Optimization of the lead peptide E1P47 by designing a retro-enantio and stapled analogues
Previous work of our group defined a new peptide lead, namely E1P47, as an entry inhibitor with a broad spectrum activity against HIV-123. Particularly, E1P47 derives from the region 139–156 of the E1 protein of the Human Pegivirus which can be considered as commensal of humans since infections provide a beneficial effect on survival in HIV-1 positive subjects24. Several experimental data demonstrated that E1P47 can be considered as an inhibitor of HIV-1 Env fusion; (1) the peptide inhibited HIV-1 Env mediated cell fusion23; (2) the peptide was not able to inhibit an amphotropic vesicular stomatitis virus (VSV) Env pseudotyped on an HIV-1 core23 (3) peptide–peptide titrations and diffusion NMR spectroscopy demonstrated the specific interaction of E1P47 peptide with its viral target site, the HIV-1 fusion peptide25. In addition, reported structural work in DPC micelles demonstrated the importance of specific structural features for the anti-HIV-1 broad-spectrum potency of E1P4725. Two α-helices in both N- and C-terminal regions of the inhibitor peptide separated by a hinge region featured by a Pro residue were identified as structural elements required for maintaining the antiviral activity. Based on this structural background, in this work we propose the design and synthesis of new optimized analogs for increasing functionality in terms of efficiency and stability (Fig. 1). Since the arrangement of d-amino acids in a reverse sequence to the l-parent peptide can lead to a conformation that achieves a good mimicry with the l-peptide8, the retro-enantio version of the E1P47 (RE-E1P4) peptide was synthesized (Fig. 1B). Alternatively, to reinforce the α-helical conformation either on N- or C-terminus of L-E1P47, we designed two stapled peptides based on the previous structural study of the parent peptide25 which clearly demonstrated the existence of salt bridges between the Glu-4 and Lys-8 residues as well as between the Asp-12 and Arg-15 residues that favored the helical structures in the N and C-terminal segments. Thus, a stapled peptide was synthesized with a lactam bridge between Asp-12 and a Lys-15 that substituted the original Arg residue in order to stabilize the C-terminal α-helix (StP1-E1P47) and a lactam bond was formed between the residues Glu-4 and Lys-8 to obtain a cyclized version of the parent peptide, namely StP2-E1P47 (Fig. 1C,D).
First, we comparatively studied the conformational features of the synthesized analogues. Structural qualitative information for the stapled peptides was obtained by 1D 1H NMR (Fig. 2A) on DPC micelles. The 1H NMR spectrum of StP2-E1P47 resembles that of E1P47, with similar dispersion of NH amide (between 7.5 and 9.1) and indole protons (two NH indole Trp side chains are overlapped in proton spectrum, Fig. 2A), whereas the spectrum of StP1-E1P47 shows lower dispersion of amide protons.
Our previous studies showed that the E1P47 structure is stabilized by a quadrupole–quadrupole interaction between the two aromatic side chains of Phe-13 and Trp-14, giving a characteristic upfield shift of Phe-13 aromatic protons. The presence of this characteristic aromatic upfield shift was also observed in StP2-E1P47 and the RE-E1P47 but not in the StP1-E1P47 proton spectra. This lack of a shielding effect in the stapled peptide could be related to the disappearance of the electronic effect on the chemical shift due to the change of the relative orientation between Phe-13/Trp-14 aromatic side chains, due to the stiffness of the peptide backbone.
In addition, the secondary structure of stapled peptides and RE-E1P47 was studied by circular dichroism (CD) on DPC micelles and compared to that of the l-parent peptide E1P47. As described previously25 the CD spectrum of E1P47 peptide in DPC micelles showed bands located between 225 and 230 nm that could be attributed to interaction of the aromatic rings of Trp with the peptide backbone and stacking of aromatic rings26. In agreement to the NMR study, the band located at 229 nm could be related to the quadrupole–quadrupole interaction between the two aromatic chains of Phe-13 and Trp-14. As described by Woody et al.27 this band can be either positive or negative depending on the orientation of the Trp aromatic rings relative to the peptide backbone.
The αR conformation is predicted to give rise to moderately strong positive bands located near 225 nm. Thus, the CD spectrum of E1P47 exhibited a positive CD band at 229 nm that can be attributed to the contribution of the Trp indole rings when these are located close to a αR region. Moreover, the negative band at 208 nm reinforces the hypothesis of an α-helix as the main secondary structural component in the peptide. Similarly, the CD spectrum of StP2-E1P47 exhibits a stronger positive band at 229 nm and a more negative band at 208 nm (Fig. 2B). Accordingly, stabilization of the α-helix at the N-terminus did not seem to change the stacking of the two aromatic chains of Phe-13 and Trp-14, leading to a similar conformational profile as E1P47. In contrast, the CD spectrum of StP1-E1P47 while conserving the negative band at 208 nm, it did not exhibit a positive band at 229 nm, suggesting that although the peptide retained its helical conformation, the relative orientation between Phe-13/Trp-14 aromatic side chains was different to that observed in E1P47 and StP2-E1P47. Thus, the tethering of Asp-12 and Arg-15 at the C-terminus represented a more substantial change for the aromatic side chains orientation in the micellar environment. These CD results were totally consistent with the above chemical shift differences obtained by RMN. In addition, the CD spectrum of StP1-E1P47 exhibited characteristics associated with a 310-helix conformation with a ratio between 222 and 207 nm bands of approximately 0.428,29. The tethering between residues i (Asp12) and i + 3 (Lys15) on StP1-E1P47 could favor the formation of an hydrogen bond between i carbonyl and i + 3 amide NH, compatible with the observation of a 310-helix on CD spectrum, which is slightly tighter than an α-helix. Also, this could affect the relative orientation of Phe13 and Trp14 aromatic side chains. Lastly, the CD spectrum of the RE-E1P47 showed characteristics of α-helical conformation but in its specular image (left-handed α-helix) since it corresponds to a d-peptide (Fig. 2B).
In addition to the analysis of conformational features of the peptides, their susceptibility to human plasma proteases was also comparatively analysed. After incubation of each peptide with human plasma at 37 °C, RE-E1P47 remained practically unaltered for almost 24 h, meanwhile the L-parent peptide was about 50% degraded at 8 h. The cyclic nature of stapled peptides was expected to hamper the action of proteases30 and in fact, in our hands they presented higher stability compared to the parent peptide but lower than that of the retro-enantio peptide and after 24 h of incubation they exhibited even more dramatic protease lability (Fig. 3). Although the stapled peptides showed higher stability at the initial hours of incubation, after 24 h they were degraded by around 80%.
To better understand the metabolically vulnerable points of the peptides, we carried out an ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) analysis of the major degradation products at different times of serum incubation. According to the predicted cleavage sites by the PROSPER bioinformatics program31 (Supplementary Table S1) the main metabolites that were identified by mass spectrometry suffered from enzymatic hydrolysis of Glu-4 (cathepsin K), Gly-16 (matrix metallopeptidase), Trp-1 and Phe-11 (chymotrypsin A) and Trp-7 (cathepsin G). The degradation product of cathepsin G was observed in StP1-E1P47 but not in StP2-E1P47. This differential behavior could be attributed to the presence of the intramolecular cycle established between positions Glu-4 and Lys-8 in StP2-E1P47, which could confer inaccessibility to degradation by cathepsin G of Trp-7. Similarly, the product of degradation of Phe-11 by chymotrypsin A was only observed in StP2-E1P47, since in StP1-E1P47 the cleavage site is next to the cycle established between Asp-12 and Lys-15. It should be noted that all of these metabolites were observed for the L-E1P47; however, its retro-enantio analogue did not suffer most of these proteolytic degradation. The introduction of non-natural amino acids would hamper the action of proteases clearly contributing to RE-E1P47 enhanced stability.
Peptides candidates inhibit HIV-1 infection in pre-clinical cellular and mucosal tissue models
Epidemiological and genetic studies have shown that > 95% of sexually transmitted infections world-wide are due to R5-tropic viruses32,33,34. Hence, taking into account the predominant transmission of R5-tropic isolates compared with X4-viruses during sexual intercourse and our previous studies evaluating the potency of prototype fusion inhibitor T2035, we assessed the potency of the fusion inhibitor candidates against an R5-tropic isolate commonly used in pre-clinical studies, HIV-1BaL. The wild type (E1P47) and derivative peptides (RE-E1P47, StP1- E1P47 and StP2- E1P47) were first tested in TZM-bl cells. The three derivative peptides strongly inhibited HIV-1Bal infection (Table 1, Supplementary Figure S1). RE-E1P47 showed about 19-fold improvement over L-E1P47 antiviral activity while StP1-E1P47 and StP2-E1P47 showed about seven and fivefold improvement, respectively. Thus, the RE-E1P47 was the most active analogue with an average IC50 value in the range of nanomolar concentration.
The inhibitory activity of the four peptides was then assessed in a mucosal model based on ex vivo HIV-1 challenge of colorectal tissue explants36,37. Mucosal tissue explant models are becoming an important tool for pre-clinical screening of pre-exposure prophylaxis (PrEP) candidates and are increasingly used in early clinical trials38,39,40,41. These models assess the anti-viral potency of drugs candidates at the mucosal portal of HIV-1 transmission. Colorectal explants were treated with peptides before and during viral exposure (3 h) as a “pulse” condition to mimic drug dosing immediately prior to intercourse and during exposure; or throughout the culture as a “sustained” exposure to mimic the activity of the peptides when delivered from a sustained release formulation, such as an intravaginal ring. A dose–response curve was observed for all peptides in colorectal tissue explants against HIV-1BaL with both treatment conditions, pulse and sustained increased the inhibitory potency of the four peptides with a decrease of IC50 values (Table 1).
RE-E1P47 tended to reach higher levels of inhibition in both dosing conditions at the highest concentration tested; however, no significant differences were observed in the inhibitory potency of the four peptides in this tissue model (Supplementary Figure S1). Sustained exposure to peptides increased the inhibitory potency of the four peptides with a decrease of IC50 values (Table 1). One limitation of our study is the absence of peptide controls. Due to the SARS-CoV-2 pandemic new experiments are a significant undertaking that requires access to laboratories that are currently closed. This includes containment level 3 laboratories (CL3) for live HIV infection experiments of cells and tissue samples.
The restriction of the conformational flexibility on N- or C-terminal end of L-E1P47 by the tethering of the residues involved in the formation of salt bridges stabilizing α-helical conformations led to an increase of the anti-viral potency of the parent peptide. Moreover, RE-E1P47 that showed a characteristic left-handed α-helix by circular dichroism was the most active peptide in cellular assays and demonstrated a tendency to reach higher levels of viral inhibition in the tissue model. In addition to the peptide conformation, peptide susceptibility to proteases also determined the inhibitory potency of the analogues. Unlike, L-and stapled E1P47 peptides, the RE-E1P47 was highly resistant to human proteases. Thus, both conformation features and proteolytic stability mostly determines the functionality of the RE-E1P47.
RE-E1P47 assembles into the membrane maintaining the active conformation needed for its interaction with the HIV-1 target
Once selected RE-E1P47 as the most active and proteases-resistant peptide, a further structural characterization of the retro-enantio peptide as well as its assembly into the membrane and its subsequent ability to recognize the HIV-1 target were studied.
Structural studies of the RE-E1P47 in DPC micelles were carried out by NMR spectroscopy and computational modelling using the experimental NOE data as distance restrains. Proton chemical shift assignments of RE-E1P47 in DPC micelles (2 mM, at 308 K in aqueous solution with DPC-d38 at a detergent/protein ratio of ~ 100) were obtained using standard methods of peptide NMR spectroscopy as 2D NMR 1H–1H TOCSY, NOESY and ROESY experiments (Supplementary Table S2). The summary of the sequential and medium-range NOEs for the RE-E1P47 peptide in a solution of H2O/D2O with DPC-d38 micelles is shown in Supplementary Figure S2.
NMR experiments revealed several sets of conformations in slow exchange in chemical shift timescale for the retro-enantio peptide; however the broad resonances due to the presence of and peptide interaction with DPC micelles prevented the complete assignment of the NOE crosspeaks from minor conformations due to resonance overlapping.
For some amino acids, these resonances were clearly resolved, like the three components of Trp-12 (Fig. 4) or the tripling of Val-10 methyl protons resonances, whereas methyl proton’s from other residues showed no additional set of signals (Supplementary Figure S3). The chemical shifts of the additional signals could be reproduced in several sample preparations and were not concentration-dependent (1H NMR spectra of 2.0 and 0.2 mM RE-E1P47 in DPC-d38 micelles were identical, data not shown). One explanation for the observation of several sets of spin systems may be cis–trans isomerization of RE-E1P47 central proline. One of the new characteristics of RE-E1P47 compared to the parent peptide is the location of a phenylalanine (Phe-8) just before the proline residue. It is known that aromatic and prolines residues can interact locally and stabilize cis-prolyl-amide bonds, in particular aromatic–proline sequences, via both the hydrophobic effect and aromatic–proline interactions (CH–π interactions)42, and could increase the concentration of cis-proline conformation in percentages > 20% for peptides in aqueous solution43.
In a subsequent step, strong and medium NOEs measurements were used as an input to derive a structure of the peptide using restrained molecular dynamics (MD). Specifically, a total of 42 strong and 148 medium no redundant NOEs were used to impose distance constrains in the interval 1.8–2.8 Å and 1.8–3.8 Å, respectively. Supplementary Table S3 lists the pairs of atoms involved in the diverse NOE measurements, grouped by their intensity. Unfortunately, simultaneous use of the 190 distance restrictions did not provide any reasonable structure, suggesting that the peptide adopts several conformations in this environment. We also proceeded to carry out diverse restrained MD simulations using different subsets of restrains that did not provide any reasonable structure either.
Since the restricted MD did not provide a reasonable structure compatible with the experimental results, we proceeded to compute a 1 µs MD trajectory of the peptide unrestricted that was used to compute the values of a set of diverse atomic distances that could be contrasted with the NOEs measurements obtained from the NMR studies. In order to analyze the conformational features of the peptide along the MD trajectory, the structures sampled were clustered using the Linkage-average algorithm44. This analysis showed that the most populated structure (~ 85%) can be described as a helix–turn–helix similar to the structure exhibited by E1P4725 as shown in Fig. 5A. Analysis of a set of selected distances attained during the sampling process showed that although a majority complied with the NOE measurements (Supplementary Figures S4–S5), there were a few that did not. In order to increase the number of distance constraints fulfilled we also investigated the conformational features of the peptide when the peptide bond previous to the central proline is in cis. For this purpose we performed a short MD trajectory of the peptide at 900 K and annealed some of the structures through energy minimization. Subsequently, we selected one of the structures with the peptide bond before Pro in cis and ran a 1 µs MD simulation. After clustering the structures, a main structure (~ 35%) with a α-helical C-terminal segment and a turn at the Pro-9 (in cis) was found (Fig. 5B).
Interestingly, this structure fulfils distances 11 and 63 of the experimental NOE derived distances that were not fulfilled by the trans structure. Accordingly, we considered that the two structures coexist in solution. However, there were other NOE derived distances not fulfilled by any of the two structures. Hence, we proceeded to analyze the time evolution of those distances not satisfied by any of the two structures along the MD trajectory. These results are shown in Supplementary Figure S6. Analysis of these Figures allowed us to identify specific structures that satisfied the rest of the NOE derived distances. These conformations retained the helical structure at the C-terminus, but were more extended at the N-terminus.
Putting all this information together, the composed picture that emerged from the structural analysis carried out suggests that the peptide adopts a helix–turn–helix or an unstructured-turn–helix conformation with the peptide bond before Pro-9 in trans or in cis conformation, being the former preferred since during the 1 µs MD trajectory, the peptide bond was not isomerized. Moreover, the N-terminus appeared to be more flexible than the C-terminus (Fig. 5C). Actually, the latter was kept basically in a helical conformation, whereas the former alternated between a helix and more extended conformations.
Furthermore, peptide assembly on the membrane as well as recognition of its viral target on the membrane mimetic environment was studied by biophysical assays since it has been demonstrated that fusion inhibitor peptides specifically interfering with the N-terminal region of gp4121, need to be embedded into the membrane in order to interact properly with their viral target22.
Partitioning isotherms estimated from the fractional change in Trp fluorescence intensity upon addition of increasing amounts of palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) liposomes demonstrated that the affinity of both peptides for the lipid bilayer was almost equal (Fig. 6A). In addition, collisional quenching of Trp residues by brominated lipid vesicles showed a similar accessibility of both peptides to the hydrophobic inner part of the lipid vesicles (Fig. 6B). The quenching of Trp residues in the RE-E1P47 peptide by brominated atoms at positions 9 and 10 of the hydrophobic tails of the phospholipid implied that these Trp residues point to the micelle core in almost the same manner that in L-E1P47. Thus, peptide assembly on the membrane was a requirement that the RE-E1P47 peptide fulfilled to maintain the recognition of the viral target site.
In addition, the interaction between the RE-E1P47 and HIV-1 FP peptides in a membranous environment was studied by means of Förster Resonance Energy Transfer (FRET) assay. The HIV-1 FP labelled with 6-(7-nitrobenzofurazan-4-ylamino)hexanoic acid (NBD) was the donor peptide and the RE-E1P47 labelled with 5(6)-carboxy-tetramethyl-rhodamine (TAMRA) was the acceptor peptide. As shown in Fig. 6C, the FRET percentage efficiency between the donor (NBD-HIV-1 FP) and the acceptor (TAMRA-RE-E1P47) at a 1:1 peptide concentration ratio was the same that the previously demonstrated with the acceptor (TAMRA-E1P47)23. A FRET control experiment was carried out between donor NBD-HIV-1 FP and acceptor TAMRA-SCR-E1P47 peptides demonstrating that there was not interaction between them.
To further verify this interaction, RE-E1P47–FP peptide complex formation was monitored using 1D 1H NMR spectra. The results showed a number of chemical shift changes during the titration. These changes were more easily followed using the Tryptophan’s NH indole protons from RE-E1P47 peptide (Figs. 7a,b) and were similar to those previously observed with L-E1P4725. The set of results indicated that the RE-E1P47 peptide interacted with the fusion peptide of HIV-1 gp41 similarly as the E1P47 peptide did.
On the other hand, a titration of the scrambled control peptide (SCR-E1P47) with the fusion peptide was also followed by NMR experiments and previously published25. The results of the titration experiment indicated that SCR-E1P47 falls off the DPC micelles after HIV-1 FP addition, perhaps due to a minor affinity for lipid micelles that HIV-1 FP and/or different location in micelles (on the surface versus inserted). The diffusion coefficients values measured corroborated these results proving that SCR-E1P47 and HIV-1 FP could not bind simultaneously to lipid micelles due to the much higher affinity of HIV-1 FP to membrane environment. This could explain the lack of inhibitory activity for SCR-E1P47.
These results led us to the assumption that the conformational flexibility of the RE-E1P47 mainly in the N-terminal end did not hinder the recognition and its interaction with the viral target and, therefore, did not compromise its anti-viral potency. Thus, RE-E1P47 was able to assemble into the membrane and subsequently to interact with the HIV-1 target similarly to the lead peptide did.
The overall results reflect the importance of designing optimized peptides from the structural knowledge of the parent peptide. The peptide affinity and its subsequent assembly into the lipid bilayer for the correct recognition of the target sites are also key issues in the design of targeted inhibitors of protein–protein interactions that take place in the cell membrane.