Deletions in variable region 1 of the HIV envelope accelerate V3-glycan bNAb induction in SHIV-infected rhesus macaques

Significance

An effective HIV-1 vaccine will need to elicit broadly neutralizing antibodies (bNAbs) that protect against the extensive diversity of globally circulating strains. Potent bNAbs rarely develop during natural infection, in large part due to inefficient precursor priming. In this study, we show that shortening the V1 region of the CH848 envelope glycoprotein enhances priming of V3-glycan bNAb precursors in the context of SHIV infection and that this leads to faster development of plasma neutralization breadth. Thus, CH848-based immunogens with shortened V1 loops are promising candidates to prime V3-glycan bNAb precursors in humans, with a clinical trial currently underway (HIV Vaccine Trials Network protocol 307), and may represent a generalizable strategy for inducing V3-glycan bNAbs.

Abstract

Elicitation of broadly neutralizing antibodies (bNAbs) is a major goal of HIV-1 vaccine research. V3-glycan bNAbs were previously identified in a person living with HIV-1, CH848, and in two of six rhesus macaques (RMs) infected with a simian-HIV (SHIV) expressing the CH848 transmitted/founder envelope (Env) glycoprotein. In all cases, neutralization breadth developed after years of infection and only after the variable region 1 (V1) of Env had shortened in response to neutralizing antibodies. These observations suggested that V3-glycan bNAb elicitation might be accelerated if RMs were infected with SHIVs bearing CH848 Envs with truncated V1 loops. To test this hypothesis, we analyzed NAb responses in 10 RMs infected with SHIV.CH848 variants containing V1 deletions that mimicked those observed in natural infection. In three RMs, potent V3-glycan bNAbs developed within 1 y of infection, nearly three times faster than in wildtype CH848 infection. Longitudinal sequencing of B cell immunoglobulin genes revealed a single bNAb lineage in each RM that was primed as early as 8 to 12 wk postinfection and that largely recapitulated plasma breadth. A meta-analysis of the frequency of V3-glycan bNAb induction in RMs infected by SHIVs expressing CH848-derived Envs vs. 15 other primary HIV-1 Envs demonstrated a substantial enhancement in the CH848 group (5 of 16 vs. 2 of 109; P = 0.0003). We conclude that CH848 Envs have a propensity for inducing V3-glycan bNAbs and that truncations in their V1 loops accelerate precursor priming, findings that may guide HIV-1 vaccine design and inform current human vaccine trials.

A central objective of HIV-1 vaccine design is the elicitation of broadly neutralizing antibodies (bNAbs) that protect against the extreme diversity of globally circulating strains (1, 2). Potent bNAbs arise in only 10 to 20% of people with HIV-1, and usually only after years of infection (3–7). For most bNAb specificities, inefficient activation of naïve B cell precursors by wildtype envelope (Env) glycoproteins contributes to the low rate of bNAb elicitation, as bNAb epitopes are occluded by hypervariable regions, a dense glycan shield, and conformational masking (8–10). Consequently, a principal strategy in HIV-1 vaccine design aims to develop immunogens that can selectively engage germline bNAb precursors more efficiently than do naturally occurring Envs (11–21).

HIV-1 bNAbs target one of several highly conserved epitopes on Env including the V3-glycan high mannose patch. This epitope supersite contains a conserved ³²⁴GDIRQAH³³⁰ motif at the base of the V3 loop that is responsible for coreceptor binding as well as surrounding N-linked glycans at positions 301 and 332 (2, 9, 22, 23). Antibodies targeting this epitope are an attractive vaccine target as they are among the most broad and potent bNAbs isolated to date (9, 23). V3-glycan bNAbs also represent the most common specificity found in people with HIV-1 and are not restricted to specific immunogenetic or structural elements, which may make them easier to elicit (5, 7, 17, 23–27). Several V3-glycan bNAb lineages have been studied extensively, thus providing a mechanistic understanding of their derivation (25, 27–33).

One well studied example of V3-glycan bNAb development is the DH270 lineage isolated from a human participant CH848 in the CHAVI 001 acute HIV infection cohort (27, 28, 34). The most cross-reactive member of this lineage achieved over 55% neutralization breadth on a 207 strain heterologous virus panel with a geometric mean 50% inhibitory concentration (IC₅₀) of 0.13 mg/mL, making it a desirable example of bNAb elicitation (27). Heterologous neutralization was dependent on the presence of the N332 glycan and the ³²⁴GDIR³²⁷ motif in the V3 epitope. Notably, neutralization potency of early lineage members was inversely correlated with the length of the variable region 1 (V1) (27). The V1 region forms a flexible glycosylated loop that shields the underlying V3-glycan epitope by sterically occluding antibody access (28). Truncating this region in the CH848 Env was critical for DH270 precursor engagement, as early bNAb lineage members were only able to bind autologous Envs with short V1 loops (15, 27). Temporally, the DH270 lineage emerged following V1 deletions in the viral quasispecies that were selected by V1-targeted autologous NAbs. These findings suggested that Env variants containing deletions in the V1 loop primed this bNAb lineage in vivo.

Additional studies in genetically engineered mice and nonhuman primates supported the conclusion that CH848 V1 truncations facilitate V3-glycan bNAb induction. Studies of DH270 precursor or unmutated common ancestor (UCA) knock-in mice have shown that immunization with CH848.1017DT, a variant containing a 17 amino acid V1 deletion and removal of the 133 and 138 V1 glycans, could prime and partially mature this lineage (15, 35, 36). Moreover, simian-human immunodeficiency viruses (SHIVs) bearing the CH848 “wildtype” transmitted/founder Env elicited V3-glycan bNAbs in two out of six rhesus macaques (RMs), but only after V1 deletions had evolved in vivo (37–39). Together, these studies suggested that V1 loop truncations likely contributed to priming of germline precursors in both human and rhesus, although direct evidence for this was lacking.

To test this hypothesis, we sought to prospectively evaluate the kinetics of V3-glycan bNAb elicitation by CH848 V1 loop variants in the SHIV infection model. Several recent studies have shown that SHIVs bearing genetically engineered Envs designed to engage bNAb precursors for the V2 apex, CD4 binding site, fusion peptide, and V3-glycan epitope supersites replicate efficiently in RMs and immunofocus B cell responses to these canonical epitopes, although none used the strategy to study derivatives of the CH848 Env (20, 32, 40–42). We hypothesized that RMs infected with SHIVs bearing CH848 Envs containing shortened and hypoglycosylated V1 loops would develop V3-glycan bNAbs at a higher frequency compared to other primary HIV-1 Envs and more rapidly than in animals infected with SHIVs bearing the CH848.WT Env. We tested this in a cohort of RMs infected with SHIVs encoding V1-truncated CH848 Envs designed from sequences that evolved naturally in human subject CH848 and in SHIV.CH848.WT-infected macaques that developed V3-glycan bNAbs.

Results

SHIVs with V1 Deletions Rapidly Elicit bNAbs.

To test whether V3-glycan bNAbs are elicited more rapidly by SHIV.CH848 variants with shortened and deglycosylated V1 loops, we created four SHIVs bearing these mutations (Fig. 1A and SI Appendix, Fig. S1). These included CH848.1017, which was isolated from human CH848 and bound early DH270 lineage members; CH848.1017DT, which was identical to CH848.1017 but lacked potential N-linked glycosylation sites (PNGS) at positions 133 and 138 and bound the DH270 UCA; CH848.w36con, which represented the viral Env consensus sequence in a SHIV.CH848.WT-infected RM near the time of bNAb precursor priming; and CH848.dV1 which contains the same 10 amino acid V1 deletion as in CH848.w36con but in the CH848.WT Env backbone (15, 27, 39). These viruses were first tested for Env functionality (infectivity), antigenicity, and structural integrity. All four ΔV1 variants exhibited infectivity in TZM cells that was comparable to wildtype CH848. Like wildtype CH848, dV1, 1017, and 1017DT all contain a glutamic acid residue at position 169 in the V2 apex C-strand, rendering them insensitive to prototypic V2 apex bNAbs PGT145 and PGDM1400. However, the w36con Env contains a naturally occurring lysine at this position, making it extremely sensitive to PGT145 and PGDM1400 (SI Appendix, Fig. S2A). E169K substitutions in CH848 1017 and 1017DT similarly rendered them sensitive to V2 apex bNAbs (43, 44). Since these antibodies bind the apex in a quaternary-dependent manner, the neutralization and infectivity results demonstrate that the Env trimer of the ΔV1 variants is structurally and functionally intact. This conclusion was further supported by the failure of two CD4i antibodies 17b and A32, which only bind Envs that are conformationally open in a CD4-bound state (45), to neutralize ΔV1 Envs. Additionally, V2p mAbs CAP228-16H and CAP228-3D as well as V2i mAbs 830A, 697-D, and 1361 failed to neutralize further demonstrating the presence of closed Env trimers. As expected, neutralization by the CD4bs bNAb VRC01 was not affected by V1 mutations. In contrast, ΔV1 viruses were more sensitive to neutralization by five different prototypic V3-glycan bNAbs PGT128, PGT121, DH270.6, BG18, and PCDN76-38B, and to a lesser extent, the linear V3 targeted mAb 3074 (SI Appendix, Fig. S2A). These changes could be explained by greater epitope exposure resulting from the V1 deletions and hypoglycosylation.

Fig. 1.

CH848 V1 modifications accelerate bNAb development. (A) Amino acid alignment of Env V1 regions. Potential N-linked glycosylation sites are highlighted blue and deletions are highlighted orange. Dots indicate sequence identity, dashes indicate deletions. (B) Viral load in 10 RMs infected with SHIV.CH848 variants indicated on the Right. (C) Reciprocal plasma ID₅₀ against a panel of 16 heterologous HIV strains in RMs that developed bNAbs. Dotted line represents the limit of detection. Infecting SHIV is shown in parentheses. (D) Time to develop breadth in RMs infected with SHIV.CH848 variants compared to RMs infected with SHIV.CH848.WT and human participant CH848 (shown as an open diamond). P value calculated with a one-tailed t test of log transformed values. Mean with SD shown.

Next, we infected 14 macaques with these SHIVs and followed virus replication over time (Fig. 1B and SI Appendix, Table S1). Acute and chronic plasma replication dynamics, as determined by sequential plasma vRNA measurements, were typical of HIV and SHIV infections (37). Four RMs infected with SHIV.CH848.1017DT experienced very high peak and setpoint plasma virus loads and progressed rapidly to AIDS, requiring euthanasia between weeks 16 and 32 (SI Appendix, Table S1) (46). These animals were similar to other SIV- or SHIV-infected rapid progressors in that they developed little or no anti-SHIV antibodies (32, 47–49) and thus were omitted from further analysis. Ten RMs were followed for up to 120 weeks postinfection (WPI) (mean 95 wk, median 100 wk) with frequent blood sampling (Fig. 1B and SI Appendix, Table S1). The setpoint viral load and length of evaluation of these animals did not differ significantly from RMs previously infected with SHIV.CH848.WT (39) (SI Appendix, Fig. S2 B and C). Each macaque potently neutralized the autologous virus as well as one or more heterologous tier 1A strains as early as 16 to 24 WPI, which was maintained for the duration of the infection (Dataset S1). These results demonstrated typical antibody responses to SHIV infection (39, 49).

RMs were tested for plasma neutralization breadth using a panel of 16 heterologous tier-1B/2 HIV strains. Our predetermined criteria for breadth were neutralization of at least 25% (4 of 16 viruses) at titers >1:100 50% inhibitory dilution (ID₅₀). Plasma from three RMs—44715 (SHIV.CH848.w36con), 44645 (SHIV.CH848.dV1), and BV57 (SHIV.CH848.1017DT)—reached this threshold (Fig. 1C and Dataset S2). If we relaxed the threshold for neutralization to ID₅₀ of 1:20, then RMs 44715, 44645, and BG57 exhibited neutralization breadth of 56%, 50%, and 37% against the panel with geometric mean titers of 1:87, 1:86, and 1:236, respectively. In contrast, none of the other seven RMs neutralized more than one heterologous virus strain at an ID₅₀ of 1:100 and therefore did not meet our criteria for neutralization breadth (SI Appendix, Fig. S3 and Dataset S2). When compared to a previous study of SHIV.CH848.WT-infected RMs (39), we observed no differences in the proportion of animals with breadth or the degree of breadth measured on the virus test panel (SI Appendix, Fig. S2D). However, we did find that heterologous tier 1B/2 neutralization was detected much earlier in the plasma of macaques infected by SHIVs expressing V1 truncated Envs (24 WPI in BV57, 40 WPI in 44715, and 48 WPI in 44645) compared to two RMs infected with SHIV.CH848.WT (80 WPI for RM6163 and 64 WPI RM6167) (39) or the human CH848 (160 WPI) (27). Overall, the time to first detection of heterologous neutralization in the plasma of RMs infected with V1 truncated variants was on average 37 wk postinfection (range 24 to 48) compared with rhesus or human infected with wildtype CH848 viruses where the average was 101 wk (range 64 to 160 wk; P = 0.025, one-tailed t test of log transformed values) (Fig. 1D). We thus concluded that deletions in the V1 loop accelerated bNAb development in SHIV.CH848 infected RMs by nearly threefold.

Broadly Neutralizing Responses Target the V3-Glycan Epitope.

Having established that CH848 ΔV1 variants elicited bNAbs in a subset of macaques at an accelerated pace, we sought to identify the epitopes that were targeted. We initially pursued two complementary strategies to determine bNAb epitope specificity. First, we inspected intact gp160 env sequences representing the virus quasispecies at sequential timepoints for sites of antibody-mediated selection (42, 49–51), since this is a sensitive method to identify target epitopes of neutralizing antibodies even before they can be detected in the plasma by the TZM assay (42, 49–51). We performed env single genome sequencing (SGS) of plasma viral RNA (Dataset S3) followed by LASSIE analysis, which identifies amino acid residues under selection (52) (SI Appendix, Fig. S4). We looked for mutations in linear sequences of Env known to contribute to canonical V3-glycan (HXB2 residues 324-334), V2 apex (166-171), fusion peptide (FP, 515-525), or membrane proximal external region (MPER, 656-683) bNAb recognition. We found significantly more selection in the V3-glycan region of Env in RMs that developed plasma breadth than in monkeys that did not (Fig. 2 A, Top Left panel; P = 0.01, one-tailed Mann–Whitney test), while there were no differences in epitope variation at the V2 apex, FP, or MPER sites (Fig. 2A and SI Appendix, Fig. S4). These results suggested that the plasma breadth in RMs 44715, 44645, and BV57 targeted the V3-glycan epitope.

Fig. 2.

Plasma breadth targets the V3-glycan high-mannose patch. (A) Average mutations in four common bNAb epitopes from longitudinal Env SGS data. Mean with SD shown as columns with error bars, individual RMs shown as dots. P values calculated using a one-tailed Mann–Whitney U test. For the V3 region, the K327R mutation was excluded as it was shown to be a fitness reversion unrelated to neutralizing antibodies (SI Appendix, Fig. S5). (B) Env residues 324-334 from SGS per RM. Mutations from the sequence of the infecting SHIV are shown, dots represent identity. The number of SGS amplicons matching each sequence are indicated on the right-hand side. (C) Reciprocal plasma ID₅₀ against Env point mutants in heterologous backbones. Colors represent relative change in ID₅₀ from the unmutated SHIV. Murine leukemia virus (MLV) was included as a negative control in all assays and was not neutralized.

To corroborate these findings, we next looked at individual mutations within the V3-glycan epitope in each macaque to determine their kinetics of appearance (Fig. 2B) and their effects on neutralization sensitivity to the developing bNAbs. In RM 44715, we observed early appearance (week 24) and then fixation of D325N and I326T mutations, as well as transient detection of S334N and H330Y variants. In RM 44645, we again observed early appearance (week 24) and then fixation of I326T and transient detection of D325N and S334A. In RM BV57, we observed early fixation of a K327R mutation as well as D325N and I326T mutations (week 12) that became more prevalent later in infection. The K327R mutation is a reversion to global consensus, appeared in nearly every RM infected with SHIV.CH848.1017 and SHIV.CH848.1017DT, did not impact neutralization sensitivity to the bNAb lineage in RM BV57, and therefore likely impacts viral fitness independent of escape from neutralizing antibodies (SI Appendix, Fig. S5). However, in BV57, we also observed early introduction of potential N-linked glycan sequons at residues 133 and 138 at week 12 (SI Appendix, Fig. S4), which conferred escape from bNAb recognition (see below) and likely explains the absence of V3 mutations between week 12 and 96. Thus, we found strong selection in the V3 and/or V1 region in all three RMs, suggesting the presence of V3-glycan bNAbs in the plasma.

We also mapped phenotypically the polyclonal neutralizing antibody responses against heterologous viruses and their isogenic V3-glycan epitope mutants in RMs 44715, 44645, and BV57. Heterologous plasma neutralization in RM 44715 was abrogated or substantially reduced by mutations in the D325, I326, and N332 (PNGS) residues in multiple Env backgrounds (Fig. 2C and SI Appendix, Fig. S6). Similar results were obtained for RM 44645 and BV57. Altogether, the phenotypic mapping indicated all three RMs developed V3-glycan bNAbs. To determine the kinetics of appearance of V3-glycan bNAb responses in the three RMs, we characterized longitudinal plasma neutralization against a panel of V1 and V3 point mutants in heterologous Env backgrounds (Fig. 2C and SI Appendix, Fig. S6). At all time points tested, plasma neutralization for the three RMs was enhanced by removal of the N133 and N138 V1 glycans. However, removal of the glycan at residue 156 enhanced bNAb recognition in 44645, diminished it in BV57 and had no effect in 44715. This phenotype was maintained across all time points except for 44715 week 80 where neutralization was enhanced twofold. In BV57 and 44715, V3 bNAb activity was strictly dependent on the glycan at residue N332, but this was not the case in 44645 where elimination of N332 led to enhancement or no difference in neutralization potency. Opposite results were observed for the N301 glycan, with plasma from 44645 showing strict dependence and plasma from 44715 and BV57 exhibiting enhancement or indifference. Finally, in RMs 44715 and 44645, neutralization was dependent on the D325 or I326 residues at early but not later timepoints, while BV57 plasma neutralization was consistently dependent on I326 across all timepoints. Overall, these findings confirm the V3-glycan supersite specificity of the bNAbs in 44715, 44645, and BV57, but at the same time, highlight differences in amino acid-level specificity of the responses.

Monoclonal Antibodies Recapitulate Plasma Breadth.

From all three RMs, we identified antibody lineages responsible for plasma neutralization breadth and determined when in the course of infection they were primed. Monoclonal antibodies (mAbs) representing bNAb lineages were identified by single cell sorting of memory B cells that bound heterologous Env trimer probes (Fig. 3A and SI Appendix, Fig. S7). A single IgG+ lineage member was identified from RM 44715 at week 64 and from RM 44645 at week 120. These were designated 44715-B.01 and 44645-A.01, respectively, with letters indicating the lineage and numbers specifying the mAb. Eight lineage members were identified from BV57 at week 88 post infection, represented by BV57-A.01. These mAbs largely recapitulated the plasma breadth detected against the 16-strain heterologous panel (Fig. 3B). Two of these antibodies (44715-B.01 and 44645-A.01) exhibited immunogenetic features typical of prototypic V3-glycan bNAbs, including VH3 or VH4 family genes and long (19 to 20 amino acid) HCDR3s containing multiple tyrosine residues (Fig. 3A) (23, 32). In contrast, BV57-A.01 from RM BV57 was less typical. It had a short HCDR3 of 14 amino acids, akin to the recently described human EPTC112 V3-glycan bNAb (53), and utilized a VH2 family gene (Fig. 3A). Thus, there were few immunogenetic features shared among all three antibody lineages, suggesting that different germline precursors were primed, expanded, and affinity-matured to breadth.

Fig. 3.

Monoclonal V3-glycan bNAbs recapitulate plasma breadth. (A) Immunogenetic summary of isolated mAbs. DH RF = D heavy chain reading frame, HCDR = heavy chain complementarity determining region, LCDR = light chain complementarity determining region. (B) Comparison of neutralization breadth between plasma (ID₅₀) and isolated mAbs (IC₅₀) for each RM. Autologous CH848 strains are shown at the top and heterologous strains are shown under these. (C) Neutralization curves for each mAb against a heterologous virus and corresponding point mutants. WT virus is shown in black, isogenic mutants that are fivefold more sensitive to neutralization are shown in yellow.

We confirmed that all bNAbs targeted the V3-glycan epitope using isogenic epitope-specific mutants. The bNAb isolated from RM 44715 was dependent on D325 and H330 residues as well as the N332 PNGS to neutralize CAP256SU, consistent with plasma mapping (Fig. 3C). The bNAb isolated from RM 44645 was also dependent on D325, H330, and the N332 PNGS in this backbone, again consistent with plasma mapping (Fig. 3C). Last, the BV57-A.01 mAb was dependent on the N156 and N332 PNGS for neutralization in the Ce1176 backbone, and neutralization titers were reduced by mutations in the D325 and I326 residues, recapitulating plasma neutralizing activity (Fig. 3C). These data confirmed that the isolated mAbs represented plasma neutralization breadth, targeted the conserved V3 supersite, and bound canonical amino acids and glycans in this epitope.

To further characterize the neutralization breadth of the bNAb lineages, we tested each of them against a multiclade 119 strain heterologous virus panel (54, 55). 44715-B.01 and 44645-A.01 each neutralized 23% of the heterologous strains with geometric mean IC₅₀ titers of 0.18 mg/mL and 2.96 mg/mL, respectively (SI Appendix, Fig. S8A and Dataset S4). This corresponded to 33% and 30% of heterologous strains with an N332 glycan, with neutralization sensitivity fingerprinting revealing high levels of similarity to human V3-glycan bNAbs in contrast to other specificities (SI Appendix, Fig. S8B). Surprisingly, BV57-A.01 neutralized just 3 heterologous viruses on the large 119-strain panel, which precluded neutralization fingerprinting analysis. We suspect the limited breadth of this mAb is due to its short HCDR3 and its dependence on the V1 region for binding (see below).

Structural and Isotypic Diversity in bNAb Lineages.

To further define the features of V3-glycan bNAbs elicited by V1 truncated versions of SHIV.CH848, we performed negative stain electron microscopy (NSEM) of mAbs 44715-B.01, 44645-A.01, and BV57-A.01 in complex with the CH848 Env to compare their binding angles and footprints. Consistent with the neutralization data, bNAb Fabs bound to both autologous and heterologous Env trimers, and showed clear binding in the V3 region (Fig. 4 A–C, Left two panels). Fitting a model of the CH848 Env into the NSEM maps (Fig. 4 A–C, Middle), we found that the 44715-B.01 mAb footprint overlapped with the conserved ³²⁴GDIRQAH³³⁰ region as well as the N301 and N332 glycans (Fig. 4A). The binding angle and orientation of this mAb were very similar to the human V3-glycan bNAb DH270, an NSEM structure of which was nearly superimposable (Fig. 4 A, Right two panels). 44645-A.01 had a similar binding footprint that overlapped with the ³²⁴GDIRQAH³³⁰ motif as well as N332 and N301 glycans but exhibited a slightly different angle of approach and was more similar to the human bNAb BF520 (Fig. 4B). Unlike bNAbs 44715-B.01 and 44645-A.01, the density of BV57-A.01 was shifted toward the V1 loop and overlapped more with the N156 glycan, likely explaining its dependence on this glycan for neutralization (Fig. 4C). This angle of approach of BV57-A.01 was structurally dissimilar from previously isolated V3-glycan bNAbs and closer to strain-specific antibodies that have previously been reported (56). The overlap with the V1 loop suggests possible contacts in the V1 region, which is poorly conserved between HIV strains, possibly explaining why this lineage was the least broad of the three bNAb lineages. Nonetheless, BV57-A.01 is an example of the wide variety of responses that can be expected to be elicited by CH848 Env-based immunizations. Our combined analyses of these antibody lineages thus reinforce how breadth can be achieved in structurally diverse ways by antibodies targeting the conserved V3-glycan supersite (32, 33).

Fig. 4.

Isolated bNAbs target the V3-glycan epitope in structurally diverse ways. Left two panels show NSEM reconstructions of mAb-Env complexes in orthogonal views, trimers are colored gray, Fabs in shades of pink. Middle panels show the mAb footprint on a model of the CH848 trimer to indicate location of key features in the V3 region: red = GDIR motif; black = N332 glycan; blue = N301 glycan; green = N156 glycan; gold = V1-loop. Right two panels show orthogonal orientations of the Fab-bound protomer, overlaid with known bnAbs or nonbroad NAbs that have a similar angle of approach. (A) 44715-B.01 mAb (pink) bound to the heterologous 25710 Env trimer (gray), and overlaid with the human V3-glycan bNAb DH270 (purple). (B) 44645-A.01 (pink) bound to the heterologous 25710 trimer (gray), and overlaid with the human V3-glycan bnAb BF520. (C) BV57-A.01 (pink) bound to the autologous CH848.1017DT trimer (gray), and overlaid with the nonbroad rabbit-derived mAb 43A2 (EMDB 42464). Scale bars represent 10 Å.

In addition to structural diversity, we also observed isotype variation among the lineages. While the bNAb lineage members from both 44715 and 44645 began as IgG, those from 44645 class switched to IgA over time (SI Appendix, Fig. S9A). We also observed a large shift from IgM, IgG, and IgA to predominantly IgA in the BV57 bNAb lineage. These changes occurred independently along multiple branches of the lineage phylogenetic trees and so could not be explained simply by expansion of a single clade (SI Appendix, Fig. S9B). Interestingly, the plasma IgA fraction neutralized the infecting SHIV more potently than did the plasma IgG fraction in both RMs, suggesting that they arose as a consequence of positive selection for binding during affinity maturation and that they contributed substantially to plasma neutralization (SI Appendix, Fig. S9D). The IgA fractions from both animals also neutralized multiple heterologous strains which further demonstrated their contribution to overall plasma breadth (SI Appendix, Fig. S9D). These findings demonstrate functional isotypic diversity of the bNAb responses, explain why we identified so few IgG⁺ mAbs from each lineage, and suggest that IgA responses should be considered in future bNAb isolation strategies.

Early Priming of bNAb Lineages.

To elucidate the timing of bNAb lineage priming, we performed longitudinal B cell receptor (BCR) sequencing. We amplified IgM, IgD, IgE, IgG, and IgA mRNA from peripheral blood mononuclear cells and created BCR libraries for next generation sequencing (NGS). We used a recently published rhesus immunoglobulin reference database (macaque unified set of alleles—MUSA) (57) together with the IgDiscover program (58) to create an individualized germline reference dataset for each macaque using the IgM sequences. We then used this reference to annotate the longitudinal BCR sequences using the SONAR pipeline (59). We identified sequences of lineage members from multiple timepoints and used IgPhyML-assisted phylogenetic analyses to infer each unmutated common ancestor (UCA) sequence (SI Appendix, Fig. S10 and Dataset S5) (49, 60). All lineage members were confirmed to be clonally related to isolated bNAbs based on shared mutations from germline genes and similarities in the nontemplated regions. Using this approach, we were able to identify lineage members from a minimum of five timepoints prior to bNAb isolation for heavy chains and six timepoints for light chains (SI Appendix, Fig. S10).

Our longitudinal NGS data enabled us to estimate the timepoint of lineage priming and expansion with high confidence. As expected, we observed an increase in somatic hypermutation (SHM) over time, with early lineage members being minimally diverged from their assigned germline genes (Fig. 5A). Additionally, these data combined with the coalescence of early lineage members into a single clade provides strong evidence that we captured the lineage shortly after its initial expansion (SI Appendix, Fig. S10). Based on this analysis, we could determine that the 44715 and BV57 lineages were likely primed between 8 and 12 WPI, while the 44645 lineage was primed 12 to 16 WPI. These timepoint estimates predate the first detection of mutational escape in V3-glycan ³²⁴GDIR³²⁷ and N332 epitopes in the evolving viral quasispecies by at least 0 to 12 wk (mean 6.7) and the first detection of neutralization breadth in plasma by at least 12 to 32 wk (mean 24), illustrating the value of longitudinal BCR sequencing in the determination of timing of bNAb precursor priming. Given that the human DH270 lineage was first detected in PBMCs more than 3 y postinfection (27), these data strongly support our hypothesis that CH848 Env V1 truncations accelerate V3-glycan bNAb induction.

Fig. 5.

V3-glycan bnAb lineages were primed early during infection by Envs with shortened and deglycosylated V1 loops. (A) Nucleotide divergence of lineage members from the respective UCA over time for the heavy chain (colored points, closed circles) and light chain (gray points, open circles). (B) Env V1 alignments of SGS sequences from the first timepoint of bnAb lineage detection. Dots indicate identity, dashes indicate deletions, PNGS are highlighted blue, and insertions or deletions are highlighted orange. The number of SGS amplicons matching each V1 loop sequence is shown on the Right. (C) Summary of V1 characteristics from panel B. Each SGS sequence is shown as an open symbol, while the infecting Envs are shown as closed symbols. Sequences from 44715, 44645, and BV57 are shown as circles, triangles, and squares, respectively. A small amount of noise has been added to prevent overplotting. The CH848.WT Env is shown as a diamond for reference, and histograms represent V1 characteristics in the most recent HIV-1 Env Super Filtered Web Alignment from www.hiv.lanl.gov.

To investigate what circulating CH848 Envs primed these lineages, we first expressed each bNAb lineage UCA as an IgG mAb and tested it for binding and neutralization of relevant Env variants. While the mature bNAbs from 44715, 44645, and BV57 readily bound CH848 V1 deletion variants expressed as soluble trimers or gp120s, we did not find evidence of binding of ΔV1 Envs to any UCA (SI Appendix, Fig. S11). Similarly, while the mature bNAbs were able to neutralize virus bearing these Envs, the corresponding UCAs did not (SI Appendix, Fig. S11). These findings are consistent with several prior studies that showed most V3-glycan bNAb UCAs fail to bind detectably or neutralize unmodified primary Envs, even those hypothesized to have primed the lineages in vivo (12, 16, 17, 27, 32). We therefore examined SGS data from the timepoint when each lineage was first detected to determine the predominant viral forms in the circulating quasispecies. At these early timepoints, quasispecies diversity was low and there was minimal divergence from the infecting Env (SI Appendix, Fig. S4). Minimal selection was evident in the V1 loop in RM 44715 at week 12, and the vast majority of sequences had no changes from CH848.w36con (Fig. 5B). Slightly more V1 selection was apparent in 44645 by week 16, although none of these mutations increased the V1 loop length or glycosylation (Fig. 5 B and C). In BV57 there was strong selection for restoration of V1 glycans by week 12, but most sequences contained only one PNGS and only one contained an insertion (Fig. 5 B and C). Thus, at the time when the three V3-glycan bNAb lineages were primed, the virus quasispecies in each monkey retained the V1 features of the initial infecting Env. We conclude that either the infecting Env or a closely related variant with a short and hypoglycosylated V1 loop likely primed these bNAb lineages as no other Envs were apparent at these timepoints. Overall, these findings of early lineage priming coinciding with a relatively homogeneous ΔV1 virus quasispecies provide a plausible mechanistic explanation for accelerated V3-glycan bNAb induction in SHIV.CH848 ΔV1 infected macaques.

Frequent Induction of V3-Glycan bNAbs by CH848 Envs.

Having demonstrated accelerated V3-glycan bNAb induction by CH848 Envs with short V1 loops, we sought to determine whether these Envs elicited V3-glycan bNAbs more frequently compared with other HIV-1 Envs. We thus performed a meta-analysis comparing the frequency of V3-glycan bNAb induction in 16 RMs infected with wildtype or ΔV1 SHIV.CH848 variants vs. 109 RMs infected by SHIVs expressing any one of 15 other primary HIV-1 Envs (49) (Fig. 6A). Thirty-one percent of the former group vs. 2% of the latter developed V3-glycan bNAbs (P = 0.00033, Fisher’s exact test). We next analyzed the proportion of individuals in the CH848 group that developed V3-glycan bNAbs vs. other bNAb specificities and compared this with three published cohorts of 19, 46, and 109 humans who also developed bNAbs (5, 7, 61) (Fig. 6B). These results showed that among those individuals who developed neutralization breadth, CH848 Envs exhibited a striking propensity for eliciting V3-glycan bNAbs compared with participants in the other groups. When the results of the three human studies were combined, we found that 44 of 174 (25%) of human subjects with potent bNAbs contained antibodies that targeted the V3-glycan bNAb supersite compared with one human and five macaques infected with CH848-derived viruses. Thus, we could reject the null hypothesis that infections with the CH848 Env have the same probability to elicit V3-glycan bNAbs as other primary Envs (P < 0.00001; Fisher’s combined probability test) and conclude that V1 shortened CH848 Envs have an innate propensity for eliciting V3-glycan bNAbs, a finding that may be advantageous for HIV-1 vaccine design.

Fig. 6.

Enhanced induction of V3-glycan bNAbs by CH848 Env derivatives. (A) Fraction of 16 RMs infected with SHIVs expressing CH848.WT or CH848 variants that developed V3-glycan bNAbs compared with 109 RMs infected with SHIVs expressing one of 15 other primary HIV-1 Envs (49). P value calculated using Fisher’s exact test. (B) Fraction of individuals with V3-glycan bNAbs, excluding those that did not develop neutralization breadth. The CH848 group is composed of human CH848 and 5 RMs infected with CH848.WT or CH848 variant Envs. P values calculated using proportion tests compared to the CH848 group and adjusted for multiple comparisons using the Benjamini–Hochberg procedure.

Discussion

The elicitation of V3-glycan bNAbs alone (12, 15–17, 21, 32, 35, 36, 62, 63) and in combination with other epitope specificities (44, 64–66) is a priority for investigators focused on HIV-1 vaccine development. Much of this work has utilized the BG505.N332 Env as a platform to develop immunogens that bind and activate V3-glycan bNAb precursors (12, 16, 17, 21). In each case, either the V1 loop length, number of glycans, or both were reduced relative to BG505.N332, suggesting this region played an important role in germline-targeting. Similarly, Skelly (32) and Clark (33) reported that naturally occurring deletions in the V1 loop of BG505.N332 and its closely related derivative 5MUT were temporally associated with priming of V3-glycan bNAbs. They also showed that these V1 truncated variants bound the corresponding V3-glycan bNAb UCAs. Similar results have been found for the CH848 Env where naturally occurring deletions in the V1 loop in a chronically infected human (27) and in two SHIV-infected macaques (39) were temporally associated with the elicitation of V3-glycan bNAbs. In a DH270 V3-glycan bNAb UCA KI mouse model, the CH848.1017DT Env effectively primed lineage precursors that then could be affinity-matured to acquire breadth and potency by sequential immunizations with Envs containing longer and more glycosylated V1 loops (15, 35, 36). Altogether, these findings provide strong circumstantial evidence that the length and glycan coverage of the V1 loop can be key determinants of efficient V3-glycan bNAb precursor binding. In the present study, we designed a prospective trial to test directly if truncation and hypoglycosylation of the CH848 Env V1 loop leads to an acceleration of V3-glycan bNAb induction and an overall increase in the frequency of bNAb elicitation relative to other Envs. Using different V1 deleted SHIV variants, we found that this was indeed the case. Importantly, we demonstrated that early bNAb induction was due to accelerated priming of bNAb precursors based on three independent analyses: i) earlier detection of heterologous neutralization activity in plasma (Fig. 1); ii) earlier detection of virus quasispecies escape from bNAb activity (Fig. 2); and iii) earlier detection of bNAb lineage members in circulating B cells (Fig. 5). We further demonstrated by meta-analysis that CH848 Envs as a group elicited V3-glycan bNAbs at a higher frequency than did other primary Envs (Fig. 6), all findings highly relevant to germline-targeting immunogen design.

Other findings of importance for HIV-1 vaccine development from this study included the diverse immunogenetic (Fig. 3) and structural (Fig. 4) features of V3-glycan bNAbs that were responsible for neutralization breadth and potency in the infected monkeys. IGHV family 2, 3, and 4 genes were represented in the three isolated bNAb lineages along with diverse IGHD and IGHJ genes. HCDR3 lengths varied widely from 14 to 20 amino acid residues and both kappa and lambda light chains were represented. Furthermore, structures of the rhesus V3-glycan bNAbs were conserved in essential features of epitope recognition but were otherwise diverse. For example, the binding footprints of all three antibodies overlapped the conserved ³²⁴GDIRQAH³³⁰ motif as well as the N301 and N332 glycans, whereas the binding angles and orientations of the three bNAbs were distinct (Fig. 4). This was consistent with plasma neutralization mapping data (Fig. 2) where we observed different amino acid and glycan dependencies in each RM. These findings, along with recent characterizations of V3-glycan bNAbs from SHIV.5MUT and SHIV.BG505 infected macaques (32, 33), indicate that the immunogenetic features of V3-glycan bNAbs can be extremely variable. Altogether, the immunogenetic, structural, and phenotypic properties of the three rhesus V3-glycan bNAbs characterized in this study highlight the diversity of antibodies that can be expected to result from CH848.1017DT and CH848.1017 immunizations in the HVTN307 clinical trial, which is testing these particular immunogens, and the types of antibodies one could expect to elicit in a clinically protective HIV-1 vaccine.

An emerging theme in recent studies showing bNAb induction in RMs is that efficient priming of germline bNAb precursors is paramount. We found that the efficiency of priming of naïve germline B cell precursors is the primary obstacle to V2 apex bNAb induction (49, 67), and that bNAb elicitation could be markedly enhanced by engineering Envs to exhibit much greater affinity (68) or avidity (69) to naïve germline bNAb precursors. Evidence reported from the current study, and recent findings by others (32, 33, 70), suggest the same is true for V3-glycan bNAbs. In each case, germline-targeted Envs that exhibited enhanced binding to bNAb precursors led to bNAb induction and subsequent neutralization breadth in plasma more rapidly and in a far higher proportion of RMs than what had been observed previously with more native-like Env immunogens (49). Thus, going forward, a balanced emphasis on further improvements in priming of diverse germline precursors, and development of boosting immunogens that can efficiently affinity mature them, will be important.

SHIV infection can uniquely inform boost immunogen designs since the evolving quasispecies can efficiently mature antibody responses to breadth, as demonstrated here and elsewhere (20, 32, 33, 39–42). Nonetheless, there are likely differences in the immune microenvironment between infection and immunization that alter B cell responses in important ways. On this point, a surprising finding in this study was evidence of class-switching to IgA by the majority of bNAb lineage members in two of the three macaques with V3-glycan bNAbs. Anti-HIV-1 IgA⁺ bNAb lineages have been described in humans, but not to our knowledge in the rhesus model (71, 72). IgA class switched B cell responses may be a desirable component of vaccine elicited immune responses due to their association with mucosal immunity (73). Additionally, in some cases mAbs expressed as IgA have been found to exhibit superior breadth or potency compared to their counterparts expressed as IgG due to differences in the constant regions (74, 75). In our study, IgA plasma fractions neutralized the autologous viruses more potently than their IgG counterparts, and overall, these data suggest that eliciting similar responses through vaccination could be of benefit. Although beyond the scope and focus of the current investigation, IgA responses to HIV-1 infection of humans, SHIV infection of macaques, and vaccinations in both species represent a promising and untapped area of investigation.

The current study has certain limitations. First, only 30% of evaluable animals developed bNAbs. The reason the remaining RMs did not is unknown. In the three animals that did develop bNAbs, only a single antibody lineage was found to explain serum breadth in each. This, along with the absence of antibody-mediated selection in the V3-glycan bNAb epitope supersite, suggest that priming efficiency remains a substantial obstacle. This is in contrast to three other SHIV models of bNAb induction where priming of bNAb precursors was substantially more efficient (32, 42, 68). The extent of neutralization breadth exhibited by polyclonal serum antibodies and by isolated mAbs from the three animals with bNAbs in this study was also not optimal. This is likely due in part to the particular features of each activated germline precursor and to the Env-Ab coevolutionary pathways that ensued with SHIV infection, eventually leading to complete virus escape from bNAb recognition. New Env designs that lead to a much higher bNAb precursor priming efficiency together with nonevolving prime and boost immunogens to replace replicating SHIVs as a means of antigen delivery, could overcome these problems.

It has been only in the last few years that reports have begun to appear describing the consistent elicitation of high-titer, potent HIV-1 bNAbs in the plasma of outbred animals other than ungulates, whose unique HCDR3 repertoires make translatability to humans limited (32, 42, 49, 63, 68–70, 76–80). The current report adds to this growing body of research and enables iterative studies of candidate vaccines, delivery platforms, and adjuvants in a clinically relevant model. This work complements human studies that are currently underway, including HVTN protocol 307 (NCT05903339), which is evaluating CH848.1017DT Env as a prime followed by CH848.1017 as a boost. With the first results from that trial expected later this year, we anticipate an opportunity for synergies in the iterative design and testing of CH848-based immunogens in rhesus and humans going forward.

Materials and Methods

Nonhuman Primates.

Indian rhesus macaques were housed at Bioqual, Inc., Rockville, MD, in accordance with guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care standards, as previously described (32). Careful attention was given to the physical and mental well-being of the primates. All procedures involving animals followed written standard operating procedures developed in accordance with guidelines of the American Veterinary Medical Association, including anesthesia and euthanasia (46). All experiments were approved by the Institutional Animal Care and Use Committees of the University of Pennsylvania and Bioqual, Inc. All animals used in this study were healthy and naïve to HIV and SHIV exposure. SHIV stocks used for infection were quantified by measuring p27 antigen using an ELISA (ZeptoMetrix, 0801169). Infections were performed by intravenous administration of 50 ng of p27 from SHIV stocks diluted in DMEM (Gibco, 10566-016) with 10% FBS passed through a 0.2 μm filter. Two RMs (T944 and T945) were inoculated with a 1:1 mixture of SHIV.CH848.1017 and SHIV.CH848.1017DT, while all others were inoculated with a single SHIV (SI Appendix, Table S1). All animals also received an intravenous infusion of 25 mg/kg anti-CD8β (Nonhuman Primate Reagent Resource CD8b255R1) 2 to 3 d prior to SHIV inoculation to transiently deplete CD8 T cells. After completion of this study, RM 44715 was not euthanized and went on to develop bNAbs targeting the V2 apex epitope that have previously been reported (49, 81). Due to the lack of selection in the V2 apex epitope, inability to detect the V2 apex bNAb lineage in B cell NGS, and lack of phenotypic mapping of heterologous neutralization to the V2 apex, we concluded this lineage was primed after 120 WPI and therefore did not affect the results reported in this study.

Processing and Storage of Clinical Samples.

Blood samples were collected from infected RMs into sterile vacutainers containing Anticoagulant Citrate Dextrose Solution A (ACD-A), as previously described (32). 40 mL of anticoagulated blood was combined into a sterile 50 mL polypropylene conical tube and centrifuged at 1,000 × g for 10 min at 20 °C. The plasma phase was collected, clarified by centrifugation, aliquoted into 1 mL cryovials, and stored at −80 °C. The white blood cell (WBC) and red blood cell (RBC) fractions were combined, resuspended in an equal volume of Hanks’ Balanced Salt Solution (HBSS) without Mg²⁺/Ca²⁺ and containing 2 mM EDTA, and distributed evenly into four sterile 50 mL tubes. The volume of each tube was adjusted to 30 mL with HBSS-EDTA (2 mM), underlaid with 14 mL of 96% Ficoll-Paque (diluted with HBSS-EDTA), and centrifuged at 725 × g for 25 min at 20 °C with slow acceleration and braking. Cells at the phase interface were collected, transferred to a sterile 50 mL tube, diluted to a final volume of 50 mL with HBSS-EDTA (2 mM), and centrifuged at 500 × g for 15 min at 20 °C. The cell pellet was resuspended in 30 mL of HBSS with Mg²⁺/Ca²⁺ and 1% FBS, then centrifuged at 200 × g for 15 min at 20 °C. If heavy RBC contamination was observed, lysis was performed using RBC Lysis Buffer (Biolegend, 420301). The cell pellet was resuspended in 30 mL of HBSS with Mg²⁺/Ca²⁺ and 1% FBS, and cell count and viability were determined with acridine orange/propidium iodide (AOPI) staining. Cells were centrifuged at 300 × g for 10 min at 20 °C, resuspended at 5 to 10 million cells/mL in CryoStor CS5 cell cryopreservation media (Sigma, C2999), and aliquoted into 1 mL cryovials (ThermoFisher, 374503). Cells were stored in a CoolCell container (Corning, 432008) at −80 °C overnight and then transferred to vapor-phase liquid nitrogen for long-term storage.

Viral Stock Generation.

The plasmids encoding SHIV.CH848.1017, SHIV.CH848.dV1, and SHIV.CH848.w36con were constructed by individually cloning the synthetic vpu-env gene cassettes into the SIV/HIV proviral backbone plasmid pCRXLTOPO.SHIV.v2.backbone1 (38). Subsequently, the N133D and N138T substitution mutations were introduced using the Q5 Site-Directed Mutagenesis Kit (NEB, Catalog No. E0554S) to generate the allelic derivative SHIV.CH848.1017DT. All isogenic and combinatorial mutations in the SHIV plasmids were generated by this method. The SHIV plasmids were amplified in MAX Efficiency Stbl2 Competent Cells (ThermoFisher, 1026819) for large-scale DNA purification, and the entire viral genome sequence of each SHIV construct was confirmed by sequencing. SHIV stocks were generated by transfection of HEK 293T/17 cells (ATCC, CRL-11268). Briefly, five million cells were plated in 10 cm tissue culture treated dishes in DMEM supplemented with 10% FBS (Hyclone, SH30071.03) and 1% Penicillin (100 U/mL)/Streptomycin (100 μg/mL) (Gibco, Catalog No.15140-122) and incubated overnight at 37 °C with 5% CO₂. Six μg of SHIV plasmid DNA was then transfected using FuGENE 6 transfection reagent (Promega, E2691) according to the manufacturer’s instructions. Supernatants were collected after 48 h, centrifuged at 500 × g for 5 min, aliquoted, and stored at −80 °C.

Plasma Viral RNA Quantification.

Plasma viral load measurements were performed by the NIH/NIAID-sponsored Nonhuman Primate Core Virology Laboratory at the Duke University, as previously described (42). This core facility is CLIA certified and operates highly standardized, quality-controlled Applied Biosystems Real-Time SIV and HIV vRNA PCR assays. QIAsymphony SP and QIAgility automated platforms (QIAGEN) were used for high-throughput sample processing and PCR setup. Viral RNA was extracted and purified from plasma, annealed to a target-specific primer, and reverse transcribed into cDNA. The cDNA was treated with RNase and added to a custom real-time PCR master mix containing target-specific primers and a fluorescently labeled hydrolysis probe. Thermal cycling was performed on a QuantStudio3 (ThermoFisher) real-time qPCR instrument. Viral RNA copies/mL was interpolated using quantification cycle data. The lower limit of quantification of this assay was 62 RNA copies/mL.

Neutralization Assays.

Neutralization assays were performed using TZM-bl indicator cells (BEI Resources/NIAID/NIH, HRP-8129), as previously described (32). Approximately 1.5 × 10⁴ TZM-bl cells were plated in flat-bottom 96-well plates in 100 μL of DMEM supplemented with 10% FBS and 1% penicillin/streptomycin and cultured at 37 °C with 5% CO₂. After 24 h, plasma samples were heat-inactivated at 56 °C for 1 h and serially fivefold diluted starting at 1:20. To maintain a constant plasma concentration of 5% v/v across all wells, media was supplemented with 10% human serum (Millipore Sigma, H5667) at all but the highest dilution. Alternatively, monoclonal antibodies were serially fivefold diluted starting at 50 or 250 μg/mL in media that was not supplemented with human serum. Virus was diluted to achieve a final multiplicity of infection (MOI) of 0.3 when added to the TZM-bl cells. Virus was incubated with plasma or mAb dilutions at 37 °C for 1 h, and this mixture was then added to the adherent TZM-bl cells and cultured at 37 °C with 5% CO₂. Forty-eight hours later, cells were lysed with 0.5% Triton X-100 in PBS and luciferase activity was quantified using a luciferase assay reagent (Promega, E1501) and a BioTek Synergy Neo2 plate reader (Agilent). The same 16 heterologous virus screening panel was selected to match the one used to detect neutralization breadth in SHIV.CH848.WT-infected RMs (39).

SGS.

The SHIV rev-vpu-env cassette was sequenced from viral RNA extracted from plasma samples as previously described (49). Reverse transcription of RNA was performed using reverse transcriptase (SuperScript III) and reverse primer 5′CACTAGCTTACTTCTAAAATGGCAGC-3′. The rev-vpu-env cassette was amplified by nested PCR using the following primers: first round forward 5′-CGAAAGGCTTAGGCATCTCCTATG-3′, second round forward 5′-TTAGGCATC TCCTATGGC AGGAAGA-3′, first round reverse 5′-CACTAGCTTACTTCTAAAATGGCAGC-3′, and second reverse 5′-TACTTCTAAAATGGCAGCTTTATTGAAGAGG-3′. Single-genome template PCR products were sequenced via the Illumina MiSeq platform. Sequence alignment and analysis were conducted using Geneious Prime software (v2026.1.1).

Fluorescence Activated B Cell Sorting.

Broadly neutralizing antibody (bNAb) lineages were derived from memory B cells that were identified by staining with antigen-specific probes and isolated by single-cell index sorting directly into 96-well PCR plates. Briefly, 5 to 10 million PBMCs from macaque BV57 (week 88), 44715 (week 64), and 44645 (week 120) were thawed and washed in RPMI 1640 supplemented with 10% FBS and 50 U Benzonase (Millipore Sigma, 70664-3). Following centrifugation at 300 × g for 5 min at 4 °C, cells were washed once with PBS and stained with LIVE/DEAD Aqua (ThermoFisher, L34957) for 15 min at room temperature. After two PBS washes, the cells were stained with a cocktail of CD3-PerCP-Cy5.5 (BD, 552852), CD4-BV785 (Biolegend, 317442), CD8-BV711 (Biolegend, 301044), CD14-PE-Cy7 (Biolegend, 301814), CD20-BV605 (Biolegend, 302334), IgD-FITC (Dako, F0189), and IgM-BV650 (BD, 740595) for 15 min at room temperature. Cells were washed twice more with PBS and then stained with fluorophore-conjugated SOSIP probes for 15 min at room temperature: macaque BV57 was stained with heterologous probes CAP256SU-PE and Ce1176-APC, while macaques 44715 and 44645 were stained with heterologous probes Ce1176-PE and Ce1176-BV421. Memory B cells (CD3^– CD4^– CD8^– CD14^– CD20⁺ IgD^– IgM^–) that simultaneously bound both CAP256SU-PE and Ce1176-APC (for BV57) or both PE and BV421-labeled Ce1176 probes (for 44715 and 44645) were index-sorted at one cell per well into 96-well plates using a BD FACSAria II flow cytometer operated with BD FACSDiva software (BD Biosciences). Cells were sorted into wells containing 20 μL of lysis buffer containing 0.5 μL RNAse OUT (40 U/μL, Invitrogen, 10777019), 5 μL 5× First Strand Buffer (Invitrogen, 18080044), 1.25 μL DTT (0.1 M, Invitrogen, 18080044), 0.0625 μL IGEPAL CA-630 detergent (Sigma, I8896), and 13.25 μL nuclease-free water.

Single B Cell Receptor Sequencing.

Immunoglobulin heavy (IgH) and light (IgK, IgL) chain genes were amplified from mRNA as previously described (32). Briefly, single B cell RNA was reverse transcribed by adding 2.3 μL random hexamers (200 ng/μL, Thermo SO142), 2 μL dNTP mix (10 mM, Invitrogen 18427088), 1 μL Superscript III reverse transcriptase (200 U/μL, Invitrogen 18080044), and 0.7 μL nuclease-free water per well and incubating in a thermocycler at 42 °C for 10 min, 25 °C for 10 min, 50 °C for 1 h, and then 94 °C for 5 min. IgH, IgK, and IgL genes were amplified by two rounds of nested PCR, each using 3 μL of template, 5 μL 10× buffer (Qiagen 203207), 1 μL dNTP mix (10 mM, Qiagen 201901), 0.5 μL MgCl₂ (25 mM, Qiagen 203207), 0.5 μL carrier RNA (1 mg/mL, Qiagen 1017647), 0.75 μL forward primer mix (50 μM per primer), 0.5 μL of each reverse primer (25 μM), 0.4 μL HotStart Taq polymerase (5 U/μL, Qiagen 203207), and 39.7 μL nuclease-free water. PCR conditions were as follows: 95 °C for 15 min, then 50 cycles of 95 °C for 30 s, 51 °C (Round 1) or 58 °C (Round 2) for 30 s, 72 °C for 30 s, with a final extension of 72 °C for 7 min. Amplicons were visualized by agarose gel electrophoresis and Sanger sequenced (Genewiz).

Monoclonal Antibody Expression and Purification.

Immunoglobulin heavy and light chain variable regions of interest were synthesized and cloned (Genscript) into rhesus IgG, IgK, or IgL expression vectors immediately upstream of the constant region using AgeI/NheI, AgeI/BsiWI, or AgeI/ScaI restriction sites, as previously described (32). Recombinant mAbs were expressed by cotransfecting paired heavy and light chain plasmids into 293F cells (Thermo, R79007) at a 1:2 ratio using ExpiFectamine transfection reagent (Gibco, A14525), purified using rProtein A/Protein G Sepharose Gravitrap kit (Cytiva, 28985256), concentrated in 30 kDa centrifugal filters (Sigma, UFC903024), and buffer exchanged into PBS. Antibody concentration was measured fluorometrically using a Qubit Protein Assay kit (Invitrogen, Q33212).

Negative Stain Electron Microscopy.

Frozen aliquots of SOSIP and IgG were thawed from −80 °C in a RT aluminum block and stored at 4 °C as needed. SOSIP was mixed with IgG at a ratio of 1.5:1, IgG:Trimer. Mixtures were diluted to 400 µg/mL SOSIP by adding buffer (150 mM Na₂SO₄, 2.5 mM NaCl, 2.8 mM NaN₃, 20 mM HEPES pH 7.4) or not diluted if already <400 µg/mL, and then incubated 1 h at RT. A portion of the mixture was then diluted 1:1 in the same buffer with 16 mM glutaraldehyde added, fixed 5 min, quenched by adding 1 M Tris (pH 7.4) to give 80 mM final Tris concentration, and incubated for 5 min. Quenched sample was then diluted with the same buffer to 100 µg/mL (or used undiluted) and applied to a glow-discharged carbon-coated EM grid for 10 to 12 s, blotted and stained with 2 g/dL uranyl formate for 1 min, then blotted and air-dried. Grids were examined on a Philips EM420 electron microscope operating at 120 kV and nominal magnification of 49,000x, and ~100 images were collected on a 76 Mpix CCD camera at 2.4 Å/pixel. Images were analyzed by 2D class averages and 3D reconstructions calculated using standard protocols with Relion 3.0 (82).

Bulk B Cell Receptor Sequencing and Analysis.

Vials of 5 to 10 million cryopreserved PBMCs were thawed and cells were resuspended in 10 mL of RPMI supplemented with 10% FBS, 1% penicillin/streptomycin, and 50 U Benzonase (Millipore Sigma, 70664-3). The cell suspension was centrifuged at 300 × g for 5 min at 4 °C, resuspended in 600 μL of cold RNAzol RT (Molecular Research Center, RN190), and RNA was extracted according to the manufacturer’s instructions. Rhesus heavy and light chain NGS libraries were prepared using a customized SMART-Seq Human BCR kit (Takara, 634778) that substituted human primers for rhesus primers that are currently in precommercial status. All isotypes (IgM, IgD, IgG, IgA, IgE, IgK, and IgL) were amplified and sequenced according to the manufacturer’s instructions. IgM BCR sequences were used to construct individualized germline references for each RM (58) and all isotypes were analyzed for sequences clonally related to bNAb lineages using the SONAR pipeline (59). Lineage UCAs were inferred with the help of IgPhyML (60) as has been described in detail (49). Improbable amino acid mutations were identified using ARMADiLLO with the respective bNAb and UCA sequences as input (83).

Enzyme-Linked Immunosorbent Assay.

Half-area 96-well (Corning, Cat. #3690) plates were coated overnight at 4 °C with anti-His tag mAb (Thermofisher, Cat. #PA1-983B) then washed three times with 0.2% (v/v) Tween 20 supplemented phosphate-buffered saline (PBST). Plates were then blocked for 1 h with 5% (w/v) fat-free milk powder dissolved in PBST. 0.25 μg of recombinant Env trimers or gp120 subunits diluted in PBS were added to the wells for 1 h and washed as above. Fourfold serially diluted mAbs were added to wells in duplicate for 1 h at room temperature and washed as above. Antibody binding was detected with Peroxidase-conjugated AffiniPure Donkey Anti-Human IgG (H+L) (Jackson ImmunoResearch Laboratories, RRID: AB_2340495) followed by a PBST wash. HRP detection was quantified with addition of 3,3′,5,5′-tetramethylbenzidine (TMB) (Sigma Aldrich, Cat. #T0440) for 10 min and this reaction was stopped by the addition of 1 N sulfuric acid (Ricca, Cat. #8300-32). HRP activity was determined as the optical density at 450 nm subtracted by the background optical density at 570 nm.

Data Analysis and Visualization.

Data were analyzed and plotted in either GraphPad Prism version 11.0.2 or R version 4.3.1. Statistical tests were performed as described in figure legends. R packages dplyr (v1.1.4), ggplot2 (v3.5.2), ggtree (v3.8.2), ggseqlogo (v0.2), pheatmap (1.0.13), and patchwork (v1.3.1) were used for data analysis and visualization. Tables were created and formatted in Microsoft Excel.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

All env sequences are available in Dataset S3 and all monoclonal antibody sequences are available in Dataset S5. NSEM density maps are available at the Electron Microscopy Data Bank: EMD-77690, EMD-77691, and EMD-77692 (84–86). Lineage tracing NGS datasets are available as NCBI BioProject PRJNA1499782 under accession codes SRR39784243–SRR39784282 (87). All other data are included in the manuscript and/or supporting information.