Mapping Polyclonal Antibody Responses in Non-human Primates Vaccinated with HIV Env Trimer Subunit Vaccines

Humoral Responses in Naive Macaques after SHIV_BG505 Infection

The SHIV_BG505 virus challenge study enrolled 12 immunized and 6 naive control animals (Pauthner et al., 2019). As previously described, the challenge virus contained a S375Y mutation that confers replication capacity in rhesus macaques and results in AIDS-like pathology (Pauthner et al., 2019). Animals received up to 12 intrarectal challenges at weekly intervals with a break after 6 challenges for those animals not infected early. Five of six naive animals became infected after the first challenge, and one after the second. In the low-titer NHP group, 2/6 animals became infected after the first challenge, and 4/6 after the second. Finally, in the high-titer NHP group, 1/6 animals became infected after the first challenge, 1/6 after the sixth, 1/6 after the tenth, 1/6 after the twelfth, and 2/6 never became infected (Pauthner et al., 2019).

Using EMPEM, we first analyzed the antibody responses in week 20 postchallenge (p.c.) sera (i.e., 20 weeks after first challenge) of the six non-immunized animals by forming complexes with BG505 SOSIP.664v5.2 Env glycoprotein trimers that were sequence matched to immunogen and the Env of the SHIV_BG505 challenge virus. Figure 1 summarizes the EMPEM results wherein each dataset was reconstructed into a global 3D average and then subjected to subsequent 3D classification that sorted data into unique classes (Figure S1). Two of the six serum samples analyzed, BZ15 and BZ31, were not amenable to 3D reconstruction. The epitopes recognized in these animals, however, could be discerned from the 2D class averages when compared with known NHP monoclonal 2D class references (Figures S1D and S1E). One animal, CB00, showed no detectable binding to the BG505 trimer probe (Figure S1F), yet this was not unexpected because this animal showed waning BG505 S375Y pseudo-virus neutralizing titers 12 weeks p.c. (Pauthner et al., 2019), suggesting viral evolution away from the infecting virus and concomitant antibody evolution. The remaining three animals exhibited antibody responses in and around the N289/S241 glycan hole region (GH-1/GH-2) that is a highly immunogenic epitope in rabbits (McCoy et al., 2016; Bianchi et al., 2018; Klasse et al., 2018) and in the C3/V5 region of gp120 that we have previously described as the “C3/T465 epitope” in NHPs. Thus, considering a sample of only six animals and a time span of just 20 weeks under viral challenge pressure, the infected control macaques exhibited a relatively narrow repertoire of antibody epitope responses to the founding virus, with four different 3D Fab phenotypes detected against two regions of Env (Figures 1B and 1C). Notably, and not unexpectedly, virus challenges did not induce any antibodies that bound to the base of the trimer (Figure 1), which in rabbits is typically immunodominant on the soluble trimer and may limit elicitation of neutralizing immune responses (Bianchi et al., 2018; Cirelli et al., 2019; McCoy et al., 2016).

Low-nAb-Titer-Vaccinated Animals Show Increased Diversity of Epitopes Compared with Infected Naive Controls

The next group that we examined were the low-titer animals subjected to a total of four identical immunizations with BG505 SOSIP.664 variants, including a booster injection 4 weeks prior to the first SHIV_BG505 challenge. We performed EMPEM analysis on these animals 4 weeks after this last booster, immediately prior to the first virus challenge (i.e., week 68). These samples showed an overlapping set of epitopes but with increased diversity relative to the infected naive controls (Figures 1 and 2). The primary differences were that immunized animals also mounted an antibody response to the FP region, the gp120/gp120 inter face, and the base of the trimer. Despite the diversity of epitopes recognized, all animals from this low-titer group became infected within the first two virus challenges (Pauthner et al., 2019). Second to base binding, which is present in all BG505 SOSIP-immunized animals (McCoy et al., 2016; Bianchi et al., 2018), the most common epitopes targeted in the low-titer animals were the glycan hole 1 (GH-1) epitope, which was also the most frequently recognized epitope among the unimmunized controls, and the FP (Figures 1C and 2G). Thus, despite the presence of a diversity of vaccine-induced antibodies, this group of animals was not protected from the virus challenge. We note that none of the animals in this group had nAb titers above the ~1:500 threshold determined as the main correlate of protection (Pauthner et al., 2019).

High-nAb-Titer-Vaccinated Animals Show the Most Diverse Humoral Responses

Relative to the infected naive and the low-titer-immunized macaques, the high-titer animals exhibited overlapping yet further diversified responses; their antibody response not only differed in the number and types of epitope recognized, but also in the apparent angles of approach for similar epitope clusters (Figures 1, 2, and 3). This higher diversity probably also gave rise to an increased Fab occupancy in the complexes observed in these animals. Figure 3 shows the results of the EMPEM analysis of week 68 serum samples from the high-titer macaque group that showed protection from infection and reduced peak viral loads relative to naive animals (Pauthner et al., 2017, 2019). Notably, two animals, 12–046 and 4O9, remained uninfected after 12 SHIV_BG505 challenges (Pauthner et al., 2019).

One striking difference between the low-titer- and high-titerimmunized animals was the appearance of antibodies binding to the V1/V3 region near the apex of the trimer in four out of six high-titer animals compared with none of the low-titer animals. At low resolution these responses resembled known bnAbs such as PGT121 that bind to the conserved glycine, aspartate, isoleucine, arginine (GDIR) co-receptor site and the N332 glycan (Figure 3H). Another notable phenotype for the high-titer animals was a different angle of approach by which Fabs bound to the C3/V5–3 epitope: three of six animals exhibited this phenotype, with the low-titer macaques favoring the GH-1 in lieu of this putative high-titer correlate (compare Figure 2 with Figures 3A, 3B, and 3F). This observation was consistent in additional unchallenged low- and high-titer animals from the immunization study (Figure S3D; Cirelli et al., 2019; Pauthner et al., 2017). Overall, the high-titer group displayed an increased diversity in the targeted epitopes, especially with the appearance of V1/V3 binding antibodies, and an associated increase in trimer occupancy (Figure 3), with apparently unique angles of approach as well, factors that may have contributed to these animals’ resistance to infection.

Evolution of Humoral Immune Responses

Taking advantage of the intensive schedule of blood draws throughout the immunization experiment, we applied EMPEM across multiple time points to study the evolution of antibody responses; we performed EMPEM analysis on available samples from high-titer animals after each booster immunization in an attempt to discern hierarchical signatures of developing protective antibody responses. The week 10 neutralization titers were 1:48, 1:83, and 1:404 for animals 12–137, 11M-088, and 4O9, respectively (Pauthner et al., 2017). Within this subset of animals there is no clear phenotype that correlated with future ability to protect from an autologous virus challenge. For example, both 12–137 and 4O9 displayed similar epitope recognition, namely, C3/V5–3 and FP, but 12–137 was infected after six intra-rectal challenges, whereas 4O9 remained uninfected after all 12 challenges. Serum samples from weeks 26–28 showed an increase in the diversity of epitopes recognized for animals 11M-088 and 4O9, which was not the case for 12–137. In fact, 12–137 and another member of the high-titer group that showed only marginal protection, 12–143, did not diversify the epitopes targeted after the first boost (see also Figure 3A).

Notably, fully protected animal 4O9 mounted V1/V3-specific antibodies as early as week 26 (post-boost 2), and these were not found in 11M-088, even after extensive 2D and 3D classification (Figure S4). Interestingly, we did not detect V1/V3-specific antibodies in 12–046, the other fully protected animal, at this stage, although this may have been a consequence of sample limitation and thus substantially lower Fab concentration during complex formation (see Method Details and Figure S4D). However, the weeks 66 and 68 (post-boost 3) 3D reconstructions demonstrate that both 11M-088 and 12–046 ultimately generated diverse antibody responses, with the V1/V3 loop now present in these animals’ polyclonal antibody responses and neutralization titers increased relative to week 26 (Figures 3E and 4B; Pauthner et al., 2017, 2019).

Thus, based on these EMPEM analyses, the ability to diversify the repertoire of epitopes recognized following continued booster immunizations generally coincided with higher neutralization titers and protection from infection.

Polyclonal Antibody Imaging Suggests Responses to the FP Are Responsible for Neutralization Breadth

Pauthner et al. (2017) reported sporadic neutralization breadth in a subset of the animals’ week 26 immunization sera using the 12-virus global pseudovirus panel, with the 398F1 virus strain being the most sensitive to neutralization (deCamp et al., 2014). Notably, this virus, like BG505, also exhibits a GH around position 289, which was shown to be very immunogenic and which was recognized by antibodies observed in 12–046 (Figure 3E), among others. To determine the cross-reactive specificity(ies), we prepared complexes of HIV_398F1 SOSIP.664 envelope trimers with week 25 or 26/28 Fabs from animals 12–137, 12–046, and 12–084 (this animal exhibited exceptionally high titers but was not among the challenge group) (Pauthner et al., 2017) and subjected them to EMPEM analysis (Figure 5A). Complexes with HIV₂₅₇₁₀ Env trimers and animal 12–046 week 26 serum were also examined (Figure S5B). These macaques had HIV_398F1 neutralization titers of 1:32,620, 1:256, and 1:399, respectively, with 12–046 titers against 25710 at a more modest 1:160. Excluding antibodies to the trimer base, we observed only a single epitope specificity binding to heterologous trimers, namely, the FP region, being contacted at two slightly different angles of approach (Figures 5A and 5B; Figure S5B). These phenotypes are reminiscent of bnAbs PGT151, VRC34, and ACS202, all of which include the FP as part of their epitope (Figure 5C; Kong et al., 2016; Lee et al., 2016; Yuan et al., 2019). Interestingly, the FP sequence is identical between BG505, 398F1, and 25710 (Figure S5A). Thus, based on these EMPEM analyses of a limited subset of heterologous viruses that were sensitive to neutralization by immunized macaque polyclonal sera, it appears that antibodies to the FP may be one of the main sites responsible for the breadth observed in several animals (Pauthner et al., 2017).

3D Cryo-EM Reconstruction Reveals Structural Details of the Macaque “V1/V3” and “C3/V5” Targeting Antibodies

Given the similarity of the apical antibodies that we detected in 4/6 high-titer animals to N332 bnAbs, we attempted further characterization of this specificity by conducting high-resolution cryo-EM analysis. We used week 26 serum from the fully protected animal 4O9 (Figures 3F and 4C) to make complexes with BG505 SOSIPv5.2 to assess the structural details of the epitope recognition in high-titer animals (summarized in Figure 6). The data produced a ~3.9-Å global resolution 3D reconstruction (Figure 7A). Notably, relative to our more typical cryo-EM sample preparation conditions using monoclonal antibodies (mAbs), we prepared the polyclonal cryo-EM sample using slightly modified conditions to maximize the likelihood of observing as many of the epitopes as possible (see Discussion and STAR Methods). Consequently, by attempting to maximize the Fab:trimer ratio, we generated complexes in which the Env trimers were saturated with Fabs, and we were not able to classify out trimers occupied by less than the 9 Fabs depicted in Figure 7.

In the global 3D reconstruction, we observed clear density for Fabs corresponding to the trimer base, V1/V3, and C3/V5–3 epitopes. The density for the trimer base Fabs was diffuse and poorly resolved, suggesting a heterogenous mixture of base antibodies were being averaged together. Conversely, the V1/V3 and, to a lesser extent, the C3/V5–3 Fab densities were remarkably well resolved (Figure 7A). These data enabled more in-depth analyses of the molecular interactions of these two paratopeepitope interfaces. The V1/V3 Fabs were so well resolved that they likely corresponded to a single clonal lineage and enabled us to determine the CDR loop lengths to be ~10, ~14, and ~5 residues long for CDRH3, CDRH2, and CDRH1, respectively (Kunik et al., 2012). Further, the less variable regions of the Fab were resolved to sufficient resolution to semi-confidently model side chains.

The primary interaction between the V1/V3 Fab and the apex of the trimer is between the CDRH3 of the Fab, inserted between glycans at positions N133 and N137 to make peptide contacts at residues N137 and I138, T139 of the trimer V1 loop, as well as the CDRL3, which contacts R143 within the same region (Figures 7B–7D). We previously reported virus neutralization data that implicated an epitope that includes residues N133 and N137, which are part of the V1 loop, as part of these antibodies binding (Klasse et al., 2018). Further, Pauthner et al. (2017) reported up to nearly 100-fold decreases in neutralization titer in this animal upon amino acid insertions in the vicinity of these residues. Finally, the CDRH1 loop makes side-chain contacts with R327 (Figure S7H), which is a residue in the GDIR motif of the co-receptor binding site.

The V1 loop of Env in the 4O9 Fab-complexed structure is in a nearly identical conformation relative to the uncomplexed BG505 SOSIP.664 trimer control, which we deem the ground-state conformation that effectively masks the conserved GDIR co-receptor binding site (Sok et al., 2016). Conversely, both PGT121 and PGT128 bnAbs binding displaces the V1 from this conformation for efficient recognition of N332 and the GDIR motif (Figure 7E; Lee et al., 2015). The cryo-EM map shows that although the main interaction is between the V1 loop and the CDRH3 of the Fabs, some contacts between the V3 loop and the CDRH1 and CDRH3 of the Fabs are present, with the N332 clearly pushed out of the way (Figure S7G). This type of interaction is also observed in a rabbit nAb recently described (Nogal et al., 2019). Given the high sequence and glycosylation variability in V1 among global HIV isolates, this epitope is likely strain specific, with limited potential for broadening. The substantial epitope overlap of the NHP V1 antibodies with the N332 supersite may constitute a roadblock for the development of bnAbs to this region.

Although the less well-resolved C3/V5 epitope precluded contact residue assignment, we discerned that the primary interaction is between the BG505 variable loop 5 and the light chain of the antibody, with the glycan at position 465 shifted out of the way to accommodate the contacts (Figure S7B). The CDRH3 appears to be relatively short, possibly in order to accommodate a glycan at position 355 (Figures S7B and S7C). Notably, negative stain reconstructions of monoclonal nAbs isolated from this same animal also bound to this site (Figure 7F), with the same angle of approach as in both the cryo-EM and negative-stain reconstructions of weeks 10, 26, and 64 (Figures 4C and 5A). These nAbs only partially account for the serum neutralization, further emphasizing the potential role of other epitope specificities, namely, the V1 response described above.

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Andrew B. Ward (andrew@scripps.edu). This study did not generate new unique reagents.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

Due to the small animal subject sample sizes and thus insufficient power to detect significant polyclonal response differences among the unimmunized naive, low titer, and high titer animals, no statistical tests were applied for the purpose of describing the trends in responses and their diversity. Furthermore, as the serum samples were themselves sometimes a limiting factor, accurate quantitation and subsequent quantitative comparison among individual animals was not possible.

DATA AND CODE AVAILABILITY

3D EM reconstructions have been deposited in the Electron Microscopy Databank (http://www.emdataresource.org/) or Protein Databank (http://www.rcsb.org) under the accession numbers listed in the Key Resources Table.