30 mm i.d. 5 μm packing diameter) at ambient temperature.

DNA-Encoded Macrocyclic Peptide Library

283

The solvents employed are: A ¼ 0.1% v/v solution of trifluoroacetic acid in water B ¼ 0.1% v/v solution of trifluoroacetic acid in acetonitrile The flow rate is 40 (mL/min). The gradient employed is as follows: Method 2 Time (min)

%A

%B

0.0

60

40

3.5

60

40

25

40

60

32

40

60

35

1

99

41

1

99

The DAD detection is 210–350 nm. MS: Waters ZQ Ionization mode: Positive electrospray Scan range: 100–1000 AMU Scan time: 0.50 s Inter-scan delay: 0.2 s 6. Method 3: The UPLC analysis was conducted on an Acquity UPLC BEH C18 column (100 mm  2.1 mm i.d. 1.7 μm packing diameter) at 50
C. The solvents employed are: A ¼ 0.1% v/v solution of trifluoroacetic acid in water B ¼ 0.1% v/v solution of trifluoroacetic acid in acetonitrile The flow rate is 0.8 (mL/min). The gradient employed is as follows: Method 3 Time (min)

%A

%B

0.0

97

3

8.5

0

100

9.0

0

100

9.5

97

3

10

97

3

284

Zhengrong Zhu et al.

The UV detection is a summed signal from wavelength of 210–350 nm. MS: Waters ZQ Ionization mode: Positive electrospray Scan range: 300–1900 AMU Scan time: 0.4 s Inter-scan delay: 0.1 s References 1. Goodnow RA, Dumelin CE, Keefe AD (2017) DNA-encoded chemistry: enabling the deeper sampling of chemical space. Nat Rev Drug Discov 16:131–147 2. Yuen LH, Franzini RM (2017) Achievements, Challenges, and Opportunities in DNA-Encoded Library Research: An Academic Point of View. Chembiochem 18:829–836 3. Arico-Muendel C (2016) From haystack to needle: finding value with DNA encoded library technology at GSK. Med Chem Commun 7:1898–1909 4. Zimmermann G, Neri D (2016) DNA-encoded chemical libraries: foundations and applications in lead discovery. Drug Discov Today 21:1828–1834 5. Zhu Z, Shaginian A, Grady LC, O’Keeffe T, Shi XE, Davie CP, Simpson GL, Messer JA,

Evindar G, Bream RN, Thansandote PP, Prentice NR, Mason AM, Pal S (2018) Design and Application of a DNA-Encoded Macrocyclic Peptide Library. ACS Chem Biol 13:53–59 6. Kleiner RE, Dumelin CE, Liu DR (2011) Smallmolecule discovery from DNA-encoded chemical libraries. Chem Soc Rev 40:5707–5717 7. Thansandote P, Harris RM, Dexter HL, Simpson GL, Pal S, Upton RJ, Valko K (2015) Improving the passive permeability of macrocyclic peptides: Balancing permeability with other physicochemical properties. Bioorg Med Chem 23:322–327 8. Johnson TW, Dress KR, Edwards M (2009) Using the Golden Triangle to optimize clearance and oral absorption. Bioorg Med Chem Lett 19:5560–5564

Chapter 13 Peptide Display Technologies Anthony Pitt and Zeke Nims Abstract With an increased interest in the use of peptides as therapeutics comes the need for strategies to allow for the discovery of novel hit candidates, in high-throughput manners, from highly complex peptide libraries. Early development of peptide therapeutics arose from the deployment of natural peptides and subsequent modification to enhance medicinal properties. Here, the implementation of synthetic peptide libraries of low complexity was sufficient, but these low-diversity starting points are an obvious limitation to the discovery of novel peptides. This limitation is compounded by the almost unarguable desire to explore unnatural amino acid chemical space, moving away from the reliance upon natural peptides. Key words Peptide display, Peptide discovery, Macrocycle, Cyclic peptide, Unnatural amino acids, Phage display, Yeast display, Ribosome display, mRNA display, Bead display

1

Peptide Display Technologies The discovery of novel peptide therapeutics requires methods by which highly complex peptide libraries can be rapidly screened against a target. In such screens, any hit candidates should be easily captured and their sequences elucidated. Ideally the method would directly link peptide genotype to phenotype. A wish list for a peptide discovery technology would also include the ability to present a wide range of peptide lengths (typically 5–30 amino acids), to capture peptides with good binding affinities to a target, and while some of the properties that go toward making a peptide hit a useful therapeutic (e.g., serum stability and oral delivery) are commonly addressed at a later stage after hit discovery, the sooner those properties can be directly tethered to the primary screen, the more valuable a technology becomes.

1.1 Desirable Properties of a Peptide Display Technology

It is known that the therapeutic usefulness of a peptide can be significantly improved by alterations of a peptide structure away from a simple linear form to constrained forms that introduce rigidity, but also lend beneficial properties to macrocyclic

Gilles Goetz (ed.), Cyclic Peptide Design, Methods in Molecular Biology, vol. 2001, https://doi.org/10.1007/978-1-4939-9504-2_13, © Springer Science+Business Media, LLC, part of Springer Nature 2019

285

286

Anthony Pitt and Zeke Nims

structures. Furthermore, the inclusion of amino acids outside of the natural 20-amino acid repertoire may improve peptide properties. Such modifications of the peptide structure have been shown to improve peptide stability, increasing resistance to protease activities, to improve affinity of a peptide to a target and even to bestow properties such as cell permeability to a peptide [1, 2]. It has been shown that a linear peptide may show no binding to a target in its linear form, but in a simple cyclic form binds to that same target [3–5]. Likewise, peptides composed of natural amino acids show poor binding to a target but when some of those amino acids are replaced with unnatural counterparts (such as their N-methylated forms) the peptide now binds or shows significantly improved binding to that target [6]. It is clear therefore that any method solely presenting linear peptides composed of natural amino acids may altogether miss potentially valuable hits that may lead to viable therapeutics. Table 1 summarizes the key features that are desired in a peptide display technology to allow for the discovery of novel peptide therapeutic hits in a high-throughput manner from highly complex libraries. Ideally, a method for the discovery of peptide hits would therefore be based on a peptide display technology capable of linking peptide genotype to phenotype, would be rapid and allow for the use of highly complex peptide libraries, and would be flexible to enable the display of modified peptides that include macrocyclizations and other constraints, inclusion of nonnatural amino acids, and other posttranslational modifications. Peptide display technologies may be characterized in several ways: in vivo or in vitro replication, RNA- or DNA-based encoding, copy number of the displayed peptide, natural or synthetic carriers, and size of the carrier. Table 1 Desirable features of a peptide display technology Desired feature

Details

Link genotype to phenotype

Required to couple encoding material to peptide to allow for simple and quick determination of peptide sequence

Peptide size

Allow for display of peptides up to 50 amino acids in size

Library complexity

Capture full sequence diversity of a given size of peptide

Screening complexity

Capture as much of the library diversity in a screen

Constraints

Allow for the introduction of constraints into the peptide, i.e., cyclizations

Unnatural amino acids Capability to introduce unnatural amino acids into the peptide, i.e., methylated amino acids Posttranslational modifications

Allow for posttranslational modification such as glycosylation

Peptide Display Technologies

2

287

Peptide Display Technologies The first technology to successfully attempt to address these needs and to come to widespread use is phage display, where a filamentous bacteriophage is genetically modified to display proteins or peptides fused to coat proteins on its surface with improvements allowing for the display of both linear and simple cyclic peptides. This timetested method is free of intellectual property protection, is sold as commercial kits, and is a simple procedure to identify novel binders. Further, it is a robust technique, capable of accommodating posttranslational modifications, and able to explore sequence space of up to 1010 variants in any given library. Newer technologies have been introduced that address some of the shortcomings of phage display and expand upon the desired feature set of a peptide display technology. Most notably, the creation of ribosome and mRNA-based display technologies introduced the incorporation of unnatural amino acids into the displayed peptide for the first time, allowing for the selection of peptides with enhanced characteristics to be selected at the primary screening stage. These methods are also basically free of intellectual property constraints but are fastidious systems that require additional trade secret-type knowledge to execute the platforms, which are scant in the public literature, but also produce novel binding candidates. Both of these methods accommodate chemical modifications and are capable of exploring sequence space of ~1012 variants. More recent advancements in peptide display technologies have seen the use of DNA-binding proteins to link encoding DNA to peptide mimicking mRNA display and the use of synthetic carriers such as microbeads where the encoding material and the peptide are linked to the bead surface.

2.1 Components of Display Technologies

The structural components of a display technology may have implications for its use which confer advantages or disadvantages when compared to alternative display technologies. Table 2 lists peptide display technologies that are currently used for the discovery of novel peptide therapeutic candidates and highlights the key technological components and how they differ. Methods of replication of the genetic and peptide components of the technology may have a number of implications. In vivo methods rely on the natural replication and protein synthesis machinery of a host organism, a bacterial host for phage and the yeast, or bacteria itself in those display technologies. In vivo methods can lead to replication bias due to some peptide sequences either suppressing or promoting the replication of its host or the peptide being lethal, which can lead to the overabundance of some peptides within a screening library and even the absence of others. Phage in particular is difficult to contain and environmental

288

Anthony Pitt and Zeke Nims

Table 2 Components of display technologies

Technology

Replication Encoding Copy number

Carrier size (μm)

Surface complexity

Phage

In vivo

DNA

Multiple (low)

Medium (0.2)

Low-medium

Yeast

In vivo

DNA

Multiple (high)

Large (3)

Very high

Bacteria

In vivo

DNA

Multiple (high)

Large (2)

Very high

Ribosomal

In vitro

RNA

Single

Small (0.002)

Low

mRNA

In vitro

RNA

Single

Small (