Diverse Murine Vaccinations Reveal Distinct Antibody Classes to Target Fusion Peptide and Variation in Peptide Length to Improve HIV Neutralization

ABSTRACT

While neutralizing antibodies that target the HIV-1 fusion peptide have been elicited in mice by vaccination, antibodies reported thus far have been from only a single antibody class that could neutralize ~30% of HIV-1 strains. To explore the ability of the murine immune system to generate cross-clade neutralizing antibodies and to investigate how higher breadth and potency might be achieved, we tested 17 prime-boost regimens that utilized diverse fusion peptide-carrier conjugates and HIV-1 envelope trimers with different fusion peptides. We observed priming in mice with fusion peptide-carrier conjugates of variable peptide length to elicit higher neutralizing responses, a result we confirmed in guinea pigs. From vaccinated mice, we isolated 21 antibodies, belonging to 4 distinct classes of fusion peptide-directed antibodies capable of cross-clade neutralization. Top antibodies from each class collectively neutralized over 50% of a 208-strain panel. Structural analyses – both X-ray and cryo-EM – revealed each antibody class to recognize a distinct conformation of fusion peptide and to have a binding pocket capable of accommodating diverse fusion peptides. Murine vaccinations can thus elicit diverse neutralizing antibodies, and altering peptide length during prime can improve the elicitation of cross-clade responses targeting the fusion peptide site of HIV-1 vulnerability.

IMPORTANCE The HIV-1 fusion peptide has been identified as a site for elicitation of broadly neutralizing antibodies, with prior studies demonstrating that priming with fusion peptide-based immunogens and boosting with soluble envelope (Env) trimers can elicit cross-clade HIV-1-neutralizing responses. To improve the neutralizing breadth and potency of fusion peptide-directed responses, we evaluated vaccine regimens that incorporated diverse fusion peptide-conjugates and Env trimers with variation in fusion peptide length and sequence. We found that variation in peptide length during prime elicits enhanced neutralizing responses in mice and guinea pigs. We identified vaccine-elicited murine monoclonal antibodies from distinct classes capable of cross-clade neutralization and of diverse fusion peptide recognition. Our findings lend insight into improved immunogens and regimens for HIV-1 vaccine development.

INTRODUCTION

A long-standing immunological challenge facing the HIV-vaccine field is how to elicit antibodies capable of neutralizing the divergent and neutralization-resistant strains of HIV-1 that are circulating in the almost 40 million humans that are currently infected (1). While multiple antibodies have been identified from natural infection that recognize much of the exposed surface of the HIV-1 envelope (Env) trimer (2), only a few vaccine-elicited monoclonal antibodies from wildtype standard-vaccine test species have been reported, capable of cross-clade neutralization.

In one report, rabbits primed with a prefusion-stabilized Env trimer with 4 glycans surrounding the CD4-binding site (CD4bs) removed and boosted sequentially with CD4bs glycans-restored on heterologous and/or wildtype trimers could elicit broadly neutralizing antibody responses (3); two antibodies were isolated, one against the CD4-binding site with 25% neutralizing breadth on a 208-HIV-1 strain panel, and a second against a gp41-gp120 interface region with 87% breadth on the same panel. In a second report, rhesus macaques were primed with glycan-altered Env trimer immunogens, designed to induce responses against the V3-glycan site; after a prime and 4 boosts, 9 monoclonal antibodies were identified with weak heterologous breadth on a 19-strain panel (4, 5). In a third report, rhesus macaques were primed with the N-terminal 6 to 8 residues of the fusion peptide (FP) conjugated to keyhole limpet hemocyanin (KLH) and boosted with DS-SOSIP stabilized Env trimers (6); dozens of FP-directed antibodies were isolated, with the best neutralizing 59% of the 208-HIV-1 strain panel. Further, in an earlier report (7) utilizing C57BL/6 mice and the same FP immunogens as the NHP study, multiple FP-directed murine antibodies were isolated, with the best neutralizing 32% of the 208-strain panel.

These reports begin to establish the ability of HIV vaccines to elicit broadly neutralizing antibodies against the CD4-binding site, against the V3-glycan site, and especially against FP. To identify immunogens and regimens capable of improving breadth, potency, or consistency of the FP-directed neutralization response, we used C57BL/6 mice for screening and tested 17 different fusion peptide-based immunization regimens. In addition to assessing serum responses and correlating outcomes with immunogens and regimens, we also isolated antibodies from immunized mice. While isolated antibodies could not provide statistically validated insight into vaccination regimens, genetic and structural analyses of the antibodies did provide insight into the diversity of the murine FP-directed antibody responses. Overall, our results demonstrate that mice can be a useful model for FP-vaccine studies and that variation in FP length during priming can positively impact the final vaccine outcome.

RESULTS

Diverse immunization regimens incorporating FP-conjugates and Env trimers induce FP-directed immune responses in mice.

The N-terminal 6 to 10 residues of FP conjugated to carrier proteins, such as keyhole limpet hemocyanin (KLH), have been demonstrated to be capable of eliciting FP-directed antibodies (6–8), and the N-terminal eight residues of FP conjugated to recombinant tetanus toxoid heavy chain (FP8-rTTHC) has proven to be a promising candidate immunogen for clinical assessment (9). But is there a specific immunogen or combination with Env trimer that might lead to especially high titers of neutralizing antibody? To explore the effect of altering immunogens and immunization regimens on improving the FP-directed immune responses, we assembled a panel of immunogens that included 11 FP-conjugates of varying FP length, sequence, and carrier protein (Fig. 1A and Fig. S1), along with 15 prefusion-stabilized HIV-1 Env trimer variants from clades A and C with varying FP (Fig. 1A and Fig. S2), and we immunized 17 groups of C57BL/6 mice (n = 6/group) using 17 prime-boost regimens (Fig. 1).

FIG 1.

Diverse immunization regimens incorporating FP carriers and Env trimers induce robust FP-directed immune responses in mice. Seventeen groups of mice were immunized with distinct prime-boost regimens using diverse fusion peptide-carrier and Env trimer immunogens. Geometric mean ID₅₀s displayed in red font are from one week after final immunization as highlighted with red arrows. (A) Variables incorporated into immunogens. (B) Eight 12-week regimens with seven immunizations at identical intervals. (C) Eight extended regimens using multiple boosts with diverse trimers of matching FP sequences. (D) An extended regimen with alternating FP-carrier and trimer. Additional details for immunogens are described in Figs. S1 and S2.

These 17 prime-boost regimens comprised ‘short’ (12 weeks) (Fig. 1B) and ‘long’ (up to 32 weeks) (Fig. 1C and D) schemes. These schemes tested different sequential combinations of the immunogens, with a focus on variation of the fusion peptide itself. Our hypothesis was that a specific sequential order of immunogens might be efficacious, and we tested specific regimens such as focusing on the highly conserved N-terminal six residues of the fusion peptide or alternating FP-immunogen and Env trimer (Fig. 1D) that have been suggested by theoretical models of epitope focusing or affinity maturation may be especially efficacious.

We monitored immune response at a final time point at one week following the last immunization for all groups by testing serum neutralization of a BG505 variant missing a glycan at position 611 (BG505.N611Q), which is 10 to 100 times more sensitive to FP-directed neutralization than the wildtype BG505 virus (7, 10). Final geometric mean BG505.N611Q neutralizing titers ranged from 82 to 1819 ID₅₀, indicating that some regimens did indeed elicit much higher titers than others.

Among the short regimens, the groups immunized in weeks 2, 4, and 6 with a combination of rTTHC and KLH conjugated FPs of variable lengths (Fig. 1B, Groups 4 and 5) elicited the highest immune response as evidenced by the endpoint neutralization titers of BG505.N611Q (geometric mean titers of 1208 for Group 4 and 684 for Group 5). Among the long regimens, the group boosted with a FP8_I513-KLH variant followed by diverse trimers with matched FP sequences (Fig. 1C, Group 10) elicited the highest immune response with a geometric mean titer of 1819 against BG505.N611Q. Group 12, with a related regimen but with a FP8_I513/L515 sequence, exhibited a geometric mean neutralization titer of 1207, suggesting that FP sequence with I513 might be effective in eliciting strong FP-directed immune response in mice. However, we did not observe a similar trend in groups 14 and 16 (geometric mean titers of 229 and 181, respectively) that also incorporated boosts with multi-clade trimers with I513, suggesting other factors may contribute to the immune response. We observed a weak trend between the number of immunizations (as well as the length of regimen) and BG505.N611Q neutralization titers, with geometric mean ID₅₀ titers against BG505.N611Q increasing from 255 for short regimens (12 weeks) to 400 for long regimens (20-32 weeks) (P value = 0.1284).

Assessment of FP-directed responses reveals moderate tolerance for FP sequence variation.

To assess more comprehensively the FP-directed response as well as the recognition of diverse FP sequences, we compared sera collected one week following the final immunization for neutralizing activity against wildtype BG505 and BG505.N611Q incorporating the three most prevalent FP sequences (referred to as FP_v1, FP_v2, and FP_v3) (Fig. 2A). Sera from at least 2 animals in each of the 17 groups could neutralize the BG505.N611Q (FP_v1) pseudovirus, and in each of 13 groups, sera from at least 1 animal could neutralize fully glycosylated wildtype BG505. While the majority of responses were directed against FP_v1, sera from 1 or more animals in 12 of these groups could also neutralize BG505.N611Q viruses with the 2nd and 3rd most prevalent FP sequences. In a majority of the immunization groups, at least 1 animal had BG505.N611Q-neutralizing ID₅₀ titers of over 1000 that were at least 5-fold higher than that against wildtype BG505, indicative of the elicitation of FP-directed responses.

FIG 2.

Elicited serum neutralization and comparison of murine responses reveal ways to improve neutralizing response. (A) Serum neutralization (ID₅₀) of BG505 WT, BG505.N611Q_FP_v1, FP_v2, and FP_v3 variants, and SVA-MLV viruses at 1 week after the final immunization is shown for each of the 17 groups (n = 6 mice/group). Animals that were selected for antibody isolation are highlighted in light blue. (B) Comparison of geometric mean ID₅₀ titers against three BG505.N611Q viral variants obtained after the final immunization is shown for the following immunization variables: FP length and carrier variation in conjugates used for priming (B, left), FP sequence variation in conjugates and trimers used for boosting (B, center), and FP sequence variation in trimers used for boosting (B, right). For calculating the geometric mean ID₅₀ for the three BG505.N611Q_FP variants, entries with ID₅₀ < 50 were given a value of 50. BG505 WT and BG505.N611Q titers after SVA-MLV subtraction were used for statistical analyses. Geometric means are shown by horizontal lines and numerical values. P values were calculated with two-tailed, non-parametric Mann-Whitney t test.

Sera of selected top responders were collected following terminal boosts performed 4 weeks after the last immunization and assessed for neutralization against a panel of 10 wildtype cross-clade HIV-1 strains (8) containing diverse FP sequences (Table S1). Overall, there was limited heterologous neutralization detected, with sera from 8 of the 19 selected animals exhibiting neutralizing activity against at least 2 strains, and 2 animals from Group 10 neutralizing 4 wildtype strains (Table S1). One strain did prove especially sensitive to neutralization with 13 of the 19 selected animals neutralizing TH023.6, a clade AE Tier 1A neutralization-resistant strain containing the fifth most prevalent FP sequence.

Comparison of serum neutralization reveals that multiple variables lead to enhanced immune responses.

We compared groups to evaluate the effect of immunization variables on the outcomes related to breadth and potency of neutralizing responses. We did not observe a specific immunogen to be associated with especially high titers. Some specific regimens, however, did appear to elicit significantly higher neutralizing titers.

Regimens that used FP-conjugates incorporating sequential variation in FP length and diverse carriers prior to boosting with trimers (Groups 4 and 5) elicited higher neutralization responses against BG505.N611Q viruses than regimens with FP8-rTTHC as an immunogen for weeks 2 to 6 immunizations (Groups 1 to 3) (P = 0.0049) (Fig. 2B, left panel), suggesting that FP length variation and/or mixing carrier proteins could enhance FP-directed responses. Mice boosted with varying FP sequence in both FP-conjugates and trimers showed higher neutralizing titers against BG505.N611Q than those with a similar regimen but without FP sequence variation (Groups 10 to 12 versus Group 9, P = 0.0183), whereas there was no significant effect on neutralization when only boosting with trimers containing variation (Groups 13 to 16) (Fig. 2B, middle and right panels). Several groups, primarily from longer regimens incorporating multiple, consecutive trimer boosts (Groups 13 to 16), showed high background neutralizing activity against the non-HIV virus, SVA-MLV, thereby complicating the analysis for these groups. Overall, our results indicated that variation in FP length and sequence and mixing of carrier proteins in FP-conjugates could potentially improve elicitation of FP-directed neutralization responses.

FP length variation in priming expands neutralization breadth in guinea pigs.

To verify the effect of varying FP lengths and mixing carrier proteins in FP-conjugates on immunization outcomes, we immunized 5 groups of guinea pigs (n = 8/group) using FP priming regimens incorporating variation either in FP length or in both FP length and carrier protein (Fig. 3A). Groups 2 and 5 evaluated the FP10-8-6 regimen, which showed higher responses in mice. We also included an FP8-7-6 prime regimen (Group 3), which has shown previously to elicit cross-clade neutralizing responses in multiple animal models (7). Guinea pigs were immunized at weeks 0, 4, and 8 with either FP-rTTHC conjugates (Groups 1 to 3) or mixtures of FP-rTTHC and FP-KLH conjugates (Groups 4 and 5). All animals were boosted twice with BG505 DS-SOSIP.664 at weeks 12 and 16, followed by a heterologous trimer boost with ConC-FP8v2 RnS-3mut-2G-SOSIP.664 trimer at week 20 and a trimer cocktail at week 24. ConC-FP8v2 RnS-3mut-2G-SOSIP.664 is a consensus clade C, prefusion-stabilized HIV-1 trimer containing an FP8_v2 substitution corresponding to the second most prevalent fusion peptide sequence (AVGLGAVF) (11).

FIG 3.

FP length variation enhances breadth and potency of FP-vaccine-elicited responses in guinea pigs. (A) Guinea pig immunization regimen and groups. (B) Anti-FP (B, top) and anti-BG505 trimer (B, bottom) ELISA responses are shown for each group. Immunization time points are indicated by vertical dotted lines. (C) Week 26 serum neutralization titers (ID₅₀) are shown for each group against a panel of 9 heterologous wild type strains containing the FP8_v1 sequence. (D) Statistical comparison of magnitude (D, left) and breadth (D, right) of week 26 serum neutralizing responses (ID₅₀) against the heterologous 9-isolate panel used in (C) are shown. Neutralization was considered positive if ID₅₀ ≥ 20. For each group, data for animals with SIV background neutralization ID₅₀ ≥ 20 are shown as open symbols and these values were excluded from the analysis. A non-parametric permutation test was used for the magnitude analysis and a Mann-Whitney U test was used for the breadth analysis (see methods for additional details).

Longitudinal development of immune responses to FP and trimer was assessed by enzyme-linked immunosorbent assay (ELISA) (Fig. 3B and Fig. S3). Overall FP ELISA titers increased through the first 2 FP immunizations and generally plateaued after week 6 through the BG505 trimer boosts. FP ELISA responses decreased after boosting with the heterologous ConC-FP8v2 trimer before increasing and reaching peak titers following the final trimer cocktail immunization for all groups. Anti-BG505 trimer ELISA titers increased after the second FP immunization and continued to rise steadily through subsequent immunizations. Anti-trimer ELISA responses peaked 2 weeks after the final immunization for all groups except for Group 3 (FP-8-7-6 prime), which showed the highest titers at week 18 after the second BG505 trimer boost. In addition, all groups generated immune responses against heterologous trimers derived from Q23, BL01 and 25710 strains, as assessed by ELISA at 2 weeks after the final immunization (week 26), with overall higher responses detected in groups immunized with FP-rTTHC conjugates (Groups 1 to 3) compared to groups immunized with a mixture of FP-rTTHC and FP-KLH conjugates (Groups 4 and 5) (Fig. S4A). Sera from all groups competed with FP-directed antibody VRC34.01 for binding to trimers (Fig. S4B), indicating that immunizations elicited responses targeting the FP site of vulnerability.

Neutralizing responses against wildtype BG505 and BG505.N611Q viruses were detected in all groups at week 10 after FP priming (Fig. S5A), with groups using variation in FP length and/or carrier proteins (Groups 2 to 5) eliciting higher titers against BG505.N611Q compared to no variation in FP priming (Group 1) (Fig. S5B). To evaluate breadth of elicited neutralization, sera at week 26 were assessed for neutralization against a diverse panel of 10 cross-clade, wildtype HIV-1 pseudoviruses containing the FP8_v1 sequence (Fig. 3C and Fig. S6). All groups elicited consistent, robust neutralizing responses against the autologous wildtype BG505 virus, and sera from a majority of animals in each group neutralized at least 5 out of 10 cross-clade, wildtype viruses. Neutralization potency was significantly higher in Groups 2 and 3, which were primed with FP-rTTHC conjugates using sequential FP length variation (FP10-8-6 or FP8-7-6), compared to Group 1, which was primed with FP8-rTTHC (P = 0.006 and P = 0.028 for ID₅₀, respectively) (Fig. 3D and Fig. S4). Sequential immunizations with FP-rTTHC conjugates containing varying FP lengths (Group 2, FP10-8-6) also elicited significantly higher neutralization breadth against the 9 heterologous viruses compared to immunizations with identical FP8-rTTHC (Group 1) (P = 0.016 for ID₅₀). Immunizing with a mixture of FP-conjugates with different carrier proteins without FP length variation (Group 4) showed a modest improvement in neutralization potency compared to immunization with FP8-rTTHC (Group 1) (P = 0.029). By contrast, priming with a mixture of FP-conjugates containing a combination of FP length variation and different carrier proteins (Group 5) did not significantly improve neutralizing responses (Fig. 3D and Fig. S6). Frequencies of antigen-specific B cells positive for both FP and BG505 were measured at weeks 10 and 27 using PBMC (Fig. S5C and Fig. S6B). Groups 3 to 5 showed significantly higher FP/BG505 double positive B cell frequencies than Group 1 following three FP primes at week 10, consistent with the significantly higher serum neutralization titers elicited against BG505.N611Q at week 10 (Fig. S5). FP/BG505 double positive B cell frequencies at week 27 showed a similar trend (Fig. S6B). These guinea pig immunization results confirm the observation from the mouse studies that sequential immunization with FP-conjugate immunogens containing varying FP lengths could improve FP-directed immune responses.

Isolated murine monoclonal antibodies reveal multiple classes of fusion peptide antibodies.

To identify and characterize vaccine-elicited murine antibodies, hybridomas were generated from top responding animals from each group (Fig. 2A), as assessed by neutralization potency and breadth against BG505.N611Q variant viruses. Hybridomas were selected for antibody isolation based on ELISA binding to both FP8-1M6T peptide-scaffolds and BG505 DS-SOSIP trimer variants, and 21 antibody sequences were selected for further characterization (Fig. 4A and Fig. S7). Antibodies were named by mouse-lineage.clone, with sequence analysis revealing 16 different lineages; we named these vFP45-vFP60, continuing the lineage numbering of prior murine vaccine-elicited fusion peptide-directed antibodies (7). We assigned these antibodies based on their V-gene usage and heavy and light chain third complementary determining region (CDR H3 and CDR L3) sequences to 4 ‘antibody classes’, with antibodies of the same class utilizing similar germ lines and having similar recognition (12, 13).

FIG 4.

Characterization of vaccine-elicited murine antibodies reveals diverse classes of fusion peptide antibodies. (A) Genetic characteristics of vaccine-elicited murine antibodies exhibiting recognition of specific FP sequences, as assessed by ELISA against FP-1M6T scaffolds and BG505 DS-SOSIP.664 trimer variants. Non vFP1 antibody classes are highlighted in color and bold. (B) Distribution of V-genes used by vaccine-elicited FP-directed antibodies, as displayed on the phylogenetic trees of heavy and light chain V gene sequences.

Vaccine-elicited antibodies isolated from Groups 6 and 7 belonged to 2 new classes, vFP48 and vFP49 (named for the potent antibody first discovered), while the remaining antibodies were assigned to previously reported vFP1 or vFP5 classes (7), based on their usage of the same V-genes and conservation of contact amino acids (Fig. 4A and Table S4).

The newly identified vFP48 class used germ line genes VH1-55 and KV1-135, which were close phylogenetically to the germ line V-genes utilized by the vFP1 class (Fig. 4B). By contrast, the newly identified vFP49 class used germ line genes VH5-9 and KV5-43, which were more distal phylogenetically from germ line V-genes used by other murine neutralizing antibodies (Fig. 4B).

Overall, the appearance of different neutralizing antibody classes appeared to be stochastic in nature, and too rare to correlate statistically with a particular immunogen or immunization regimen feature. The varied immunization regimens employed in this study thus appear to have elicited primarily antibodies of the vFP1 class, though several other murine neutralizing antibody classes were isolated.

Vaccine-elicited FP-directed murine antibodies collectively neutralize > 50% of a cross-clade panel of 208 strains.

We assessed the ability of the 21 FP-directed antibodies to neutralize three BG505.N611Q variant viruses and a cross-clade panel of 19, tier 1 and 2 wildtype neutralization resistant and sensitive viruses (8) containing the 6 most prevalent FP8 sequences (v1:AVGIGAVF, v2:AVGLGAVF, v3:AIGLGAMF, v4:AVGTIGAMF, v5:AVGIGAMI, and v6:AVGIGAMF) (Fig. 5A). The newly identified antibodies exhibited cross-clade neutralization on a 19-strain diagnostic panel (8); 13 of the 21 antibodies could neutralize greater than 7 strains. The antibodies of the vFP48 class neutralized all three BG505.N611Q variants with IC₅₀ values of 0.003 to 0.011 μg/mL, albeit with limited breadth (2 to 5 out of 19 wildtype viruses). One antibody of the vFP49 class, 8495-vFP49.02, also neutralized the 3 sensitive BG505.N611Q variants with a preference for the FP8_v2 sequence with IC₅₀ of 0.08 μg/mL and exhibited greater breadth (8 out of 19 wildtype viruses). Interestingly, vFP52 lineage antibodies, which appeared based on V-gene and CDR H3 and L3 analyses to be members of the previously identified weakly neutralizing vFP5 class, neutralized BG505.N611Q variants with IC₅₀ values of 0.08 to 3.32 μg/mL and 8 out of 19 wildtype viruses of the 19-strain panel, much broader and more potently than the previously identified vFP5-class antibodies. Of the newly identified antibodies of the vFP1 class, 7291-vFP45.01, 7321-vFP53.02, and 7339-vFP56.01 exhibited the greatest breadth, neutralizing 13 out of 19 wildtype viruses, with 7321-vFP53.02 being the most potent among the three (Fig. 5A).

FIG 5.

Vaccine-elicited FP-directed antibodies collectively neutralize >50% of a cross-clade panel of 208 strains. (A) Neutralization on a representative 19-strain panel with diverse FP sequences. (B) Neutralization dendrograms for vaccine-elicited antibodies on the 208-strain panel. Viral strains are represented by individual branches, colored according to neutralization potency (non-neutralized branches shown in gray). (C) Dendrogram shows the predictive collective breadth of vFP48.01, vFP49.02, vFP52.02, vFP53.02, vFP20.01, and vFP16.02. (D) Neutralization-fingerprint dendrogram calculated from 208-strain panel neutralization data. Antibodies vFP48.01, vFP49.02, and vFP52.02 (italicized) have <20% breadth and, thus, their fingerprint analysis is likely to be less accurate. (E) Analysis of FP sequence preference for murine, human, and NHP antibodies; greater than 50% sequence preference for a FP sequence is highlighted in pink.

Three antibodies, 7551-vFP48.01, 8495-vFP49.02, and 7314-vFP52.02, each from a distinct class, and the best vFP1 class antibody identified in this study, 7321-vFP53.02, were further assessed on a panel of 208 Env pseudoviruses, encompassing diverse strains of primary isolates from 8 major clades (Fig. 5B and Table S2) (14). At 50 μg/mL, antibody 7551-vFP48.01 neutralized 27 strains (13% breadth), 8495-vFP49.02 neutralized 36 strains (17% breadth), and 7321-vFP53.02 neutralized 47 strains (23% breadth). Notably, antibody 7314-vFP52.02, which was a member of the vFP5 antibody class, neutralized 31 strains (15% breadth), many more than the first member of the class, 1868-vFP5.01, which neutralized only 2 strains of the 208-strain panel (7). As in the case of previously identified murine FP-directed antibodies (2712-vFP16.02 and 2716-vFP20.01), neutralization by the newly identified antibodies extended to multiple clades, including clades A, B, and C. For 7551-vFP48.01, neutralization also extended to clades D, CD, and G, while 8495-vFP49.02 could neutralize several clade AG strains (Fig. 5B and Fig. S8). Overall, each of the 4 antibodies neutralized a distinct set of viruses with some overlap. Notably, strains neutralized by 7551-vFP48.01, 8495-vFP49.02, 7314-vFP52.02, and 7321-vFP53.02 antibodies identified in this study and previously identified 2712-vFP16.02 and 2716-vFP20.01 antibodies (7) combined to a total of 107 of the 208-strain panel, corresponding to ~51% neutralization breadth, collectively, at 50 μg/mL for the vaccine-elicited murine antibodies (Fig. 5C).

To understand the neutralization pattern of the antibodies, we analyzed their neutralization fingerprints (15) relative to those of other known FP-directed antibodies and representative antibodies targeting other sites of vulnerability (Fig. 5D). While most FP-directed antibodies clustered together, 7551-vFP48.01 clustered with V3-glycan and CD4bs antibodies instead, likely a consequence of its low (13%) neutralization breadth resulting in a less accurate fingerprint analysis. Further, 7321-vFP53.02 clustered closely with previously identified vFP1-class antibodies, 2712-vFP16.02 and 2716-vFP20.01, and NHP 0PV lineages antibodies, and together, they were related closely with the new member of the vFP5-class antibody, 7314-vFP52.02. Antibody 8495-vFP49.02 was more distantly associated with a large cluster encompassing murine, NHP, and human FP-directed antibodies. In general, the neutralization fingerprint dendrogram was consistent with the preferences of antibodies for viruses with the 5 most prevalent FP sequences (Fig. 5E). Notably, antibody 7551-vFP48.01, displayed a distinctive neutralization fingerprint and could neutralize viruses containing four of the five most prevalent FP8 sequences in the 208 strain panel, somewhat analogous to the infection-elicited antibody PGT151 (16), which also neutralizes four out of the five most prevalent FP sequences (Fig. 5E). Antibody 8495-vFP49.02 showed a preference for viruses with FP_v2 sequence, with some neutralization activity against viruses with FP_v1 or FP_v5. Antibody 7321-vFP53.02 exhibited a preference for FP_v1 and FP_v5 sequences similar to other vFP1-classes antibodies, whereas antibody 7314-vFP52.02 neutralized FP_v1 and FP_v2 viruses. Overall, the diverse immunization schemes yielded multiple FP-directed antibodies, with top antibodies from each of the different antibody classes collectively neutralizing over 50% of our 208-strain panel.

Structural characterization of vaccine-elicited antibody complexes reveals vFP48 antibody class to be capable of accommodating diverse fusion peptide sequences.

To gain structural insight into FP recognition by the new vFP48-antibody class, we determined the crystal structure of the antigen-binding fragment (Fab) of 7551-vFP48.03 in complex with the FP8_v2 peptide (AVGLGAVF) at 2.4-Å resolution (Fig. 6A and Table S3). The first 5 residues (AVGLG) of FP had well-defined electron density, whereas the last 3 residues of FP lacked electron density, likely indicating that they were not engaged with the antibody and were disordered in the crystal. FP adopted a hook conformation and nestled in a binding pocket lined with hydrophobic residues from CDRs H1, H2, H3, L1, and L3 (Fig. 6A). The N terminus of FP (Ala512_FP) was anchored via a salt bridge and hydrophobic interactions to Asp27D_vFP48 and Tyr32_vFP48 side chain in CDR L1 (for clarity, we list residue numbers with molecule or antibody lineage, with antibodies numbered according to Kabat nomenclature [17]). FP residues Val513_FP and Leu515_FP, were stabilized by a series of hydrophobic interactions with CDR residues: Val513_FP with Phe94_vFP48 and Pro95_vFP48 of CDR L3 and Trp100D_vFP48 of CDR H3; Leu515_FP with Ala100A_vFP48 of CDR H3. Additional interactions that stabilized the bound form of FP included several hydrogen bonds between main-chain amide and carbonyl oxygens of FP and Gly91_vFP48 in CDR L3 and Trp33_vFP48, Thr98_vFP48, and Trp100D_vFP48 in CDRs H1 and H3 (Fig. 6A).

FIG 6.

vFP48 antibodies have binding pockets that accommodate diverse fusion peptide sequences. (A) Crystal structure of vFP48.03 Fab bound to FP8_v2 peptide (AVGLGAVF). Heavy chain is shown in light blue, light chain in gold, and peptide in gray. Hydrogen bonds are depicted as dotted lines. (B) Cryo-EM reconstruction of vFP48.02 Fab bound to BG505 DS-SOSIP (FP8v1) trimer showing overall complex and closeup of FP-trimer binding interface. (C) Sequence alignment of vaccine-elicited vFP48 class antibodies and germ line genes, with FP contacts (red boxes), trimer contacts (light blue), and glycan contacts (underlined) highlighted.

A cryo-EM reconstruction at 4 Å was obtained for the vFP48 class antibody, 7551-vFP48.02, in complex with BG505 DS-SOSIP.664 trimer, which has the FP8_v1 sequence (Fig. 6B and Fig. S9, and Table S3). The model was refined using C3 symmetry and showed three 7551-vFP48.02 Fab molecules bound to the trimer, with an approach angle of 103° relative to the trimer axis (Fig. S10). As observed in the crystal structure described above, FP adopted a hook conformation, with its N-terminal five residues (AVGIG) buried within the hydrophobic binding pocket formed by the CDRs before extending from the antibody into the main body of the trimer. Glycans at Asn611_BG505 and Asn88_BG505 extended on either side of the heavy and light chains of 7551-vFP48.02 (Fig. 6B). Overall, antibody binding interactions for the vFP48 class involved CDR residues from both heavy and light chains, with CDRs H2 and H3 playing dominant roles (Fig. 6C).

Structural characterization of a vFP49-class antibody reveals another mode of FP recognition.

The crystal structure of 8495-vFP49.02 Fab in complex with FP8_v2 peptide at 1.7-Å resolution revealed 2 nearly identical complexes in the asymmetric unit (Fig. 7A and Table S3). Electron density was well-defined for all 8 residues of FP8_v2, including an acetyl moiety at the N terminus of the peptide. FP extended along a groove formed by CDRs H1, H2, H3, L1, and L3 (Fig. 7A) The N terminus of FP was anchored by hydrogen bonds between the main-chain amide of Ala512_FP and Asn32_vFP49 in CDR L1, while the N-terminal acetyl group of FP interacted with Tyr50_vFP49 (CDR L2), Thr91_vFP49 (CDR L3), and Ser100B_vFP49 (CDR H3). There were 6 hydrogen bonds from the main-chain carbonyl oxygen and amide nitrogen atoms of FP to CDR residues. FP residues Val513_FP, Leu515_FP, Ala517_FP, and Phe519_FP extended into the binding groove, stabilized by a series of hydrophobic interactions with CDR residues: Val513_FP with Thr91_vFP49 and Trp94_vFP49 of CDR L3; L515_FP with Trp94_vFP49 of CDR L3, Thr50_vFP49 of CDR H2, and His95_vFP49 of CDR H3; Ala517_FP with Tyr32_vFP49 of CDR H1; and Phe519_FP with Phe31_vFP49 of CDR H1 (Fig. 7A). Unlike the hooked conformation observed in the co-crystal structure of FP8_v2 and 7551-vFP48.03, the extended conformation of the FP8_v2 when bound to 8495-vFP49.02 results in the N terminus of FP interacting primarily with CDRL residues and the C terminus of FP interacting primarily with CDRH residues.

FIG 7.

Structural analysis of vFP49.02 antibody reveals different modes of peptide recognition. (A) Crystal structure of vFP49.02 Fab bound to FP8_v2 peptide is shown as surface (A, left) and ribbon (A, right) with heavy chain in pink, light chain in gold, peptide in gray, and hydrogen bonds as dotted lines. (B) Cryo-EM reconstruction of vFP49.02 Fab bound to BG505 DS-SOSIP (FP8v1) trimer showing overall complex and closeup of FP-antibody binding interface. (C) Sequence alignment of vaccine-elicited vFP49 class antibodies and germ line genes, with FP contacts (red boxes), trimer contacts (light blue) highlighted.

Cryo-EM reconstructions for 8495-vFP49.02 Fab in complex with BG505 DS-SOSIP.664 Env trimer revealed 3 distinct particle classes ranging from 4.3 to 4.7 Å resolution with C1 symmetry (Fig. S11 and Table S3). The cryo-EM maps enabled docking of the 8495-vFP49.02 Fv domain at 1 site in each map, while density for the antibody at the other 2 FP sites was weak and details of the binding interface between FP and antibody 8495-vFP49.02 were not well-resolved. Considerable variation in the approach orientation of antibody to Env trimer was observed among the three 3D-reconstruction classes, with approach angles ranging from 19° to 115°, relative to the trimer axis (Fig. S10 and 11). In 1 reconstruction at 4.7 Å, glycan 88_BG505 extended between the 8495-vFP49.02 heavy chain and gp41 while glycan 448_BG505 approached the other side of the heavy chain (Fig. 7B). FP was bound primarily by interactions with 8495-vFP49.02 CDRs L3, H1, H2, and H3, consistent with that observed in the crystal structure described above. Very few direct contacts were observed between 8495-vFP49.02 and glycans or other parts of the Env trimer (Fig. 7B), allowing potentially for substantial flexibility of the binding orientation of the antibody. Reconstructions from 2 other classes revealed 8439-vFP49.02 to interact directly with residue Asn611_BG505 of gp41, with no extending glycan density observed, and indeed there was little space for any glycan. This observation indicated Asn611_BG505 to be variably glycosylated, as indicated by mass spectrometry (18, 19).

Overall, crystal and cryo-EM structures revealed vFP49-class antibodies to recognize FP in a relatively extended conformation, interacting with all 8 N-terminal FP residues. Similar to vFP48-class antibodies, vFP49-class binding involved exclusively CDR residues, with CDR H3 dominating the binding (Fig. 7C). However, no direct interactions between vFP49-class antibody to glycan and other parts of the trimer were observed, resulting in a substantially flexible mode of recognition.

Structural characterization of new members of the vFP5 and vFP1 classes.

We also determined cryo-EM structures of new members of the vFP5 and vFP1 classes. A cryo-EM reconstruction for the vFP5-class antibody, 7314-vFP52.02, in complex with BG505 DS-SOSIP.664 trimer was obtained at 4 Å resolution (Fig. 8A and Fig. S12, and Table S3). The model was refined using C1 symmetry and showed three Fab molecules binding with slight variation in approach angles to the trimer, ranging from 108° to 118°, relative to the trimer axis (Fig. S10). Asymmetric binding to Env trimer was observed previously for the vFP5 class antibody, 1868-vFP5.01 (7), which showed a larger variation in the range of approach angles (60°, relative to trimer axis) (7). 7314-vFP52.02 light chain played an important role in binding FP, with its framework region 2 (FR L2) also involved, along with the CDR L3 (Fig. 8B).

FIG 8.

Structural analyses of vFP52.02 and vFP53.02 antibodies reveal recognition of fusion peptide with different conformations. (A) Cryo-EM reconstruction of vFP52.02 Fab bound to BG505 DS-SOSIP (FP8v1) trimer showing overall complex and closeup of FP-antibody binding interface. (B) Sequence alignment of vaccine-elicited vFP52 antibodies and germ line genes, with residues contacting FP highlighted by red boxes, those contacting other trimer protein residues in light blue, and those contacting glycan underlined. (C) Cryo-EM reconstruction of vFP53.02 Fab bound to BG505 DS-SOSIP (FP8v1) trimer showing overall complex and closeup of FP-antibody binding interactions. The right panel is a zoom-in at the FP with 180° rotation of the middle panel about the horizontal axis. Hydrogen bonds are shown as yellow dashed lines, and a salt bridge is shown as a red dashed line. (D) Sequence alignment of vaccine-elicited vFP53 antibodies and germ line genes, with contact residues highlighted as in (B).

A cryo-EM structure of a new member of the vFP1 class, 7321-vFP53.02, in complex with BG505 DS-SOSIP.664 trimer was determined at 2.5 Å with C3 symmetry (Fig. 8C and Fig. S13, and Table S3). The structure revealed vFP53.02 to bind at the FP site of vulnerability, with a binding mode similar to that of vFP16.02 (7), another vFP1-class antibody with known structure in complex with Env trimer, and a similar approach angle but a slight shift in the recognition of trimer relative to the position of FP (Fig. S13F). The FP conformations, however, were similar in both structures, with residues 515 to 520 forming a helix. Light chain of vFP53.02 interacted mainly with the N-terminal 2 residues of FP, with E34_vFP53 side chain having a charge interaction with the N-terminal amine of FP (Fig. 8C, right panel). Heavy chain interactions involved all 3 CDRs and FP residues 512 to 519 (Fig. 8C and D, and Fig. S13E). Even though 7321-vFP53.02 was isolated from a group-10 mouse immunized with FP8-I513 immunogens, the structure revealed tight interactions of V513_FP with the light chain residues Y27D_vFP53, Y32_vFP53, and Y96_vFP53. However, with some slight shifts of antibody side chains, the binding pocket is likely to accommodate an I513 in the FP sequence of the trimer.

Overall, the structures of the new members of previously known murine antibody classes recapitulate the binding modes with some variations and confirm recognition of the distinct FP conformations for different classes of FP-directed antibodies. The structures suggest varying immunization regimens to elicit antibodies of varying modes of recognition of FP and other segments of Env trimer.

Distinct new classes of antibodies recognize diverse fusion peptide sequences.

To provide insight into overall sequence requirements for binding and neutralization, we measured free FPv1-v3 peptide affinity, analyzed epitope substitution scans performed with PEPperMAP peptide microarrays, and utilized a panel of Ala and Gly mutants (7) to screen for binding against all four of the vFP antibody classes (Fig. 9). Alterations to residues 512 to 516 and 519 affected recognition in varying degrees by all 4 classes. vFP48-class antibodies exhibited micromolar affinity to free FPv1-v3 peptides (0.10-0.18 μM) (Fig. 9A and Fig. S14). PEPperMAP epitope analysis for the vFP48-antibody class highlighted a conserved N-terminal core motif A₅₁₂VGIG₅₁₆. Specifically, specific vFP48-class recognition was observed for Ala512_FP and Val513_FP and only Ile was tolerated at position 513_FP. Gly514_FP was essential, and conservative substitutions by Leu, Val or Met at position 515_FP were tolerated, a result corroborated in the neutralization pattern observed in the 208-strain panel (Fig. 5B) and supported in Ala-Gly mutation analysis (Fig. 9A).

FIG 9.

Biophysical analyses of fusion peptide. Functional analyses of FP N-terminal sequences by PEPperPRINT epitope mapping for vFP48.01, vFP49.02, vFP52.02, and vFP53.02 against peptide AVGIGAVFLGFLGAA (left); buried surface area calculated from cryo-EM structures of Fabs bound to BG505 DS-SOSIP trimer (second from left); single Ala/Gly substitutions of the FP8_v1 peptide on the binding affinity to vFP Fabs (third from left), and vFP Fabs binding affinities to FP_v1, FP_v2, and FP_v3 peptides assessed by biolayer interferometry analysis (right). Binding dissociation constant for vFP49.02 to FP_v1 could not be determined due to weak binding (Fig. S14) and is denoted as ND.

Antibodies of the vFP49 class preferred the 7 amino acid core FP motif A₅₁₂VGIGAV₅₁₈ (Fig. 9B). Interestingly, this class exhibited ~ 5-fold preference for Leu515_FP over Ile515_FP, whereas any change at A517_FP abrogated binding and a similar pattern was also observed in Ala-Gly analysis (Fig. 9B). The preference of vFP49-class antibodies for a Leu at position 515_FP was also reflected in the observed weak binding for peptide FPv1 (Fig. S14). Binding of 8495-vFP49.02 Fab was assessed to FP8_v1 and FP8_v2 variants of diverse Env trimers, revealing slightly weaker affinity to BG505 DS-SOSIP trimer containing FP8_v1 sequence and diverse trimers containing FP8_v2 (Fig. S15). However, negative-stain EM analysis showed well-formed complexes for 8495-vFP49.02 Fab bound to either FP8_v1 or FP8_v2 Env trimer variants (Fig. S15). The strong preference of the vFP49 class for Leu at position 515 was further validated by the neutralization pattern observed in the 208-strain panel (Fig. 5B).

Both vFP48 and vFP49 classes exhibited strong preferences for Val513_FP along with a long hydrophobic side chain at position 515_FP in PEPperMAP and Ala-Gly mutant analysis. A similar preference was observed for the new vFP5-class member 7314-vFP52.02 for the 7 N-terminal residues of FP, A₅₁₂VGIGAV₅₁₈ with a strong preference for Ile515_FP and Phe519_FP (Fig. 9C). Alterations to residues 513 to 516_FP and 519_FP affected recognition, like the previously identified 1868-vFP5.01 antibody (7). Antibody 7314-vFP52.02 exhibited nanomolar affinity to free FPv1 and FPv2-v3 peptides (Fig. 9C and Fig. S14) and unlike previously identified 1868-vFP5.01 antibody, could neutralize a few FP_v2 and FP_v3 containing strains in the 208-strain panel (Fig. 5B). Meanwhile, new members of the vFP1 class, such as 7321-vFP53.02 showed patterns of recognition similar to previously identified vFP1 class antibodies, 2712-vFP16.02 and 2716-vFP20.01, with preferential recognition of the N-terminal 5 residues of FP, A₅₁₂VGIG₅₁₆ and a requirement for conservation for the first 4 FP residues (Fig. 9D). Furthermore, antibody 7321-vFP53.02 bound free FP_v1-v3 peptides with nanomolar affinity (Fig. 9D and Fig. S14).

Overall, all 4 murine classes of FP-directed neutralizing antibodies showed strong preferences for N-terminal residues of FP, though with different requirements for specific residues, allowing for differential binding and neutralization by each class, and the collective accommodation of more divergent FP sequences. The preference for N-terminal residues of FP may provide a structural rationale for the observation that sequentially reducing FP length during priming enhanced the elicitation of neutralizing responses.

DISCUSSION

Prior studies revealed FP-directed neutralizing antibodies to be dominated by those of the vFP1 class, and our analysis of 21 additional FP-directed neutralizing antibodies showed two-thirds of these to be of the vFP1 class, and the other 7 antibodies to be members of three other classes: vFP5, vFP48, and vFP49. While origin genes of the vFP48 class were phylogenetically more similar to vFP1, no strong relationship between origin gene and breadth or mode of recognition was observed.

Collectively, antibodies from these four vaccine-elicited classes could neutralize over 50% of a panel of 208 HIV-1 strains, providing proof of principle that the mouse immune system can generate antibodies capable of broad HIV-1 neutralization at the FP site of vulnerability. In general, elicited antibodies all focused on the N-terminal exposed residues of FP, and substantial antibody flexibility was observed in the context of the trimer. Some of this flexibility appeared to be a consequence of variable glycosylation at residue Asn611_BG505. We note that the murine antibodies resembled the recognition of vaccine-elicited antibodies from rhesus macaques, which also showed a focus on the N terminus of FP (6), and to differ from infection-elicited human antibodies, PGT151 and ACS202.

Here, we used the ease of immunization in C57BL/6 mice to test multiple immunogens and immunization regimens. Overall, it appeared that most of the variables tested in the 17-murine immunization regimens did not substantially increase neutralization, similar to the many variables tested to improve elicitation of autologous BG505 neutralization, where only slow release showed the ability to substantially increase the vaccine outcome (20). Our specific finding that focusing on the N-terminal 6 residues of FP (10-8-6 and 8-7-6) during the priming phase could positively impact the overall potency and breadth of the final trimer-boosted neutralizing titers was found not only with mice, but was also confirmed with guinea pigs. This finding may relate to the observation that alteration of FP during trimer boosting could also improve the vaccine outcome in guinea pigs (11), suggesting that variation of FP in either peptide or trimer focuses the immune recognition on the conserved N terminus of FP, enabling antibodies to recognize - and neutralize - diverse strains of HIV-1. It will be interesting to determine whether cocktails of FP peptides of different length will also improve the vaccine outcome, similar to the positive impact of cocktails of FP-carrier and Env trimer during the priming phase (21).

MATERIALS AND METHODS

Preparation of fusion peptide-carrier conjugates and epitope-scaffolds.

HIV-1 FPs used for conjugation were synthesized (GenScript, LifeTein) with a free amine group on the N terminus and an extra cysteine added to the C terminus. To prepare peptide-carrier protein conjugates, each fusion peptide was conjugated to the carrier protein keyhole limpet hemocyanin (KLH) (Thermo-Scientific) or recombinant tetanus toxoid heavy chain fragment (rTTHC) (Vaccine Production Program, VRC) using a sulfo-SIAB cross-linker, as described previously (9). FP8-1M6T FP epitope-scaffold proteins with varying FP sequences used for binding assays were expressed and purified as described previously (10).

Expression and purification of Env trimers.

BG505 DS-SOSIP.664 trimer and ConC-FP8v2 RnS-3mut-2G-SOSIP.664 trimer used in guinea pig immunizations were produced in stable CHO-DG44 cell lines and purified using non-affinity chromatography (VRC Production Program) (18). BG505, Cap256.Chim DS-SOSIP, ConC, and CH505 DS-SOSIP trimer variants were expressed by transient transfection in 293 Freestyle cells and purified as described previously (22). Briefly, 600 μg of Env trimer plasmid was co-transfected with 150 μg furin plasmid DNA per L of cells. After culturing for 6 days, supernatants were harvested and purified using either VRC01 or 2G12 affinity chromatography, followed by size exclusion chromatography using a Superdex 200 (16/600) column. Purified Env trimers were assessed for structural integrity by negative-stain electron microscopy and antigenicity by Meso Scale Discovery (MSD) platform.

Antigenic characterization of Env trimers by MSD-ECLIA.

Env trimers were characterized using standard 96-well bare MULTI-ARRAY MSD Plates coated with a panel of HIV neutralization antibodies targeting different antigenic sites, following previously published protocols (23). Briefly, antibodies were coated in duplicate and incubated overnight at 4°C, washed and blocked with MSD Blocker A for 1 h, followed by a washing step. HIV-1 Env trimers were titrated in serial dilution in MSD blocker buffer A + 0.05% Tween 20, transferred to the MSD plates and incubated for 2 h with shaking at 650 rpm. The plates were washed following the incubation phase and read using 1X MSD Read Buffer on the MSD Sector Imager 2400.

Negative-stain electron microscopy.

HIV-1 Env trimers were characterized by negative-stain electron microscopy. Samples were diluted with buffer containing 10 mM HEPES, pH 7.0, 150 mM NaCl, adsorbed to a freshly glow-discharged carbon-coated grid, washed with the above buffer, and stained with 0.7% uranyl formate. Images were collected semi-automatically using SerialEM (24) on an FEI Tecnai T20 electron microscope equipped with a 2k x 2k Eagle CCD camera and operated at 200 kV. The nominal magnification was 100,000, corresponding to a pixel size of 0.22 nm/px. Particles were selected from micrographs using e2boxer from the EMAN2 software package (25). Reference-free 2D classification was performed using EMAN2 and Relion 1.4 (26).

Mouse protocols and immunizations (GenScript).

Female mice (C57BL/6) around 8 weeks old were immunized in 2-week intervals with either 25 μg HIV-1 Env trimer or 25 μg FP-KLH or 25 μg FP-rTTHC conjugates, using 20% (vol/vol) Adjuplex as adjuvant (Adjuplex equivalent formulated based on US patent 6,676,958 B2). Intraperitoneal (IP) route was used for all mice immunizations. Sera samples were collected either 7 days or 14 days after each immunization for ELISA and other analyses.

All experiments were performed in accordance with protocols reviewed and approved by the Genscript’s Institutional Animal Care and Use Committee. All mice were housed and cared for in a facility in GenScript accredited by Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC International).

Hybridoma creation and monoclonal antibody selection.

The top responding mice from each immunization scheme, as assessed by ELISA against the immunogens and neutralization titers against wildtype BG505 and BG505.N611Q incorporating the 3 most prevalent FP sequences (referred to as FP_v1, FP8_v2, and FP8_v3), were selected for cell fusion. Terminal boosts were performed 3 weeks following the last immunization. Mice spleens were harvested 3 days post terminal boost, and hybridomas were generated for monoclonal antibody selection following standard procedures at GenScript. Monoclonal antibody selection was based on binding to FP-scaffolds (FPv1-v3-1M6T) and BG505 Env trimers as measured by ELISA (7).

Production of mouse antibodies and Fab fragments.

Antibody variable heavy chain and light chain sequences were codon optimized, synthesized, and cloned into a VRC8400 (CMV/R expression vector)-based IgG1, IgG2a or IgG2b vector as previously described (27). The constant regions of IgG1, IgG2a and IgG2b were synthesized to match their isotype amino acid sequences, respectively. These variants were expressed by transient transfection in Expi293 cells (Thermo Fisher) using Turbo293 transfection reagent (SPEED BioSystems) according to the manufacturer’s recommendation. A total of 500 μg plasmid encoding heavy chain and 500 μg plasmid encoding light chain variant genes were mixed with the transfection reagents, added to 1000 mL of cells at 2.5 × 10⁶/mL, and incubated in a shaker incubator at 120 rpm, 37°C, 9% CO₂. At 5 days posttransfection, cell culture supernatant was harvested and purified with a Protein G (SPEED BioSystems) column. The antibody was eluted using IgG Elution Buffer (Thermo Fisher) and was brought to neutral pH with 1 M Tris-HCl, pH 8.0. Eluted antibodies were dialyzed against phosphate-buffered saline (PBS) overnight and were confirmed by SDS-PAGE before use.

vFP48.01, vFP48.02, vFP48.03, vFP49.02, vFP52.02, and vFP53.02 antibodies used for structure determination were expressed as Fab-3C-His strep constructs, as well as IgGs with a HRV3C cleavage site on the heavy chain. Specifically, heavy chain plasmids, encoding the chimera of mouse variable domain and human constant domain, with HRV3C cleavage site in the hinge region; and light chain plasmids, encoding the chimera of mouse variable domain and human constant domain were co-transfected in Expi293F cells (Thermo Fisher) using Turbo293 transfection reagent (SPEED BioSystem) according to the manufacturer’s protocol. Transfected cells were incubated in shaker incubators at 120 rpm, 37°C, 9% CO2 overnight. On the second day, one tenth culture volume of AbBooster medium (ABI scientific) was added to each flask of transfected cells and cell cultures were incubated at 120 rpm, 33°C, 9% CO2 for an additional 5 days. Following 6 days posttransfection, cell culture supernatants were harvested. IgGs were purified from the supernatants using protein A chromatography: after PBS wash and low pH glycine buffer elution, the eluate was immediately neutralized using 10% volume of 1M Tris buffer pH 8.0. Fabs were obtained by cleavage using Glutathione-S-transferase (GST)-HRV3C protease. The fragmented Fabs were further purified by passing over Glutathione resin to remove any uncleaved antibody and further purified by size exclusion chromatograph (SEC) in a Superdex 200 column (GE) with a buffer containing 5 mM HEPES, pH 7.5, 150 mM NaCl or PBS. Fab-3C-His strep constructs used for crystallization and cryo-EM were purified using IMAC and Strep affinity purification, followed by cleavage with HRV3C protease and SEC to obtain purified Fab fragments.

Neutralization assays.

A single round of entry neutralization assays using TZM-bl target cells were used to assess murine monoclonal antibody neutralization as described previously (28). Briefly, the murine monoclonal antibodies were tested via 5-fold serial dilutions starting at up to 50 μg/mL. On day 1, monoclonal antibodies were mixed with virus stocks in a final volume of 50 μL and incubated at 37°C for 1 h. Subsequently, 20 μL of TZM-bl cells (0.5 million/mL) were added to the mixture and incubated at 37°C overnight. On day 2, 80 μL cDMEM was added, and on day 3, cells were lysed and assessed for luciferase activity (RLU). Hill slope regression analysis was used to determine the 50% and 80% inhibitory concentrations (IC50 and IC80) as described previously (28). Monoclonal antibody neutralization on a panel of 208 HIV-1 Env-pseudotyped viruses was assessed using an automated 384-well microneutralization assay as described previously (29).

Serum neutralization was assessed in the single round of entry neutralization assays using TZM-bl target cells, as described above. Before evaluation, animal sera from immunized and control animals were heat-inactivated at 56°C for 1 h and tested via 5-fold serial dilutions starting at 1:20 dilution. The ID₅₀ and ID₈₀ neutralization titers were reported as the serum dilutions required to inhibit infection by 50% and 80%, respectively.

Genetic assignment of antibodies.

IMGT-Vquest server was used to identify V(D)J germ line genes of isolated mouse antibody (30).

Guinea pig protocols and immunizations.

All animal experiments were reviewed and approved by the Animal Care and Use Committee of the Vaccine Research Center (VRC), NIAID, and NIH. Animals were housed and cared for in accordance with local, state, federal, and institute policies in an American Association for Accreditation of Laboratory Animal Care-accredited facility at the VRC.

Two-month-old female Hartley guinea pigs with body weights of 300g (Charles River Laboratories) were used for immunization studies. For each immunization, 25 μg of a single immunogen, 25 μg of an FP-conjugate mixture (12.5 μg of each FP-conjugate) or 25 μg of a trimer cocktail (12.5 μg of each trimer) were diluted in PBS with 20% (vol/vol) Adjuplex (80 μL of research-grade Adjuplex, Advanced BioAdjuvants LLC of Omaha [ABA]) for a final immunization volume of 400 μL. Immunizations were administered intramuscularly as 2 separate injections of 200 μL into each quadriceps muscle. Blood samples were collected as sera for serological analyses 2 weeks postimmunization or as whole blood for PBMC isolation and B cell analysis 3 weeks postimmunization.

B cell analysis.

For B cell analysis, 3 to 5 mL of whole blood was centrifuged over 2 mL of Lympholyte in a 15 mL conical tube at 900 g for 15 min and the PBMCs at the interface was collected, washed and stained sequentially with ViViD and a staining mix containing anti-GP IgG-Ax594, GP T cell marker-FITC, BG505.DS.SOSIP-PE, and FP9-Peg12-APC. The stained cells were washed and analyzed on flowcytometer for frequency of ViViD-, T cell marker-, IgG+, BG505+, and FP+ cells among total IgG+ B cells.

Anti-trimer ELISA.

Anti-trimer ELISA was performed as previously described (8). Snowdrop lectin from Galanthus nivalis (Sigma-Aldrich) in PBS was coated on costar half plate (Costar High Binding Half-Area; Corning). Following lectin capture, plates were blocked with 5% skim milk (FisherScientific). Env trimer (BG505 DS-SOSIP, Q23, BL01, or 25701) at 2 μg/μL was coated for 2 h at room temperature. Between each subsequent step, plates were washed 5 times with PBS-T (PBS plus 0.05% Tween). Guinea pig serum was serially diluted (7 point 5-fold) and incubated for 1 h at room temperature. Goat anti-guinea pig IgG- Horseradish peroxidase conjugated secondary antibody (Jackson Immunoresearch) was diluted at 1:5000 and incubated for 1 h at room temperature. Next, plates were developed with tetramethylbenzidine (TMB) substrate (SureBlue; KPL) for 10 min at room temperature, and reaction was stopped with 1 N sulfuric acid (Fisher Chemical). Plates were read at 450 nm (SpectraMax using SoftMax Pro, version 5, software; Molecular DevicesA). Optical densities (OD) were analyzed following subtraction of the nonspecific horseradish peroxidase background activity. The endpoint titer was defined as the reciprocal of the greatest dilution with an OD value above 0.1 (2 times average raw plate background).

Anti-fusion peptide ELISA.

Anti-FP ELISA was performed as previously described (8). Streptavidin coated plates (Thermoscientific) were coated with a biotinylated 8 amino acid linear peptide FP8-PEG12-biotin (AVGIGAVF-PEG12{Lys[Biotin]}, synthesized at GenScript) at 2 μg/mL. Following blocking with in-house B3T Blocking solution for 1 h at 37°C, plates were washed with PBS-T (PBS plus 0.05% Tween). All subsequent steps were preceded with 5X wash with PBS-T. 7-point 5-fold dilutions of guinea pig sera was added to the plates and incubated for 1 h. Goat anti-guinea pig-HRP conjugated secondary antibody (Jackson Immunoresearch) at 1:5000 dilution was added and incubated for 1 h. Plates were developed with tetramethylbenzidine (TMB) substrate (SureBlue; KPL) for 10 min at room temperature and reaction was stopped with 1 N sulfuric acid (Fisher Chemical). Plates were read at 450 nm (SpectraMax using SoftMax Pro, version 5, software; Molecular Devices). OD were analyzed following subtraction of the nonspecific horseradish peroxidase background activity. The endpoint titer was defined as the reciprocal of the greatest dilution with an OD value above 0.1 (2 times average raw plate background).

VRC34.01 ELISA competition assay.

VRC34.01 competition ELISA was performed by using Snowdrop lectin from Galanthus nivalis (Sigma-Aldrich) coated on Costar half plate (Costar High Binding Half-Area; Corning). Following overnight coating with lectin, plates were blocked with 5% skim milk (FisherScientific). The lectin coated plates were then incubated with Env trimer (BG505 DS-SOSIP or BL01 or 25710) at 2 μg/mL for 2 h at room temperature. Between each subsequent step, plates were washed 5 times with PBS-T (PBS plus 0.05% Tween). Guinea pig serum was added at a fixed dilution of 1:100 as the competitor and incubated for 1 h at room temperature. Without washing the plate, VRC34.01 antibody was serially diluted (7 point 5-fold, 0.1 μg/mL starting concentration) and incubated for 30 min at room temperature. Goat anti-human IgG- Horseradish peroxidase conjugated secondary antibody (Jackson Immunoresearch) was diluted at 1:5000, and incubated for 1 h at room temperature. Next, plates were developed with tetramethylbenzidine (TMB) substrate (SureBlue; KPL) for 10 min at room temperature, and reaction was stopped with 1 N sulfuric acid (Fisher Chemical). Plates were read at 450 nm (SpectraMax using SoftMax Pro, version 5, software; Molecular Devices). OD were analyzed following subtraction of the nonspecific horseradish peroxidase background activity. The endpoint titer was defined as the reciprocal of the greatest dilution with an OD value above 0.1 (2 times average raw plate background).

PEPperPRINT epitope mapping using FP microarray.

Epitopes of FP-directed murine antibodies were analyzed for their FP sequence analysis by the standard commercial PEPperPRINT analysis (Heidelberg, Germany) with binding to diverse peptides on a chip. Briefly, analysis was performed using substitution scans of wild type peptide AVGIGAVFLGFLGAA, wherein each amino acid position in FP was exchanged with the standard 20 amino acids. Peptide microarrays included a total of 300 different FP peptide variants and 78 HA control peptides (YPYDVPDYAG). Murine antibodies vFP48.01, vFP49.02, vFP52.02, and vFP53.02 at 1 μg/mL were incubated with shaking at 140 rpm for 16h at 4°C. Secondary goat anti-mouse or anti-human IgG DyLight680 and mouse anti-HA DyLight860 at 0.2 and 0.5 μg/mL respectively, were incubated with each microarray and scanned using a LI-COR Odyssey Imaging System for intensities of 7/7 (red = 700nm/green = 800nm). Microarray image analysis was done with PepSlide Analyzer. PEPperPRINT software algorithm was used to generate intensity maps with intensity color codes with red for high and white for low interactions (31). Intensity plots of wild type and variant FP peptides and their interaction with the murine antibodies were subsequently generated, with red for preferred and blue for less preferred amino acids.

Alanine/glycine scanning analysis.

Binding of the vaccine-elicited murine vFP antibodies to 16 His-tagged fusion peptide (residue 512 to 521), including wildtype and alanine/glycine mutants, was assessed using a fortéBio Octet HTX instrument (7). His-tagged fusion peptides were synthesized (GenScript) with a 6-histidine residue tag at the C terminus of FP. Briefly, the 16 peptides at 50 μg/mL in PBS were loaded onto Ni-NTA biosensors using their C-terminal histidine tags for 60 s. Typical capture levels were between 1.1 and 1.3 nm and variability within a row of 8 tips did not exceed 0.1 nm. These peptide-bound biosensors were equilibrated in HBS-EP+ buffer for 60 s followed by capture of the antigen-binding fragments (Fabs, 250 nM in HBE-EP+) of the vaccine-elicited vFP antibodies, VRC34.01, and an RSV F antibody Motavizumab for 120 s and a subsequent dissociation step in HBS-EP+.

In all Octet measurements, parallel correction to subtract systematic baseline drift was carried out by subtracting the measurements recorded for a loaded sensor incubated in HBS-EP+. Data analysis was carried out using Octet software, version 12.0. The normalized responses obtained from one or triplicate data set were plotted using PRISM (PRISM 9 GraphPad Software).

Biolayer interferometry analysis.

All assays were carried out with agitation set to 1000 rpm in HBS-EP+ 1%using solid black 96-well plates (Geiger Bio-One). Briefly, FP_v1-v3 peptides at 10 μg/mL in PBS were loaded onto Ni-NTA biosensors using their C-terminal histidine tags for 60 s. Typical capture levels were between 1.2 and 1.4 nm, and variability within a row of 8 tips did not exceed 0.1 nm. Biosensor tips were then equilibrated for 60 s in HPS-EP+ buffer prior to assessment of binding to the vaccine-elicited murine Fab molecules in solution (0.00625 to 0.4 μM). Association was allowed to proceed for 180 s followed by dissociation for 300 s. Dissociation wells were used only once to prevent contamination. Parallel correction to subtract systematic baseline drift was carried out by subtracting the measurements recorded for a sensor loaded with FP v1-v3a incubated in HBS-EP+ buffer.

For HIV-1 Env trimer binding studies, 2G12 (60 μg/mL) was immobilized onto an anti-human capture sensor for 300 s (typical loading levels 1.0 nm) followed by incubation with either the BG505.DS-SOSIP(FP8_v1), BG505 DS-SOSIP (FP8_v2), or Conc FP8_v2 and 6540-FP8_v2 for 300 s. The 2G12: HIV-1 Env complex was then allowed to associate with Fab molecules (0.4 μM − 0.00625 μM) in HBS-EP+ buffer for 300s followed by dissociation for 600s. Data analysis and curve fitting were carried out using Octet software, version 12.1.

Crystallization of antibody Fab and fusion peptide complexes.

Antibody Fab and fusion peptide (residues 512 to 519) complexes were prepared by first dissolving fusion peptide in 100% dimethyl sulfoxide (DMSO) at a 50 mg/mL concentration and then mixing with Fab in a 5:1 molar ratio to a final protein complex concentration of 15 mg/mL. Antibody Fab and fusion peptide (residues 512 to 519) complexes were screened for crystallization with Hampton Research, Wizard, and Qiagen crystal screening kits using a Mosquito crystallization robot. Crystals initially observed from the wells were reproduced manually. Crystals of the vFP48.03/FP_v2 complex grew in 0.1 M Tris, pH 8.5, 25% wt/vol PEG3350; crystals of the vFP49.02/FP_v2 complex grew in 0.2 M ammonium sulfate, 0.1 M HEPES, pH 7.5, 20% wt/vol PEG8000, and 10% wt/vol isopropanol.

X-ray data collection, structure solution, model building, and refinement.

Crystals were cryoprotected in 20% glycerol and flash-frozen in liquid nitrogen. Data were collected at a temperature of 100K and a wavelength of 1.00 Å at the SER-CAT beamline ID-22 (Advanced Photon Source, Argonne National Laboratory). Diffraction data were processed with the HKL2000 suite (32). Structure solution was obtained by molecular replacement with Phaser using the vFP16.02 Fab structure (PDB ID: 6CDO) as a search model. Model building was carried out with Coot (33). Refinement was carried out with Phenix (34). Data collection and refinement statistics are shown in Table S3B.

Cryo-EM data collection and processing.

To prepare Env complexes, BG505 DS-SOSIP.664 at a final concentration of 1 mg/mL was incubated with 3–fold molar excess of the antibody Fab fragments per protomer. Each sample was supplemented with 0.085 mM dodecyl-maltoside (DDM). Complexes were vitrified by applying 3 μL of sample to freshly plasma-cleaned C-flat holey carbon grids (CF-1.2/1.3, protochip). Grids were vitrified with an FEI Vitrobot Mark IV with a wait time of 30 s, blot time of 3 s, and blot force of 1 at a temperature of 20°C with 90% humidity. Data was acquired using Leginon software with a Titan Krios electron microscopes operating at 300 kV and fitted with Gatan K2 Summit direct detection device. The dose was fractionated over 50 frames during a 10 s exposure. Individual frames were aligned and dose-weighted with MotionCor2 (35), and the CTF was estimated using CTFFind4 (36). Particles were picked using DoG Picker within the Appion pipeline (37). Then, 2D classifications, 3D reconstruction, and final refinements were performed using cryoSPARC (38).

For the vFP53.02-BG505 DS-SOSIP complex, the antibody Fab was mixed with trimer at ~ 4.5 molar ratio of Fab to trimer and a total protein concentration of ~ 5 mg/mL. DDM (stock concentration of 1 mM) was added to a final concentration of 0.1 mM. 2.7 μL of the mixture was pipetted to Quantifoil R 2/2 gold grids and vitrified with no wait time, 2 s blot time, and blot force of -5 at 4°C with 95% humidity. Data were acquired with a Titan Krios fitted with K3 summit DED operated in the super-resolution mode (0.415 Å/pixel before binning). Data were processed using cryoSPARC 3.3 for patch motion correction, patch CTF estimation, particle picking with Blob picker, 2D classifications, ab initio 3D reconstructions, homogenous, and non-uniform refinements. The final reconstruction map was calculated with non-uniform refinement at C3 symmetry.

Cryo-EM model fitting.

Initial rigid body fits of HIV-1 trimer (PDB ID: 5FYL, 6OT1, or 6CDI) and Fab to the cryo-EM reconstructed maps were performed using Chimera (39). For antibody fitting, we used the fusion peptide-bound coordinates where available. The coordinates were further refined to the electron density by an iterative process of manual fitting using Coot (33) and real space refinement within Phenix (34). Molprobity (40) and EMRinger (41) were used to evaluate the structures at each iteration step. Figures were generated in UCSF Chimera and PyMOL.

Neutralization fingerprinting analysis.

The neutralization fingerprint of a monoclonal antibody or polyclonal plasma is defined as the potency pattern with which the antibody/plasma neutralizes a set of diverse viral strains. The neutralization fingerprints of 35 monoclonal antibodies, including the here isolated FP-targeting antibodies were compared and clustered according to fingerprint similarity, as described previously (15). A set of 208 HIV-1 strains was used in the neutralization fingerprinting analysis for Fig. 5D.

Statistical analysis for guinea pig serum neutralization at week 26.

For the statistical analysis, ID50 inequalities were converted to numerical values in the form “<20” → 19. Animals with background values greater than 20 for SIVmac251.30.SG3 were removed from the analysis. Homologous strains (SVA-MLV.SG3, BG505.W6M.C2.SG3, BG505.W6M.C2.N88Q.SG3, BG505.W6M.C2.N611Q.SG3, and BG505.W6M.C2.N88Q.N611Q.SG3) were also removed from the analysis.

For the magnitude analysis, a non-parametric permutation test was used. Briefly, the geometric mean for each animal across all ID50 virus values was calculated (n =#animals per group). The baseline metric $T_0$ was defined as the difference of geometric means of ID50s for a given group pair. A metric $T$ was calculated from 10000 permutations samples by randomly shuffling the group labels and the P value was defined by counting the number of times where $T>T_0$:

$$T_0=\left|\mathit{gmean}\left(X^{\prime }\right)\hspace{0.17em}-\hspace{0.17em}\mathit{gmean}\left(Y^{\prime }\right)\right|$$ (1)

$$p\left(T\ge T_0\right)=\frac{1}{m}{\displaystyle \sum }_{j=1}^{m}I\left(\left|\mathit{gmean}\left(X^{\prime }\right)\hspace{0.17em}-\hspace{0.17em}\mathit{gmean}\left(Y^{\prime }\right)\right|\ge T_0\right)$$ (2)

X′ = Group A ID 50 values for each animal (averaged across viruses) with reshuffled group label from {A, B}

Y′ = Group B ID 50 values for each animal (averaged across viruses) with reshuffled group label from {A, B}

m = 10000,

where $I(\cdot )$ is the indicator function.

For breadth analysis, the number of heterologous strains neutralized was counted per animal and the pair of groups was tested with Mann–Whitney using a two-sided scipy.stats.mannwhitneyu implementation with continuity.

Data and materials availability.

Atomic coordinates and structure factors for the vFP48.03+FP8_v2 and vFP49.02+FP8_v2 crystal structures were deposited in the Protein Data Bank under accession codes 6X78 and 6X7W, respectively. Cryo-EM maps and fitted coordinates were deposited to the EMDB and RCSB, with EMDB accession codes EMD-29396, EMD-29836, EMD-29880, EMD-29881, EMD-29882, and EMD-29905, and PDB codes 8FR6, 8G85, 8G9W, 8G9X, 8G9Y, and 8GAS. Heavy chain and light chain-variable sequences of vaccine-elicited murine monoclonal antibodies were deposited with GenBank under accession numbers: OP242458 through OP242499. All other data are available in the main text or the supplementary materials.