Rapid boosting increases germinal center responses to sequential vaccines

Reimmunization during ongoing germinal centers increases VRC01^gHL cell frequencies

Class-switched memory B cells infrequently participate in GCs following conventional boost immunizations in mice. In an effort to bypass this problem, we explored reimmunization of mice during an ongoing GC response. Mice were first seeded with 1,000 VRC01^gHL B cells (1 in 10⁶ precursor frequency) and primed with BG505-GT3.1 protein mixed with ISCOM adjuvant (Extended Data Fig.1a,b). A stabilized Env trimer, BG505-GT3.1 has germline targeting mutations grafted onto the CD4bs, giving it affinity (86–240 nM see methods) for the germline VRC01 BCR^17,51. A second immunization was given on day 6, 8, 10, or 12, with or without adjuvant. Three days after subcutaneous injections (d+3), the draining inguinal lymph nodes (iLNs) were harvested for analysis. No significant differences were observed in the outcomes of d6–8 reimmunization versus d10–12 for the protein-plus-adjuvant reimmunization groups (Extended Data Fig.1, Lognormal Welch’s t test). Germinal center B (B_GC) cells, plasma (B_PC) cells, and VRC01^gHL GC B cell numbers all increased after reimmunization as compared to singly immunized animals (Extended Data Fig.1c-e). B cells increases after reimmunization depended primarily on antigen inclusion, as protein-only reimmunizations showed equivalent outcomes to protein-plus-adjuvant (Extended Data Fig.1c-g). For all protein-plus-adjuvant reimmunizations, total B_GC numbers increased 5- to 7-fold compared to single immunizations. VRC01^gHL B_GC numbers increased 30- to 84-fold (d10–12 and d6–8, respectively). VRC01^gHL cell frequencies in GCs increased from 4–8% to ~33%. In some mice, VRC01^gHL cells became the majority of B_GC cells (Extended Data Fig.1f). CD4⁺ T cell increases after reimmunization appeared to depend on adjuvant, while plasma cell (B_PC) numbers were increased with protein-plus-adjuvant as well as protein alone (Extended Data Fig.1d,h). These findings are consistent with slow delivery and escalating dose priming investigations in other experimental contexts^65,66. Overall, reimmunization with homologous protein at 6 to 12 days post-prime resulted in large increases in VRC01^gHL B_GC cell numbers three days later.

Heterologous reimmunization at 1-week influenced VRC01-class BCR specificity

Heterologous booster immunizations are required for germline targeting strategies. Therefore, we tested the impact of heterologous early immunizations during ongoing GCs (Fig.1a-g)(Extended Data Fig.2a-l). To the best of our knowledge, no investigations of this kind have been reported. An Env trimer from 001428–2.42 clade C HIV with an N276 glycan mutation was chosen as the sequential antigen, referred to here as 001428-N276Q^25,26. Naive VRC01^gHL cells have no measurable affinity for 001428-N276Q, confirmed by probe staining (Fig.1b) and surface plasmon resonance (SPR) (Extended Data Fig.2a). Reimmunizations using Env trimer and ISCOM adjuvants were given at 1- or 6-weeks and evaluated for GC responses seven days post boost (d+7) (Fig.1a-g). The B_GC and CD4⁺ T cell counts significantly increased after 1-week reimmunization compared to the more conventional 6-week reimmunization (P ≤ 0.001, Fig.1c)(Extended Data Fig.2c). B_GC cell totals were 2-fold higher for heterologous reimmunization and 3.3-fold higher for homologous reimmunization when using 1-week reimmunizations (Fig.1c).

Homologous reimmunization at 1-week boosted VRC01^gHL cell frequencies and numbers to levels 24-fold higher than what conventional 6-week reimmunization elicited (Fig.1d). Animals receiving 1-week homologous reimmunization had the largest numbers of 001428-N276Q⁺ VRC01^gHL B_GC cells amongst all groups, despite being 001428-N276Q antigen naïve (Fig.1e). Endogenous antigen-specific B_GC cells were not boosted by 1-week immunizations (Fig.1f, Extended Data Fig.2f). Endogenous mouse B cells have been shown to predominantly target the non-neutralizing, Env-trimer base⁶⁷.

Heterologous immunization at 1-week increased total numbers of B_GC cells compared to 6-week reimmunizations (Fig.1c). VRC01^gHL B cells were comparable in number to heterologous reimmunization at 6-weeks (Fig.1d-f). Strikingly, heterologous 1-week reimmunization increased the frequency of 001428-N276Q probe-binding (001428-N276Q⁺) VRC01^gHL cells compared to non-boosted or homologously boosted animals at matching timepoints (70% vs 30%, P < 0.0001, Fig.1g). Comparing heterologous to unboosted animals after 2 weeks, the boost is also seen to expand the 001428-N276Q⁺ VRC01^gHL cell numbers reaching significance (P = 0.019 Fig. 1e).

In sum, although naive VRC01^gHL cells do not bind 001428-N276Q, 1-week heterologous booster immunization resulted in rapid increases in 001428-N276Q⁺ VRC01^gHL B_GC cell numbers and frequencies 7 days after boosting.

GC B cells rapidly responded to heterologous antigen

To investigate the kinetics of the B_GC cells and their antigen-binding specificities following reimmunization, we studied GC responses over a period of three weeks after homologous (BG505-GT3.1) or heterologous (BG505-N276D) reimmunization (Fig.2a). Similar to 001428-N276Q, BG505-N276D has no detectable affinity for the naïve VRC01^gHL BCR^17,51. Upon reimmunization with homologous or heterologous protein at d12, B_GC cell frequencies rapidly (d+3) increased for both groups (Fig.2b). VRC01^gHL B_GC cells expanded the greatest amount after homologous reimmunizations (Fig.2c). BG505-N276D-binding VRC01^gHL cells were detected after homologous or heterologous reimmunization (Fig.2d), with the most rapid increase in BG505-N276D⁺ VRC01^gHL B_GC cell frequencies observed after heterologous reimmunization (Fig.2e). BG505-N276D⁺ VRC01^gHL B_GC cells were sustained for multiple weeks (Fig.2f).

Homologous reimmunization at 1-week had resulted in higher VRC01^gHL B_GC cell numbers, 8- and 15-fold over controls for d+3 and d+7 respectively (Extended Data Fig.1e, Fig.1d). It was unknown whether these increases were the result of newly primed VRC01^gHL cells or expansion of existing VRC01^gHL B_GC cells. We therefore performed a series of experiments to determine the ontogeny of the expanded B_GC cells. Immunofluorescence histology was performed on draining iLNs excised before or after reimmunization at d7 (Fig.2g). VRC01^gHL-GFP cells were transferred at day −1 prior to the priming immunization and detected using an anti-GFP fluorescent antibody. Large GCs with many VRC01^gHL-GFP cells were observed at d+3 after reimmunization (Fig.2h,i, yellow). The rapidity indicated that expansion of existing VRC01^gHL-GFP B_GC cells was likely responsible for the GC increases, rather than new priming and differentiation of naive VRC01^gHL-GFP cells.

A second experiment was designed to examine this further. A transfer of CD45.1⁺ VRC01^gHL cells before priming was followed by a second transfer of naive GFP-VRC01^gHL cells (constitutively expressing GFP), the day before reimmunizations (Fig.2j). Reimmunizations used 001428-N276Q, as in Figure 1. Heterologous reimmunization resulted in no observable naive VRC01^gHL cells entering GCs. Mice with no reimmunization did not have GFP⁺ VRC01^gHL cells detectable in their GCs (Fig.2k).

Reimmunization with homologous antigen did result in rare detectable GFP⁺ VRC01^gHL; however, 99% of VRC01^gHL cells in the GC were GFP^–. Thus, the vast majority of responding VRC01^gHL cells were already in GCs prior to reimmunization (Fig.2k). Additionally, GFP⁺ VRC01^gHL cell frequencies in GCs after reimmunization were no higher than GFP⁺ VRC01^gHL cell frequencies in GCs of mice receiving their first immunization (Controls: single transfer, single-immunized on d7, Fig.2k). This result indicated that the naive VRC01^gHL cells present at the time of reimmunization received no net advantage in activation and differentiation into B_GC cells compared to a naïve priming environment. Thus, homologous or heterologous reimmunization at 1-week during a primary response expanded previously activated VRC01^gHL B_GC cells.

Antibody titers regulate VRC01^gHL cell responses to immunization

Antibodies present at the time of immunization influence B cell responses to antigen^39–42,68. Due to different B cell responses after 1-week and 6-week homologous boosts, we sought to measure and then model the antibody landscape at these times of reimmunization. Serum ELISAs for CD4bs-specific IgG and total BG505-GT3.1-specific IgG were performed on immunized mouse serum collected over six weeks (Fig.3a,b). These benchmarks were compared to serum IgG titers of mice that received IgG transfers from immunized animals (Fig.3c-h). To recapitulate pre-existing circulating IgG conditions similar to those present immediately prior to the 1-week reimmunizations described in Figure 1, polyclonal IgG from immunized mice possessing no VRC01^gHL cells was used (Fig.3c,e,f, Extended Data Fig.3a). Transfers of low quantities of serum IgG (10–40μg) reconstituted recipient mice with moderate total anti-BG505-GT3.1 IgG titers but minimal CD4bs-epitope specific IgG (Fig.e,f). Presence of pre-existing anti-BG505-GT3.1 IgG without CD4bs specificity resulted in 10-fold enhancement of VRC01^gHL cell expansion after BG505-GT3.1 immunization compared to mice with no serum transfer (P < 0.001, Fig.3i,j, Extended Data Fig.3b-e).

In a second set of mice, serum IgG was transferred from immunized mice possessing VRC01^gHL cells (Fig.3d). These IgG transfers reconstituted recipient mice with high CD4bs-epitope IgG titers, matching circulating IgG titers 6 weeks after immunization (Fig.3b,g,h). Presence of pre-existing anti-CD4bs IgG resulted in 30-fold impairment VRC01^gHL B_GC cell numbers after BG505-GT3.1 priming compared to mice with no serum transfer (P = 0.015, Fig.3l). IgG transfers similar to 6-weeks increased the number of endogenous BG505-GT3.1-binding B_GC cells, cells that do not target the CD4bs (Fig.3e,g,i,k. Extended Data Fig.3f-i). In conclusion, VRC01-class responses were enhanced by pre-existing IgG not targeting the CD4bs but substantially impaired by pre-existing high anti-CD4bs IgG. These data indicate that early reimmunization is both a strategy to leverage low early antibody titers to increase B_GC cell frequencies and simultaneously avoid the negative impacts of high circulating antibody titers on B_GC cell specificities^65,66,69.

Antibody enhancement of vaccine responses has been previously examined in repeated dosing studies, with observations of increased antigen deposition in GCs^63,65,66,70. We hypothesized that enhancement of B cell responses herein by rapid boosting may be driven by immune complexes captured by complement receptors, which are highly abundant on follicular dendritic cells (FDCs). To test this, antibody transfers were repeated with the addition of cobra venom factor (CVF) to deplete complement (Extended Data Fig.3j). GC enhancement was lost upon complement depletion (P < 0.01 and < 0.001. Extended Data Fig.3j-l), demonstrating a critical requirement for complement and indicating immune complex formation and deposition on FDCs as the most likely mechanism of action of the antibody-mediated enhancement of rapid development of WT Env trimer-binding B_GC.

Rapid evolution of VRC01^gHL B cells capable of binding WT BG505 Env

To determine if rapid reimmunization could cause more extensive affinity maturation, sequential immunizations included stabilized wild-type BG505 Env (BG505 MD39), which will be referred to as “BG505-WT”^18,71,72. BG505-WT includes the 276 glycan and no germline targeting mutations. This was given as the third and final immunogen on day 14, one week after 001428-N276Q reimmunization (Fig.4a). Vaccine-induction of WT-Env binding VRC01-class B cells has long been acknowledged as a very difficult process, with few reported successes^17,22,52. Thus, it was unsurprising that no BG505-WT binding (BG505-WT⁺) VRC01^gHL B_GC cells were observed on d+7 after immunization (3 weeks after priming) (Fig.4a-e).

In the next experiment, the immune response was allowed to continue for ~6 weeks after BG505-WT immunization (Fig.4a, Extended Data Fig.4a-d), given observation of sustained GCs (Fig. 2), and considering the large antigenic distance between BG505-GT3.1 to BG505-WT. Substantial numbers of VRC01^gHL B_GC cells were detected 6 weeks after the last immunization (Fig.4b,c). Strikingly, high numbers of these VRC01^gHL B_GC cells bound BG505-WT (~3,600 cells per mouse. Fig.4d,e). Probe staining also tested VRC01^gHL binding to 001428-WT and 001428-CD4bs.KO, which confirmed VRC01^gHL cells bound the CD4bs epitope (Extended Data Fig.4e). In contrast, endogenous B cells almost exclusively bound non-CD4bs epitopes (Extended Data Fig.4e).

To more precisely determine GC requirements for induction of BG505-WT binding VRC01^gHL cells in 56 days, additional immunization experiments were performed (Fig.4f-h, Extended Data Fig.4f-j). First, combinations of homologous and homologous-plus-WT-Env reimmunizations were compared. Homologous BG505-GT3.1 reimmunization at 1-week followed by BG505-WT immunization, one or six weeks later, both generated robust numbers of WT-binding VRC01-class B_GC cells (Fig.4h). Notably for the 1-week delivery this occurred in the absence of detectable BG505-WT-binding VRC01^gHL B_GC cells close to immunization times (Fig.4e). Nonetheless, BG505-WT immunization had a significant impact as it significantly surpassed the homologous GT reimmunization alone (~7,000 vs. ~400 cells median. P = 0.002. Fig.4h). Moreover, a BG505-GT3.1 prime followed by a conventional 7-week BG505-WT boost—a regimen lacking reimmunization—was not sufficient to robustly mature WT-binding VRC01-class cells (Fig.4h).

Next, combinations including heterologous reimmunization were examined (Fig.4h). Strikingly, heterologous reimmunization with 001428-N276Q Env trimer at 1-week was sufficient to robustly generate BG505-WT VRC01^gHL cells 7 weeks later (Fig.4h). Thus, immunization with BG505-WT was not required to evolve BG505-WT binding VRC01^gHL B cells. Finally, longer periods of maturation after late BG505-WT immunization (+4 weeks) were tested; overall patterns were maintained (Extended Data Fig.5a,b). In sum, by utilizing rapid reimmunizations it was possible to evolve large populations of BG505-WT binding VRC01-class cells with only two immunogens and 56 days.

Differential VDJ SHM outcomes

To better understand the swift development of WT Env binding by VRC01-class B cells after rapid reimmunizations, an in depth examination of SHM and affinity maturation was initiated. VRC01^gHL B_GC cells (CD45.1⁺) from 6 experimental groups were FACS sorted and single cell sequenced for VDJ content (Fig.5a-d, Extended Data Fig.5c). A DNA oligo-barcode was conjugated to the SA-BV421 used for the BG505-WT probe⁷³. The total numbers of VRC01-class WT-binding B cells were similar across each rapid boosting regimen (Fig.5b,c), as in the full data set shown in Figure 4.

Microfluidics emulsion-based single cell RNA sequencing yielded 8,000 individual VRC01-class VDJ sequences across all groups (Fig.5d). The heavy chains (HC)s averaged 9–13 amino acid (a.a.) mutations for reimmunized groups and light (LC) chains averaged 3–5 a.a. mutations per group. Plots show the average of each gel-in-bead emulsion (GEM) originating from distinct pooled animals (Fig.5d). Groups immunized with 001428-based Env exhibited differential accumulation of LC and HC mutations compared to groups exclusively receiving BG505-based Env proteins (P = 0.0005, Fig.5e), and the ratio of LC to HC was inversely correlated (r = −0.76). SHM varied across LC and HC VDJ a.a. positions, with pronounced enrichments in L-CDR1 and H-CDR1 and H-CDR2 (Fig.5f).

When viewing VDJ sequences by group with germline residues masked, marked trends appeared among BG505 scaffolded-trimer regimens and mixed scaffold regimens (Extended Data Fig.6a). To quantify this, 7,000 complete VDJ’s across the six immunization groups were analyzed for effect-size correlations of individual a.a. mutations to either BG505-only or mixed-scaffold regimens (Extended Data Fig.6b,c). 355 significant (adjusted P < 0.05) mutations were identified (Extended Data Fig.6b). The top differential mutations were analyzed in detail (Fig.5g). Among LC mutations, immunization regimens using only BG505 selected CDR1 S31R and S30R mutations in 10% to 18% of clones. This was likely due to the presence of the negatively charged sidechain N276D in BG505-GT3.1 and not in 001428-N276Q available for L-CDR1 charged contacts⁵¹. BG505-only regimens mutated E1 in their first position with 53% frequency (P = 0). The mutation of the nucleotides in this codon interrupted our annotation pipeline so the E1A/G assignment was inferred post analysis (Extended Data Fig.6c).

For H-CDR1 in BG505-only regimens Y32H was a commonly selected mutation (73%) (Fig.5g). In contrast, in 001428-containing immunizations Y32H was only modestly selected (15%) and H35N increased from 15% to 76% instead. Either Y32H or H35N can create H-CDR3 loop-stabilizing hydrogen bonds, which may promote different H-CDR3 orientations⁵¹. In the H-CDR2, immunization regimens using 001428-N276Q resulted in multiple strongly selected mutations (Fig.5g). This included S54R (S55R) – main notations refer to Kabat numbering, italics refer to direct residue number – a mutation which occurred in 91% of sequences after 001428-N276Q, whereas in BG505-only groups either R or G mutations were commonly selected (Fig.5g. Extended Data Fig.6c). The H-CDR2 position T57I (T58I) was found in 45% of BG505-only regimen BCRs, while T57I was almost never detected after 001428-N276Q immunization (1% occurrence, Fig.5g). Residues Q61 (Q62) and K62 (K63) were observed to be frequently mutated only after 001428-containing immunization regimens (Fig.5f). The combination of H-CDR2 T57, Q61, and K62 mutation data indicate that a strong contact by H-CDR2 may be differentially selected between the two immunogen choices. Overall, deep BCR mutational analysis of HC and LC revealed many mutations highly correlated to reimmunization immunogen exposure.

Correlations of mutations to WT Env binding

DNA oligo-WT-probe UMI counts per cell by scRNAseq paralleled BG505-WT BV421 MFIs by flow cytometry (Fig.6a), indicating probe counts were a proxy for BCR affinity to BG505-WT. Conversely, probe UMI counts did not directly correlate with average total mutations per group, indicating differential impacts of distinct mutations on the ability of the VRC01^gHL BCR to bind WT Env (Fig.6b). WT Env probe binding did not significantly correlate with two VRC01-class scoring metrics (VRC01-class HC score from Briney et. al., Fig.6a, nor VRC01 “key residue” score averages shown in Extended Data Fig.7a), but VRC01 key residue score did correlate with total HC mutations (Extended Data Fig.7a)^17,23,25.

Next each individual VDJ mutation was correlated to probe UMI count increases. Separate analyses were performed for BG505-only and 001428-containing regimens. The top positive and negative correlations by FDR corrected P-values were examined (Fig.6c,d and Extended Data Fig.7c,d). Mutations identified by probe-UMI-correlation were largely consistent with differential mutations observable in sequence logos (Extended Data Fig.7b).

LC mutations associated with WT Env binding were examined. For VRC01-class human bnAbs, selection of small residues or deletions in L-CDR1 is important to allow for accommodation of the N276 glycan^55,74. In this study, S28, S30, and S31 with G/T mutations were positively correlated with BG505-WT binding (Fig.6d). Additionally, L-CDR1 deletions were present in all reimmunized groups, with the highest frequencies observed in group D (6.4%. 124 sequences with L-CDR1 deletions in group D. Extended Data Fig.6a).

In HCs strong Y33F correlations (P = 10^–22 and 10^–13) corroborate prior structure-function studies of VRC01-class antibodies demonstrating that reduction of the Y33 bulky side chain improves virus neutralization^22,75. S54 (S55) was mutated in nearly 100% of BCR sequences becoming S54R or S54R/G/H/N for the two different regimens (Extended Data Fig.6a). Many human VRC01-class bnAbs rely on this H-CDR2 position to mimic F43 of CD4 using bulky side chains (F/T/Y/R/W)⁴⁸. The commonality of S54R may explain why our correlations did not link it to WT binding. Less common, S54H (S55H) was significantly enriched among high WT Env-binding BCRs in the BG505-only regimens (Corrected P = 8.7e-25, Fig.6d). S54H has been previously observed as a key residue in the human VRC01-class bnAb PCIN63–71I⁷⁴.

T57V (T58V) in the HC was mutated in the majority of WT Env probe-binding sequences and exhibited some of the highest single position correlations with WT Env binding (Fig.6d). T57 is mutated, most commonly to a valine, in every known VRC01-class bnAb thus matching this correlation. Further, Chen et al. identified T57V as a mutation of high importance for HIV neutralization by VRC01-class bnAbs²². Our correlations conversely found T57I (T58I) had a negative association with BG505-WT binding (Fig.6c) which echoes its absence in any known VRC01-class bnAb. Significant correlations were observed for mutations at N100a (N105) for both BG505-only and 001428-containing immunization groups, N100aT and N100aD respectively (Fig.6d). This residue of the H-CDR3 can exhibit both main and side chain contacts with Env^48,53. N100aD is observed in three human VRC01-class bnAbs, and N100aT appears in 12A12 (Extended Data Fig.7c)^53,74. Overall, a number of known and important LC and HC VRC01-class mutations were observed and correlated with acquisition of WT Env binding, in addition to identification and interpretation of novel or lesser studied mutations. Thus, analyses of mutation frequencies and probe UMIs identified mutations relevant for VRC01-class B cell affinity maturation, including mutations previously validated in human VRC01 bnAb analyses.

Binding and neutralization of fully glycosylated Env viruses

To further validate the inferred WT binding of sequenced BCRs, a subset was selected for expression as mAbs. ~1,000 additional VDJs were added to the dataset from a smaller experiment. The now 9,351 gHL-VRC01 sequences were down selected to 25 mAbs. For down selection, clonal families were inferred using LCs and HCs with near-matching mutational patterns (Extended Data Fig.8a). Families with consistently high BG505-WT UMI counts were prioritized and a representative member was progressed. Sequences were next examined for unique complementary determining region (CDR) mutations, particularly in L-CDR1 (Fig.7a-c. Extended Data Fig.8b). The 25 expressed mAbs, notated as RPD01–25, included sequences from each immunization group (Extended Data Fig.9a).

21 of 25 expressed mAbs bound BG505-WT (ELISA, Extended Data Fig.9b). Two of the non-binding mAbs were from the prime-only animal group, consistent with the absence of BG505-WT-binding VRC01^gHL B cells by flow cytometry from that group (Fig.4g, Extended Data Fig.5c). Affinity measurements determined that all mAbs bound BG505-WT with 2 × 10^–5 to 4 ×10^–7 K_D, except for the sequences from the prime only animals, which had no binding (Fig.7b. Extended Data Fig.9c,f). No mAbs bound Env with CD4bs knock out mutations, confirming the binding modality was VRC01-type, as expected (Extended Data Fig.9d).

The 13 mAbs with the best WT Env binding were then evaluated for pseudovirus (PSV) neutralization on a CD4bs-sensitive panel of HIV viruses (Fig.7b,c). Aside from the 001428-N276Q HIV PSV, the six other PSV used were fully glycosylated, including at residue N276. BG505 and 191084 B7–19 are HIV clade A strains, while 001428 is clade C and 1HD2 clade B. All 13 mAbs efficiently neutralized 001428-N276Q PSV (Fig.7b). 11 of 13 mAbs neutralized 001428 WT PSV with an IC₅₀ below 11 μg/mL with 8 mAbs neutralizing with an IC₅₀ < 1 μg /ml. Three of the mAbs neutralized 001428 WT comparably to the potency of mature VRC01. Six of 13 mAbs neutralized BG505-WT PSV, the best of which neutralized BG505-WT with an IC₅₀ of 1.1 μg/ml. Of the remaining viruses tested for neutralization, 1HD2 and 191084 B7–19 were neutralized by approximately half of the mAbs (Fig.7b). Thus, using rapid reimmunizations and only 8-weeks total vaccination time, VRC01-class precursor B cells evolved the ability to neutralize multiple HIV strains bearing the N276 glycan.

Sequences of the six mAbs that neutralized BG505-WT were compared (Fig.7c). All six mAbs possessed VRC01-class HC mutations Y33F, M34L/I, N53Y/T, S54R, T57V, K62N/Q and Y32H or H35N. These mutations are consistent with well-described requirements for H-CDR1 and H-CDR2 affinity maturation to bind and neutralize HIV by VRC01-class bnAbs ^22,29,74. Each of these six mAbs also had multiple H-CDR3 mutations including Y100 (Y104) or N100a (N105). Overall, the HCs showed key mutations in H-CDR1 and H-CDR2 and showed that H-CDR3 could be mutated to enhance stabilization or BG505-WT binding.

L-CDR1 SHM is a critical component of VRC01-class bnAb development, with key mutations demonstrated to accommodate the N276 glycan that shields the Env CD4bs⁵⁵. Multiple distinct L-CDR1 mutations were observed among the 6 mAbs that neutralized WT BG505, and the broader set of 11 mAbs that neutralized WT 001428 (Fig.7c). Those L-CDR1 mutations included instances of L-CDR1 single residue deletions (mAb RPD10D) and double residue deletions (RPD16E), as well as multiple different side-chain size reducing mutations, including a single G mutation (RDP07C) and double GG mutation (RPD08D). Nevertheless, mAbs RPD05C and RPD12D neutralized four of the WT viruses without a.a. deletions or glycine mutations in L-CDR1 (Fig.7b). Of note, both RPD05C and RPD12D shared all seven of the aforementioned H-CDR1 and H-CDR2 VRC01-class mutations as well as additional HC mutations that together confer substantial affinity gains over germline VRC01 (Fig.7c). Interestingly, RPD19E, which came from a large clonal family, maintained a near-germline LC and still neutralized BG505 virus (Extended Data Fig.9a). It appears that key HC mutations provided sufficient binding potency for moderate BG505-WT neutralization in the absence of LC affinity maturation, in line with outcomes from VRC01-class motif reversion and swapping experiments done for minVRC01 and PCIN63^29,72.

Germline-reverted or mutation-swapped versions of RPD mAbs were tested for changes in affinity to Env (Extended Data Fig.10a-c). L-CDR1 reversion did not change affinities of high performing mAbs, except RPD20F, which did not possess H-CDR3 mutations. Reversion of the L-CDR1 showed RPD20F depends on it for affinity (Extended Data Fig.10a). AlphaFold3 modeling of RDF20F suggested L-CDR1 interactions between N31 and N32 and the N276 glycan base. Separately, reversion of H-CDR3 105D (N105) to germline was found to greatly reduced affinity to WT BG505 Env, but not 001428 (Extended Data Fig.10a,c). H-CDR3 loop modelling indicated that G95R (G99R) and N100aD (N105D) may result in an H-CDR3 loop conformational change, reorienting the Y104 residue of RPD20E differentially to BG505 versus 001428 Env (Extended Data Fig.10d). This intra-loop interaction could explain why full H-CDR3 reversion of RPD08 maintained more robust Env binding than N105D reversion (Extended Data Fig.10a). Overall, the observations of a necessity for L-CDR1 or H-CDR3 affinity maturation but not both for WT Env binding is consistent with prior VRC01-class bnAb findings^{29,55,74,76,77}.

The most potent mAb identified post-vaccination, RPD08D, neutralized four WT viruses and had the lowest geometric mean IC₅₀ (Extended Data Fig.9e). RPD08D possessed 18 a.a. mutations in total, with 7 a.a. mutations fulfilling previously defined VRC01-class bnAb mutation criteria, including two glycines in the L-CDR1 and canonical Y32H, Y33F, T57V, Q61H, and S54R mutations (Extended Data Fig.9a). Furthermore, the RPD08D H-CDR3 contained four mutations of which G95R (G99R), K96R (K100R), and N100aD (N105D) were highly correlated with increased BG505-WT binding in our sequence analysis when considering both regimen analyses (Fig.6d). RPD08D came from group D, which had been given a BG505-GT3.1 rapid boost followed by BG505-WT rapid boost (Fig.5a). Thus, transition from GT to WT Env binding occurred in these rapidly boosted germinal centers. Altogether, the mAbs with WT HIV Env neutralization activity possessed known VRC01-like mutations. These data demonstrate that neutralization of multiple tier-2 HIV strains can rapidly evolve within 56 days.

Protein expression and purification

Immunogens used in this study were generated as in Lee et. al. 2022⁵¹. Trimers used the MD39 SOSIP backbone to achieve stabilized closed confirmations. Sequences originated from the BG505 strain or 001428–2.42 strains as noted²⁵. CD4 binding site knock out trimers were generated using the G473T mutation⁸¹. Trimers were transiently transfected into HEK-293F cells (Thermo Fisher Scientific: R79007) using polyethylenimine. Cells were harvested after 7 days and supernatant was purified by Galanthus nivalis lectin or 2G12 antibody affinity chromatography. Trimers were polished by SEC on Superdex 200 16/600 size column (Cytvia: 28–9893-35). Molecular weight and homogeneity were confirmed by size-exclusion chromatography-multi-angle light scattering (SEC-MALS) in phosphate-buffered saline (PBS) using Superdex 200 Increase 10/300 GL column (Cytiva) operating with an isocratic flow of 0.5 mL/minute followed by DAWN HELEOS II and Optilab T-rEX detectors (Wyatt Technology).

His-Avi-tagged eOD-GT5 monomer and His-Avi-tagged trimers were used for FCM probe and ELISA assays. They were produced using transient transfection of HEK-293F cells and purified as previously described using immobilized metal ion affinity chromatography followed by size-exclusion chromatography (SEC)²⁵. Molecular weight and homogeneity were confirmed by SEC-MALS. Antigens were biotinylated with BirA (Avidity) and re-purified via SEC to remove excess biotin.

Monoclonal antibodies were made by GenScript. Amino acid sequences for LC and HC were codon optimized and inserted into plasmids by GenScript. TurboCHO-HT 2.0 pipeline was used where plasmids are transiently transfected into CHO-S cell lines in 10mls and then supernatant is purified with Protein A Sepharose Fast Flow resin (Cytiva). CE-SDS and SEC-HPLC were used for quality control. Proteins were frozen at −80 in elution buffer (NaAC Buffer,0.2 M L-Arginine,pH5.5) or exchanged into PBS via Zeba 40k MWCO columns (ThermoSci: A57763) for pseudovirus (PSV) neutralization assays.

Mice and Immunizations

Mouse experiments took place at the La Jolla Institute for Immunology (LJI). Mice were housed under standard laboratory conditions in a temperature-controlled (69–75 °F) and humidity-controlled (30–70%) ventilated facility with a 12 h light/12 h dark cycle, with ad libitum access to food and water. Experiments and procedures were approved by the IACUC committee of LJI (IACUC protocol AP00001129). Experiments used age and sex-matched mice between 6–12 weeks of age. Both males and females were used. C57BL/6-J, B6.SJL-Ptprc^a Pepc^b /BoyJ (B6.CD45.1) and CByJ.B6-Tg(UBC-GFP)30Scha H2b/J (B6.GFP) mice from Jackson Laboratory were used in this study (JAX: 000664, 002014, 004353). VRC01-gH/VRC01-gL mice were kept homozygous in B6.GFP or B6.CD45.1 backgrounds. After one backcross, heterozygous VRC01-gH/VRC01-gL (VRC01^gHL) were used as all donors for experiments.

For transgenic B cell transfers, one VRC01^gHL heterozygous mouse and one B6 mouse were euthanized via CO² chamber and spleens were collected. Spleens were mashed and resuspended in 2mls of RPMI with 10% FBS (R10). AKT lysis was performed, cells were filtered through 45uM, and returned to R10. Mouse B cell isolation (StemCell EasySep: Cat. 19854) was done on 700ul of the VRC01-gHL heterozygous spleen following manufacturer’s instructions except for maintaining reaction on ice and in R10. After isolation VRC01^gHL B cells were stained with an abbreviated FCM panel to verify CD45.1⁺/CD45.2^–/B220⁺/CD4^–; this panel included an eOD-GT5 60mer with non-specific AF647 labeling to denote all VRC01gHL-BCR containing cells (Kato et. al., 2020). Purified cells were >95% live via MUSE cell counter and additionally hand counted for verification. VRC01^gHL cell concentrations were adjusted based on eOD-GT5 60mer staining, normally 35–45% of isolated B cells carried the correct VRC01^gHL BCR. Exactly 1000 VRC01^gHL BRC⁺ cells per 200ul injection of RPMI were given per mouse. To increase cell concentrations and cell viability of transfers, we included 500,000 “buffer” splenocytes from B6 spleens^57,59. Cells were transferred via retro orbital injections one day prior to priming immunizations.

Mice were primed subcutaneously, bilaterally, at the tail base with 10μg BG505-GT3.1 or other Env protein and 5U ISCOM adjuvant total. General anesthesia was used and tail-base was shaved before immunization. Unless stated reimmunizations contained 10μg protein and 5U ISCOM. After reimmunizations (d+1 through d+28) draining iLNs and spleens were resected. For serum kinetics blood was collected from submandibular vein puncture.

Soluble immune stimulating complex (ISCOM)-like saponin nanoparticle adjuvant (ISCOMs) was used for all immunizations. ISCOMs is a self-assembling nanoparticle containing cholesterol phospholipids, and quillaja saponin. It was prepared as previously described^59,63. 0.2μg cholesterol and 1μg saponin constitute 1 Unit ISCOMs. ISCOMs is soluble in PBS and does not interact with protein antigens. 10 Units of ICOMs was used per mouse, mixed in PBS with antigen before immunization.

Complement depletion with cobra venom factor (CVF, Complement Technology Inc.: Cat. A150) was performed on two consecutive days with administration of 20ug CVF in 200ul PBS, given intra peritoneally split bilaterally. C3 ELISA (Abcam: Cat. Ab263884) was used to confirm depletion of complement in serum.

IgG purification and transfer

Blood was collected via external cardiac puncture. After clotting for 30+ minutes and centrifugation at >6,000 rpm serum was collected and re-spun for purity. Serum was frozen at −30C until use. For IgG purification Thermo Nab SpinKit Protein A Columns were used (ThermoSci: 89979) followed by concentration in Amicon Ultra-4 4ml Filters 50k MWCO (Millipore: UFC805024) and finally protein was buffer exchanged into PBS with Zeba 40k MWCO columns (ThermoSci: A57763). For transfer, mice were injected intraperitoneally with 100ul in PBS bilaterally one day before immunization.

Flow Cytometry

Inguinal lymph nodes (iLN) and spleens were resected from mice and mashed through a 70μm filter into R10 supplemented with 0.1ul/ml benzonase (MinneBio: MBPN01–50KU). Spleens were processed with AKT lysis (Gibco: A10492–01). ~1/10^th of iLN or 1/50^th spleens were plated for staining in round bottom plates (1–3 million cells). Staining buffer was 3% FBS, 1mM EDTA in PBS (Corning: 2104CV). Staining proceeded with pre-stain blocking, probe stain, and antibody master mix added sequentially in intervals of 15 min, 30 min, 45min. See Supplementary Tables 2 and 3 for details. After staining, cells were washed and 30ul of sphero count beads at predetermined concentrations were added (BDBiosciences: 556296). Cells were collected in staining buffer on a Cytek Aruora 5L instrument.

B cell receptor (BCR) sequencing and analysis

Cells prepared as normal for FCM, with the exchange of SA-BV421 for SA-BV421-biolegend-DNA-oligo (Biolegend: Custom TotalSeqTM-C). Using a Becton, Dickinson and Company S6, cells were sorted for Singlets/Live/B220⁺/GL7⁺/CD38^lo/IgD^– /CD45.1⁺/CD45.2^– into chilled FBS. Cells were then loaded onto a Chromium chip and controller (10X Genomics) following manufacturer’s protocol. V(D)J and Feature hashtag libraries with indexes were prepared following the 10X Genomics protocol for Single Cell 5’ Reagent Kits v2 (Dual Index). Sequencing was performed on NovaSeq 6000 (Illumina). Cell Ranger v7.1 (10X Genomics) was used for demultiplexing and FASTQ generation. VDJ reads were aligned de novo using cell ranger and counts was used for feature oligo counts. All QC metrics were left as in programs. IgBlast was run with the Immcantation Framework™ using a custom reference genome of mouse with additional VRC01 gHL sequences included⁵¹ . AIRR frames of all sequences as database were generated from Immcantation. Python was used to convert oligo-probe counts to barcoded dataframe⁵¹ . Cell barcodes were used to join dataframes in R studio. Light chain and heavy chain dataframes with correct IgH and IgK calls for VRC01 were joined and translateDNA (Immcantation) was used to translate to amino acid sequence. Excel database was used to extract amino acid position, compare to germline, and count mutations. Jupyter Notebooks and Python were used to plot ggplot logos, 2D histograms, extract mutation matrices, run correlation analyses (see Statistical Analysis Methods) and to make volcano plots of correlations.

A VRC01 scoring algorithm was used to grade each complete VDJ sequence. Conventionally, counting of heavy chain residues to those matching positions in six representative VRC01-like bnAbs generated the “VRC01-class HC score” in Fig.6, and a more refined HC key residue count and LC key residue score were used in Extended Data Fig. 5 as published previously^25,26.

Analyses included at times VDJ sequences extracted from live and frozen experiments combined. Following successful thawing of pooled iLN cells from immunized mice, cells were stained and sequenced identical to fresh samples. These sequences within their groups (C and E) and are denoted as coming from GEM numbers 01 and 02.

Enzyme-linked immuno-sorbent assay (ELISA)

BG505-GT3.1 Env was used to assay total IgG against Env and eOD-GT5 monomer was used to readout IgG reactivity for the CD4bs of Env. Corning half area plates (Corning: 3690) were coated overnight with 2.5 μg/mL streptavidin (Biolegend: 280302). Plates were washed (5x in PBS with 0.05% v/v/ Tween20). Plates were incubated in 2.5μg/mL Env-biotin trimer or 0.42 μg /ml eOD-biotin monomer (molar equivalents). For mAb analysis 2 μg/mL streptavidin and 2 μg/mL Env-biotin was used. Plates were washed again, blocked with 3% BSA solution, and serum samples in 1% BSA were added to wells. All incubations were 1hr and followed by washing. Secondary detection used 1:2,000 dilution of anti-mouse HRP (Jackson: 115–035-166) in 1% BSA. After final washes 30 μL of TMB substrate (ThermoSci: 34021) and 30 μL sulfuric acid was used to quench reaction. Plates were read at 450nm and EC50 was calculated using top constrained to plate maximum for standard sample and bottom constrained to highest blank well.

Surface plasmon resonance (SPR)

Affinity and kinetics measurements were performed using a ProteOn XPR36 biosensor (Bio-Rad) with HC30M sensor chips (XanTec) and HBS-EP+ buffer (pH 7.4, Teknova) supplemented with 1 mg/ml BSA. about 6000 RUs of Goat Anti-Mouse IgG Fc (SouthernBiotech 1013–01) were immobilized via EDC/NHS coupling, and IgG ligands were captured at 0.3 µg/ml (100–200 RUs/channel). Antigens were injected at 30 µl/min for 180 s, followed by 300 s dissociation. Surfaces were regenerated with 1.7% phosphoric acid (3 × 60 s). Data were processed in ProteOn Manager (Bio-Rad) with interspot and double referencing, and binding analyzed using a 1:1 Langmuir model.

Neutralization assay

Pseudoneutralization analysis was performed exactly as stated in Cottrell et al. 2024. On a similar but abbreviated N276-glycosylated Envpanel. Briefly, pseudoviruses were produced by co-transfecting HEK293T cells with Env-expressing plasmids and an Env-deficient HIV-1 backbone (pSG3ΔEnv) using Fugene 6 or PEI MAX. Virus-containing supernatants were harvested at 72 hours and stored at –80°C. On the day of the assay, serially diluted mAbs were incubated with pseudovirus (pre-treated with DEAE-dextran) for 1 hour at 37°C before being added to pre-plated TZM-bl cells. After 72 hours of incubation, luminescence was measured following cell lysis and substrate addition. IC50 values were calculated using nonlinear regression in GraphPad Prism. All cells were maintained in D10 medium under standard conditions.

Statistical Analysis

Statistical tests were performed using GraphPad Prism 10. Where all groups had over eight data points one-way ANOVA tests assuming Gaussian distribution and unequal standard deviations (using Brown-Forsythe and Welch ANOVA tests) with Dunnett T3 multiple comparison correction was used. In other cases, when any groups compared had fewer than eight points, Kruskal-Wallis nonparametric uncorrected tests were used. P-values are NS > 0.05, * <0.05 ,** <0.01, *** <0.001, **** <0.0001.

For VDJ mutation statistics, aligned and gapped sequences were filtered for complete only, then joined to one aa string. In Jupyter Notebooks, Python (version 3.9.2) with pandas, NumPy, SciPy, and statsmodels packages were used to analyze mutation-phenotype relationships. For each mutation identified relative to the VRC01 germline sequence, statistical associations with BG505-WT-probe UMI hashtag output counts were evaluated. Spearman's rank correlation coefficient was calculated to assess monotonic relationships between mutation presence and output values. Mann-Whitney U tests were performed to compare output distributions between sequences with and without each mutation. Multiple testing correction was applied using the Benjamini-Hochberg false discovery rate (FDR) procedure to control for false positives across all tested mutations. Mutations were considered statistically significant when both FDR-corrected Spearman and Mann-Whitney p-values were <0.05 and occurred in >1% of sequences.

For group-level comparisons of mutations aquired, two-sided Chi-squared tests with Benjamini-Hochberg FDR correction were performed to assess associations between mutation presence and group membership. Effect sizes were quantified using Cliff's Delta, a non-parametric measure ranging from −1 to +1, where negative values indicate the mutation is depleted in the group of interest, positive values indicate enrichment, and 0 indicates no systematic difference