CHAPTER 1: Renaissance in Peptide Drug Discovery: The Third Wave

Peptide drug discovery has evolved from highly focused efforts on specific receptors and proteases to a plethora of targets spanning receptors to enzymes and protein–protein interactions, and shattering a long-lived challenge to penetrate into cells to modulate intracellular targets in promising ways to expand “druggable” target space. Since the turn of the new millennium (2000) there is no doubt among peptide scientists that there is a genuine renaissance of peptide drug discovery. So, what has really inspired and propelled such a renaissance? Scholarly passion and intellectual perseverance have been absolutely essential, as this has been the proverbial long and winding road. And, of course, science and technology are empowering key advancements. In particular, synthetic chemistry (e.g., novel amino acid building blocks and peptide secondary structure mimetics) and super-diverse phage-display, mRNA-display and DNA-encoded libraries are expanding peptide chemical space in amazing ways. Likewise, molecular genetics, structural biology and computational chemistry are continuing to play powerful roles to unveil extraordinary target space opportunities. Furthermore, our increased understanding of the complex pharmacology of disease mechanisms and peptide drug delivery, including cell permeability to prosecute intracellular targets, is re-defining the term “druggability”. Beyond such convergent, multidisciplinary science and enabling technology, the steady growth of a pipeline of marketed and clinically investigated peptides, as well as a competitive resurgence of peptide research and development in both pharma and biotech are propelling this renaissance of the peptide therapeutic modality.

Historically, one of the earliest archetypal flags in the ground for peptide chemistry can be traced to that of Emil Fischer and the synthesis of the simplest dipeptide Gly-Gly at the very beginning of the twentieth century. In retrospect the sheer number of milestones that deserve recognition are well beyond those highlighted in this chapter (Figure 1.1), albeit they exemplify some quite significant achievements in peptide science and technology. Specifically, these include the discoveries of insulin, penicillin, oxytocin, pepstatin, thyrotropin-releasing hormone, gonadotropin-releasing hormone (GnRH), somatostatin, melanocyte-stimulating hormone (MSH), enkephalin and, more recently, tri-cystine knotted cyclotides. They include the development of potent peptide and peptidomimetic analogs thereof that have provided working models for bioactive conformations, agonist/antagonist pharmacophores and cellular receptor signaling mechanisms. Likewise, and with respect to intracellular targets, the natural product macrocyclic peptide cyclosporine A, HIV Tat (a progenitor of the first cell-penetrating peptides), synthetic peptidomimetic HIV protease inhibitors, designed non-peptide Src homology-2 antagonists and macrocyclic α-helical proteomimetic antagonists of MDM2/X collectively illustrate the scope of both past and current peptide drug discovery approaches to overcome the challenge of cell permeability. Lastly, some key past and emerging disruptive innovations that have leveraged the power of molecular biophysics and molecular diversity to enable peptide drug discovery include X-ray crystallography (i.e., identification of canonical secondary structures such as the α-helix and β-sheet), solid-phase peptide synthesis, synthetic peptide/peptoid libraries, phage-display peptide (monocyclic/bicyclic) libraries and mRNA-displayed macrocyclic peptide libraries. In the case of macrocyclic peptides, there is no doubt that the impact of stapled helical peptides and non-helical macrocyclic peptides having varying size and inclusion of N-methyl amino acids, d-amino acids and other unique amino acid building blocks is both driving the generation of novel lead molecules and expanding druggable target space. Lastly, varying chemical modification of clinically investigated peptides that exemplify improvement in pharmacokinetic (half-life) properties have been achieved (see later), and include lipidation, pegylation and, more recently, macrocyclization (e.g., amphipathic α-helical stapled peptides).

Since the 1980s, several hundred peptide and peptidomimetic candidates have advanced into clinical trials for a wide range of therapeutic indications,^1–3  including endocrine, metabolic, cardiovascular, cancer, immune and central nervous system diseases, and more than 50 such agents have been approved by the United States Food and Drug Administration (Table 1.1). Several of these drugs have achieved major commercial success, including Lupron, Zoladex, Sandostatin, Byetta and Forteo. Unquestionably, a strong understanding of the structure–activity relationships of such peptides at their specific targets and translation to in vivo preclinical disease models has been critical for their drug development. Likewise, overcoming challenges such as the metabolic instability (owing to rapid degradation by proteolytic enzymes) and generally poor pharmacokinetic properties of peptides has been strategic for their optimization relative to in vivo efficacy and route of administration.

Peptide oral bioavailability remains elusive, although what was once thought to be the exceptional case of cyclosporine A is changing as a result of a deeper analysis of macrocyclic peptides to understand the relationship of the structural and conformational impact of backbone modifications, ring size and side-chain lipophilicity to passive transport (see later). Consequently, a majority of marketed peptide therapeutics leverage subcutaneous and injectable routes of administration. Such modified peptides and peptidomimetics exemplify classic hierarchical strategies^4–20  to achieve an effective combination of high affinity to target and proteolytic stability, including (i) backbone amide N-alkylation; (ii) backbone amide replacement with non-hydrolyzable surrogates; (iii) amino acid Cα-stereoinversion and/or Cα-alkylation; (iv) β-amino acids; (v) cyclic α-/β-amino acids; (vi) dipeptide replacements that mimic canonical secondary structural motifs such as α-helix, β-strand/β-sheet or β-/γ-turns; and (vii) macrocyclization designed to stabilize α-helical, β-strand/β-sheet, β-/γ-turns and/or other conformationally restricted peptide/peptidomimetic chemotypes (e.g., monocyclic or multicyclic). A few examples of the pioneering and contemporary chemistry that has contributed to modified peptides and peptidomimetics are described below.

For receptor-targeted peptide therapeutics, long-lasting exposure levels in vivo have been achieved using varying approaches,^21–27  including both chemical conjugation (e.g., fatty acids, polyethylene glycol, antibodies and related recombinant proteins and serum albumin) and sustained-release formulations applicable to parenteral routes of administration (e.g., subcutaneous). Specific examples include the palmitoyl-modified glucagon-like peptide-1 (GLP-1) agonist liraglutide,²⁸  a pegylated GLP-1 agonist/glucagon antagonist hybrid peptide,²⁹  the pegylated granulocyte colony-stimulating factor drug pegfilgrastim³⁰  and a human Fc domain–thrombopoietin peptide agonist conjugate romiplostim.³¹  Although absolute exposure levels, in terms of both time and concentration, are case-specific, the relatively short half-lives (typically a few minutes) of most endogenous (native) peptides provide an opportunity to advance viable drugs with substantially improved metabolic and pharmacokinetic properties.

Indubitably, the greatest impact of peptide drug discovery so far has been that focused on receptor target space, especially the G protein-coupled receptor (GPCR) group (e.g., class A and B GPCRs), which has been determined using human genome sequencing to be one of the largest protein families^32–34  (Figure 1.2), and this has translated to the first wave of peptide therapeutics.^{9,11,14,19,35}  Pioneering studies on class A GPCR peptides may be traced to oxytocin, vasopressin, α-MSH, GnRH, somatotropin-release inhibiting factor (somatostatin) and the opioid peptides (e.g., enkephalin, β-endorphin, dynorphin) during the 1970s–1990s. Likewise, but more recently, a second wave of peptide drug discovery has successfully extended to class B GPCRs as exemplified by glucagon-like peptide-1 (GLP)-1, islet amyloid polypeptide (amylin), GLP-2 and parathyroid hormone.

Importantly, these early class A and B GPCR peptide agonist/antagonist structure–activity studies provided insight to understand the intrinsic peptide conformational properties as well as predictive 3D-pharmacophore models.^14,19,35  Unfortunately, such pioneering studies were not empowered by high-resolution X-ray crystallographic structures of class A and B GPCRs until more recently.^36,37  Instead, a systematic analysis of peptide structure–activity relationships (e.g., analog modifications by d-amino acids, N-alkyl-amino acids, Cα-alkyl-amino acids and/or macrocyclization) as well as biophysical characterization (e.g., nuclear magnetic resonance (NMR) spectroscopy and circular dichroism) revealed that β-turn and α-helix secondary structures often correlated with GPCR molecular recognition for such peptides (Figure 1.3). As exemplified by Scenesse, Lupron and Sandostatin, incorporation of a d-amino acid regiospecificially within their peptide sequences has been conceptualized to stabilize β-turn conformations of their respective class A GPCR agonist pharmacophores.^38–40  In contrast, Byetta, Forteo and Gattex share relatively high propensities for α-helix conformations that have been correlated with their molecular recognition of and binding to and activation of their respective class B GPCRs.^41–43 

As a personal reflection on the first wave of peptide drug discovery, my first scientific foray at the University of Arizona contributed to the discovery of class A GPCR superagonist peptides for the melanocortin receptor MC1R^38,44–46 , namely Scenesse™ (Figure 1.3) and cyclo[Cys⁴, Cys¹⁰]α-MSH (Figure 1.4). Indisputably, these two molecules have inspired the design of numerous linear and macrocyclic α-MSH peptide analogs,^47,48  including the recent clinical development macrocyclic α-MSH peptide analog setmelanotide (Figure 1.4) from Rhythm Pharmaceuticals.⁴⁹  It is noteworthy that the d-Phe⁷ modification and macrocyclization about the central pharmacophore tetrapeptide within these peptide agonists has been consistent with stabilizing a predicted β-turn (Figure 1.4). Furthermore, the second-generation macrocyclic α-MSH superagonist Ac-cyclo[Nle⁴, Asp⁵, d-Phe⁷, Lys¹⁰]α-MSH_4-10-NH₂ ⁴⁶  and the structurally related α MSH antagonist analog Ac-cyclo[Nle⁴, Asp⁵, d-Nal(2′)⁷, Lys¹⁰]α-MSH_4-10-NH₂ ⁵⁰  (Figure 1.4) from the University of Arizona exemplify significant benchmarks to enable the design of MSH ligands for the MC1R, MC3R, MC4R and MC5R. Most recently, MCR relationships to several key diseases (e.g., energy homeostasis, inflammation and neurodegeneration) have been described^51–53  and implicate new opportunities to leverage MCR-specific peptide and non-peptide agonists/antagonists.^47,48,54 

The development of peptide agonists for the GLP-1 receptor has been advanced in what may be the most competitive worldwide efforts focused on class B GPCRs. Such efforts have been focused on type 2 diabetes mellitus and obesity, including glucose homeostasis and regulation of gastric motility and food intake.^55–57  Currently, several GLP-1 receptor-targeted peptide agonists have reached the market and representative examples include exenatide, liraglutide, lixisenatide and semaglutide (Figure 1.5). Such GLP-1 peptide agonists illustrate regiospecific amino acid modifications to confer metabolic stability as well as the incorporation of fatty acid or other types of conjugation to enhance their in vivo pharmacokinetic and pharmacological properties.^58–60  Of historical significance to class B GPCRs, the first X-ray crystallographic structures of GLP-1 receptor extracellular domain complexes with GLP-1 peptide analogs have provided the 3D molecular maps and insight to their receptor binding mechanism.⁶¹  Furthermore, the deeper biological study of GLP-1 peptides (including truncated, modified and chimeric analogs) as well as small-molecule modulators of the GLP-1 receptor activation have provided further understanding of molecular recognition and biased cellular signaling.⁶²  Importantly, biased cellular signaling is being found in an increasing number of GPCRs by both peptide and non-peptide agonists.⁶³ 

Beyond GPCRs, there has been steady progress in the development of peptide modulators of growth factor and cytokine receptors, integrins and ion channels to expand the scope of receptor target space for peptide drug discovery. Noteworthy for such receptors are the more structurally complex peptide agonists and antagonists, including those having multiple disulfide bridges such as insulin, linaclotide, ziconotide and ProTx-II, and which exemplify high specificity for the insulin receptor, guanylate cyclase-C, N-type Ca²⁺ channel and voltage-gated Na¹⁺ channel targets, respectively.^64–67 

A significant opportunity for the peptide drug modality is emerging relative to a third wave that is focused on modulating intracellular targets relative to stapled α-helical peptides,^68–73  structurally/conformationally diverse macrocyclic peptides inspired by cyclosporine A^74–81  and both linear and macrocyclic peptides incorporating cell-penetrating peptide motifs.^82–85  Both β-strand and α-helix secondary structures are widely found in at the interfaces of protein–protein interactions, and have been described^86–91  with respect to comprehensive analysis of the Protein Data Bank (www.rcsb.org). Examples of β-strand protein–protein interactions include PDZ domains, PTB/PI domains, Mad2:Cdc20, NS3 protease:NS4A and PKA:Rbα.^89,90  Representative α-helical protein–protein interactions of >1600 non-redundant, unique high-resolution 3D structures derived from the Protein Data Bank^87–91  include BIM BH3:Mcl-1,⁹²  BAD:Bcl-Xl,⁹³  p53:MDM2,⁹⁴  MAML:Notch,⁹⁵  HIF-1α:p300,⁹⁶  Myc-Max,⁹⁷  eIF4G:eIF4E,⁹⁸  TIF2:GCCR,⁹⁹  Scr-1:RORγ¹⁰⁰  and IKKβ:NEMO¹⁰¹  (Figure 1.6). As detailed later, α-helical secondary structures have inspired significant peptide drug discovery efforts to modulate such protein–protein interactions, with a specific focus on intracellular targets of therapeutic interest.

Peptide drug discovery leverages the typical high affinity and/or selectivity properties that endogenous peptides have for their cognate targets (e.g., ≤10⁻⁹ M range for most GPCR-targeted peptides). We understand that the high-fidelity molecular recognition between peptides and their targets is achieved through intermolecular interactions by means of a dynamic orchestration of specific hydrophobic, electrostatic and hydrogen-bonding forces. Nevertheless, mimicking critical target interactions effectively is a key challenge in peptide drug discovery, owing to the molecular size, conformational flexibility and functional group diversity of the peptides. Our early knowledge of the molecular recognition of peptide hormones with their cognate GPCR targets evolved from studies focused on 3D pharmacophore models of peptide agonists or antagonists generated from structure–activity relationships and conformational analysis from NMR spectroscopy.^{4–6,9,11,14,19}  Our current knowledge is rapidly expanding by way of high-resolution 3D structures of peptide–target complexes, leveraging X-ray crystallography and/or NMR spectroscopy, as well as increasingly sophisticated computational modeling studies.^{6,7,12,15,36,37,61,86–90,102–117} 

Peptide drug discovery has evolved from hierarchical drug design strategies,^{4–20,68–79,91,118}  integrating synthetic chemical modifications of the native peptide or protein, including amino acid substitutions, amide bond replacements, peptide scaffold modification (e.g., peptidomimetic analogs), macrocyclization, secondary structure mimicry and non-peptidic templates (e.g., de novo designed small-molecule replicas of peptide ligands) (Figure 1.7). This peptide drug design and synthetic chemistry has expanded our understanding of β-turns, γ-turns, β-strands, β-sheets, α-helices and 3₁₀-helices in terms of both their 3D structural properties and molecular recognition at cognate targets (see later).

The 3D structural and conformational properties of peptides are defined in terms of torsion angles (ψ, ϕ, ω and χ) between the backbone amine nitrogen (N^α), the backbone carbonyl carbon (C′), the backbone methine carbon (C^α) and the side-chain hydrocarbon functionalization (e.g., C^β, C^γ, C^δ, C^ε of Lys) derived from the amino acid sequence (Figure 1.8). A Ramachandran plot (ψ versus ϕ) may be used to analyze the preferred combinations of torsion angles for ordered secondary structures (i.e., conformations) of peptides, such as α-helix, β-turn, γ-turn or β-strand. With respect to the amide bond torsion angle (ω), the trans geometry is more energetically-favored for most natural dipeptide substructures; however, when the C-terminal partner is Pro- or another N-alkylated (including cyclic) amino acid, the cis geometry is probable and may also contribute to β-turn or γ-turn stabilization.

Molecular flexibility is directly related to covalent and/or non-covalent bonding interactions within a specific peptide. In this regard, a single replacement of hydrogen by a methyl moiety within amino acids (i.e., N^α-methyl, C^α-methyl or C^β-methyl) may have significant consequences on the local conformational properties of a peptide. The peptide N^α–C^α–C′ and/or C^α–C^β scaffold may also be transformed to create novel amino acid building blocks^{4,20,119–128}  such as dehydro-amino acids (e.g., X^α–X^β → X=C), vinylogous amino acids (e.g., C^α–C′ → C^α–C=C–C′), β-amino acids (e.g., C^α–C′ → C^α–CH₂–C′ or C^α–CHR-C′ wherein R is a side-chain independent or intramolecularly cyclized to the C^α-carbon), aza-amino acids (e.g., C^α → N) and achiral N-substituted Gly or “peptoid” (e.g., N^α–C^α–C′ → NR–CH₂–C′ wherein R is a side-chain). Furthermore, the C^β-carbon may be substituted to create “chimeric” amino acids. Overall, such N^α–C^α–C and/or C^α–C^β scaffold modifications provide significant opportunities to enhance peptide-like chemical diversity relative to peptide ϕ–ψ–χ space.

The peptide backbone has been the focus of significant chemical modifications over the past three decades to impart varying desired properties, including (i) increased metabolic stability against proteolytic degradation; (ii) enhanced molecular recognition for active site binding in designing peptidomimetic inhibitors of protease therapeutic targets; and (iii) improved bioavailability. Examples of amide bond replacements that have been described within the scope of peptide and peptidomimetic drug discovery^{6,20,129–132}  include aminomethylene (CH₂NH), ketomethylene (COCH₂), ethylene (CH₂CH₂), olefin (CH=CH), ether (CH₂O), thioether (CH₂S), tetrazole, thiazole, retroamide (NHCO), thioamide and ester. Other non-hydrolyzable amide bond isosteres have been leveraged for protease inhibitor drug design,^{6,13,20,33,133,134}  including hydroxymethylene (CH[OH]), hydroxyethylene (CH[OH]CH₂), dihydroxyethylene (CH[OH]CH[OH]), hydroxyethylamine (CH[OH]CH₂N) and related C₂-symmetric hydroxymethylene and dihydroxyethylene moieties. Some marketed HIV protease inhibitors (e.g., amprenavir, indinavir, saquinavir, darunavir, atazanavir and nelfinavir^133,134 ) have successfully exploited such non-hydrolyzable amide bond isosteres (Figure 1.9) and further underscore the greater scope of the second wave of peptide drug discovery, in terms of both peptidomimetic design and tackling intracellular target space.

An understanding of 3D structural and conformational properties of peptides that correlates with their binding and modulation of therapeutic targets has inspired creativity in peptide drug design.^{6–8,10,12,15,16,18,20,68–79,135–152}  Historically, regiospecific substitution of d-amino acids, N^α-Me-amino acids, C^α-Me-amino acids, dehydro-amino acids and cyclic amino acids within a peptide lead compound have provided insights relative to their propensities to induce or stabilize β-turn, γ-turn, β-strand, α-helix and 3₁₀-helix secondary structures. Furthermore, the progression of versatile synthetic peptidomimetics and designed non-peptide templates that attempt to incorporate key side-chain and/or backbone elements of such peptide secondary structures is noteworthy (see later).

The β-turn was often inferred from structure–activity analysis to be a critical secondary structural element within class A GPCR peptide pharmacophores (e.g., somatostatin, α-melanotropin and enkephalin).^{4–6,9–11,16,19,38–40,118,136,141,147,148,153}  The β-turn motif exists within tetrapeptide sequences in which the first and fourth Cα atoms are separated by ≤7 Å and a 10-membered intramolecular H-bond exists between the i and i + 4 amino acid residues. Noteworthy are monocyclic or bicyclic templates (1–4) (Figure 1.10) from early efforts focused on designed β-turn peptidomimetics.^154–157 

The β-strand motif has been recognized with respect to the molecular recognition of both peptide substrate-based inhibitors of proteases and at the interfaces of numerous protein–protein interactions.^{16,89,90,109,118,137,145,148–150}  Noteworthy are the peptidomimetic scaffold and non-peptide templates (5–8) (Figure 1.11) that have been advanced as designed β-strand mimetics.^{149,158–161} 

The α-helix is a major secondary structure found in nature, and exists at the interface of a plethora of protein–protein interactions.^{12,16,18,36,37,68–73,86–88,91,118,151,152}  Such α-helices may be categorized as linear, kinked and curved, of which the latter is the predominant type. Noteworthy are peptidomimetic scaffold and non-peptide templates (9–12) (Figure 1.12) that have been advanced as designed α-helical mimetics.^162–165  Furthermore, peptidomimetics generated from other building blocks, including N-substituted Gly (“peptoid”) and β-amino acids (“β-peptides”), have been shown to exhibit helical-type secondary structures.^166–169 

Over recent years, peptide macrocyclization has resurged as a key driving force of the third wave of peptide drug discovery in terms of leveraging both structure-based design and super-diverse library screening of synthetic peptides to accelerate their generation, optimization and development.^{68–79,84,85,91,109,113–118,170–175}  The scope of chemical space is remarkable and provides significant opportunity for peptide drug discovery focused on intracellular targets such as those exemplified by known macrocyclic peptide natural products (e.g., cyclosporine A or CsA, 13¹⁷⁶ ), designed macrocyclic peptidomimetic inhibitors of proteases (e.g., hepatitis C virus (HCV) protease inhibitor 14¹⁷⁷ ), designed stapled α-helical peptides (e.g., dual MDM2/X antagonist 15¹⁷⁸ ), CsA-inspired mRNA display library-generated macrocyclic peptides (e.g., SIRT2 inhibitor 16¹⁷⁹ ), CsA-inspired DNA encoded library-generated macrocyclic peptides (e.g., XIAP antagonist 17¹⁸⁰ ) and cell-penetrating peptide (CPP) hybridized synthetic library-generated macrocyclic peptides (e.g., KRAS antagonist 18¹⁸¹ ) (Figure 1.13). Collectively, such macrocyclic peptides are providing novel chemotypes with unique biophysical and pharmacological properties to tackle so-called “undruggable” targets. In particular, there has been significant progress in the study of α-helical peptides,^{68–73,91,118,171–175,178}  including progression into the clinic.

Without question, the revival of macrocyclic peptide drug discovery has been led by the conception of stapled α-helical peptides and the application of varying synthetic chemistries (e.g., ring-closing metathesis, azide–alkyne cycloaddition and thioether) such as those exemplified (Figure 1.14) by macrocyclic α-helical peptides (19–22),^182–185  the first described MDM2 and/or MDMX antagonists of their particularly unique chemotype. Noteworthy is the highly potent and in vivo effective dual MDM2 and MDMX antagonist ATSP-7041 (15, Figure 1.13), which has become a benchmark stapled peptide relative to its intrinsic biophysical properties (e.g., amphipathicity, solubility, cell permeability and metabolic stability^178,186 ) that are implicated as key features to design concepts for this class of macrocyclic α-helical peptides.^71–73,171  Of note, such work has successfully translated to the first stapled α-helical peptide for an intracellular target to advance into the clinic (ALRN-6924 by Aileron Therapeutics). Collectively, and related to the first publication¹⁸⁷  of stapled α-helical peptides incorporating α-methyl amino acids as part of the macrocyclization framework, there are now many reported studies illuminating the design, chemistry and biological properties of stapled peptides for a wide range of intracellular protein–protein interaction targets (e.g., Bcl-2 family, Notch, estrogen receptor, vitamin D receptor, HIV capsid, HIV integrase, HIF-1, KRAS, Rab, p53, β-catenin, replication protein A, eIF4E, WASF3 and aurora-A kinase^{178,182–221} ). Nevertheless, there is more to understand with respect to cell permeability and drug delivery of peptides (see later) to fully exploit this macrocyclic α-helical peptide modality class relative to tackling intracellular target space and the translation to clinical development.

Beyond the aforementioned macrocyclic α-helical peptide drug discovery efforts there has been extraordinary progress over recent years with respect to exploiting super-diverse mRNA display,^74–78  phage display,^79,80  DNA-encoded⁸¹  and synthetic macrocyclic (e.g., CPP) peptide libraries.^84,85  Such super-diverse libraries may explore virtually all possible canonical and non-canonical peptide conformational space, generally relative to within a 8–20 amino acid size range, although with the inclusion of multiple cyclization designs there is opportunity to introduce further constraints to molecular flexibility (Figure 1.13). In the case of mRNA display, DNA-encoded peptide and other synthetic macrocyclic peptide libraries, the inclusion of unnatural amino acids (e.g., N-methyl-amino acids, Cα-methyl-amino acids, d-amino acids, β-amino acids and further novelties relative to side-chain functionality) and/or other structurally related building blocks may be incorporated.^{74–78,81,84,85}  In the case of phage display libraries, the use of thioether bridging of Cys residues⁷⁹  provides the opportunity to explore diversity (e.g., peptide sequence, ring-size and varying linking moieties), and more recently the synthesis of phage-generated glycopeptide libraries⁸⁰  illustrates further innovative chemistry methods. Lastly, but not exactly fitting into the realm of super-diverse libraries, is the more recent focused study of smaller macrocyclic peptides (i.e., five- and six-membered rings) to explore CsA-like passive permeability relative to systematic modification of both backbone and side-chain functionalities, and then leverage such structure–property relationships to identify peptide scaffolds showing cell permeability (see later) may then be further designed to achieve molecular recognition for therapeutic targets.^222–227 

As the saying goes, “this is where the rubber hits the road”; it is a fitting introductory comment, especially in discussing peptide cell permeability and drug delivery, as these have been major challenges to overcome and unleash the potential of the peptide modality. In terms of addressing this topic, peptides traverse cell membranes in two ways: passively or non-passively. This oversimplification of what is unquestionably complex nevertheless guides peptide drug discovery strategies focused on integrating screening tools and advancing design rules for peptide cell permeability. What is the strategic modus operandi? It begins with learning from what nature has revealed and the use of empirical methods to systematically explore peptide structure–property relationships as well as delineating the cellular mechanisms of membrane permeability. With respect to passive permeability, cyclosporine A has provided a benchmark for designing macrocyclic peptides with oral bioavailability.^{176,222–230}  In contrast, with respect to non-passive permeability, Tat and Tat-related polycationic CPPs have provided benchmarks and tools to understand linear peptides exhibiting cell uptake by endocytosis and/or other translocation mechanisms.^231–251  Beyond linear CPPs, noteworthy progress is being achieved with stapled α-helical peptides, and more recently, CPP-hybrid macrocyclic peptides relative to exploring their structure–permeability relationships to optimize their cellular potency to modulate intracellular targets as well as further understand the their cell uptake mechanisms.^84,85,181  In the specific case of stapled α-helical peptides, some key properties that have been proposed^{71–73,171,174,178}  to correlate with cellular potency include amphipathicity (predominantly hydrophobic, but with preservation of solubility), lipophilicity (partitioning into membranes), α-helicity (intramolecular H-bonding), proteolytic stability and high binding affinity to their cognate target (K_d < 1–10 nM range) to their respective cognate targets (Figure 1.15). Therefore, iterative design and analysis of peptide structure–permeability relationships may address intrinsic physicochemical/biophysical properties as well as experimental permeability properties using cellular and related non-cellular model systems. It is likely that comparative analysis of linear CPPs, CPP-hybrid macrocyclic peptides and/or stapled α-helical peptides will provide insight with which to further benchmark and more deeply understand peptide cell uptake, including relationships to endocytosis and/or other translocation mechanisms. Recent studies have addressed this relative to comparative analysis of linear CPPs with CPP-hybrid macrocyclic peptides,⁸⁴  as well as comparative analysis of linear CPPs with stapled α-helical peptides.²⁵²  Furthermore, in the case of CPP-hybrid macrocyclic peptides it was demonstrated that such molecules were capable of high cytosolic delivery efficiency, so that cell uptake included direct interaction with plasma membrane phosopholipids, endocytosis and subsequent release from endosomes.⁸⁴  In addition, this study revealed that such molecules induced membrane curvature on giant unilamellar vesicles and budding of small vesicles that subsequently collapsed into amorphous aggregates of peptide and lipid. In contrast, an evaluation of structurally varied stapled α-helical peptides implicated a correlation of cell uptake via a clathrin- and caveolin-independent endocytosis pathway that was suggested to involve, in part, anionic cell surface proteoglycans.²⁵² 

To understand more deeply peptide cell permeability and drug delivery it is critically essential to have screening tools, including both computational and experimental methods, which can facilitate predictive design rules and iterative testing thereof. Unsurprisingly, for passive permeability there exists a framework of existing screening tools (e.g., transcellular permeability using cellular monolayers and proteolytic stability in varying biological matrices) as well as emerging computational and biophysical methods^{222–227,253–260}  that address exposed polar surface properties, NMR/mass spectroscopy (MS) analysis of intramolecular H-bonding, radius of gyration and aqueous/lipid phase partitioning (Table 1.2). Furthermore, cell-based screening methods^261–265  utilizing MS and fluorescence microscopy for unlabeled or fluorescently labeled peptides, respectively, may be implemented to determine cell uptake as well as provide correlation with their cellular functional properties (Table 1.2). For non-passive permeability, it is likely that some (perhaps most) of the same physicochemical and biophysical properties will be important, since peptide partitioning into the cell membrane will have similar requirements for hydrophobicity/lipophilicity and desolvation of exposed H-bonding moieties of the peptide molecule. Certainly, there is ample opportunity for the development of innovative permeability screening tools to enable peptide drug discovery and it is anticipated that such tools will continue to emerge in the future.

A compelling future of peptide drug discovery is envisaged relative to the promise of novel breakthrough medicines that are no longer restricted to extracellular or receptor targets, but rather overcome permeability barriers to enable the prosecution of intracellular targets, especially those generally perceived as undruggable. As highlighted in this chapter, there is robust knowledge that reflects more several decades of pioneering multidisciplinary science and technology on the peptide modality. This has contributed to hierarchical design strategies, expansive chemical space, sophisticated computational modeling, and a plethora of high-resolution crystal structures of peptide complexes with their cognate targets. Super-diverse peptide libraries generated by mRNA display, phage display, DNA-encoding and other synthetic chemistry methods exemplify a disruptive innovation that will impact both basic research and translation to clinical development. Collectively, several key technology platforms (Table 1.3) that are enabling such structure-based design and library-driven diversity of macrocyclic peptides are acknowledged with respect to contributing to such disruptive innovation.

To conclude, the renaissance of peptide drug discovery has reignited the torch of both chemistry and biology to converge upon intracellular target space with an armamentarium of structurally diverse macrocyclic peptides along with cell permeability screening tools and design rules to advance the peptide modality. This is the third wave, and it is growing with force and momentum to be a powerful one.

I wish to acknowledge my colleagues at Merck & Co. (Kenilworth, NJ, USA) who have been actively engaged within the Peptide Drug Hunter network, and I am grateful for their many contributions and achievements. Likewise, I wish to acknowledge our collaborations with Ra Pharmaceuticals, PeptiDream, A*STAR and IRBM to advance our peptide drug modality pipeline.