Uncoupling the Folding-Function Paradigm of Lytic Peptides to Deliver Impermeable Inhibitors of Intracellular Protein–Protein Interactions

Abstract

Here, we describe the use of peptide backbone N-methylation as a new strategy to transform membrane-lytic peptides (MLPs) into cytocompatible intracellular delivery vehicles. The ability of lytic peptides to engage with cell membranes has been exploited for drug delivery to carry impermeable cargo into cells, but their inherent toxicity results in narrow therapeutic windows that limit their clinical translation. For most linear MLPs, a prerequisite for membrane activity is their folding at cell surfaces. Modification of their backbone with N-methyl amides inhibits folding, which directly correlates to a reduction in lytic potential but only minimally affects cell entry. We synthesized a library of N-methylated peptides derived from MLPs and conducted structure-activity studies that demonstrated the broad utility of this approach across different secondary structures, including both β-sheet and helix-forming peptides. Our strategy is highlighted by the delivery of a notoriously difficult class of protein-protein interaction inhibitors that displayed on-target activity within cells.

Graphical Abstract

INTRODUCTION

Membrane-lytic peptides (MLPs) are produced by a variety of species and can be one of the host’s first responses toward pathogenesis.^1,2 Linear MLPs are nearly always unstructured in solution, and their folding is facilitated by the formation of electrostatic interactions between peptide cationic side chains and anionic groups at the cell surface, i.e., the negatively charged glycocalyx or lipid head groups. The folding into an amphiphilic bioactive conformation can account for up to 50% of the total energy associated with the peptide-cell surface interaction,^2–6 indicating that lytic activity is tightly coupled to peptide folding potential. While many MLPs evolved as antimicrobial agents, some show anticancer activity through analogous mechanisms of cell engagement and membrane disruption.^7,8 This ability to globally interact with cell surfaces makes MLPs attractive starting points for designing new cell penetrating peptides (CPPs) that can deliver otherwise impermeable cargo into cells.^9–12 However, the inherent toxicities of MLPs typically result in limited therapeutic windows for effective delivery. Engineering peptides with improved cytocompatibility without sacrificing cell entry is challenging. Our lab has been developing methods to inhibit the folding of MLPs in order to uncouple this high energy process from their ability to enter cells as a general approach to overcome this problem.^13,14 We previously reported the design of an intrinsically disordered and nontoxic CPP called CLIP6, where membrane-induced folding was inhibited by the installation of an anionic glutamate residue that disrupted the peptide’s amphiphilicty.¹³ Although this approach worked to greatly limit cytotoxicity, the negatively charged residue also reduced peptide cellular uptake. Thus, an alternate approach was needed to inhibit folding without invoking electrostatic charge so that potent cell delivery could be retained.

Backbone N-methylation is found in a variety of bioactive peptide natural products.^15,16 Drawing inspiration from nature, medicinal chemists have used N-methylation to improve the proteolytic stability and cellular uptake of synthetic peptides, while also reducing their aggregation potential.^15,17 Herein, we show that N-methylation can also be used to reduce the folding propensities of lytic peptides whose bioactive amphiphilic conformation is stabilized by amide-based hydrogen bonding. This simple modification represents a new strategy to modulate the folding-function paradigm of MLPs, resulting in cytocompatible peptides suitable for cell delivery applications (Figure 1A).

Figure 1.

(A) Using backbone N-methylation to disrupt folding and toxicity of lytic peptides while retaining cell penetrating behavior. (B) Backbone N-methylated analogs of: SVS1 (NMS), CLIPΔ (NMΔ), and M-lycotoxin (NMM).

RESULTS AND DISCUSSION

Design and Synthesis of N-Methylated Analogs of MLPs

We prepared single and dimethylated analogs of three MLPs (Figures 1B and S1): two de novo designed β-hairpin sequences previously developed in our lab (SVS1¹⁸ and CLIPΔ¹³) and a helical peptide derived from wolf spider venom (M-lycotoxin^19,20). Each of these parent peptides are known to undergo cell-surface induced folding, which accompanies their lytic activity. Folded SVS1 and CLIPΔ contain an amphiphilic arrangement of valines in each of their β-strands that form a series of intramolecular hydrogen bonds that stabilize their hairpin conformations. N-methylation at valine 15 (NMS1 and NMΔ1) was designed to disrupt a central hydrogen bond to hinder hairpin formation. However, in addition to folding, self-assembly at the cell surface might also be linked to the lytic activities of some MLPs. To examine this possibility, we designed two additional peptides, NMS2 and NMΔ2, with N-methylation at hydrophilic positions 5 and 14 intended to block the formation of intermolecular hydrogen bonds. Lastly, we designed variants of M-lycotoxin to inhibit intramolecular hydrogen bonds necessary for helix formation. NMM1 contains one N-methyl group at the central position 13, while NMM2 contains two methyl groups at the N- and C-terminal regions of the sequence. Unmodified peptides were prepared using automated microwave-assisted peptide synthesis, whereas N-methylated analogs utilized a combination of automated and manual synthesis to install the N-methyl groups using Fukuyama-Mitsunobu conditions (Figure S2).^21,22 The completed peptides were then purified by prep-HPLC (Figure S3).

Investigating Effects of Folding on Lytic Activity and Cellular Uptake.

To determine the folding potential of each peptide, we performed circular dichroism (CD) spectroscopy. Studies were done in buffer alone and in the presence of model liposomes to assess how cell surface charge influences peptide folding. As expected, SVS1, CLIPΔ, and M-lycotoxin fold in the presence of negatively charged liposomes (Figure 2A), demonstrating typical spectra for β-sheet¹⁸ and α-helical²⁰ structure. Likewise, NMS1 and NMΔ1 show random coil CD spectra, indicating that the incorporation of one N-methyl group to block an intramolecular hydrogen bond completely abolished folding for the hairpin-forming sequences. Dimethylated NMS2 and NMΔ2 were designed to inhibit self-assembly, but unexpectedly, there was no folding observed in the presence of the negatively charged liposomes. This suggests that these two processes are linked and that inhibiting self-assembly can affect peptide folding.^23,24 For the M-lycotoxin analogs, a single centrally positioned N-methyl group partially reduced helix formation (NMM1), while two distally spaced N-methyl groups completely inhibited folding (NMM2). All of the peptides were unstructured in buffer alone or in the presence of neutral liposomes (Figure S4), highlighting the requirement for an oppositely charged surface for MLP folding. Peptide toxicity toward a panel of cells correlated nicely with folding potential (Figures 2B,C, S5), with all of the N-methylated analogs being less toxic than their parent MLPs. The correlation between folding and lytic behavior was particularly striking for the M-lycotoxin analogs, where a clear step function in activity was observed between the folded, intermediately folded, and unfolded states. In general, the NMS analogs were the most cytocompatible, followed by the NMΔ and NMM derivatives, respectively.

Figure 2.

(A) CD spectra of lytic peptides and their N-methylated analogs in the presence of model negatively charged liposomes that mimic surface charge of cancer cells. Peptide concentration was 50 μM and total lipid content was 2.5 mM. A 1:1 ratio of POPC:POPS lipids was used for hairpin-based sequences, whereas a 4:1 ratio was used for helical peptides. (B) Cytotoxicity of peptides against HeLa cells and (C) various cell lines.

Next, total cell uptake was measured by flow cytometry, using fluorescently labeled peptides, and coplotted against cytotoxicity. Figure 3A shows that the N-methylated analogs are capable of entering cells (green data) to similar extents as their corresponding parent MLPs (SVS1, CLIPΔ, and M-lycotoxin), while displaying drastically decreased cytotoxic behavior (gray data). Since folding is more critical to lytic activity than to cell uptake, N-methylation uncouples the folding-function requirement with respect to cell delivery. We also investigated how altering the folding potential of the peptides affects their mechanism of uptake (Figure S6A). The N-methylated analogs enter cells through a mixture of mechanisms, as shown by partial loss of cell entry when inhibiting active uptake processes using ATP depletion conditions. However, the N-methylated peptides appear to be less dependent on clathrin-mediated endocytosis than their parent lytic sequences, with hyperosmolar sucrose conditions showing limited or no effect on cell entry. We further probed uptake mechanisms by performing studies with selected peptides at 4 °C (Figure S6B), which also limits energy dependent uptake. These experiments largely mimicked the ATP depletion conditions by indicating a partial reliance on active uptake, although in HeLa cells, CLIPΔ, M-lycotoxin, and NMM1 all had a more pronounced loss of uptake at 4 °C than under ATP depletion. Since these peptides are known to undergo surface-induced folding, we hypothesize that in addition to reducing active uptake, the lower temperature conditions might also inhibit the ability of these peptides to engage with and fold at cell surfaces (NMM1 is the only N-methylated peptide in this study that still has some capacity to fold). We then performed additional mechanistic experiments with NMΔ1 and NMM1 in the presence of inhibitors of other forms of endocytosis but observed that macropinocytosis or caveolae-mediated endocytosis were not major factors in the uptake of these two peptides for HeLa or A549 cells (Figure S6C).

Figure 3.

(A) Normalized cellular uptake (green) coplotted with toxicity (gray). (B) Total uptake for selected fluorescently labeled CPPs. (C) Confocal microscopy images of Flu-NMΔ1 treated cells.

In terms of cell delivery potential, we were pleased to see that both fluorescently tagged NMΔ1 and NMM1 displayed comparable uptake to SVS1 (Figure 3B), the previous gold standard CPP from our lab that is limited by its general toxicity.^12,25,26 Furthermore, both of these N-methylated analogs have improved uptake over CLIP6,¹³ the nontoxic CPP previously reported from our lab, and the commonly used TAT²⁷ domain (Figure S7A). We also assessed the cell uptake of these peptides using the chloroalkane penetration assay,^28,29 which determines effective concentrations for cytosolic delivery. Transfected HeLa cells expressing Halo-GFP-Mito,³⁰ which is displayed in the cytosol, were treated with chloroalkane-tagged peptides (Figure S7B). The cells were subsequently treated with Halo-TAMRA and analyzed by flow cytometry for levels of red fluorescence. High levels of red fluorescence indicate weak cytoplasmic delivery at a given concentration, while lower levels reflect more potent peptide entry. Again, we observe that NMΔ1 and NMM1 have improved delivery over CLIP6 and TAT (Figure S8A). The determined CP₅₀ values indicate that NMΔ1 is 2x and 3x more effective at entering the cytoplasm than TAT and CLIP6, respectively. These results are in good agreement with the observed improvement for NMΔ1 over these peptides in the fluorescently labeled peptide uptake experiments (Figure 3B). Confocal microscopy studies show that NMΔ1 predominantly accumulates at the surface of cell nuclei with some diffuse cytoplasmic distribution (Figure 3C), with similar behavior also observed for NMM1 (Figure S8B). While it is not clear at this time why the peptides accumulate at the nucleus, similar behavior has been observed for other reported CPPs.^31,32 Given NMM1’s relative cytotoxicity (Figure 2), we decided to focus our remaining cell delivery efforts on NMΔ1, which has both potent uptake and greatly reduced lytic behavior.

Intracellular Delivery of Protein-Protein Interaction Inhibitors.

To showcase the potential of NMΔ1 as an effective CPP, we focused our attention on delivering two different cell-impermeable peptide ligands that disrupt intracellular protein-protein interactions (PPIs). Peptides are potentially promising therapeutics,^33,34 but one major limitation is that many cannot effectively enter cells. We first chose the BH3 domain of the pro-apoptotic protein Bim as a model ligand because targeting the mitochondrial Bcl-2 pathway has been well established for initiating apoptosis.^35–39 NMΔ1-BH3 was synthesized by conjugating NMΔ1 to the N-terminus of Bim-BH3 through a -GGSGS- linker (Figure S9). We assessed the apoptotic activity of NMΔ1-BH3 in U937 and A549 cells, along with the unconjugated Bim-BH3 ligand and NMΔ1. Gratifyingly, NMΔ1-BH3 displayed low μM dose-responsive activity, whereas free Bim-BH3 was relatively inactive and NMΔ1 had lytic behavior at higher concentrations (Figures 4A and S10). We then conducted Western blot analysis for cytochrome C release in U937 cells to evaluate whether treatment led to cell death via the mitochondrial pathway as designed.³⁷ In response to cell damage or stress, activation of pro-apoptotic Bax/Bak causes mitochondrial outer membrane permeabilization (MOMP) and the release of several proteins (including cytochrome C) that can further activate downstream apoptotic signaling.⁴⁰ The intracellular delivery of the BH3 ligand would directly induce Bcl-2 family dependent MOMP without the need for other stimuli to initiate the apoptosis cascade. After treatment with 10 μM peptide, cells were subjected to a mild lysis of the outer cellular membrane to reveal the soluble cytoplasmic contents (S fraction), followed by the solubilization of the pelleted components (P fraction). When cells were not undergoing apoptosis, as was the case with NMΔ1 and BH3 treatment, Western blot staining showed cytochrome C only within the P fraction, because the mitochondria have not been compromised (Figure 4B). Treatment that induced MOMP showed cytochrome C within the S fraction and was only observed here with NMΔ1-BH3. Further, treatment of cells with a lytic concentration (100 μM) of NMΔ1 failed to induce MOMP (Figure 4C), indicating that rapid cytochrome C release for NMΔ1-BH3 is likely due to on-target intracellular activity.

Figure 4.

(A) U937 cytotoxicity for BH3 peptides. (B) Western blot for cytochrome C release in U937 cells after 2 h peptide treatment (10 μM). S-fraction = soluble cytoplasmic contents; P-fraction = pelleted contents. Cytochrome C release into the S fraction, indicative of MOMP, was only observed with NMΔ1-BH3 treatment. Bak staining was used as a loading control. (C) Toxic concentrations of free NMΔ1 (100 μM) still does not lead to cytochrome C release, indicating that nonspecific methods of cell death can be distinguished from Bcl-2 pathway mediated apoptosis.

Finally, we wanted to determine whether we could use NMΔ1 to deliver a promising ligand for an even more challenging PPI. We chose the ligand 4j* (Figure S11),^41–43 which contains a phospho-mimicking functionality and is a selective polo-box domain-binding inhibitor of the nuclear-localized polo-like kinase 1 (Plk1).⁴⁴ Since it is required for cell division and is often overexpressed in cancer, Plk1 is an attractive target for therapeutic development.⁴³ Peptide 4j* has potent in vitro Plk1 binding affinity (Figure S12), but like many phospho-containing ligands, it has limited cellular uptake and activity (Figures 5A and S13).⁴¹ The NMΔ1–4j* conjugate retained in vitro binding to Plk1 and displayed low μM activity against U937 cells, which demonstrates the first example of a 4j*-like ligand having such potent cellular activity. Similar behavior was observed in A549 and HeLa cells (Figure S13). To probe whether the mechanism of action involved Plk1 inhibition, we tested the activity of NMΔ1 conjugated to 4j*S4A, a 4j* control with reduced in vitro Plk1 binding due to a Ser to Ala substitution (Figures S11 and S12). The attenuated cellular activity of NMΔ1–4j*S4A compared to NMΔ1–4j* is consistent with weaker Plk1 binding, supporting an on-target mechanism of cell death. We then performed cell cycle analysis of treated U937 to further validate Plk1 inhibition as the primary cause of activity. Healthy cells transition between several different phases during the division process, namely cellular growth and organelle duplication (G1), DNA replication (S1), a second growth stage (G2), and finally mitosis (M). Plk1 largely performs its function within the later stages of cell division,⁴⁴ therefore on-pathway targeting would stall cell cycle at the G2/M phases and enrich these populations. Treatment with the cell impermeable 4j* ligand or unconjugated NMΔ1 did not alter the cell cycle from that of untreated control cells, whereas treatment with NMΔ1–4j* led to a 10-fold increase in the G2/M phases and subsequent reduction in both the G1 and S phases (Figures 5B,C and S14). The NMΔ1–4j*S4A control showed a similar trend of reduced activity as the viability experiments. Lastly, NMΔl–4j* treatment did not induce rapid cytochrome C release (Figure S15), signifying that the delivered ligand was not causing apoptosis before disrupting cell cycle. Collectively, these results demonstrate that NMΔl can deliver impermeable 4j* into cells with on-target PPI activity toward Plkl.

Figure 5.

(A) U937 cytotoxicity for Plk1 inhibitor compounds. (B) Quantification of U937 cell cycle phases. Difference between control and NMΔ1 or 4j* were not significant. ****P < 0.0001; **P < 0.01. (C) Representative U937 cell cycle analysis dot plots for untreated control and NMΔ1–4j*.

CONCLUSION

We’ve developed a new approach that can transform cytotoxic lytic peptides into cytocompatible CPPs by inhibiting their cell-surface folding through backbone amide N-methylation. This strategy was applied to both β-hairpin and α-helical lytic peptides and should be broadly applicable to other linear sequences that fold at cell surfaces. The hairpin-derived peptide NMΔ1, containing one N-methyl group, is cytocompatible and was effective at delivering two impermeable peptide ligands. Notably, NMΔ1 conjugated to 4j* could inhibit the nuclear-localized enzyme Plk1, halt cell division, and induce cell death. This first in class Plk1 inhibitor with cellular activity highlights the potential for N-methylated CPPs to deliver ligands for challenging intracellular PPI targets.