Novel Structures of Self-Associating Stapled Peptides

Abstract

Hydrocarbon stapling of peptides is a powerful technique to transform linear peptides into cell-permeable helical structures that can bind to specific biological targets. In this study, we have used high resolution solution NMR techniques complemented by Dynamic Light Scattering to characterize extensively a family of hydrocarbon stapled peptides with known inhibitory activity against HIV-1 capsid assembly to evaluate the various factors that modulate activity. The helical peptides share a common binding motif but differ in charge, the length and position of the staple. An important outcome of the study was to show the peptides share a propensity to self-associate into organized polymeric structures mediated predominantly by hydrophobic interactions between the olefinic chain and the aromatic side-chains from the peptide. We have also investigated in detail the structural significance of the length and position of the staple, and of olefinic bond isomerization in stabilizing the helical conformation of the peptides as potential factors driving polymerization. This study presents the numerous challenges of designing biologically active stapled peptides and the conclusions have broad implications for optimizing a promising new class of compounds in drug discovery.

INTRODUCTION

Peptides offer a versatile platform for drug design and have proven to be highly effective for targeting protein-protein interaction surfaces without using small molecules. Traditionally, peptide based drugs have suffered from lack of stability, limited cellular uptake and rapid proteolysis in vivo. These issues outweigh some of the advantages such as higher specificity and fewer side effects from accumulation in tissues. Over the years several techniques have been developed to stabilize α-helical peptides ¹ and in this context the unique advantages of hydrocarbon stapling have been applied very successfully to transform biologically active peptides into viable therapeutic agents ^2,3 against cancer ⁴ and HIV-1 capsid assembly ^5,6.

The genesis of the hydrocarbon stapling technique can be traced to the ruthenium based Grubb’s catalysis used for ring closing metathesis ⁷. The stapling, cross-links two α,α disubstituted amino acids bearing olefinic chains of variable length at positions “i” and “i+4” or “i+7” in the peptide sequence ^2,8. The presence of the hydrocarbon cross-link has been shown to reinforce the α-helical conformation of peptides, render it cell-permeable and improve its ability to resist proteases in vivo ⁹. The increased helicity of the peptides alter cellular uptake and eventual release using the endocytotic pathways for efficient inhibition of the target in cells ¹⁰. The rationale for selecting specific sites for inserting the cross-link is dictated by the need to preserve the key interactions at the binding site. The recent structures of the stapled MCL-1 BH3 helix bound to the anti-apoptic MCL-1 ¹¹ and peptides complexed with estrogen receptor ¹² have shown supplementary interactions involving the staple are an important consideration from the design perspective ¹³.

Atomic level details of the bound form of select peptides are available from high resolution three dimensional complex protein structures ^11,12,14 and there is interest in developing a unified model that can accurately predict the helical propensity and the optimum length of the staple required to stabilize a peptide fragment in complexes. Experimentally determined structures and theoretical simulations have furnished valuable insight into the underlying physical forces stabilizing S_i,i+4S(8) cross-linked peptides, and simplified the cumbersome task of deciding the optimal site for staple incorporation ^10,15. Such studies have successfully validated experimental data for p53 peptide ^13,16 and characterized the conformational space populated by the stapled RNase A and BID BH3 peptides ¹⁰ with variable staples. However the structural properties of helical stapled peptides with longer R_i,i+7S(11) cross-links are not well understood. The impact of increasing the effective length and the configuration of the olefinic chain on the conformational flexibility of the peptide backbone are of particular interest when designing the stapled peptides.

While the intrinsic hydrophobicity of the olefinic chain limits the overall solubility of the stapled peptides and is an issue for longer staples, an unresolved question is whether the lipophilicity in concert with the helical structure of the peptides promote further self association in solution. To address these outstanding and relevant questions, in this study we have focused on a small family of rationally designed stapled peptides with inhibitory activity against HIV-1 virus capsid assembly ^5,6. To date the most successful inhibitor targeting CA-CTD in vitro is the linear CAI peptide ¹⁷ and the stapled analogue NYAD-1 with a S_4,8S(8) cross-link ⁵ in cells. Recent studies have identified a second hydrocarbon stapled peptide NYAD-203 designed specifically to disrupting the dimerization of CA-CTD with modest activity during virus particle assembly in cellular assays ⁶. Until recently there has been no report on the extent of non-specific aggregation or self-assembly of stapled peptides with possible deleterious effects on activity ⁶.

Therefore the principal objective of this study was to expand these preliminary studies of select biologically active stapled peptides to a detailed evaluation of a larger family of peptides with variable sequence and staples to improve the overall solubility, inhibit self association, and enhance target activity. The goal of this study was twofold, first to apply NMR and dynamic light scattering to investigate the impact of the hydrocarbon staple and the net charge on solubility and the propensity of the free peptides to polymerize. We then address the question of whether this property of the peptides to self-associate modulates the biological activity and design a strategy to modify the self association, with the potential for improvement of the structure activity relationships.

The extensive experimental data on the NYAD family of stapled peptides presented in this study offers a valuable perspective of their structural diversity that has not been reported previously. Given the potential for developing stapled peptides into viable therapeutic agents the conclusions of this study have broad implications for optimizing side-chains in stapled peptides for improving binding affinity and activity against other established biological targets.

Results

Self Association of Stapled Peptides

In a previous study based on the linear peptide CAI ¹⁷, we presented the structural basis for a stapled peptide NYAD-1 that is cell permeable with a tenfold improved affinity for mCA-CTD (K_d ~ 1 µM) ⁵. The low solubility of the first NYAD-1 peptide was a limiting factor in the preliminary studies and necessitated further modifications of the sequence with a short poly-lysine tag at the C-terminal end of the peptide NYAD-13. Although highly soluble, in course of our analysis of NMR spectra of free NYAD-13 peptide, we observed concentration dependent changes in chemical shift seen in the overlay of natural abundance ¹H-¹³C HSQC spectra (Figure 1a). Likewise, the change in the diffusion constant with concentration is consistent with self-association of the peptide into polymeric entities (Figure 1b). The chemical shift changes (Figure S1) fit to an isodesmic model of indefinite association yields a K_d of 2.5 ± 1.4 mM ^18,19. A single polymeric species with variable radius of gyration (R_g) in the range of ~15 nm was detected by DLS (Figure 1d) excluding the possibility of the formation of small oligomers of fixed stoichiometry such as dimers or trimers. At lower peptide concentration, the equilibrium shifts in favor of the monomer with an effective radius of gyration equal to 1 nm (Figure 1c). The transferred NOE effect manifested by fast exchange of the peptide with the relatively large polymeric species is evident from the changes in the NOE cross-peak intensity with concentration (Figure 2). The chemical shifts of the free peptide were assigned at multiple concentrations and based on the comparison of observed and random coil Cα chemical shifts ²⁰ we conclude there is essentially no change in the length of the helical structure of the peptide at higher concentration (Figure 3c,d).

Figure 1.

(A) Overlay of the aliphatic region of ¹H-¹³C HSQC spectra of NYAD-13 peptide dissolved in D₂O at two concentrations, 10.8 mM (blue) and 0.5 mM (black) respectively. The peaks are annotated in corresponding colors. The data were acquired at 298 °K and pH 7.0. (B) Diffusion constants as a function of NYAD-13 peptide concentration in 99.9% D₂O at 298 °K. Bar plots of normalized intensity versus radius of gyration (nm) at two different concentrations of NYAD-13, (C) 100 µM and (D) 1 mM obtained from DLS measurements. The hydrodynamic radius and polydispersity index of major peaks are annotated in the figure.

Figure 2.

Amide region of 2D-NOESY spectra of NYAD-13 peptide acquired at different concentrations in 90%H2O/10%D2O at 298K. Mixing times reported in brackets. (A) 200 µM (200ms) (B) 1 mM (200 ms) (C) 6.5 mM (100 ms) (D) 45 mM in D2O (100 ms).

Figure 3.

(A) The lowest energy representation of peptides arranged in the proposed model of the smallest asymmetric unit from an extended polymeric chain calculated from NMR data acquired on a 45 mM peptide sample dissolved in 99% D₂O at 298 °K. Residues along the two surfaces of the amphiphilic helix are color coded based on standard definition of polarity: green for polar, red for negatively charged, blue for positively charged residues and yellow for hydrophobic residues respectively. The cis olefinic linker between residues 4 and 8 is represented in orange-red. (B) The interactions involving N- and C-terminal residues of the peptides. The schematic representation of the NYAD-13 peptide is accompanied by plots of the ¹³Cα chemical shifts difference from the “random coil” values ²⁰ at two different concentrations: (C) 45 mM and (D) 1 mM peptide dissolved in D₂O and referenced to DSS.

To model the structure of the polymeric species, NOESY spectra were acquired on a concentrated peptide sample at 45 mM dissolved in D₂O. The data is consistent with a pair of helices aligned in an anti-parallel fashion as the smallest asymmetric unit from a longer polymeric chain. Extensive contacts between Phe3, Leu7 and Tyr10 from one helix and the hydrocarbon chain from a second helix leave the aromatic side-chains free to capture a third peptide (Figure 3a). This particular stacking arrangement effectively buries the hydrophobic surface of the amphiphilic helix maximizing the solvent exposure of the polar side-chains of Thr2, Asp5 and Tyr9. Although we see no evidence of secondary structure in the C-terminal poly-lysine tag (Figure 3c,d), it is likely that the positive charge engages nearby aromatic side-chains in cation-π interactions between the peptides. The peptide-peptide interface shares remarkable similarities to the target bound conformation of the peptide where the side-chains of Ile1, Phe3, Leu6 and Tyr9 are buried in the hydrophobic pocket of the protein (Figure 7). The hydrocarbon staple essentially serves as a substitute for the protein by recruiting the freely accessible aromatic and hydrophobic side-chains driving the polymerization. The organization of the polymeric chains into higher order nano-assemblies is a strong possibility and a topic for further investigation by other techniques suitable for studying such systems ²¹.

Figure 7.

(A) A comparison of 1D proton NMR spectra of NYAD-13 with (Upper Panel) and without (Lower Panel) homo-decoupling of the H^δ protons at a chemical shift of 2.15 ppm during acquisition. The 1D spectra were acquired independently on a 10.8 mM sample of uncomplexed NYAD-13 dissolved in 100% D₂O, pH 7.0 and 298 °K. (B)–(C) Views of the intermolecular contacts at the binding site of the complex between mCA-CTD (cyan) and the peptide NYAD-13 (pink) (PDB deposition 2L6E). Interactions involving key peptide side-chains are annotated in color.

With the goal of establishing the general propensity of stapled peptides to self associate and identify the relevant parameters that drive this phenomenon, we initially focused on a group of CAI cross-linked peptides that contains a consensus sequence specifically targeting the distal hydrophobic pocket of CA-CTD dimer previously identified from phage display libraries ¹⁷. The residues critical to this interaction have been highlighted in the CAI sequence in Table 1 and fine-tuning the motif is the focus of a separate study. Therefore based on the presence of the CAI consensus sequence alone the first group of peptides studied in Table 1, are all expected to bind CA-CTD and possess nominal activity. These peptides essentially differ from each other by mutations at non-interface sites, C-terminal charge or the position of the hydrocarbon cross-link, factors that can be readily manipulated to optimize the biophysical properties of the original peptide without altering the activity significantly. Although of limited interest from the activity stand point, several peptides with mutations in the binding motif were also investigated for their ability to self-associate (Table 1).

Table 1.

The sequences and activity of NYAD type peptides

Peptide	Sequence	Antiviral Activitya	Self-Association	Maximum Solubility	Net Charge
CAI		−	−		−3
NYAD-1		++	Yes	<100 µM	−1
NYAD-13		−	Yes	>1 mM	+2
NYAD-13-I2A		−	Yes	>1 mM	+2
NYAD-14		−	No	>1 mM	0
NYAD-30		+	Yes	<50 µM	−1
NYAD-35		++	Yes	600 µM	−2
NYAD-36		++	Yes	<100 µM	−3
NYAD-43		+	Yes	<100 µM	−4
NYAD-44		−	Yes	>1 mM	0
Peptides with mutated CA-CTD Binding Motif
NYAD-16		−	Yes	200 µM	−2
NYAD-26		−	No	300 µM	−2
NYAD-31		+	Yes	200 µM	0
NYAD-32		+	Yes	500 µM	−1
NYAD-42		+	Yes	<100 µM	−4
NYAD-59		+	Yes	>1 mM	0

^aThe antiviral activity of the peptides was measured in MT-2 type cells using established protocols described in previous publication ⁵. The values indicated as: ++ (IC₅₀ <10 μM); + (10 µM < IC₅₀ <100 μM); − (IC₅₀ >100 μM). Details on activity will be published separately.

^bAlthough the peptide binds to the protein with high affinity ~ 1uM in vitro but is cytotoxic in cells owing to the lysines ⁵.

^cThe peptide does not bind to the protein.

X = (S)-2-(4’-pentenyl)alanine

Z = (R)-2-(7’-octenyl)alanine

Interestingly with the exception of NYAD-14 and NYAD-26 all the peptides with modest solubility (>100 µM) in water shared the propensity to self-associate to different extents (Figure 5 and Figure S4) when monitored by concentration dependent DLS measurements and NMR chemical shift perturbation. For example the functionally redundant sites suitable for mutations in the S_4,8S(8) cross-linked CAI peptides include Thr2, Leu7 and Gly11 of which perhaps Leu7 is of most interest in the context of the polymeric structure of NYAD-13 (Figure 3a). However the single mutation in NYAD-13-I2A is not effective in eliminating self-association in the presence of the aromatic side-chains (Figure S3). Similarly, increasing the negative charge of the peptides by eliminating the C-terminal positive charge in NYAD-30 or replacing the lysines with the neutral ‘SGS’ solubility tag in NYAD-35 offers no significant advantage.

Figure 5.

Normalized intensity versus hydrodynamic radius (nm) measured at maximum solubility of stapled peptides in H₂O by DLS. Each peak is labeled with the hydrodynamic radius and polydispersity index. Peptides were dissolved in 95% H2O/5%D2O for DLS experiments. (A) 274 µM NYAD-16 (B) 409 µM NYAD-32 (C) 2.0 mM NYAD-59 (D) 543 µM NYAD-35 (E) 54.3 µM NYAD-35 (F) 100 µM NYAD-36 with <1% DMSO.

We next turned our attention to the S_8,12S(8) cross-linked NYAD-14 peptide to test the hypothesis whether shifting the staple, reorganizes and possibly destabilizes the polymeric structure (Figure 4a). Based upon the Cα difference plots (Figure 4a), we conclude the helix in NYAD-14 is not as well formed as the remaining NYAD peptides. With the N-terminal backbone essentially random coil, the free peptide forms a truncated helix spanning residues 8–12. The partial structure with the staple straddling across the Tyrosine side-chains evidently disrupts the peptide stacking because polymerization is suppressed at even millimolar concentrations (Figure S2). Although monomeric with the binding motif intact, NYAD-14 is biologically inactive because it lacks the correct helical structure required to bind CA-CTD. The example of NYAD-14 also suggests that the staple itself has limited ability to drive the self association but requires an ally in the hydrophobic side-chains.

Figure 4.

Schematic representation of stapled peptides (right) and corresponding ¹³Cα chemical shift difference (left) from the “random coil” values ²⁰: (A) 3 mM NYAD-14, (B) 2 mM NYAD-59, (C) 1.3 mM NYAD-36 with 50 µM deuterated SDS, and (D) 5.0 mM NYAD-203 in 90% H₂O, 20 mM sodium phosphate at pH 6.5. The NMR samples in (A)–(C) consist of peptides dissolved in D₂O and referenced to DSS. Using established nomenclature ² the stapled peptides NYAD-14, NYAD-59 and NYAD-203 include a S_i,i+4S(8) linkage and NYAD-36 a R_i,i+7S(11) cross-link.

It is evident from the previous examples that mutating the hydrophobic residues in the CA-CTD binding motif and in particular the aromatic side-chains would be the most effective approach to minimize the self-association of the peptides with the caveat that it would have a significant impact on activity as seen in Table 1. Hence by replacing Phe3 and Leu7 pair by a S_3,7S(8) cross-link in NYAD-16 and Leu6 and Tyr10 by a S_6,10S(8) cross-link in NYAD-59 we observed the monomeric state is favored at even high concentrations (Figure 5a,c). Unlike NYAD-14, by altering the position of the staple the helical structure of NYAD-59 is still intact as verified from the Cα difference plot (Figure 4b). Increasing the negative charge of the peptide NYAD-42 otherwise similar to the neutral NYAD-59 impairs its solubility but fails to completely suppress self-association (Table 1 and Figure S4b).

Besides the aromatic residues we have looked at other S_4,8S(8) cross-linked peptides where Ile1 and Leu6 in the CAI binding motif have been mutated independently to explore the effect of these residues on polymeric assembly. While NYAD-26 with the N-terminal Ile1->Lys substitution is monomeric, it is also inactive. Based on the model structure of the NYAD-13 peptide, we hypothesize the N-terminal Lysine disrupts key hydrophobic interactions between Ile1 and Tyr10 while favoring intra helical side-chain interactions with Asp5 (Figure 3b). Concurrently, the Gly11->Glu mutation at the C-terminal end of the peptide NYAD-26 has the potential to disrupt inter-peptide H-bonds involving Asp5 and lysine side-chains.

Based on the NYAD-13 model structure (Figure 3a) Leu6 appears to be unimportant for peptide interactions in the polymer although critical for establishing hydrophobic contacts in the mCA-CTD complex (Figure 7b). Hence the Leu6->Trp mutation in the peptides NYAD-31/32 does not exacerbate the tendency to self-associate as one might have expected from increasing the side-chain hydrophobicity (Figure 5b). Furthermore by fine tuning the C-terminal charge to slightly negative values in concert with the small increase in polarity of Trp improves the overall solubility of the peptide.

To summarize these results the hydrocarbon staple in conjunction with a folded and stable helix with a sequence rich in hydrophobic residues such as that of the original CAI peptide, can be organized very effectively into polymeric chains. The veracity of this hypothesis was further tested by characterizing an unrelated stapled peptide NYAD-203 with a S_6,10S(8) cross-link designed as a structural mimic of the helix at the dimer interface of CA-CTD ⁶. NYAD-203 dissociates CA-CTD dimers with modest affinity in the ~50 µM range and functions as an active inhibitor of gag assembly in cells ⁶. The helical structure of the eighteen residue peptide extends between residues 1 and 14 (Figure 4d) and from the concentration dependent NMR (Figure S5a) and DLS experiments (Figure S6) we conclude the stapled peptide associates into much larger macromolecules (<1µ) of variable size. Interestingly a linear peptide with a sequence identical to the dimerization helix from CA-CTD was shown to self-associate with a K_d ~ 17 ± 10 µM comparable to that of the full length CA-CTD ^22,23. Consequently the stapled peptide self-associates with even greater efficiency at much lower concentrations, but the application of NMR data to build a unified structural model of the peptide interactions in the oligomeric state was complicated by the ambiguity of intramolecular side-chain NOEs that could not be distinguished readily from those arising from intermolecular contacts at the CA-CTD dimer interface ²⁴ (Figure S5b). Extensive NOEs between the hydrocarbon chain and the aromatic ring of Trp7 are reminiscent of the stacking of the olefinic chain and the aromatic ring of phenylalanine in NYAD-13. Collectively the NMR and DLS evidence would suggest the peptide NYAD-203 like NYAD-13 can associate in more than one arrangement including the possibility of dimerization as an intermediate structure during the assembly of polymeric chains.

Structural Characterization of longer Staples

In theory stapled peptides with a R_4,11S(11) cross-link are expected to form a highly stable helix with three turns ² but atomic level details of representative structures are not available. In this context one of the most active peptides from the library NYAD-36 was of particular interest for further characterization (Table 1). Notwithstanding the negative charge, the intrinsic hydrophobicity of the eleven carbon R_4,11S(11) cross-link drives the spontaneous aggregation of the peptides NYAD-36 (Figure 5f) and the analogous NYAD-43 (Figure S4c) at relatively low concentrations precluding characterization of the monomeric/oligomeric states. The redesigned soluble analogue NYAD-44 appears to be aggregated over a wide concentration range judging from the broad methyl resonances (Figure S4d). This would also account for its inability to bind CA-CTD (data not shown). However in the presence of small amounts of SDS (<100 µM) the peptide NYAD-36 is monomeric at even high concentrations (~3 mM) and the superior quality of the spectra in H₂O facilitated the complete analysis of the secondary structure (Figure S7).

Based on the chemical shift difference plot the α-helical region of NYAD-36 extends from Phe3 to Ser13 with three turns of a helix (Figure 4c). Two sets of cross-peaks from the olefinic protons in the ¹H-¹³C HSQC spectra were observed and assigned to the major trans (16 Hz) and minor cis isomer (<10 Hz) based on the ³J_HH coupling constant of the olefinic protons (Figure 6c) and NOE cross-peal patterns (Figure S7). Although the major product of the olefinic metathesis of a S_i,i+4S(8) hydrocarbon staple by Grubb’s catalysis is the cis isomer ^2,25, with increasing length of the staple the metathesis reaction favors formation of both isomers in equal amount ^2,25. NYAD-36 is a special example where the higher stability of the trans over the cis configuration of the double bond in the R_i,i+7S(11) peptide drives the reaction almost exclusively in favor of the former (~95% by HPLC).

Figure 6.

(A) The lowest energy structure from an ensemble of NYAD-36 structures calculated from NMR data acquired at 298 °K on a 2.7 mM peptide sample dissolved in 90% H₂O/10%D₂O with 50 µM SDS. Residues along the two surfaces of the amphipathic helix are color coded based on standard definition of polarity described in Figure 3. The trans olefinic R_i,i+7S(11) linker is represented in yellow. (B) ¹H-¹⁵N natural abundance HSQC spectra acquired with 512 scans at 500 MHz. (C) Comparison of 1D proton NMR spectra of NYAD-36 with (Lower Panel) and without (Upper Panel) selective homo-decoupling of the methylene H^η/H^δ protons at a chemical shift of ~2.00 ppm during acquisition. The data were acquired on a 1.25 mM sample of NYAD-36 dissolved in 100% D2O with 50 µM SDS at 298K.

The two sets of cross-peaks with distinct chemicals shift in the ¹H-¹⁵N-HSQC spectrum suggests the presence of two distinct peptide conformations resulting from the isomerization of the double bond (Figure 6b). From the calculated NMR structures of the trans isoform we deduced the presence of four pairs of hydrogen bonds between the “i” and “i+4” residues (F3 CO..H^N L7, X4 CO‥H^N D8, L7 CO‥H^N X11, D8 CO‥H^N E12) respectively, that is consistent with three turns of a helix (Figure 6a). Coincidentally the large chemical shift perturbation of specific hydrogen bond donor-acceptor pairs mapped to residues X4, Leu7, Asp8 and Glu12 in the cis isoform would imply changes in the H-bond network. The long-range carbon-carbon distances across the olefinic bond calculated from the lowest energy NMR structures provide an approximate measure of the effective length of the staple. Compared to the trans isomer these distances are consistently shorter in the cis isomer and the difference less than 1Å. The slightly shorter staple suggests the helix maybe strained in the cis isomer, necessitating the deviation of the backbone from standard H-bond geometry.

Refined mCA-CTD bound Structure of NYAD13

In light of the peptide aggregation issues and the significance of the double bond isomerization on the stapled peptide conformation, we revisited the previously published complex structure of the trans isoform of NYAD-13 peptide bound to mCA-CTD¹⁴. Our first objective was to verify the major isomer of the peptide by measuring the ³J_HH coupling across the olefinic bond. The 11 Hz separation is consistent with a cis double bond being the major product (Figure 7a). This prompted us to refine the structure of mCA-CTD complexed with the cis isoform of NYAD-13 using published protocols ¹⁴ with new data acquired at relatively lower concentration of protein and peptide to maximize the available monomeric peptide. Since the peptide affinity for the protein (<10 µM) is relatively high compared to its propensity for self-association (>1 mM) the effect of concentration is not reflected in the binding site interactions. However the consequence of the cis isomer of the double bond is a change in the pitch of the two turns of the α–helical peptide. The calculated distance between the X4 and X8 Cα atoms in the trans isoform (6.8–7.8 Å) is slightly larger than the 6.0–6.5 Å range in the cis isoform. The shorter pitch of the peptide helix with a cis olefinic bond is in close agreement with the published 6.35 Å reported for the X-ray structure of the MCL-1 BH3 peptide. Pairwise hydrophobic side-chain interactions anchoring the helical peptide to the binding surface involving Helix-I and Helix II from mCA-CTD are intact in the refined structure and nearly identical to those observed for the linear CAI peptide ²⁶ and the trans peptide bound to the protein. Phe3 is a critical anchor, inserting itself into a hydrophobic pocket formed by Leu172, Phe168 from Helix I and Thr186 from Helix II in Figure 7b. Likewise Ile1, Leu6 and Tyr9 from the peptide are buried within a second hydrophobic pocket encircled by Val165, Leu211, Leu212 and Met215 respectively (Figure 7c). As concluded previously there is no evidence in the NMR structure of the complex for interactions between the hydrocarbon staple and the mCA-CTD in solution. Recent model structures of the mature capsid lattice would suggest otherwise, with likely contacts between the peptide and CA-NTD disrupting a crucial interface between the two domains ^27,28.

DISCUSSION

In conclusion the most significant finding of this study is the ability of stapled peptides that are soluble to assemble into polymeric entities with limited impact on biological activity. The nature of the polymerization is different from the non-specific association attributed to the intrinsically hydrophobic linker which limits the overall solubility of the peptides and becomes an issue when longer staples are incorporated into the peptide. Particularly in stapled peptides rich in aromatics, the hydrocarbon cross-link stabilizes the secondary structure of the peptide and also mediates stacking interactions between the peptide molecules to assemble into multimeric higher order structures. The examples of NYAD-36 and NYAD-203 suggest increasing the overall length of hydrophobic stapled helices to more than two turns may further enhance the intermolecular interactions.

The consequence of self-association of the stapled peptides is to reduce the available concentration of the monomers and block access to critical side-chains. The apparent immunity of the anti-viral activity of the NYAD peptides to polymerization/aggregation has two explanations. (i) All the peptides used in this study were designed from the CAI binding motif and therefore will have nominal activity against CA-CTD. The examples of NYAD-16 and NYAD-59 illustrate the intolerance of the sequence to mutations in the functionally important sites. (ii) Owing to sequence similarity and using the analogy of NYAD-13 we hypothesize essentially all the peptides within this family associate weakly (>1 mM) and successfully compete for the much higher affinity binding to CA-CTD at lower concentrations. This reasoning may not apply to NYAD-203 which dimerizes with affinity (~17 µM) comparable to the affinity of the peptide for CA-CTD dimers (~40 µM). Similarly the R_i,i+7S(11) peptides prone to aggregation are likely to compromise the affinity at even lower concentrations.

The exhaustive study of different NYAD peptides presents an interesting conundrum for rationale design. The aromatic residues in the binding motif are critical to the specificity of the peptide for its biological target but it also drives self-association through hydrophobic interactions. Therefore mutating key hydrophobic residues to mitigate polymerization is mutually incompatible with enhancing activity without forfeiting specificity as seen in the examples of NYAD-16 and NYAD-59. The impact of mutations at functionally redundant sites and the C-terminal tag appears to have nominal impact on eliminating self-association but influences the overall solubility of the peptides. As seen in the examples of NYAD-13 and NYAD-44 the C-terminal positively charged poly-lysine tag enhances solubility but at the same time favors cation-π stacking interactions between the peptides. The self association is not entirely reversed by increasing the negative charge which lowers the overall solubility of the peptide instead. It would appear then there is little or no real advantage to increasing the solubility of the stapled peptides at the cost of driving self association by the hydrophobic effect. Altering the position of the staple is of limited scope for short peptides with potential to destabilize the helix and eliminate key interactions with the target.

In summary the design of stapled peptides involves more than just achieving the desired helical structure guaranteed by the hydrocarbon staple. To function as an effector molecule against cellular targets requires a moderately soluble monomeric stapled peptide obtained by exploring a limited matrix of parameters to improve solubility and activity. At present the structural origin of the activity of the R_4,11S(11) cross-linked peptides is not completely understood given its strong propensity to aggregate and the effect of olefinic bond isomerization on helix stability has opened up avenues for future investigation.

The present study underlines the inadequacy and significance of extending the existing theoretical models used to simulate stereochemical properties of stapled peptides in the context of their structural diversity that extends beyond the formation of a monomeric helix ^10,15. Such information could be crucial for the iterative optimization of the lead peptides. Modifying the hydrocarbon staple itself by incorporating polar functional groups maybe a more permanent solution to overcoming the potential shortcomings of working with stapled peptides in drug design and clearly the way forward. The stapled peptide assemblies are of independent interest in potential applications related to nanobiotechnology ^21,29.

Materials and Methods

NMR Samples

The stapled peptides with greater than 90% purity were synthesized by Anaspec, CA and CPC Scientific, INC, CA. NMR samples were prepared by dissolving 1–5 milligrams of peptide in 100% D₂O and 90%H₂O/10% D₂O to achieve a final concentration in the millimolar range. The concentration of the peptides was measured from optical absorbance at a wavelength of 280 nm.

Peptide NMR Experiments

NMR-data for the free peptides were acquired at 25 °C on Bruker AVANCE spectrometers equipped with CryoProbes at various field strengths ranging from 500 to 900 MHz. The data were processed in Topspin 2.1 from Bruker Biospin and analyzed using CARA1.5 ³⁰. A standard suite of homonuclear two dimensional experiments DQF-COSY, NOESY with mixing time 100–200 ms and TOCSY with mixing time 70 ms were acquired for chemical shift assignments ³¹. To complete the heteronuclear carbon chemical shift assignments, natural abundance ¹H - ¹³C/¹⁵N HSQC and long-range ¹H -¹³C HMBC spectra were acquired. Standard DOSY experiments (LED) with compensation for eddy currents with bipolar gradients were acquired to measure the diffusion constants ³².

Peptide Structure Calculations

Free peptide structures were calculated from distance and dihedral restraints using the programs CYANA 2.1 ³³ and refined using the CNS force field in ARIA 2.2 ³⁴. The topological parameters for the non-standard amino acid with cis and trans olefinic bond were generated using a previously published protocol ¹⁴. Five hundred structures were calculated and the statistics of the twenty lowest energy structures reported in the Supplementary Table 2.

Dynamic Light Scattering Experiments

The dynamic light scattering of the peptides at various concentrations was measured using the DynaPro NanoStar from Wyatt Technology. Typically concentrated peptide samples were diluted into D₂O or H₂O and spun at 6,000–8,000 g for 10 m before loading 50 µl into disposable plastic cuvettes. One hundred scans of 5 seconds were acquired on each sample at ambient temperature. The data was processed using default settings in the Wyatt software corrected for solvent signal.

Refined Structure of NYAD-13 complex

Uniformly ¹⁵N/¹³C-enriched protein samples of a mutant monomeric form of CA-CTD (W184A, M185A) complexed with unlabeled NYAD-13 peptide (a soluble analogue of NYAD-1) were prepared by using published protocols ^5,14. The concentration of the protein in the complex is 680 µM and that of the peptide 700 µM respectively dissolved in 20 mM sodium phosphate, 90% H₂O/ 10% D₂O and 10 mM DTT at pH 6.9. Chemical shift assignments for the bound NYAD-13 peptide in complex with mCA-CTD were obtained from heteronuclear filtered NOE experiments ^35,36. The data was analyzed and restraints generated for the peptide bond structures of mCA-CTD using methods published previously ¹⁴. Statistics of the 20 best structures selected from a larger ensemble of 1000 structures sorted based on energy and restraint violations is presented in Supplementary Table 1. The coordinates of the refined structure has been deposited in the RCSB database (PDB code 2L6E) and chemical shift information in BMRB (17307).

Abbreviations

CA

Capsid

CA-CTD

wild type C-terminal domain of capsid

CD

Circular Dichroism

COSY

Correlation Spectroscopy

DLS

Dynamic Light Scattering

HIV-1

Human Immunodeficiency Virus type 1

HSQC

heteronuclear single-quantum coherence

mCA-CTD

monomeric and mutant form (W184A/M185A) of C-terminal domain of HIV-1 capsid protein

NMR

Nuclear Magnetic Resonance

NOESY

Nuclear Overhauser Spectroscopy

TOCSY

Total Correlation Spectroscopy

RCM

Ring Closing Metathesis

SDS

Sodium dodecyl sulfate