Targeting Intracellular Protein-Protein Interactions with Macrocyclic Peptides

Abstract

Intracellular protein-protein interactions (PPIs) are challenging targets for traditional drug modalities. Macrocyclic peptides (MPs) prove highly effective PPI inhibitors in vitro and can be rapidly discovered against PPI targets by rational design or screening combinatorial libraries but are generally impermeable to the cell membrane. Recent advances in MP science and technology are allowing for the development of “drug-like” MPs that potently and specifically modulate intracellular PPI targets in cell culture and animal models. This article reviews the recent progress in generating cell-permeable MPs that enter the mammalian cell by passive diffusion, endocytosis followed by endosomal escape, or yet unknown mechanisms.

MPs as PPI Inhibitors and Challenges

It is estimated that current drugs (i.e., small molecules and biologics) are effective against only ~20% of all disease relevant human proteins [1]. The other ~80%, which are predominantly proteins involved in intracellular PPIs (see Glossary), are “undruggable” because they do not contain well-defined hydrophobic pockets for small molecules to bind and are inaccessible to conventional biologics (e.g., antibodies), which are too large to enter the cell by passive diffusion. Over the past decade, MPs have emerged as an exciting modality for inhibition of PPIs. MPs in the molecular weight range of 700-2000 are 3-5 times larger than small-molecule drugs and capable of binding to flat protein surfaces with antibody-like affinity and specificity [2]. Moreover, powerful combinatorial library technologies are now available, allowing researchers to generate highly potent and specific MP ligands against essentially any protein target [2-5]. Proteolytic instability, previously considered as a major limitation of peptidyl drugs, is largely mitigated by cyclization and the incorporation of non-proteinogenic amino acids (e.g., N^a-methylated and D-amino acids). In general, monocyclic peptides of ≤10 amino acids and bicyclic peptides of ≤20 amino acids are relatively stable against proteolytic degradation.

The “Achilles’ heel” of MPs is that they are generally impermeable to the cell membrane and, therefore, suffer from the same limitation as proteins with respect to intracellular targets. Although it has been known for decades that some peptides of both natural and synthetic origins, as well as proteins, can enter the cytosol of mammalian cells with varying efficiencies, their mechanism of cell entry was not well understood. This knowledge gap has historically hampered the rational design of cell-permeable MPs as PPI inhibitors. However, as described below, the recent discovery of highly active cell-penetrating peptides (CPPs) and elucidation of their cell entry mechanism have enabled researchers to develop cell-permeable, metabolically stable, potent, and specific MPs against intracellular PPIs. This review provides an overview of the key advancements in the science and technology of CPPs and MPs, the most notable examples of MP inhibitors against intracellular PPIs reported during the past four years (since 2017), and the outstanding questions/future directions of the MP field. For studies prior to 2017, the readers are referred to two previous reviews [6, 7].

Recent Advancements in CPP and MP Science/Technology

CPPs are small peptides (typically 5-30 aa) that enter the cell without causing significant damage to the cell membrane (Box 1). Initially, all CPPs were linear peptides, which undergo rapid proteolytic degradation in vivo and have low cytosolic entry efficiencies (e.g., ~2% for Tat [8]). A breakthrough was the discovery of cyclic CPPs, which are metabolically stable and have vastly improved cytosolic entry efficiencies (e.g., 120% for CPP12 [9]). CPPs enter the cell by endocytic mechanisms followed by endosomal escape, although at high concentrations some of them also directly translocate across the plasma membrane [10,11]. The greater cytosolic entry efficiency of cyclic CPPs stems primarily from their unique ability to efficiently escape the endosome [9, 12]. The latter in turn facilitated the mechanistic investigation of endosomal escape, which had been a long-standing mystery in the field of cell biology [13]. Pei and co-workers discovered that CPPs escape the endosome by inducing budding and collapse of small vesicles from the endosomal membrane [9,13]. Another key advancement is the development of robust assay methods that directly measure the relative and/or absolute cytosolic entry and endosomal escape efficiencies of CPPs, MPs, and other biomolecules [8, 9, 14-17].

Box 1. CPPs and Their Mechanism of Action.

The CPP field emerged three decades ago, when transactivator of transcription (Tat) protein of human immunodeficiency virus (HIV) (in 1988) [89, 90] and Antennapedia homeodomain of Drosophila metanogaster (in 1991) [91] were found to enter the cell. Subsequent studies identified minimal peptide sequences responsible for cellular entry, which evolved into first CPPs – Tat (GRKKRRQRRRPQ) [92] and penetratin/Antp (RQIKIWFQNRRMKWKK) [93]. Since then, hundreds of CPPs have been discovered and applied to transport a variety of cargos, including small molecules, peptides, proteins, and nucleic acids, into the cell both in vitro and in vivo. However, therapeutic application of the linear CPPs is limited by low cytosolic delivery efficiency, susceptibility to proteolytic degradation, and poor PK properties [40]. The discovery of cyclic CPPs, such as CPP9 [cyclo(phe-Nal-Arg-arg-Arg-arg-Gln)] and CPP12 [cyclo(Phe-phe-Nal-Arg-arg-Arg-arg-Gln)] (Figure Ia), has greatly accelerated the therapeutic applications of CPPs, because they not only have vastly improved cytosolic delivery efficiencies (e.g., 120% for CPP12 compared to 2% for Tat) [9], but also are metabolically stable and exhibit good bioavailability and broad biodistribution.

Despite of intense research over three decades, the mechanism of action of CPPs has been a subject of contention since the inception of the CPP field. It is generally established that at lower concentrations, CPPs enter the cell primarily via energy-dependent endocytosis, followed by endosomal escape, while at high concentrations (≥10 μM), some of the CPPs can also directly translocate across the plasma membrane in an energy-independent manner [10-12]. Following endocytosis, CPPs must escape from the endosomal/lysosomal compartments to deliver their cargo into the cytosol (or nucleus). Pei and colleagues recently demonstrated that Tat, cyclic CPP12, and CPM3 (a non-peptidic cell-penetrating molecule) all exit the endosome by a novel vesicle budding-and-collapse (VBC) mechanism [9, 13]. In this mechanism, CPPs bind to the endosomal membrane and cluster into CPP-enriched lipid domains, which subsequently bud off as small vesicles (Figure Ib). The budded vesicles spontaneously and rapidly disintegrate into amorphous lipid/peptide aggregates, thereby releasing the vesicular content into the cytosol. This mechanism reconciles how CPPs can deliver macromolecular cargos (e.g., proteins and nucleic acids) into the cytosol without compromising the endosomal integrity. CPPs of high endosomal escape efficiencies are often amphipathic and conformationally constrained (e.g., by cyclization or α-helix formation). Conformational rigidity increases the binding affinity of CPPs for the endosomal membrane, while amphipathic structures allow CPPs to bind selectively to and stabilize the budding neck (and promote vesicle budding) [12]. It appears that highly efficient CPPs (e.g., cyclic CPPs) escape from the early endosome, whereas CPPs of poor endosomal escape/cytosolic entry efficiencies (e.g., Tat) exit from the later endosome. Early endosomal escape is advantageous for intracellular delivery of biomolecules, as it avoids acid denaturation and enzymatic degradation which may occur inside the late endosome and lysosome.

To achieve cell-permeability, MPs are often conjugated to or integrated with CPPs (see below). In addition to endocytosis/endosomal escape and direct translocation, MPs may also enter cells by passive diffusion or binding to membrane transporters [12]. Still other MPs enter the cell, but their entry mechanisms have not yet been well understood. Below we shall classify the recently developed cell-permeable MPs based on their mechanisms of cell entry.

MPs That Enter the Cell by Passive Diffusion

Most bioactive MPs violate Lipinski’s rules (Ro5), which are a set of criteria for achieving passive membrane permeability and oral bioavailability [18]. Despite violating those rules, some MPs (e.g., cyclosporine A) passively diffuse across cellular membranes and are orally bioavailable. Such MPs usually balance lipophilicity and polar surface area by various combinations of peptide backbone modification (e.g., N^α-methylation), nonpolar side chains, nonproteinogenic building blocks (e.g., D-amino acids), and formation of intramolecular hydrogen bonds [2, 12, 19]. N^α-Methylation and formation of intramolecular hydrogen bonds reduce the number of hydrogen bond donors (HBDs) and/or acceptors (HBAs) that usually impede membrane permeability. These strategies, unfortunately, also limit the structural diversity for target recognition and aqueous solubility [12]. As such, recently reported MPs with passive permeability continue to be dominated by model peptides devoid of any biological activity [20-24] but with a few notable exceptions. Hayashi et al. [25] designed cycloheptapeptidyl inhibitors of nicotinamide N-methyltransferase (NNMT). The most potent inhibitor (compound 1, Table 1) inhibited NNMT activity in HEK293 cells (IC₅₀ = 0.77 μM). Compound 1 features four proline analogs and an N^α-methylation to minimize the number of backbone -NH’s. Although NNMT is not a PPI target, this work nevertheless demonstrates that relatively large MPs (MW > 1200) can be rendered cell permeable by passive diffusion and biologically active. Another class of cell-permeable macrocycles is cyclic peptoids and other peptidomimetics that have a reduced number of backbone - NH’s. Using in silico design, Schneider et al. [26] developed a cyclic peptoid (compound 2) targeting the β-catenin/TCF interaction, a critical modulator of Wnt signaling pathway. The peptoid inhibited Wnt pathway in a luciferase reporter assay (IC₅₀ = 0.11 μM) as well as in a zebrafish model. Sulfono-γ-AApeptides, oligomers of N-sulfono-acylated N-aminoethyl amino acids, represent yet another class of cell-permeable and proteolytically stable N^α-substituted peptidomimetics. They fold into rigid and structurally defined helices with unique helical parameters that may be employed to disrupt α-helix-mediated PPIs [27]. Sulfono-γ-AApeptides demonstrate cell permeability owing to their pronounced helical conformation and a decrease in availability of amide bonds. Cai and co-workers rationally designed left- and right-handed sulfono-γ-AApeptidyl inhibitors of the p53/MDM2 (compound 3) [28, 29] and BCL9/β-catenin PPIs [30]. Given the demand of balancing multiple and often conflicting parameters, it is unclear whether passively permeable MPs will provide a general modality for intracellular targets.

Table 1.

Representative cell-permeable MPs reported since 2017. CPP motifs are highlighted in green color.

Compd No.	Compound structure	Target (KD or IC50 value for target)	Biological activity	Cell entry mechanism	Refs
1		NNMT(IC50 = 1.9 nM for human NNMT)	In vitro (IC50 = 0.77 μM for NNMT inhibition in HEK293 cells)	Passive diffusion	[25]
2		β-catenin (IC50 = 5.4 μM)	In vitro; in vivo (rescued eye development in zebrafish model)	Passive diffusion	[26]
3		MDM2 (KD = 26 nM)	In vitro by stapled analog (activated p53 in U2OS cells at 30 μM)	Potentially passive diffusion	[28]
4		TEAD (KD = 0.7 pM for human TEAD)	In vitro (activated Hippo-associated genes in cardiomyocytes at 30 μM)	Endocytosis	[33]
5		RbAp48 (KD = 8.6 nM)	In vitro (increased p53 level in U2OS cells)	Endocytosis	[34]
6		Tankyrase (KD = 440 nM)	In vitro (decreased Wnt reporter activity in HEK293T cells [IC50 ≈ 50 μM])	Endocytosis	[36]
7		Jun subunit of transcription factor activator protein-1 (KD =2 μM)	In vitro (antiproliferative activity against MCF-7 and ZR75-1 breast cancer cells)	Endocytosis	[37]
8		HDAC (IC50 = 99-303 nM for different HDAC isoforms)	In vitro; in vivo (TGI @ 50 mg/kg in PA-1 and NTERA-2 mouse xenograft models)	Endocytosis	[38]
9		Calcineurin (KD = 16 nM)	In vitro; in vivo (prevented acute lung injury @ 3-5 mg/kg in a mouse model)	Endocytosis	[41]
10		Keapl (IC50 = 153 nM)	In vitro (inhibited Keap1/Nrf2 PPI in ARE reporter-HepG2 cell line [EC50 ≈ 1.1 μM])	Endocytosis	[43]
11		Pin1 (IC50 = 220 nM)	In vitro (inhibited the proliferation of HeLa cervical cancer cells)	Endocytosis	[46]
12		Grb2 SH2 domain (IC50 = 400 nM)	In vitro (inhibited Grb2/Ras signaling in MDA-MB-468 breast cancer cells [IC50 ≈ 15 MM])	Endocytosis	[47]
13		NEMO(IC50 = 1.0 μM)	In vitro (inhibition of TNFα-induced NF-κB activation [IC50 = 10 μM])	Endocytosis	[49]
14		K-, H-, and N-Ras (KD = 21 nM for KRasG12V-GppNHp)	In vitro; in vivo (TGI @ 1-5 mg/kg in A549 and H358 mouse xenograft models)	Endocytosis	[50]
15		PDZ domain of CFTR-associated ligand (KD = 6 nM)	In vitro; ex vivo (improved ion channel activity of F508del-CFTR by 3-fold [EC50 ≈ 10 nM])	Endocytosis	[51]
16		Ras (KD = 2 μM for HRas)	In vitro (antiproliferative activity in Ras mutant cancer cells)	Endocytosis	[53]
17		Bcl2A1 (apparent Ki <0.1 nM)	In vitro (covalent modification of Bcl2A1 in U937 lymphoma cells)	Unknown	[68]
18		β-catenin (KD = 123 nM)	In vitro (antiproliferative effect in Wnt-addicted colorectal cancer cells DLD-1 and SW480)	Unknown	[69]
19		MDM2/X (KD = 0.87 and 3.9 nM for MDM2 and MDMX)	In vitro (antiproliferative effect in HCT116 cells [IC50 ≈ 25 μM])	Unknown	[71]
20		β-catenin (KD = 91 nM)	In vitro (inhibition of Wnt signaling in HEK293T reporter cell line [IC50 = 8 μM])	Unknown	[74]
21		Keap1 (KD = 35 nM)	In vitro (increased the expression of Nrf2-regulated genes in HEK293T cells)	Unknown (potentially endocytosis)	[79]
22		UEV domain of TSG101 protein (KD = 11.9 μM)	In vitro (viruslike particle budding assay in HEK293T cells [IC50 ≈ 2 μM])	Unknown	[83]
23		KRasG12D (EC50 = 22 nM)	In vitro; in vivo (TGI @40 mg/kg in PANC-1 mouse xenograft model)	Unknown	[84]
24		K48Ub tetramer (KD = 9 nM)	In vitro; in vivo (TGI @1 mg/kg in mouse model)	Unknown	[85]
25		LC (KD = 120 and 192 nM for LC3A and LC3B, respectively)	In vitro (inhibited autophagy in HeLa cervical cancer cells)	Unknown	[87]
26		MyD88	In vivo (reduced disease severity @ 10 mg/kg in sclerosis mouse models)	Unknown	[88]

MPs That Enter the Cell by Endocytosis

To increase their cell-permeability, MPs have been covalently attached to CPPs to form CPP-MP conjugates or integrated with CPPs to produce CPP-MP hybrids (Figure 1) [2, 6, 7]. The advantage of CPP-MP conjugates is that they are straightforward to design, as each unit (CPP, MP, or linker) functions independently (usually) and can be individually optimized. Moreover, the same CPP may be used to deliver many different MPs. CPP-MP hybrids, on the other hand, allow the dual use of structural elements for both cell entry and target engagement. Compared to CPP-MP conjugates, CPP-MP hybrids are more challenging to design; but once successfully designed, they usually have simpler structures, making them easier and less expensive to produce. Both CPP-MP conjugates and hybrids are expected to enter cells by endocytosis followed by endosomal escape, at least at lower concentrations [12]. However, one should always bear in mind that the combination of a CPP with a MP may mutually interfere with each other’s function or result in off-target binding.

Figure 1.

Schematic representation of cargo delivery modes of CPPs. A cell-impermeable peptidyl cargo (blue) is conjugated to a CPP motif (green) to form cell-permeable CPP-MP conjugates or hybrids that can reach and modulate their intracellular PPI targets. (a) Generation of CPP-MP conjugates by covalent attachment of a linear or cyclic CPP to a cargo (linear, cyclic, or stapled peptide). (b) Formation of CPP-MP hybrids by insertion of a small peptidyl cargo into a cyclic CPP (endocyclic delivery). (c) Fusion of a cyclic CPP with an MP cargo results in a bicyclic hybrid (bicyclic delivery). (d) Reversible cyclization of a CPP-cargo fusion peptide into a monocyclic peptide via a disulfide bond. (e) Reversible bicyclization of a CPP-cargo fusion peptide into a bicyclic peptide via two disulfide bonds.

CPP-MP conjugates

Linear CPPs including Tat, penetratin, and oligoarginines remain popular choices for intracellular delivery of MPs including stapled peptides (SPs). SPs are stabilized α-helical peptides with their side chains on the same helical face (usually at positions i and i+4 or i+7) covalently crosslinked and represent a special type of MPs [31]. Because of their increased target binding affinity, proteolytic stability, and in some cases intrinsic cell permeability, SPs have become a popular modality for inhibiting intracellular PPIs mediated by α-helical motifs (see below). However, like MPs, most SPs are either impermeable to the cell membrane or have very limited cell permeability.

Depending on the specific structure of MPs/SPs, a CPP may be covalently attached to the N-terminus, C-terminus, a side chain of the MP/SP, or in the case of SPs, to the stapling unit. For example, Tat has been conjugated to linear and cyclic peptidyl inhibitors against the dimerization of Leishmania infantum trypanothione reductase (TryR) [32], a cyclic two-helix inhibitor against the interaction between transcription factor TEAD and its co-repressor VGL4 (compound 4, Table 1) [33], and a bicyclic peptidyl inhibitor targeting the interface between RbAp48 (K_D = 8.6 nM) and scaffold protein MTA1 (compound 5) [34]. All three Tat-MP conjugates displayed biological activities in cell culture. Similarly, penetratin and nonaarginine were employed to improve the cellular entry of bicyclic peptidyl inhibitors of Grb7-SH2 domain [35] and MP inhibitors of tankyrase (compound 6) [36]. Baxter et al. [37] conjugated Tat and a nuclear localization signal (NLS) to the C-terminus of a lactam-stapled peptidyl inhibitor of Jun (K_D = 2 μM) (20-NLS-Tat, compound 7, Table 1). The combination of a CPP with NLS promoted efficient cellular uptake and nuclear localization in breast cancer cells resulting in the decrease in cell proliferation. Conjugation of oligoarginines as short as R3 to the stapling unit of SPs against protein kinase CK2 [38] or the N-termini of SP inhibitors against histone deacetylase (HDAC) [39] greatly increased their cell permeability. One of the most active HDAC inhibitors, 16cyc-HxA (compound 8), resulted in cell-cycle arrest and apoptosis of ovarian teratocarcinoma PA-1 cells with minimal toxicity to normal cells. Intraperitoneal injection of the peptide (at 50 mg/kg) inhibited tumor growth by 80-90% in PA-1 and testicular embryonic carcinoma mouse xenograft models. However, in vivo application of linear CPPs is often complicated by their poor proteolytic stability and low cytosolic delivery efficiencies (typically <5%) [8, 40].

To overcome the limitations of linear CPPs, cyclic CPPs have recently been used to deliver linear peptides (LPs) [41,42], MPs [43], and SPs [44,45]. Dougherty et al. [41] discovered a nonapeptidyl inhibitor against the NFAT-binding site of calcineurin as a novel anti-inflammatory agent. Conjugation of cyclic CPP9 to its C-terminus generated CNI103 (compound 9, Table 1) as a metabolically stable (serum t_1/2 >24 h), cell permeable, and potent inhibitor of the calcineurin/NFAT interaction in vitro (K_D = 16 nM). Intranasal or intravenous administration of CNI103 (at 3-5 mg/kg) prevented lipopolysaccharide-induced acute lung injury in a mouse model. Similarly, conjugation of a previously reported MP inhibitor of the Kelch-like ECH-associated protein 1 (Keap1; K_D = 18 nM) to CPP9 increased its cytosolic entry efficiency by 98-fold (compound 10) [43]. Finally, Dougherty et al. [44] conjugated CPP9 to a dozen different SPs, at their N-terminus, C-terminus, or stapling unit. The resulting CPP-SP conjugates were all highly cell permeable, irrespective of the intrinsic permeability of the SPs. These studies suggest that conjugation with a cyclic CPP represents a general solution to the intracellular delivery of LPs, MPs, and SPs.

CPP-MP hybrids

Small peptidyl cargos (≤5 aa) may be inserted into a cyclic CPP to form monocyclic CPP-MP hybrids (Figure 1b, endocyclic delivery). Bedewy et al. [46] designed a cycloheptapeptidyl inhibitor of peptidyl-prolyl isomerase Pin 1 as the smallest, biologically active CPP-MP hybrid known to date (compound 11, Table 1). MP 11, in which the naphthylalanine (Nal) residue performs the dual function of Pin1 binding and cell penetration. This method is, however, less effective for larger cargos, which greatly decrease the conformational rigidity and cell entry efficiency of cyclic CPPs. To accommodate larger peptidyl cargos, Wen et al. [47] introduced a D-Pro-L-Pro motif, which induces β-hairpin structures [48], into a CPP-MP hybrid. It was anticipated that a CPP containing the D-Pro-L-Pro motif would be conformationally constrained and retain high cell entry efficiency, even after the insertion of larger cargos. Through this approach, the investigators developed several cell-permeable MPs of large ring sizes, including a cyclodecapeptidyl inhibitor against the Grb2 SH2 domain (compound 12), although the generality of this approach remains to be demonstrated.

A more general approach to cell-permeable CPP-MP hybrids involves fusing cyclic CPPs and MPs to form bicyclic peptides, in which one ring is primarily responsible for cellular uptake while the other for target binding (Figure 1c, bicyclic delivery). Rhodes et al. [49] designed a bicyclic peptide library, in which one ring contained randomized peptide sequences for potential binding to targets of interest, while the other ring featured a family of CPP motifs for cell penetration. Screening of the library against NF-κB essential modulator (NEMO) identified a cell-permeable inhibitor (compound 13, Table 1) of the NEMO/IKKβ interaction (IC₅₀ = 1.0 μM). Interestingly, alanine scanning revealed that three of the CPP residues contributed significantly to NEMO binding. Buyanova et al. [50] discovered a metabolically stable (serum t_1/2 >24 h), cell-permeable, and potent bicyclic peptidyl pan-Ras inhibitor, B4-27 (compound 14; K_D = 21 nM for KRasG12V-GppNHp), in which the cell-penetrating and Ras-binding sequences are fully integrated. SAR studies demonstrated that several residues of B4-27 (e.g., D-Arg, D-Phe) were critical for both KRas binding and cellular uptake. B4-27 binds selectively to the active (GTP-bound) form of K-, H-, and N-Ras isoforms, blocks their interaction with downstream effectors including Raf and PI3K, and induces apoptosis of Ras-mutant cancer cells. In A549 and H358 lung cancer mouse xenograft models, B4-27 effectively suppressed tumor growth at ≤5 mg/kg daily doses. Since the cargo does not affect the structure of the CPP ring, bicyclic delivery can in principle accommodate cargos of any size or sequence.

Some proteins require linear peptidyl ligands and cannot be targeted by the above CPP-MP hybrids. To this end, a reversible cyclization strategy has been developed to enhance the cell permeability and metabolic stability of peptidyl therapeutics (Figure 1d) [51, 52]. In brief, a CPP motif is conjugated to the N- or C-terminus of a linear peptidyl ligand and the fusion peptide is then cyclized via an intramolecular disulfide bond. Upon entering the cytosol of a target cell, the disulfide is reduced by intracellular glutathione to regenerate the linear peptide as the active drug. Dougherty et al. [51] designed a highly potent, selective, cell-permeable, and metabolically stable peptidyl inhibitor, PGD97 (compound 15), against the PDZ domain of CFTR-associated ligand (CAL; K_D = 6 nM). PGD97 increased the cell surface expression of CFTR by blocking the CAL/CFTR interaction and CAL-mediated lysosomal degradation of CFTR. When tested against cystic fibrosis patient-derived primary lung epithelial cells, PGD97 improved the ion channel activity of F508del-CFTR by ~3-fold with an EC₅₀ value of ~10 nM. It is notable that four of the five CPP residues in PGD97 contribute significantly to CAL binding. A reversible bicyclization strategy was also employed to produce a bicyclic peptidyl inhibitor of the NEMO/IKKβ interaction (Figure 1e) [52].

Proteomimetics

Conformationally stabilized peptides mimicking secondary and tertiary protein structures have been developed as PPI inhibitors. Arora and colleagues reported proteomimetic inhibitors of various PPIs termed crosslinked helix dimers (CHDs) by generating synthetic stabilized coiled coil mimics [53-56]. Some CHD examples include CHD^Sos-5 (compound 16), which mimics Ras effector protein Son-of-Sevenless (Sos) and exhibited selective toxicity towards Ras-mutant cancer cells [53], and CHD3^NEMO, which mimics NEMO and blocks its interaction with vFLIP (K_i = 6.9 μM) [54], CHD3^NEMO significantly reduced tumor growth in a PEL mouse xenograft model at 20 mg/kg via intraperitoneal injection. The authors proposed that CHDs are taken up by macropinocytosis and the uptake rate is dependent on the macropinocytic activity of the cells [55]. The arginine-rich nature of CHDs likely promotes both cellular uptake and endosomal escape. The same group also reported a proteolytically stable cell-permeable α₃β chimeric Sos helix mimic that covalently targets RasG12C [56]. Scientists at FogPharma are also designing cell-penetrating miniproteins (CPMPs) to target intracellular PPIs such as β-catenin and Ras (https://fogpharma.com/#science).

MPs That Enter the Cell by Yet Unknown Mechanisms

Stapled peptides

Some SPs are intrinsically cell permeable, although their exact mechanism of cellular entry remains unknown. A hydrocarbon-stapled peptide against the MDM2/p53 interaction, ALRN-6924, is currently in phase II clinical trials [57]. This finding has inspired researchers to search for cell-permeable SPs against other intracellular PPI targets. These efforts have led to highly potent, biologically active (in cellular assays) hydrocarbon-stapled peptidyl inhibitors against a variety of intracellular PPI targets including Mcl-1 [58], the nuclear receptor coactivator 1 (NCOA1) [59], small GTPase Rab25 [60], cytokinesis regulating protein UNC119 [61], eukaryotic translation initiation factor 4E (eIF4E) [62], and cell-cycle regulator protein Id [63]. In general, hydrocarbon or other hydrophobic staples promote cellular uptake more effectively than hydrophilic ones [64, 65], although excessive hydrophobicity and positive charge may result in membrane leakage [66,67].

Researchers have explored innovative approaches to improving the potency, selectivity, and/or cell permeability of SPs. Fairlie and colleagues [68] appended an acrylamide warhead to an SP against oncogenic protein Bcl2A1 to increase its residence time on the target. The BimBH3 derivative (compound 17, Table 1) entered U937 lymphoma cells, became localized in the mitochondria (where Bcl2 proteins are mainly located), and covalently bound to Cys55 within the BH3-binding site of Bcl2A1. Grossman and colleagues [69] reported an integrative strategy to enhance the cellular uptake of poorly permeable SPs, including N-terminal modification with NLS and substitution of homoarginine for arginine. The optimized peptide against the β-catenin/TCF interaction, NLS-StAx-h (compound 18), bound to β-catenin with a K_D of 123 nM, showed good cytosolic exposure, inhibited Wnt signaling in cell culture, and reduced the proliferation and migration of colorectal cancer cells. Partridge and colleagues [67] generated hydrocarbon-stapled D-peptides to simultaneously improve target binding, proteolytic resistance, and cell permeability. Pentelute and co-workers [70] developed a solution-phase library screening platform to rapidly discover peptidyl PPI inhibitors containing non-canonical amino acids. This platform was applied to identify a series of high-affinity, unnatural ligands that were further engineered into macrocyclic inhibitors of the p53/MDM2 interaction or the dimerization of a HIV capsid protein.

Efforts have also been made to develop alternative, simpler (and less costly) stapling methods. Li et al. [71] reported a dithiocarbamate (DTC)-based stapling method and applied it to generate peptidyl inhibitors of MDM2/X. One such inhibitor, ^DTCPMI (compound 19), showed a 50-fold tighter binding to MDM2 and MDMX (K_D = 0.9 and 3.9 nM, respectively) than its linear counterpart. ^DTCPMI readily entered the mammalian cell and reduced the viability of cancer cells in a p53-dependent manner (IC₅₀ = 25 μM). Zhang et al. [72] developed a solid-phase synthesis of bisthioether-stapled peptides of various architectures, including single-turn, double-turn, and double-stapled peptides. Pentelute and co-workers [73] demonstrated that perfluoroarene-based macrocyclic variant of Bim BH3 SP can cross the blood-brain barrier.

β-Sheet mimetics

Compared to α-helical peptides, β-sheet structures remain relatively unexplored as PPI inhibitors. Most β-sheet mimetics are β-hairpin peptides consisting of two antiparallel β-strands connected by a turn. Grossman and co-workers [74] designed β-sheet mimicking bicyclic peptides to target the β-catenin/TCF interaction. The bicyclic β-sheet mimetic contained a D-Pro-L-Pro motif and was bicyclized using a butene bridge resulting in a pronounced β-sheet character, which may contribute to cell penetration. The most potent peptide A-b6 (compound 20, Table 1) bound β-catenin (K_D = 91 nM) and inhibited Wnt signaling in a cell-based assay. The importance of a β-strand conformation was also demonstrated by Nadal-Bufi et al. [75], who grafted peptidyl inhibitors that block the oligomerization of lactate dehydrogenase 5 (LDH5) into a cyclic cell-penetrating β-hairpin scaffold to generate a cell-permeable LDH5 inhibitor of modest potency (IC₅₀ = 2.5 μM). In addition to the β-strand conformation, the inhibitor contained five arginine residues and hydrophobic aromatic residues (e.g., L-Phe and L-Trp), which were important for its cell-penetrating properties. These structural features suggest that the LDH5 inhibitor likely enters the cell by endocytosis and the β-hairpin structure presumably constrains the CPP motif into a proper membrane-binding conformation(s), analogous to that of the Grb2 SH2 domain inhibitor (compound 12) [47].

Cyclotides

Cyclotides are plant-derived MPs of ~30 aa containing a head-to-tail cyclized backbone and multiple intramolecular disulfides [76]. Thanks to their exceptional stability and modest cellular permeability, cyclotides have been used as a scaffold for grafting of peptidyl inhibitors. Recent examples of this grafting strategy include a cell-permeable kalata B1-based cyclotide inhibitor of 14-3-3 proteins [77] and a cell-permeable MCoTI-II-based inhibitor of kallikrein-related peptidase 4 (K_D = 0.1 nM) [78]. Another interesting strategy combined MCoTI-II grafting with the introduction of a CPP motif, FΦRRRR, resulting in a cell-permeable inhibitor of the Keap1/Nrf2 interaction, MCNr-2c (compound 21) [79]. Although the specific mechanism of cellular uptake of cyclotides remains unknown, there is some evidence that internalization of cyclotides occurs via multiple endocytic pathways and is dependent on the overall positive charge of the native sequences [78, 80-82].

Other cell-permeable MPs

Screening of combinatorial libraries has resulted in many highly potent MPs against PPI targets, some of which are cell permeable by yet unknown mechanisms. For example, Tavassoli and colleagues [83] reported a cell-permeable cyclooctapeptidyl inhibitor of the p6/UEV interaction (compound 22, IC₅₀ = 5.4 μM) which reduced the release of virus-like particles in a cell-based assay. Through screening of a phage-display library and optimization by medicinal chemistry, Sakamoto et al. [84] developed a cell-permeable bicyclic peptide (KS-58, compound 23) which bound KRasG12D with an EC₅₀ of 22 nM, inhibited Ras signaling and the proliferation of KRasG12D-mutant cancer cells in vitro, and suppressed tumor growth in a PANC-1 mouse xenograft model. By employing the RaPID system, Rogers et al. [85] discovered a cycloundecapeptide (compound 24) which bound to K48-linked ubiquitin (^K48Ub) chains, caused the accumulation of Ub conjugates, induced apoptosis of cancer cells in vitro, and inhibited tumor growth in a mouse model. Similarly, Nawatha et al. [86] identified MPs that potently and specifically bind to ^K48Ub chains (e.g., K_D = 6 nM for Ub4ix) but not ^K11Ub or ^K63Ub chains or mono-Ub. The MPs entered HeLa cells, protected Ub chains from recognition by deubiquitinases or the proteasome, and induced apoptosis of HeLa cells. Gray et al. [87] screened mRNA display libraries and discovered MP inhibitors of the autophagy protein LC3. The most promising candidate, compound 25, bound LC3A and LC3B with nanomolar affinity, entered the cytosol, and inhibited autophagosome formation in HeLa cells. Finally, Dishon et al. [88] demonstrated that N-terminal myristoylation of an MP inhibitor of myeloid differentiation primary response 88 (MyD88), c(myD4-4), rendered it cell permeable and orally bioavailable. Oral administration of Myr-c(MyD4-4) (compound 26) ameliorated disease severity in a mouse model for multiple sclerosis. These MPs violate most of Lipinski’s rules [18] and are not expected to passively diffuse across the cell membrane, suggesting that they may enter the cell by endocytosis and/or binding to membrane transporter proteins [12].

Concluding Remarks and Future Perspectives

Recent advances in MP science and technology have allowed the development of cell-permeable, metabolically stable, potent, and specific MPs against a variety of intracellular PPI targets. Some of these MPs have demonstrated in vivo efficacy, bioavailability, and suitable pharmacokinetics (PK) in animal models. We expect MPs to become a general modality for inhibition of intracellular PPIs and some of the MPs to reach the clinic/market during this decade. Among the different types of MPs, CPP-MP conjugates are most straightforward to design, since highly active CPPs are already available and specific MP ligands against PPI targets can be rapidly discovered by screening MP libraries. These molecules are, however, relatively complex and require lengthy chemical synthesis and usually purification by HPLC. Future innovations in their synthesis and non-chromatographic purification would facilitate their development as therapeutics. On the other hand, CPP-MP hybrids, MPs of passive permeability, and SPs have simpler structures and are easier to produce but are more challenging to design. Elucidation of the mechanism of cell entry (for SPs and other MPs) and improved understanding of the structure-activity relationship of cell penetration should help design the latter classes of MPs. Although cyclic CPPs are orally bioavailable on their own [9], oral bioavailability has rarely been demonstrated for biologically active MPs. Achieving therapeutically relevant levels of oral bioavailability for functional MPs would further unleash the power of this class of molecules. Finally, cell-permeable MPs (by endocytosis) appear to exhibit unique pharmacokinetics [41] that is inadequately described by current PK models, which were developed for small molecules and conventional biologics. New analytical methods and PK models need to be developed for this class of drugs. For MPs of passive permeability, the challenge continues to be engineering biological function into these hydrophobic molecules.

Figure I.

Schematic representation of the structures and mechanism of action of CPPs. (a) Structures of cyclic CPP9 and CPP12, two highly efficient CPPs. (b) CPPs (green stars) enter the cell primarily via endocytosis at low concentrations (<10 μM) but can also undergo direct translocation at high CPP concentrations (typically ≥10 μM). Inside the endosome, CPPs bind to the endosomal membrane and cluster into CPP-enriched lipid domains, which subsequently bud off as small vesicles. The budded vesicles disintegrate into lipid/peptide aggregates releasing the intraluminal contents into the cytosol. The molecular mechanism of direct translocation remains poorly defined.

Outstanding Questions.

Can MPs achieve tissue selectivity?

Can MPs be orally bioavailable?

How are MPs eliminated from the body?

Highlights.

• Macrocyclic peptides are an emerging class of drug modality for PPI inhibition

• Rational design and library screening are established to generate lead peptides

• Membrane permeability can be attained by passive diffusion or endocytosis

• Integration of membrane permeability and target binding is essential

Glossary

Alanine scanning

a mutagenesis method during which an amino acid is mutated to alanine to determine its contribution to a given function

Apoptosis

a process of programmed cell death used by multicellular organisms

Combinatorial Library

a large collection of structurally diverse compounds prepared from a small set of building blocks

Cytosolic Entry Efficiency

the ratio of intracellular over extracellular concentration of a biomolecule (e.g., a CPP)

Endocytosis

an energy-dependent process occurring at the cell surface and involving generation of small vesicles that transport cargo molecules into the cell

Direct Translocation

an energy-independent process by which a molecule enters the cytosol by directly traversing the plasma membrane

Endosomal Escape

a process by which a molecule exits the endosome after endocytic uptake

mRNA display

a technology that enables in vitro selection and directed evolution of peptides/proteins that are covalently linked to their coding mRNAs

Nuclear localization signal (NLS)

a short peptide sequence (e.g., PKKKRKV) that mediates direct import of proteins (or other cargo) into the nucleus

Oral Bioavailability

fraction of an orally administered drug that reaches systemic circulation; usually positively correlated with membrane permeability

Passive Diffusion

an energy-independent process by which a molecule can cross the plasma membrane

Peptidomimetic

a compound that mimics the mode of action of a bioactive peptide but contains various modifications that improve the metabolic stability, bioavailability, and/or target binding affinity

Peptoid

an oligomer of N-alkylated glycines

Protein-Protein Interaction (PPI)

physical interaction between two or more proteins that execute and/or regulate various biological processes

RaPID system

an in vitro system that combines mRNA display with flexible in vitro translation (FIT) to generate and screen large peptide libraries