Emerging Methods and Design Principles for Cell-Penetrant Peptides

Abstract

Biomolecules such as antibodies, proteins and peptides are important tools for chemical biology and leads for drug development. They have been used to inhibit a variety of extracellular proteins, but accessing intracellular proteins has been much more challenging. In this review, we discuss diverse chemical approaches that have yielded cell-penetrant peptides and identify three distinct strategies: masking backbone amides, patterning of guanidinium groups, and amphipathic patterning. We summarize a growing number of large data sets, which are starting to reveal more specific design guidelines for each strategy. We also discuss advantages and disadvantages of current methods for quantifying cell penetration. Finally, we provide an overview of best-odds approaches for applying these new methods and design principles to optimize cytosolic penetration for a given bioactive peptide.

Graphical Abstract

1. Introduction

The cell’s defenses against foreign biomolecules have been refined by billions of years of evolution. Overcoming these defenses is challenging, but a growing number of antibodies, proteins, peptides have been delivered to the inside of the cell.^[1–3] With intracellular protein-protein interactions emerging as tantalizing targets, there has been a renewed push to find those rare molecules that are large enough to block them but can still access the cytosol.^[1,4]

Fundamental work in this area has focused on the cell penetration of short peptides, especially cell-penetrating peptides (CPPs). CPPs comprise several families of peptides that not only penetrate cells, but transport covalently attached “cargo” into cells.^[5] The specific mechanisms of direct translocation, endosomal uptake and endosomal escape for CPPs remain an active area of investigation, and their applications for cargo delivery continue to be explored. Both of these areas have been reviewed extensively.^[5–12]

Despite the relative maturity of the field of CPP-mediated transport, the independent goal of making bioactive molecules intrinsically cell-penetrant, without attachment to a discrete CPP, is still elusive. Peptides are a major focus of this effort, because they are easily synthesized and modified, and because they continue to be prominent among new bioactive molecules. Several classes of intrinsically cell-penetrant peptides have been discovered, including many that do not resemble typical CPPs. Still, the question remains: are there general principles for improving cell penetration for a given peptide?

In this review, we discuss current approaches to improving the cell penetration of bioactive peptides, and extract general principles about how charge, hydrophobicity, and structure influence cell penetration. We focus on short peptides, but similar questions can be asked of other molecules including larger drug-like molecules, macrocycles, proteins, and oligonucleotides. A major obstacle in this field has been the difficulty of measuring cell penetration quantitatively and distinguishing between cytosolic and endosomal localization. Thus, we also provide an overview of commonly used assays for comparing cell penetration.

1.1. Cell-penetrating peptides (CPPs)

While this review focuses on making bioactive peptides more cell-penetrant without attaching a discrete CPP sequence, that effort is directly informed by insights gained from over twenty years of investigations of CPPs. CPPs are short peptides, usually between 5 and 30 amino acids, that can deliver themselves and attached cargo into cells.^[10] The most commonly used and well-understood CPP is the polycationic 11-mer peptide known as Tat. Tat was derived from the HIV protein trans-activator of transcription, after investigations into how this protein was able to transduce into cells.^[13,14] Other CPPs include penetratin, pVEC, and polyarginine; the biophysical properties of prominent CPPs are summarized in Table 1.^[15–17]

Table 1.

Commonly used CPPs and their properties

Name	Sequence	Class	MW*	pI	Charge at pH 7	Ref.
Tat	YGRKKRRQRRR	Cationic	1558.9	12.8	+9	[14]
R9	RRRRRRRRR	Cationic	1422.7	14	+10	[17]
Penetratin	RQIKIWFQNRRMKWKK	Cationic	2245.8	14	+8	[15]
DPV3	RKKRRRESRKKRRRES	Cationic	2211.6	12.6	+11	[52]
pVEC	LLIILRRRIRKQAHAHSK	Amphipathic	2208.7	14	+7.2	[16]
TP10	AGYLLGKINLKALAALAKKIL	Amphipathic	2181.8	11.2	+5	[53]
Pep-1	KETWWETWWTEWSQPKKKRKV	Amphipathic	2847.2	10.8	+4	[54]
MAP	KLALKLALKALKAALKLA	Amphipathic	1876.5	14	+6	[55]
p28	LSTAADMQGVVTDGMASGLDKDYLKPDD	Amphipathic	2913.2	3.7	−3	[56]
Transportan	GWTLNSAGYLLGKINLKALAALAKKIL	Amphipathic	2840.5	11.2	+5	[57]
TP2	PLIYLRLLRGQF	Hydrophobic	1487.8	12.1	+3	[58]
C105Y	CSIPPEVKFNKPFVYLI	Hydrophobic	1993.4	9.9	+1.9	[59]
KFGF	AAVLLPVLLAAP	Hydrophobic	1146.5	14	+1	[60]
Pep-7	SDLWEMMMVSLACQY	Hydrophobic	1806.2	3.9	−1.1	[61]

^*Molecular weight was calculated with a free N-terminus and a C-terminal amide.

Extensive work has focused on understanding the chemical and biophysical properties of CPPs that affect their internalization. It was shown early on that having many (typically, 6 or more) positively charged amino acids, especially arginine, promotes cell penetration.^[17–19] Later studies showed that guanidinium groups play an important role in cell uptake by binding cell membrane components and sulfated proteoglycans. Structure-activity data have demonstrated that the most efficient CPPs have specific three-dimensional patterning of cationic and hydrophobic groups.^[19–22]

While the structure-penetration relationships of CPPs are helpful guidelines, the ultimate efficiency of a specific CPP-cargo conjugate remains hard to predict. Different cargoes, fluorophores and other attached groups have been shown to affect mechanism of uptake, subcellular localization, and toxicity. In a particularly informative case, Ottinger and co-workers showed that C-terminal fusion of negatively charged sequences to polyarginine led to vesicular localization, unlike the free CPP which showed greater cytosolic localization.^[23] A similar phenomenon was reported by Cardoso and co-workers, who observed that Tat fused to a protein, Tat-Cre, was primarily localized to endosomes, while Tat fused to peptides (less than 50 amino acids) had greater localization in the cytosol.^[24] The effects of dyes on cell penetration and subcellular localization are similarly CPP- and cargo-dependent.^[25,26]

All the findings on the properties of CPPs have recently been incorporated into databases and software for predicting extent of cell penetration for a given CPP.^[27–30] General trends that guide these predictions include bias towards high content of arginines and lysines, and low degrees of predicted secondary structure.^[28] While the effects of attached cargoes, dyes and other functional groups remain difficult to account for, one can envision incorporating more data on cargo and dye effects in the near future. Thus, prediction algorithms for CPP-mediated transport may soon accelerate the optimization process for attaching CPPs to otherwise impermeant peptides, proteins or oligonucleotides.

1.2. Mechanisms of cell penetration

Size and polarity are the major factors that determine the mechanism of cell penetration for peptides and other molecules (Figure 1).^[31] Small, relatively non-polar molecules typically penetrate the plasma membrane through passive diffusion.^[36] The hallmarks of a passive diffusion mechanism are concentration dependence, rapid kinetics, and energy independence. Some larger molecules can also enter cells by an energy-independent mechanism – this has been prominently observed for some CPPs when delivered at high micromolar concentrations.^[11] Evidence from experiments with model membranes implies that this process occurs via a direct translocation mechanism.^[32–35] In this mechanism, the molecule binds the membrane or membrane-associated components such as sulfated sugars in the glycocalyx, and then either inserts into the membrane or otherwise causes local membrane disruption.^[37,38] Several reports have provided evidence for distinct mechanisms of translocation, including a carpet-like model of membrane disruption,^[39] nucleation of a transient pore,^[40,41] or formation of an inverted micelle.^[42]

Figure 1. Mechanisms of cell penetration.

Passive diffusion is the predominant process by which smaller, non-polar molecules diffuse into the cytosol. Molecules diffuse into the cell membrane and out into the cytosol through a concentration-dependent, equilibrium-driven process.^[31] Some larger molecules can bind the plasma membrane and/or membrane-associated components such as the glycocalyx, and then enter the cell through direct translocation across the membrane.^[32–35] Endocytosis is the predominant process by which larger peptides are taken up by cells. This energy-dependent process can be summarized in three steps: first, molecules bind the cell membrane or membrane-localized receptors; next, binding deforms the membrane leading to cell uptake via one of several endocytosis pathways; and finally, molecules escape from endosomes into the cytosol.^[11]

Many larger molecules penetrate cells using an endocytosis-dependent mechanism. The current model for this mechanism of entry involves three phases: membrane binding, endosomal uptake, and endosomal release (Figure 1). First, peptides interact with the cell membrane by binding membrane components, including phospholipids, protein receptors, or cell surface proteoglycans.^[43–46] Then, peptides are endocytosed into the cell in an energy-dependent process. Different endocytosis pathways can be involved in uptake, including clathrin-mediated endocytosis, caveolin-mediated endocytosis, and micropinocytosis.^[47] Finally, peptides escape endosomes and are released into the cytosol. The critical step of endosomal release is poorly understood, largely due to the difficulty of measuring extent and timing of endosomal release.^[48] Recent work by Pei and co-workers has proposed a vesicle budding and collapsing model, where curvature of the endosomal membrane leads to the budding of microvesicles, which then collapse to release their contents into the cytosol.^[35]

Many CPPs, including Tat, penetratin, and polyarginine, are internalized through direct translocation at higher concentrations (above 5–10 micromolar), and through one or more endocytosis pathways at lower concentrations.^[11,49] However, such generalizations about mechanism are not absolute. Several factors have been shown to affect mechanisms of entry depending on the type of CPP and the cargo attached.^[23,24] The presence of serum and membrane-associated proteoglycans have also been shown to affect the mechanisms and extent of uptake.^[50,51] As we discuss different classes of cell-penetrant peptides beyond the CPPs listed in Table 1, we will broadly refer to passive diffusion, direct translocation, and endocytosis mechanisms as likely routes of entry. However, any assumptions about predominant entry mechanism must be tested explicitly for each individual cell-penetrant peptide.

2. Three distinct strategies for improving cell penetration

Beyond simple conjugation to CPPs, many chemical approaches have been used to render bioactive peptides more intrinsically cell-penetrant. These include introduction of conformational constraints such as backbone cyclization and side chain “stapling,” and systematic alteration of physicochemical properties. Conformation and physicochemical properties are interdependent and cooperative, so combinations of these two approaches have often produced the best results. Despite many individual successes, only a few approaches have matured past the point of iterative trial-and-error. While these approaches resemble traditional medicinal chemistry, measuring activity in biochemical and cell-based assays is an indirect way to understand cell penetration. Directly measuring cytosolic localization, and using these measurements to maximize cell penetration for a given molecule, is a unique area that warrants separate consideration.

Given the great variety among different chemical approaches, extracting general principles for promoting cell penetration will require integration of data across many different peptide sequences and structures (see section 3 below). But even without these larger data sets, seemingly disparate results are beginning to converge on three overall strategies (Figure 2). These are: (i) masking backbone amides to promote passive penetration, (ii) patterning of guanidinium groups to promote endocytosis, and (iii) amphipathic patterning to promote endocytosis. In this section, we discuss key examples of intrinsically cell-penetrant peptides whose development have helped to define these three distinct strategies.

Figure 2. Representative examples of three different strategies for promoting cell penetration.

a) Small peptides with minimal polar surface area can be passively penetrant. This typically requires cyclic structure and chemical modifications, such as N-methylation (shown in pink), to mask backbone amides from solvent. The compound shown is an N-methylated, cyclic hexapeptide designed as an analogue of somatostatin.^[62] b) Certain structured patterns of guanidinium groups can promote endocytic uptake and release. Depicted is a miniature protein with a helical penta-arginine motif (shown in pink).^[20] c) Larger peptides can also enter cells via endocytosis by virtue of amphipathic patterning. Prototypical examples include all-hydrocarbon stapled helices such as StAx-35 (hydrocarbon staple shown in pink). StAx-35 is a cell-penetrant inhibitor of β-catenin.^[63]

2.1. Masking backbone amides can make small cyclic peptides passively penetrant

Diffusion through cell membranes, or passive cell penetration, has long been studied by medicinal chemists as an important factor for oral bioavailability.^[64–66] Principles from medicinal chemistry have been applied to peptides for decades, including efforts to limit molecular size, polar surface area, and hydrogen bond donors. Still, recent applications of these principles to peptides and peptidomimetics have been uniquely informative.

Historically, natural product peptides with significant oral bioavailability such as cyclosporine have guided the search for general physicochemical principles that promote passive penetration.^[67] Cyclosporine is a prototypical member of a large class of passively penetrant natural product peptides that are head-to-tail cyclic, with N-methylated backbones.^[66] Thus, cyclic N-methylated peptides emerged early-on as a critical proving ground. Specific architectures and internally hydrogen-bonded structures were reported roughly ten years ago by the Kessler and Lokey groups.^[62,68] More recently, Lokey and co-workers have reported larger and more diverse data sets of N-methylated cyclic peptides, allowing extraction of more general principles for promoting passive penetration (see section 3 below). These include small size (4–7 residues, with notable exceptions), reduced polar surface area, and most critically, minimized solvent accessibility of hydrogen bond donors including backbone amides.

Notably, N-methylation is not the only chemical modification that can serve to mask backbone amides. Peptoids (oligomers of N-alkylated glycines) are, on the whole, much more passively penetrant than peptides with similar side chains,^[69] and incremental substitution of peptides with peptoid residues has been shown to increase passive penetration in selected cases.^[70] Moving beyond backbone amide modification, Yudin and co-workers have employed macrocyclization linkers with exocyclic amides that nucleate an extensive hydrogen bond network, resulting in a cyclic peptide with substantial passive penetration.^[71] A separate report described the introduction of an oxadiazole linker into macrocyclic peptides, again nucleating a specific hydrogen bond network that promoted passive penetration.^[72] Shuto and co-workers implemented a cyclopropane-containing macrocycle tether that, for some stereoisomers, promoted an internally hydrogen-bonded structure that improved passive penetration.^[73] Kodadek and co-workers reported improving cell penetration by incorporating two “cycloalanine” residues within a HIV-Revtargeting peptide, though the underlying hydrogen bond network was not identified.^[74] Incorporation of γ-amino acids (such as statine) in a manner that complemented intramolecular hydrogen bonding was also shown to improve passive penetration, and separate work showed a chiral hydrocarbon linker can organize a helical turn within a pentapeptide, promoting passive penetration.^[75,76] Also, several groups have used NMR structures to guide the design of internally H-bonded networks that promote passive penetration.^[77–81] All of these examples are independent means of organizing intramolecular H-bond networks in order to reduce solvent-exposed backbone amides, resulting in passive penetration of peptides in the tetramer-to-heptamer size range.

For each of the examples above, passive penetration is promoted by reducing exposed polar surface area. However, such compounds must also have reasonable aqueous solubility. An intriguing explanation for how these compounds balance these opposite requirements is that they have different structures in aqueous and hydrophobic environments.^[82] It remains unclear whether the majority of passively penetrant peptides are “chameleonic” in this manner, but very recent work has begun to directly test this hypothesis.^[83] Our understanding of macrocycle structures and dynamics is rapidly increasing, so it may soon be possible to learn whether all passively penetrant peptides are also structurally dynamic, or if some are able to passively diffuse through membranes without the need to adopt different structures in aqueous and membrane environments.^[84,85]

2.2. Patterning of guanidinium groups can promote endocytic uptake and release

One way to promote endocytic uptake and endosomal release of bioactive peptides is to incorporate a structured pattern of guanidinium groups. This strategy is directly informed by decades of research on CPPs, as well as work on “supercharged” or “resurfaced” proteins,^[86,87] but there are several examples that highlight its application to bioactive peptides. One example is the development of cell-penetrant miniature proteins by Schepartz and co-workers. Well-structured miniature proteins (36–40 residues) were substituted on polyproline helix or α-helix faces with arginines, producing cell-penetrant versions that maintained target binding.^[88,89] Careful mechanistic studies later revealed a helical penta-arginine motif that promoted uptake of miniature proteins and escape from early endosomes.^[20,90] This motif showed similar effects when grafted onto a small zinc finger domain and a cyclized helix-loop-helix protein,^[20,91] providing evidence that this helical penta-arginine motif may be more generally graftable onto peptides and proteins to improve penetration to the cytosol.^[92–94]

Others have similarly found that installation of specific structured oligo-arginine motifs enhance cytosolic localization. A series of papers by Pei and co-workers describe the development of cyclic and bicyclic peptides with constrained tetra-arginine motifs.^[95–97] These motifs were shown to promote endocytic uptake and release into the cytosol, both alone and as fusions to bioactive peptides. The extent of cytosolic localization was directly compared to the penta-arginine miniature proteins, finding that both accumulate in the cytosol more efficiently than conventional CPPs.^[95] Additional work has applied structured oligo-arginine motifs in other contexts, including polyproline helices, beta-peptides and other foldamers.^[98–102] While it remains to be seen whether all of these examples enter cells via the same mechanisms, one advantage of this strategy appears to be the ability to promote efficient release from endosomes.^[20,95]

It is important to note that only a subset of structured oligo-arginine motifs promote efficient uptake and endosomal release. For instance, in one series of bicyclic peptides, linear stretches of arginines promoted uptake better than arginines placed in the cyclic portion.^[103] Also, while the primary feature of this class is a structured oligo-arginine motif, extensive data support that choice and positioning of hydrophobic residues is also important.^[104] For instance, Pei and co-workers showed that the incorporation of the unnatural hydrophobic amino acid naphthylalanine enhanced cytosolic penetration of their cyclic and bicyclic peptides.^[97] Thus, while the primary property governing cell penetration for these peptides is arginine content and patterning, the choice and position of hydrophobic residues is also important for maximizing uptake and endosomal release.

2.3. Amphipathic patterning can promote endocytic uptake for constrained peptides

Cyclization of peptides through covalent cross-linking of side chains, or “stapling”, has emerged as a powerful way to promote cell penetration.^[105–107] Several different chemistries have produced cell-penetrant stapled peptides, including ring-closing metathesis, cysteine alkylation, cysteine arylation, alkyne-azide cycloaddition, and lactam formation.^[108–119] These examples have much in common. Importantly, the most cell-penetrant stapled peptides tend to have staples made exclusively of hydrocarbons or other hydrophobic groups, with hydrophilic staples such as lactams generally reducing overall cell penetration.^[108]

The best-studied examples of cell-penetrant constrained peptides are the all-hydrocarbon stapled helices. In 2004, Walensky, Verdine and co-workers showed that the BID BH3 peptide can be constrained with an (i, i+4) hydrocarbon staple to yield a cell-penetrant α-helical peptide.^[109] Penetration increased over time for 4 hours and uptake was inhibited at 4 °C, suggesting entry via an endocytic pathway.^{[109,120,121]} Over the last 13 years, extensive development of these and other BH3-derived stapled helices has revealed that optimal cell penetration requires stabilized helical structure, minimal charge, and substantial hydrophobic character. This has been recapitulated for many other stapled helices.^[105,118] For example, Verdine and co-workers found that stapled peptides with an (i, i+7) hydrocarbon staple could not penetrate Jurkat cells until they replaced negatively charged glutamates and aspartates with glutamines and asparagines.^[120] Similarly, Grossmann and co-workers found that double-stapled antagonists of Rab8a required substitutions to decrease negative charge and increase hydrophobicity to improve cell uptake and activity in cell culture.^[119]

This approach and the guanidinium-patterning approach are not completely independent. Arginines and other positively charged groups are often added to amphipathic peptides as a design strategy, with the intention of improving solubility and, potentially, cell penetration.^[123] Grossmann and co-workers decorated a cell-penetrant, stapled helix with arginine and non-natural arginine analogs, revealing some structure-activity analysis of how hydrophobicity and positive charge can cooperate to promote uptake.^[22,63] In separate work, application of a penta-arginine motif directly to hydrocarbon-stapled helices was also shown to improve their penetration to the cytosol.^[124]

It is interesting to note that, in all the examples above, the staple typically stabilizes helical structure, and extent of helical structure correlates with cell uptake (at least to a point; see section 3). A recent publication by Ulrich and co-workers used lactam stapling to show that enforcing helical structure did not promote cell uptake on its own, without a large, continuous hydrophobic surface.^[108] Brock and co-workers showed that a series of hydrocarbon-stapled Hdm2-binding peptides were taken up similarly to their linear precursors.^[125] Clearly, helical structure does not guarantee efficient uptake. Instead, the predominance of helical structures among cell-penetrant stapled peptides may reflect the intrinsic ability of α-helices to present amphipathic character and a large, continuous hydrophobic surface, which are required for efficient uptake.^[126] For example, a recent screen for cell-penetrant variants of an autophagy-inducing peptide was not biased for any specific secondary structure, yet it discovered a cell-penetrant α-helix with an unusual staple configuration.^[112] Recent work has sought to unify the effects of amphipathic patterning on cell penetration, independent from secondary structure.^[127,128] However, most efforts continue to focus on stapled α-helices. It will be an important development in the next few years if other secondary structures, such as beta-hairpins or even non-natural foldamers, can be tailored with amphipathic patterning to enable similar mechanisms of entry as stapled α-helices.^[129,130]

3. Large-scale efforts to define the properties that promote cell penetration

In the previous section, we highlighted examples of three emerging strategies for promoting cell penetration. However, these isolated examples only hint at a set of practical guidelines. New and improved assays for high-throughput quantitation of cell penetration, described in section 4, have allowed the production and analysis of much larger data sets. These data are beginning to reveal more precise guidelines for maximizing cell penetration. In this section, we discuss these efforts as they relate to two of the three strategies.

3.1. High-throughput quantitation helps define limits for passive penetration

The first work to directly address factors that affect cell penetration in a broad and systematic way was published by Kodadek and co-workers starting in 2005.^[131–133] They used a novel assay which detected the cytosolic penetration of steroid-tagged molecules using a transcriptional readout. Over the course of several papers, Kodadek’s team assessed the effects of various physicochemical properties on passive penetration, including differences between linear versus cyclic peptides and peptide versus peptoid backbones. One of the most important findings of this work was that linear and cyclic peptoids were consistently more cell-penetrant than peptides with similar side chains, reflecting the strategy of masking backbone amides to promote passive penetration.^[132] Peptoids as large as octamers were found to be cell-penetrant, but most of this work was performed on peptides and peptoids of 4–6 residues.^[131]

More recent work using libraries of cyclic, N-methylated peptides has refined our understanding of the intrinsic limits of passive penetration. This work by Lokey and co-workers has discovered dozens of examples of cyclic, N-methylated peptides with internally H-bonded structures, and systematic structure-activity relationships that illustrate the features that promote their passive penetration.^{[68,75,82,134–136]} Recently, they used this approach to define an upper size limit for passive penetration at a molecular volume of roughly 1500 ų, which translates to roughly 1000–1200 Da depending on compactness of structure.^[135] They also defined a rough lower limit on hydrophobicity for passively penetrant molecules at (log K_hc/w) > −2, where K_hc/w is the experimentally determined hydrocarbon/water partition coefficient; the upper limit was essentially bounded by solubility.^[135] Strikingly, these results match trends observed for “beyond-rule-of-5” drugs that are orally absorbed, and trends observed for cell-penetrant, non-peptidic macrocycles.^{[64–66,137]}

The determinants for passive penetration of small cyclic peptides are relatively clearly defined, but there remain several challenges. Can this chemical space be successfully mined for lead compounds, especially inhibitors of intracellular protein-protein interactions? Can backbone-shielding networks of internal hydrogen bonds be designed de novo, especially for peptides that organize a desired loop structure?^[138–140] How extensively can polar side chains be incorporated into these structured macrocycles while maintaining passive penetration?^{[70,136,141,142]} Can these peptides be engineered as substrates for endogenous peptide transporters?^[143] What additional features promote not just passive cell penetration, but oral bioavailability and even blood-brain barrier permeability, for this class of molecules?^[77]

3.2. Understanding the determinants of endocytic uptake for amphipathic helices

As described in Section 2, hydrocarbon-stapled helices are prototypical examples of amphipathic patterning to promote cell uptake via endocytosis. Recent efforts have applied large libraries of these molecules to uncover the determinants of cell penetration. In 2015, Verdine and co-workers evaluated a library of 200 peptides in which linear peptides were compared to stapled peptides and “stitched” peptides (peptides with two hydrocarbon staples).^[144] In general, cell penetration was greatly increased with one staple and further increased with two staples, but extent of penetration for any given stapled helix varied depending on staple position. Charge at pH 7.5 was also correlated to cell penetration, with maximal cell penetration observed for stapled peptides with charge between +1 and +7.

In a 2016 report, Bird, Walensky and colleagues sought to determine key factors that govern cell penetration for hydrocarbon-stapled peptides by analyzing point mutation and staple scan libraries.^[145] The internalization of 46 peptides was analyzed using high-content microscopy and custom algorithms to integrate total internalized fluorescence intensity on a per-cell basis. For each peptide whose cell penetration was quantitated, Bird et al. also characterized physical properties, including helicity, pI, and HPLC retention time at pH 7 as an experimental measure of hydrophobicity. These properties were correlated to uptake using principal component analysis, revealing that hydrophobicity was the most important contributor, followed by percent helicity and pI. Their model defined a minimum cutoff for hydrophobicity, defined practically as a minimum retention time for reverse-phase HPLC at pH 7.0, below which stapled helices had little to no cell uptake. Uptake was maximal when stapled helices were between 60 and 87% helical as measured by circular dichroism, but further helical stabilization was not correlated with increased uptake. Positive net charge promoted cell penetration, but helices with pI above 9.7 saw a decrease in cell penetration. Importantly, Bird et al. correlated violation of these upper bounds on hydrophobicity and positive charge with cell-lytic activity. This provides a useful guideline to avoid this known liability of some amphipathic helices. Beyond these overall biophysical properties, they also used a subset of peptides where (i,i+4) staple location was systematically varied to address the effect of staple position. No consistent correlation was observed with location relative to N- and C-termini, but uptake seemed to be maximal when the staple was located at the boundary between the hydrophilic and hydrophobic faces of the helix.

All together, these large data sets for stapled helices have defined some overall parameters to aim for in designing stapled helices with maximal cell uptake. They have also begun to explore the importance of three-dimensional patterning of hydrophobicity, suggesting that extending the hydrophobic surface of the peptide beyond 180 degrees may be critical for cell uptake.^[145,146] Notably, both these studies defined an upper limit to the benefits of positive charge on uptake for stapled helices.^[104,147] Though it can be hard to deconvolute effects of charge, polarity and hydrophobicity using substitution studies, the data to date indicate that the strategy of amphipathic patterning is at least somewhat independent from the strategy of presenting a structured cluster of guanidinium groups. Importantly, this also provides evidence that the underlying mechanisms of endosomal uptake and endosomal release for these two classes of cell-penetrant peptides are likely overlapping, but distinct.^[35,144]

High-throughput studies of cell penetration have been critical, but they are not without their limitations. One important limitation is the ability to distinguish between cell uptake and cytosolic penetration (Figure 1). Studies on stapled helices and other peptides that are internalized via endocytosis have most often relied on measurements of cell uptake rather than cytosolic localization. Careful microscopy and phenotypic assays can be used to correlate uptake with cytosolic localization,^[105,145] but it remains difficult to quantitate cytosolic localization independently from endosomally trapped material.

It is important to note that the predictive power of the studies described in this section required testing many dozens or hundreds of peptides at a time. The guidelines extracted from high-throughput analysis of cyclic, N-methylated peptides may not apply to all passively penetrant peptides. Similarly, guidelines extracted from high-throughput analysis of hydrocarbon-stapled helices may not translate directly to other classes of amphipathic peptides. These issues will need to be addressed by systematic studies on libraries of other types of cell-penetrant peptides. Similar high-throughput studies on peptides with clustered guanidinium groups could also provide improved guidelines for applying this distinct strategy. With the advent of new high-throughput assays for cell penetration, these and other larger data sets are likely to emerge in the next few years.

4. Methods for measuring cell penetration

Many methods have been employed to measure cell penetration. These methods can be roughly categorized according to their readouts: some measure cellular fluorescence, some use transcriptional reporters, and some use mass spectrometry (Table 2). Each method has advantages and disadvantages, as described in detail below.

Table 2:

Summary of methods for quantitating cell penetration of peptides and related molecules.

Method	Label/Tag	Throughput	Readout	Ref.
Fluorescence microscopy	Fluorescein	384-well plate	Fluorescence from dye-labelled peptide	[144,145,157]
	Azide or alkyne, followed by in-cell click reaction to attach dye	35-mm dish		[114]
Confocal microscopy and flow cytometry	Fluorescein	24-well plate		[22,113,152]
	Naphthylfluorescein	12-well plate		[35,96]
Fluorescence correlation spectroscopy (FCS)	Fluorescein	48-well plate		[90]
FACS-FCS	Alexa 488	6-well plate		[156]
Glucocorticoid-receptor-based expression assay	Dexamethasone	96-well plate	Gene reporter-based luminescence	[69,132]
Glucocorticoid-induced eGFP induction/translocation	Dexamethasone	384-well plate	Gene reporter-based fluorescence	[157]
Cell monolayer assays (Caco-2 and MDCK)	None	24-well plate	LC-MS	[175,176]
Parallel artificial membrane permeation assay (PAMPA)	None	96-well plate	UV or LC-MS	[158]
Mass spectrometry-based assay	Biotin	12-well plate	Mass spectrometry	[165]
Protein complementation assays	None	384-well plate	Reporter fluorescence, luminescence	[167,169]
Chloroalkane penetration assay (CAPA)	Chloroalkane	96-well plate	Fluorescence from dye-labelled chase compound	[112]

The most widely used assay involves treatment of cells in culture with dye-labeled molecules, followed by fluorescence microscopy (optimally, confocal fluorescence microscopy). This method has been used on all kinds of molecules, from early work on CPPs until the present day.^[148,149] Initially, methodological issues led to conflicting results. For instance, fixing cells before imaging was shown to produce artifacts including redistribution of dye-labeled peptides.^[150,151] Even with modern improvements, there are three major drawbacks to this method. First, it requires the attachment of a large, hydrophobic dye to the molecule of interest. Few studies have investigated the effects of dye attachment on cell penetration, and dyes also adversely affect solubility for most hydrophobic peptides. One solution would be to add the dye to the molecule of interest in situ, using bio-orthogonal chemistry;^[114] however, these methods have high background in many cells and have not found widespread adoption. A second drawback is that fluorescence microscopy can be low-throughput and qualitative. Recent reports, including some described above, have partially overcome this drawback using high-throughput microscopy, high-content image analysis, and custom algorithms for accurate quantitation.^[144,145] A final drawback is difficulty distinguishing between cytosolic and endosomally trapped dye. Using careful confocal microscopy and colocalization studies, one can get a relative measure of degree of endosomal escape, but these techniques cannot absolutely quantitate the amount of endosomally trapped material or the amount of material that has escaped into the cytosol. The most common complement to fluorescence microscopy is flow cytometry.^[22,113,152] This provides the high throughput and quantitation that microscopy lacks, but flow cytometry does not provide even a qualitative analysis of subcellular localization. Together, fluorescence microscopy and flow cytometry can be used to quantitatively measure cell uptake, and qualitatively assess endosomal release and cytosolic localization.

Several groups have reported improvements to the standard methods of tracking dye-labelled peptides. The first assay to measure cytosolic penetration directly was reported by Langel and co-workers in 2001.^[153] It attached a dye-labelled cargo to a quencher-labelled CPP using a disulfide bond. Upon cytosolic entry the disulfide bond was reduced, thus yielding an increase in fluorescence intensity. Wender and co-workers later reported a similar assay that linked peptides to luciferin using disulfide bonds, so luciferin delivery to the cytosol could be measured.^[154] Taking a different approach, Pei and co-workers used the pH-sensitive dye naphthylfluorescein to report on cytosolic localization using flow cytometry.^[35,96] Ratios of signal for fluorescein-labeled versus naphthylfluorescein-labeled peptides provided a quantitative measure of endosomal trapping and release into the cytosol. Schepartz and co-workers took a different approach, using fluorescence correlation spectroscopy (FCS) to determine the absolute concentration of dye-labelled peptides in femtoliter volumes within the cytosol.^[90,155] A recent approach combined FCS with fluorescence-activated cell sorting (FACS), in order to gate for healthy cells prior to measurement.^[156] Though FCS can report the absolute concentration of peptide in the cytosol, it has not been widely adopted, likely due to lower throughput and the requirement for more specialized microscopy equipment.

Assays relying on transcriptional reporters have also been developed to investigate cell penetration. Kodadek and co-workers developed a luciferase-based reporter assay linked to activation of the glucocorticoid receptor.^[131] This assay used a cell line that expressed the Gal4 DNA-binding domain fused to the glucocorticoid receptor ligand-binding domain (GR) and the VP16 transactivation domain. Molecules were labeled with dexamethasone, so if they reached the cytosol they would release the chimeric transcription factor from Hsp90, this allowing its translocation into the nucleus. A Gal4-responsive firefly luciferase gene was used for quantitation, using a constitutively expressed Renilla luciferase as an internal control.^[131] Schepartz and co-workers published two updated versions of this assay: one used eGFP instead of luciferase as the transcriptionally activated reporter, and another used a GR-fused eGFP to directly monitor translocation from cytosol to nucleus without relying on changes in reporter expression.^[157] All of the steroid-mediated assays give quantitative signals proportional to cytosolic localization rather than endosomal uptake, and were performed in a high-throughput manner (96-well and 384-well plates).

All of the above assays rely on a chemical tag on the molecules of interest, an approach with notable disadvantages. Attachment of an additional chemical group can easily perturb the cell penetration properties of the molecules being studied. Also, large dyes and steroids can also reduce the solubility of the resulting conjugate. Another disadvantage of tag-based assays is that artifacts can arise from molecule degradation. For example, if the tag is released from the molecule outside cells or inside the lysosome, this could lead to an increase in signal without actual cytosolic entry. These drawbacks are common for all tag-based assays but can be controlled for by explicitly studying the effects of the tag, and by using degradation-resistant molecules such as peptoids and D-peptides.

Mass spectrometry offers a promising, tag-free method for quantitating cell penetration – or more commonly, the extent to which molecules cross membranes or tissue monolayers. The parallel artificial membrane permeation assay (PAMPA) is used to screen for molecules that diffuse across artificial membranes, with LC-MS as the most common quantitation method.^[158] Similar quantitation can be applied to high-throughput assays for transport across tissue monolayers, including Caco-2 intestinal epithelial cells and Madin-Darby canine kidney (MDCK) cells.^[159–161] LC-MS can also be used to quantitate internalization of peptides into cultured cells, but without enrichment steps this can be technically demanding.^[162,163] For example, Chassaing and co-workers used mass spectrometry to quantitate the cell penetration of biotinylated peptides.^[164] Cells were incubated with biotinylated peptide, trypsinized to remove surface-bound material, then lysed and spiked with a deuterated analog of the biotinylated peptide as an internal standard. All biotinylated peptides were then captured using streptavidin beads and analyzed by MALDI-TOF-MS.^[164,165] This method allowed for absolute quantification of biotinylated peptide taken up by cells, but required the synthesis of deuterated standards for each molecule to be analyzed. Also, it required technically demanding separation steps to distinguish between cytosolic and endosomally trapped material.

Many protein complementation assays have been reported which can monitor disruption of PPIs in the cytosol.^[166] Several papers have used complementation assays and other two-hybrid assays to quantitate the extent to which bioactive peptides access the cytosol.^[167–170] These assays are amenable to a high-throughput format and require no tag on the molecules of interest. They also exclusively measure cytosolic material since the hybrid proteins are expressed in the cytosol. Drawbacks to using cell-based complementation assays include requiring the design, cloning and production of both PPI partners as fusions to fragments of GFP, luciferase, or another reporter. The approach also requires consistent transient transfection or stable cell lines for each target of interest, and it cannot deconvolute effects of target binding and cell penetration, which can be interrelated.

In 2017, we described a new assay that reports on cell penetration of molecules tagged with small chloroalkane groups.^[112] The Chloroalkane Penetration Assay, or CAPA, uses a stable cell line that expresses the HaloTag enzyme in the cytosol.^[171] HaloTag is a modified haloalkane dehalogenase developed by Wood and co-workers that covalently labels itself with chloroalkanes with fast kinetics and high selectivity.^[172,173] In CAPA, the cells are first pulsed with a chloroalkane-tagged molecule of interest, which covalently blocks HaloTag if it reaches the cytosol. Chasing with a chloroalkane-dye allows quantitation of unreacted HaloTag using flow cytometry in 96-well plates. CAPA is not tag-free, and thus has all the liabilities of a tag-based assay described above. However, the chloroalkane is smaller and less hydrophobic than fluorescent dyes or steroids and is easily attached to peptides and other biomolecules. Importantly, CAPA provides quantitative data on the cytosolic penetration of molecules of interest, and its low cost and high throughput allows for rapid generation of data on dose- and time-dependence.^[112,174]

5. How do I make my peptide more cell-penetrant?

Peptide drug development has seen a recent resurgence, both in academic and industrial settings. This has led more and more researchers to ask whether there might be general strategies for modifying their own bioactive peptides to make them more cell-penetrant. Direct fusion to a CPP (Table 1) remains an attractive option, especially for proof-of-principle studies. In this review, we highlighted three other strategies for modifying peptides to make them more cell-penetrant. Notably, these do not encompass all possible CPPs or strategies for improving cell penetration. For instance, CPPs discovered by Wimley and co-workers represent a means for promoting cell penetration via direct translocation.^[58,177,178] These have overall physicochemical properties that are similar to many non-penetrating peptides, and do not conform strictly to any of the three strategies described here. Still, as more and more data are compiled for these three strategies, we are gaining a better appreciation of how to apply them to a larger variety of bioactive peptides. It is important to note that none of these are simple, “plug-and-play” modifications that guarantee cell penetration. Rather, they are best-odds strategies based on the results from many research groups. They also do not take into account additional effects, such as transport kinetics, tissue selectivity, and binding kinetics at the cellular target, which medicinal chemistry has shown is critical for understanding distribution properties for all molecules.^{[125,179–182]}

The decision of how to modify a peptide to make it intrinsically cell-penetrant depends on its size, charge and hydrophobicity. A peptide might be amenable to a passive penetration mechanism if the peptide is on the smaller side, defined as molecular volume less than 1500 ų, which roughly corresponds to a constrained peptide of less than 1100–1200 Daltons.^[65,135] Passive penetration also limits the number of solvent-exposed hydrogen bond donors. These parameters can often be satisfied for cyclic peptides that have a substantial portion of non-peptidic or N-modified backbone, or that are structured in a way that shield amide protons from solvent. If this mechanism is pursued, empirical measurement of relative hydrophobicity should be an integral part of the optimization process. New analytical and chromatographic techniques have been reported that correlate well with propensity for passive diffusion, including the Exposed Polar Surface Area technique of supercritical fluid chromatography, and shake-flask partitioning using 1,9-decadiene instead of octanol.^[135,183] If the experimentally determined hydrophobicity of the peptide is close to the range known to allow passive penetration, then systematic alteration of the peptide to optimize overall hydrophobicity could produce a passively penetrant molecule. Because passive penetration is the intended mechanism, optimization could be performed using artificial membrane assays such as PAMPA or cell monolayer assays with Caco-2 or MDCK cells. Importantly, efforts to mask remaining hydrogen bond donors would be better than simply adding large hydrophobic functional groups, because the former strategy will minimize polar surface area while still maximizing aqueous solubility. Emerging work implies that specific rotatable bonds may allow for dynamic structures that better balance solubility and low polar surface area.^[83] In the next few years, this may emerge as a specific, designable feature for maximizing passive penetration.

If the peptide being considered is larger than 1200 Daltons or has too many hydrogen bond donors, one should consider a guanidinium-patterning or amphipathic patterning strategy. As noted above, these may involve overlapping guidelines and internalization mechanisms, but as a development strategy they are distinct. If activity of the peptide is not abrogated by substitution of four or five arginines, then different patterns of these arginine substitutions could be tested to test if any are as effective as the examples mentioned in Section 2.2. Optimization should also focus on hydrophobic groups near the clustered arginines, in order to maximize their effect on cell penetration. If arginines are not tolerated in the peptide, hydrophobic staples could be added in a geometry compatible with the target-bound conformation. 3D modeling of the constrained peptide could be used to design variants with the large, continuous hydrophobic surface known to be critical for this mode of entry. Empirical measures of hydrophobicity could also be a critical part of this development process, along with guidelines for overall charge and pI that maximize cell penetration and minimize cell-lytic activity.^[144,145] Finally, it is important to note that endocytic uptake is the intended mechanism for both guanidinium-patterning and amphipathic patterning strategies. In pursuing these strategies, it is critical to employ a cell penetration assay that measures cytosolic access, such as a protein complementation assay or CAPA. This will avoid overinterpretation of data that indicate efficient endosomal uptake, but inefficient delivery to the cytosol.

6. Conclusion

We are only beginning to understand the factors that affect the cell penetration of peptides and other biomolecules, but the stakes are high. Among the recently discovered classes of cell-penetrating peptides, hydrocarbon-stapled peptides are currently the most advanced in clinical trials.^[184,185] This relative success has broadened the landscape of “cell-penetrating peptides” beyond polycationic carrier molecules and natural products, and demonstrates the vast potential of cell-penetrant biotherapeutics.

In this review we discussed three distinct strategies that are emerging from different chemical approaches: masking backbone amides, presenting a structured pattern of guanidinium groups, and amphipathic patterning. Novel high-throughput and quantitative assays are producing large data sets that directly address these approaches, identifying properties that promote cell penetration in unprecedented detail. As mechanisms of entry are better elucidated, these emerging “structure-penetration relationships” will be understood in terms of specific interactions with cellular components. Until then, we can apply them as empirical, best-odds strategies for making bioactive peptides more cell-penetrant. Going forward, these new data, assays and design strategies will greatly benefit academic research and drug development, accelerating the transition of biotherapeutics from the lab to the clinic.

Biography

Leila Peraro received her BS in Biochemistry from Hobart and William Smith Colleges, and conducted research in Dr. Justin Miller’s lab focusing on the total synthesis of Spiruchostatin analogs. She is currently a PhD candidate in Dr. Joshua Kritzer’s lab in the Department of Chemistry at Tufts University. Her research projects include developing cell penetrant autophagy-inducing peptides, as well as the development of a high-throughput assay for measuring cell penetration in vitro.

Joshua A. Kritzer earned a BE in Chemical Engineering at The Cooper Union and a PhD in Biophysical Chemistry at Yale University. After NRSA-sponsored postdoctoral work in genetics at the Whitehead Institute, he started his own group at Tufts University in 2009. The Kritzer group uses peptides to solve vital chemical and biomedical problems.