While most of the surface of gp41 is thought to be hidden within the native trimer prior to fusion, some epitopes of gp41 appear to be somewhat accessible during, and more so following, receptor activation, when gp41 switches from the native configuration through a pre-hairpin intermediate to the postfusion structure (2
). Specifically, the MPER of gp41 encompasses the epitopes for three neutralizing antibodies (4E10, 2F5, and Z13). However, immunogens incorporating MPER sequences have failed to elicit antibodies with the breadth and potency associated with these existing neutralizing antibodies (2
). One explanation for the failure of at least some of these immunogens is that the peptide epitopes have been minimized to the extent that they adopt a largely unstructured conformation in solution and that immunogens based on extended MPER sequences or those with greater constraints should be better candidates (2
). Alternatively, it has been proposed that to be effective immunogens, MPER sequences may require a membrane context, since the binding affinity of 4E10 and 2F5 to gp41 peptides increases in the presence of a membrane (28
). A final concern has been raised by Haynes et al., who suggest that 2F5 and 4E10 cross-react with autoantigens, such as cardiolipin, and that MPER epitopes could mimic autoantibody epitopes (18
). In this case, the B cells making antibodies to the MPER would be clonally deleted or suppressed, and this would explain the failure of MPER immunogens. An alternative explanation of any cross-reactivity observed, at least for 4E10, is that it arises from the highly hydrophobic nature of the binding site of this antibody (7
). In any case, it is clear, however, that a detailed characterization of the 4E10 peptide epitope and the synthesis of peptide antigens that mimic the structure of this epitope are valuable steps toward the use of the design of immunogens eliciting antibodies to the MPER.
We decided to focus on the human monoclonal antibody 4E10 as it is the most broadly neutralizing antibody described to date. A recent crystallographic study shows that the peptide KGWNWFDITNWGK (called “KGND”) adopts a largely helical structure with all of the crucial amino acids for binding being presented on the same side of the helix (9
) We refer to the residues that are not involved in binding to the antibody as the “nonneutralizing face,” in keeping with the same terminology used for the trimeric envelope spike (38
) (Fig. ). The 4E10 epitope is not trivial to mimic because it is not a perfect α-helix throughout its entire length and the crucial residues “WF” are in a 310
-helical structure, which is frequently observed to terminate α-helical structures. Thus, designing a perfect α-helix might not generate the optimal candidate for immunization.
The ability of a peptide to elicit a strong immune response is not predictive of its ability to elicit neutralizing antibodies. This problem has been encountered for 2F5 (12
). Furthermore, the affinity of an antibody for a particular antigen (antigenicity) is not necessarily predictive of the ability of the same antigen to elicit that antibody (immunogenicity). To develop an effective antigen, we envision a multistep strategy. Initially, the particular epitope is characterized: in this case, by first identifying the length of the peptide that gives the tightest binding to the antibody and then performing alanine mutations to find the key amino acids. The next stage consists of restricting the peptide conformation to the one adopted when bound to the neutralizing antibody (in this case, a helical conformation). This step has been satisfactorily achieved in the present study. The last stage will consist of the replacement of unnecessary parts with less immunogenic substituents to mask the “nonneutralizing face” without perturbing the constrained conformation. This final step will ensure that only the side of the helix which is involved in the binding to the antibody will be available to the immune system (Fig. ). Masking the “nonneutralizing” face is a principle that has been suggested to focus the immune response (28
). Finer modifications will be evaluated iteratively and empirically in subsequent stages.
FIG. 4. Schematic of the vaccine design process where constraints are introduced. (a) The peptides are constrained to a helix conformation via the introduction of an Aib or tether constraint. (b) The “nonneutralizing face” is blocked with the (more ...)
In this study, we have identified the optimal length of the peptide epitope as NWFDITNWLWYIK (residues 671 to 683). A peptide containing this epitope and a solubilizing tail has an IC50 of 10 nM in peptide competition experiments and a Kd of 20 nM (as measured via BIAcore) (Table ). This peptide also blocked neutralization of different HIV-1 strains by 4E10.
In order to identify permissive sites for further modification to the 4E10 epitope, an Ala scan was performed. Alanine substitution at residues W672, F673, and T676 resulted in a major loss of binding to 4E10 (over 1,000-fold decrease) (Fig. ). Because substitution in these positions also slightly increased the helicity (CD; see the supplemental material), the loss of binding does not then appear to result from a loss of helical structure. These binding results are in agreement with the 4E10 crystal structure in complex with the “KGND” peptide (9
), where W672, F673, and T676 make intimate contacts with the antibody. Mutation of these amino acids on the virus also decreased neutralization of the mutant virus by 4E10 (40
). Taken as a whole, these studies confirm that W672, F673, and T676 are important components of the 4E10 epitope.
Surprisingly, mutation of L679 to alanine resulted in a major decrease in 4E10 binding (70-fold). The importance of this residue could not have been foreseen from the crystal structure since L679 was replaced with a glycine spacer in the 13-mer used in the crystal data (9
). The mutation L679A resulted in a small decrease in the helical character of the peptide, but not enough to account for the observed 70-fold loss in binding. Therefore, we believe that L679 makes direct contact with the antibody. During the Ala scan on the virus, L679 was not found to be critical, as the L679A virus could still be neutralized. The differences in the effect of Ala substitutions on peptide/affinity versus HIV-1 neutralization by 4E10 are discussed below (40
The substitutions I675A and W680A also resulted in major increases in IC50
(20- to 30-fold). The importance of I675 is predicted from the crystal structure, where it was found that I675 makes contact with the antibody, but less than W672, F673, and T676 (9
). The I675 substitution also resulted in a slight decrease in helical character, which could have affected the 4E10 binding. W680A resulted in a pronounced loss of binding of the peptide to 4E10, while the helical character was improved. The importance of W680 had been seen in the mutation study performed on the virus, as the W680A mutation decreased neutralization of the mutant virus by 4E10 (40
) and was also suggested from the crystallographic analysis, even though the tryptophan had been replaced by a lysine to obtain a soluble peptide for crystallization (9
). Alanine substitutions on the “nonneutralizing face” usually did not result in major increases in IC50
, with the exception of D674A. The alanine substitutions N671A and D674A resulted in a disruption of the peptide conformation (CD; see the supplemental material). These two residues do not make contact with 4E10 in the crystal structure (9
); therefore, they apparently play an important role in stabilizing the structure of the peptide in a helical conformation. Also, the Ala scan on the virus shows that N671 and D674 are not critical for neutralization (40
). Similarly, mutations of I682 and K683 strongly decrease the helical content of the peptide (CD; see the supplemental material), which may explain the lower affinity of the respective mutants. These residues also play an important role in stabilizing the peptide structure, but probably do not make contact with the antibody. Finally, N677, W678, and Y681 could be mutated to alanine with no major effect on the binding affinity to 4E10 (increase of less than twofold) or on the peptide structure.
The Ala scan allowed us to refine the synthetic peptide epitope of 4E10 as NWFDITnwLWyIK, with the uppercase letters as important residues (among them W, F, T, and L are the major residues) and the lowercase letters as replaceable ones. We believe that appropriate modifications of residues as N677 and W678 (found on the nonneutralizing face of the helix) should not affect binding to 4E10 but could result in a reduction of the immunogenicity of this side of the helix. The importance of some of the residues concurs with mutagenesis experiments performed on the virus, in which neutralization resistance occurred with the substitutions W672A, F673A, and W680A (40
). However, in general, these results show how peptides in solution may behave quite differently from the corresponding region on a folded protein that is anchored to a membrane. In our study, both faces of the helix are exposed to water, whereas the “nonneutralizing” face on the virus may be interacting with neighboring protomers of gp41 or gp120 within the trimer or with the membrane. This difference in the surrounding environment could explain the differences between the Ala scans on the peptide and those on the virus. Moreover, the Ala substitutions in the viral protein may affect the entry kinetics of the virus, causing enhanced susceptibility of the virus to 4E10 without affecting the intrinsic affinity to the membrane-proximal external region epitope.
The next step of our strategy focused on limiting the conformational diversity of the peptides by designing analogs that are constrained to adopt a conformation in solution similar to that of the peptide bound to 4E10. In the crystal structure, the 4E10 epitope peptide is in a largely helical conformation. Peptides derived from the native gp41 sequence are generally helical in PBS buffer (Fig. , squares, right panel). In order to reduce alternative peptide conformations, constraints were introduced to further enhance this helical propensity through the use of cyclothioethers, lactams, and reversed lactam bridges, as well as Aib-containing analogs. The presence of a helical conformation is generally associated with strong 4E10 binding.
We introduced constraints closer to the N terminus of the sequence, initially forming thioether tethers (residues 670 to 674 or 671 to 674). These peptides did not show significant binding to the antibody. When we moved the position of cyclization toward the center or the C terminus to constrain residues 674 to 677 or 674 to 678, we saw an increase in binding: the cyclic ether formed between residues 674 and 677 is among our best derivatives (Table ). This result is in agreement with the crystal structure, as the peptide is more α-helical toward the center and the C terminus. The incompatibility of N-terminal tethers may be due to a steric clash with the 4E10 binding pocket or the transition of the α-helix to a 310
helix at the N terminus (9
In summary, cyclic and acyclic analogs (native or Aib containing) were identified in which the tight binding to 4E10 (10 nM) was maintained (Table ) and yet the possible backbone conformations adopted by the different analogs were restricted. Although, in some cases, further enhancement of helicity or structure did not increase 4E10 binding, we anticipate that the more rigid peptides will be more specific immunogens. Compatibility of an Aib substitution with tight 4E10 binding is very promising for the use of such peptides in the design of a vaccine. Not only does the presence of an Aib residue increase the helicity, it also destabilizes alternative conformations. Such stability may be particularly useful in the presence of denaturing adjuvants. In addition, Aib introduces a minimal structural modification, reducing the chances of directing an immune response to the constraint. Therefore, the sequences described here would appear to be useful candidates for immunization studies. The best analogs from each series (an Aib-containing peptide, a lactam, and a thioether) are now being assessed in immunization studies.