Well-ordered HIV-1 envelope glycoprotein (Env) trimers are prioritized for clinical evaluation, and there is a need for an improved understanding about how elicited B cell responses evolve following immunization. To accomplish this, we prime-boosted rhesus macaques with clade C NFL trimers and identified 180 unique Ab lineages from ∼1,000 single-sorted Env-specific memory B cells. We traced all lineages in high-throughput heavy chain (HC) repertoire (Rep-seq) data generated from multiple immune compartments and time points and expressed several as monoclonal Abs (mAbs). Our results revealed broad dissemination and high levels of somatic hypermutation (SHM) of most lineages, including tier 2 virus neutralizing lineages, following boosting. SHM was highest in the Ab complementarity determining regions (CDRs) but also surprisingly high in the framework regions (FRs), especially FR3. Our results demonstrate the capacity of the immune system to affinity-mature large numbers of Env-specific B cell lineages simultaneously, supporting the use of regimens consisting of repeated boosts to improve each Ab, even those belonging to less expanded lineages.
Traditional assessments of vaccine-induced antibody (Ab) responses rely on serological assays to determine if immunization has induced the desired Ab specificity and potency. However, measurement of serum Igs does not reveal information about the specific Ab variable (V), diversity (D), and joining (J) segment gene rearrangements responsible for the antigen-specific response, nor about the underlying dynamics and maturation of the responding B cell populations. For a deeper understanding of vaccine-induced B cell responses, we developed protocols for antigen-specific single memory B cell sorting and mAb isolation from immunized rhesus macaques. These studies revealed the targeted epitopes and the mode of recognition by their cognate Abs, providing information that will help guide the design of improved immunogens and immunization protocols (Martinez-Murillo et al., 2017; Navis et al., 2014; Phad et al., 2015; Sundling et al., 2012a). However, the isolation of mAbs is low throughput and typically identifies only one or a few somatic variants from each Ab lineage, yielding limited information about the maturation of the response at the clonal level. In contrast, high-throughput Ab repertoire sequencing (Rep-seq) enables analyses of millions of B cells per sample, allowing definition of large numbers of clonally related sequences and more comprehensive understanding of Ab responses (Davydov et al., 2018; Galson et al., 2014; Georgiou et al., 2014; Jiang et al., 2013; Wiley et al., 2011; Yermanos et al., 2018). The use of Rep-seq is especially valuable if antigen-specific lineages can be identified in the data, as has been demonstrated for HIV-1 infection–induced Ab that undergo extensive affinity maturation (Bonsignori et al., 2016; Doria-Rose et al., 2014; Wu et al., 2015).
The examination of genetic properties of elicited Abs relies on the availability of comprehensive and validated reference databases of Ab VDJ germline gene segments. Even more than humans, rhesus macaques are highly diverse at both their MHC (Shen et al., 2013) and Ab VDJ loci (Corcoran et al., 2016). A comprehensive public reference database of macaque Ab germline genes is not yet available despite recent efforts (Cirelli et al., 2019; Corcoran et al., 2016; Francica et al., 2015; Ramesh et al., 2017). Definition of the specific VDJ alleles present in a given animal is necessary for correct Ab gene assignments, precise determination of somatic hypermutation (SHM), and definition of B cell clonal relationships (Yaari and Kleinstein, 2015), a prerequisite for B cell lineage-tracing studies. To meet this need, we developed IgDiscover, a computational tool allowing rapid generation of individualized Ab germline gene databases for both humans and rhesus macaques (Corcoran et al., 2016).
While the induction of cross-neutralizing HIV-1 Abs through immunization has proven extremely challenging (reviewed in Kwong et al., 2013), persistent efforts have resulted in the design of well-ordered soluble HIV-1 envelope glycoprotein (Env) trimers that closely mimic the native HIV-1 spike (Guenaga et al., 2017; Julien et al., 2013), infusing new optimism in the field. In initial studies, these trimers elicited strain-restricted neutralizing Abs against difficult-to-neutralize (tier 2) clinical HIV-1 isolates (Martinez-Murillo et al., 2017; Pauthner et al., 2017; Sanders et al., 2015), and recent work demonstrated that such responses protect against homologous chimeric simian-human immunodeficiency virus challenge (Pauthner et al., 2019). Several studies are now underway to evaluate the response elicited by a variety of well-ordered trimers, with or without epitope-focused immunogens as a prime (Kong et al., 2019; Xu et al., 2018), in small animal, macaque, or human clinical trials. A major objective of these studies is to induce Ab responses against conserved Env epitopes, such as those defined by broadly neutralizing Abs identified in HIV-1–infected individuals (Huang et al., 2012, 2014; Kong et al., 2016; Scheid et al., 2011; Walker et al., 2011; Wu et al., 2010).
To investigate the dynamics and evolution of the B cell response elicited by Env, we combined single B cell sequencing of Ab heavy chain (HC) sequences with high-throughput bulk repertoire sequencing and Ab lineage tracing analysis. We used the native-like flexibly linked (NFL) trimers derived from a clade C clinical isolate, 16055 (Guenaga et al., 2017), which we previously showed elicited autologous tier 2 neutralizing Ab responses that target the Env V2 region (Martinez-Murillo et al., 2017). We selected the animal that responded best to the vaccination, D20, for an in-depth investigation of the response, akin to Ab tracing studies performed in single HIV-1–infected patients, except that we traced a highly polyclonal response and not just one or a few Ab lineages. We generated Rep-seq data from multiple time points and immune compartments, and we used the IgDiscover software to define the specific germline alleles present in D20 to ensure correct gene assignments, clonal grouping, and SHM calculations.
Our analysis revealed broad dissemination of the response in the draining LN, spleen, blood, and bone marrow (BM), with increasing SHM levels following boosting for almost all lineages analyzed, independent of the expansion level of the lineage. Interestingly, the SHM was not restricted to the complementarity determining regions (CDRs), as numerous mutations also accumulated in the framework regions (FRs), especially in FR3. Functional analysis of variants belonging to a neutralizing Ab lineage targeting the V2 apex of the 16055 Env spike, the GM9_TH8 lineage, showed that enhanced neutralizing potency against the autologous virus was associated with increasing SHM, including residues located in the FR3. These data demonstrate the positive impact of the affinity maturation process on the evolving polyclonal Ab response following repeated boosting.
Env-specific single memory B cell sorting and sequencing identified multiple Ab clonotypes
Previously, we demonstrated the capacity of the well-ordered 16055 NFL Env trimers (Guenaga et al., 2017) to elicit HIV-1 autologous tier 2 neutralizing Abs, which mapped to the V2 region of Env (Martinez-Murillo et al., 2017). Here, we inoculated four macaques (D15, D16, D19, and D20) with trimers in the Matrix-M adjuvant i.m. at weeks 0, 4, 12, 24, 36, and 45 (Fig. 1 A, left). We first administered 16055Δgly4 NFL Env trimers lacking sites for N-linked glycosylation proximal to the CD4 binding site (CD4bs; Dubrovskaya et al., 2017), and we then boosted with WT NFL Env trimers (16055wt) as described in Fig. 1 A. Blood samples were collected 2 wk after each inoculation for analysis of neutralizing Ab activity. All four animals developed plasma tier 2 16055 autologous neutralizing titers, with D20 displaying the most potent response (Fig. 1 A, right). We selected this animal for a comprehensive analysis of the Ab response and sorted Env-specific single memory B cells from peripheral blood mononuclear cells (PBMCs) from the last time point, 2 wk after the sixth immunization. We stained Env-specific memory B cells using a panel of Abs previously shown to cross-react with macaque memory B cells (Sundling et al., 2010) and a C-terminally biotinylated Env trimer probe identical to the 16055 NFL trimer immunogen, except for the presence of a C-terminal Avi-tag. Using FACS, we sorted ∼1,000 single Env-specific CD20+IgG+CD27+ B cells (Fig. 1 B), from which we obtained 672 full-length HC VDJ sequences by Sanger sequencing. For a summary of the sorting statistics and FACS gating panels, see supplemental material (Fig. S1, A and B).
We next collected preimmunization blood samples and sequenced an IgM library from D20 to generate an individualized germline gene database using the IgDiscover software as previously described (Corcoran et al., 2016; Fig. S1 C). We assigned the 672 Env-specific HC sequences isolated from the time point after the sixth immunization (post-6) to the specific germline VH and JH alleles derived from D20 using the individualized database and calculated the level of SHM for each region (Fig. 1, C and D). This analysis demonstrated that the Env-specific B cells used a broad range of alleles, consistent with our previous studies using an earlier-generation Env trimer immunogen (Sundling et al., 2014). Several of the HC sequences were highly mutated, with SHM levels ranging from 5 to 15% at the nucleotide level (Fig. 1, C and D).
We next determined the number of unique B cell clonotypes present among the 672 Env-specific HC sequences. We defined Ab clonotypes as follows: identical V and J allele assignments, identical HCDR3 length, 80% HCDR3 amino acid homology, and at least one junction with no mismatches (called the “80%*” criterion). We included the junction matching requirement because the junctions are an important feature of clonal identity. Using the 80%* criterion, we defined 180 clonal Ab lineages among the 672 HC sequences (Fig. 1 E), with HCDR3 lengths ranging from 5 to 24 aa. There was one highly expanded clone in the single-cell data with an HCDR3 of 7 aa: GM4_TD7. The large representation of this clonal family, composed of 84 variants, was apparent when comparing the HCDR3 length distribution of the starting 672 HC sequences with that of the 180 clonotypes defined by the 80%* criterion (Fig. S1 D).
To confirm the specificity of the FACS sorting, we picked the GM4_TD7 HC sequence and 20 additional HCs from families displaying varying levels of expansion and isolated their matching light chain (LC) VJ sequences from the originally single sorted cells. Expression of the resulting mAbs demonstrated that 17 of 21 were Env-specific by ELISA, including GM4_TD7 (Fig. S1 E). Three mAbs, GM9_TH8, GM3_TC2, and GM6_TG1, neutralized the autologous tier 2 16055 virus, while GM7_TE9 and GM7_TH2 neutralized the MW965 isolate, and the remaining mAbs were non-neutralizing (Fig. 1 F). The most potent 16055-neutralizing mAb, GM9_TH8, was ∼100-fold more potent than the previously reported D11A.F9 and D11A.F2 mAbs (Martinez-Murillo et al., 2017). To determine if the GM9_TH8 mAb recognized a similar V2 epitope as the D11A.F2 mAb, we performed cross-competition ELISAs (Fig. 1 G). The results demonstrated that the two Abs competed efficiently with each other, while they did not compete with PGT145 (trimer apex), VRC01 (CD4bs), or 3BC315 (gp41 interprotomeric). We also performed a subtractive binding assay using 16055 gp120 with or without a deletion in the V2b region (gp120ΔV2b) targeted by the D11A.F2 mAb, which showed that GM9_TH8 was able to bind gp120 but was unable to bind the ΔV2b protein. The control mAb, b6, which recognizes the CD4bs, bound gp120 and gp120ΔV2b equivalently (Fig. 1 H). Finally, we compared the fine epitope specificity of D11A.F2 and GM9_TH8 by performing a neutralizing assay using viruses carrying individual alanine substitutions across the V2b region. This revealed that GM9_TH8 bound a similar but not identical epitope in the V2b region, perhaps in part explaining its increased potency compared with D11A.F2 (Fig. 1 I). We performed further ELISA binding assays to map additional mAbs and found that three of the non-neutralizing mAbs, GM5_TF7, GM11_TG8, and GM9_TH5, bound the 16055 gp140 trimers but not the 16055 gp120 monomers, suggesting that they were directed against gp41 (Fig. S1 E). The two 16055-neutralizing mAbs, GM3_TC2 and GM6_TG1, bound WT gp120 and gp120ΔV1, but did not bind gp120ΔV1V2 or gp120ΔV2b, demonstrating that they had similar epitope specificity as GM9_TH8. In contrast, the MW965-neutralizing mAb, GM7_TH2, bound all four gp120 ligands (WT gp120, gp120ΔV1, gp120ΔV1V2, and gp120ΔV2b), and it also bound the 16055 V3 peptide, demonstrating that it was V3-specific (Fig. S1 E). The remaining non-neutralizing mAbs that recognized gp120 were not further mapped.
VH gene usage in the different immune compartments was similar
In addition to the peripheral blood sample used for Env-specific FACS sorting and mAb isolation, we collected BM, spleen, draining inguinal LN (iLN), and ileum lymphoid tissue (gut) from the post-6 termination time point for library constructions and Rep-seq analysis of animal D20 (Figs. 2 and 3). In addition, libraries from blood post-1, post-2, and post-4 time points (Fig. 4) and BM post-2 and post-4 time points (Fig. S3 B) were generated for longitudinal analysis of Env-specific lineages. To establish baseline information, we sequenced the post-6 total IgG repertoire from each compartment and defined the frequencies of V and J gene usage in each library. We assigned all IgG HC sequences with unique bar codes (see Materials and methods) to the individualized germline V gene database of D20 and observed abundant usage of some germline genes and moderate-to-low frequency usage of others. There was no marked difference in V gene usage in the overall IgG repertoires in the different compartments compared with in the preimmunization blood IgM repertoire (Fig. 2 B). The similarity in V gene usage between the pre- and postimmunization samples is consistent with previous data showing that Env vaccine-induced responses comprise a minor fraction of the total memory B cell repertoire 2 wk after immunization, once the short-lived plasma cell response has declined (Martinez et al., 2015; Sundling et al., 2010). The results were similar for the J genes, with no detectable difference in gene usage between the compartments or between pre- and postimmunization samples (Fig. 2 C).
The Rep-seq library statistics, including numbers of reads, merged reads, barcode clusters, unique VDJ sequences, and numbers of clonotypes (Fig. 2 D), confirm the deep sampling of the B cell repertoire from each compartment. The clonal diversity in the BM was notably lower than that observed in the other compartments, perhaps reflecting a more stringent selection process for B cells to access this compartment. When comparing the HCDR3 lengths of the clonotypes identified in the five compartments, we found a similar length distribution in all compartments analyzed (Fig. 2 E).
Vaccine-induced B cell responses seed multiple immune compartments
The next step in our analysis was to trace the 180 Env-specific lineages in the Rep-seq data generated from the different immune compartments. We used the IgDiscover module Clonoquery (see Materials and methods) to identify clonally related sequences for each given query sequence in the Rep-seq data, using the 80%* criterion to define somatic variants for each Ab lineage. We first investigated how many of the 180 Env-specific lineages isolated from circulating memory B cells could be detected in the BM, spleen, or gut lymphoid tissue, and we also probed the draining LN, which we expected would contain abundant vaccine-induced Ab sequences. Lineage tracing identified more than half of the Env-specific Ab lineages in both the draining LN and spleen, consistent with ongoing germinal center (GC) responses in these compartments 2 wk after boosting as we previously reported using a similar immunization regimen (Martinez-Murillo et al., 2017). We detected several vaccine-induced Ab lineages in the BM, perhaps representing plasma cells destined to persist as long-lived cells. In contrast, we could not trace many Env-specific Ab lineages in the ileum, suggesting limited dissemination to this compartment following i.m. vaccine inoculation. Overall, the numbers of traced Ab lineages were 101 (56%) in blood, 111 (62%) in draining LN, 102 (57%) in spleen, 41 (23%) in BM, and 8 (<5%) in ileum (Fig. 3 A).
The level of expansion varied significantly between the lineages. Some were represented by hundreds of variants in the Rep-seq data and detected in all compartments, while other lineages were represented by only a few variants in selected compartments, and other lineages were not detected at all. We sorted for Env-binding memory B cells, but a fraction of the lineages may be nonspecific, as shown by the recognition profiles displayed by the isolated mAbs (17 of the 21 cloned Abs were Env specific), potentially explaining the lack of expansion for a subset of the lineages following boosting. The non-neutralizing GM4_TD7 lineage, which we already identified as an expanded clone from our single-cell data, was highly expanded, as were the GM2_TD2 and GM9_TH5 lineages (Fig. 3 B). One 16055-neutralizing lineage, GM6_TG1, was represented by many somatic variants in BM, while the more potent 16055-neutralizing lineage, GM9_TH8, could not be traced in BM.
For an independent confirmation that Env-specific B cells were present in the different compartments, we analyzed the samples by B cell ELISpot assay using two distinct protocols that enumerate memory B cells and plasma cells, respectively. We demonstrate that Env-specific memory B cells were abundant in blood, draining LN, and spleen but undetectable in BM and ileum (Fig. S3, left). In contrast, Env-specific plasma cells were present in BM, draining LN, and spleen but undetectable in blood and ileum (Fig. S3, right). These results are consistent with our previous observations that short-lived antigen-specific plasma cells are no longer present in blood 2 wk after boosting and that the BM compartment does not harbor vaccine-induced memory B cells (Sundling et al., 2010). Based on the ELISpot results, our interpretation is that the variants traced in the blood represent memory B cells, while the variants traced in the BM represent plasma cells. Whether these Env-specific plasma cells represent long-lived plasma cells could not be determined with the sampling time points used in the current study. The spleen and LN harbored both antigen-specific memory B cells and plasma cells, consistent with ongoing GC responses in these compartments.
When analyzing V region SHM levels of all variants belonging to a unique Ab lineage, i.e., intraclonal analysis, for a large set of Env-specific lineages, we found that the most highly mutated variants were present in the draining LN and spleen, again consistent with ongoing GC reactions in these compartments at this time point (Fig. 3 C). Because the post-6 time point was the termination time point, we could not determine if these variants eventually entered the circulation. However, these data demonstrate the capacity of the immune system to promote multiple independent vaccine-induced Ab lineages simultaneously after boosting, of which many disseminate to distant immune compartments, including to the BM.
Intraclonal analysis demonstrates increasing SHM in both CDRs and FRs following boosting
We next used Clonoquery to interrogate longitudinal Rep-seq data from animal D20. We used libraries generated from blood collected 2 wk after the second, fourth, and sixth immunizations (post-2, post-4, and post-6). We first determined the level of expansion over time, focusing on 60 of the 180 lineages that we detected most readily in blood (Fig. 4 A). Of these, some lineages increased over time, while others did not. The potent neutralizing Ab lineage GM9_TH8 was among the least expanded, with clonal variants detected only in post-2 and post-6, but not post-4, Rep-seq data. When analyzing SHM levels for individual variants and comparing them intraclonally, we found a gradual increase both between the post-2 and post-4 time points and between the post-4 and post-6 time points for most lineages, independent of their level of expansion. These data suggest an efficient affinity maturation process for a large number of lineages simultaneously in response to boosting (Fig. 4 B), an effect also observed when plotting the average SHM for all lineages traced in the post-2, post-4, and post-6 Rep-seq data (Fig. 4 C). For the lineages we could trace at more than one time point in the BM, we also observed higher levels of SHM at the post-6 time point compared with the post-2 or post-4 time point (Fig. S3 B), suggesting an active recruitment of affinity-matured variants into this compartment.
Evaluation of average SHM levels across the V gene offers only a rough estimation of the Ab affinity maturation process. For a more detailed analysis, we investigated which region of the Abs accumulated most mutations by calculating average SHM frequencies in the Ab FRs (FR1, FR2, and FR3) and the CDRs (CDR1 and CDR2). The results demonstrated that for most lineages, SHM predominantly accumulated in the CDR2, followed by the CDR1. Interestingly, many lineages also accumulated high SHM levels in FR3 and, in some cases, also in FR2 (Fig. 5 A). To examine this in more detail, we selected 10 lineages for which we identified the unmutated, or close to unmutated, precursor VDJ sequence in the Rep-seq data. This allowed us to assess SHM levels also in the HCDR3, which otherwise is difficult due to ambiguities in the nontemplated VD and DJ junctions. One caveat remained, since our criterion for clonality, 80%*, limits the number of mutations tolerated in the HCDR3. Thus, the SHM levels calculated for this region may be an underestimation. For each of the 10 lineages, we identified nine somatic variants from blood, ranging from low to high SHM, which we aligned with the corresponding precursor sequence to inspect region-specific SHM. This analysis suggested an accumulation of immunization-induced mutations in the CDRs but, in the more mutated variants, also extended to the FRs (Fig. S4). When we plotted the region-specific SHM against the overall SHM for the respective VDJ regions, we observed a clear increase in SHM in the FR3s with increasing SHM in the overall VDJ sequence for all 10 lineages (Fig. 5 B). These results suggest that SHM in multiple Ab regions, including FR2 and FR3, contributes to enhance antigen binding.
SHM levels directly impacts HIV-1 Ab neutralizing capacity
We next selected the GM9_TH8 lineage to functionally test if increases in SHM conferred improved Ab binding and neutralization. We selected 11 HCs, in addition to the HC of the cloned GM9_TH8 mAb, with levels of SHM ranging from 0 to 24% at the amino acid level. Of these HC sequences, we identified one variant with no SHM in the V gene and only one mutation in the J gene in a Rep-seq library following a single Env trimer inoculation. This sequence likely represented the germline GM9_TH8 HC, except for a J gene mutation, which allowed definition of the HCDR3 junctions. We also selected one GM9_TH8 variant from the post-2 BM sample (7.4% SHM), one variant from the post-4 BM sample (16.7% SHM), and several variants from the draining LN and spleen (15.6–24% SHM) collected at post-6 (termination) time point. The amino acid alignment of the 12 HC sequences demonstrated that most of the mutations in the low-SHM variants were located in the HCDR2. In the HCs with higher overall SHM levels, mutations additionally accumulated in the other Ab regions, especially in FR3 (Fig. 6 A). To analyze the relatedness of these 12 GM9_TH8 variants in the context of all traced GM9_TH8 variants from all compartments, we generated a maximum-likelihood tree. The unmutated common ancestor sequence of the tree was created using the assigned germline V and J, with the junction and D gene masked with ambiguous (“N”) nucleotides, which are treated as uninformative “missing data” in the phylogenetic likelihood calculation. The resulting tree revealed multiple subbranches, with the GM9_TH8 mAb sequence clustering most closely with variants from the draining LN (Fig. 6 B).
We cotransfected the 12 GM9_TH8 HC sequences with the original GM9_TH8 LC sequence and expressed each variant as both IgG and Fab. We performed ELISAs with or without a NaSCN wash to examine the relative binding avidity of the IgG variants. When the wash was performed in PBS, we observed that the post-1 GM9_TH8 variant (PB_seq681430_post-1), for which the V gene was in the germline configuration, bound the 16055 NFL trimers only weakly, while all other variants, ranging from low SHM to higher, displayed similar binding avidity (Fig. 6 C, left). In contrast, when the ELISA was performed under high salt conditions using a NaSCN wash (Fig. 6 C, right), we observed a marked reduction in binding for the post-2 GM9_TH8 variant (PB_seq24927_post-2), which had seven mutations in the V gene (7.3% SHM), while the more mutated variants of the lineage mostly retained their binding.
For an independent assessment of functionality, we measured the neutralization capacity for both bivalent IgG and monovalent Fabs for each expressed variant of the GM9_TH8 lineage. The post-1 variant of GM9_TH8 (PB_seq681430_post-1), with an unmutated V gene, lacked neutralizing capacity as both IgG and Fab, while all other variants displayed potent tier 2 neutralizing activity as bivalent IgG, including the modestly mutated PB_seq24927_post-2 variant. In contrast, the post-2 variant was only weakly neutralizing as a monomeric Fab (Fig. 6 D). By comparing the monomeric Fab to the bivalent IgG concentration resulting in 50% virus neutralization (IC50) values on a molar basis, it was clear that there was a large effect of avidity on the neutralization potency for this lineage (Figs. 6 D and S5 A). When examining the relationship between SHM and virus neutralization potency for the respective Fabs, we observed a strong correlation for both IC50 and IC80 with the level of SHM (Fig. 6 E). The most highly mutated variant of GM9_TH8 identified in the Rep-seq data, iLNG_seq732127, displayed 24% SHM at the amino acid level and was the most potent neutralizing Ab of the GM9_TH8 variants analyzed here. We named this variant Ab GM9_TH8seq732127 and obtained crystals of this Fab that diffracted x rays to a 2.26-Å resolution. We used this structure to visualize the location of the amino acid residues modified by SHM. We first analyzed SHM that were common between the GM9_TH8 and GM9_TH8seq732127 HC, which included 2 aa in the HCDR2 (Fig. 6 F, left). We then evaluated SHM that were located at the same position but changed relative to the V gene germline sequence to a different amino acid residue in GM9_TH8 (first letter) compared with GM9_TH8seq732127 (second letter; Fig. 6 F, middle). Finally, we denoted all SHM in the GM9_TH8seq732127 HC to visualize the full spectrum of SHM in this Ab HC (Fig. 6 F, right). At the present time, we do not possess a high-resolution structure of the GM8_TH9 lineage Ab with its epitope on the HIV-1 trimeric spike. Therefore, we cannot draw firm conclusions about direct paratope:epitope contacts. However, the analysis performed here suggested that most SHM, including the mutations in the FR3, affected solvent-exposed residues clustered near the HCDR2, which could be involved in GM9_TH8 recognition of its cognate epitope.
Immunization with complex multiepitope proteins results in diverse repertoires of affinity-matured Abs. For HIV-1, the elicitation of protective Abs through vaccination is extremely challenging due to the host-selected N-glycan shield that occludes neutralizing epitopes on the native Env spike (Kwong et al., 2002; Wei et al., 2003). How individual antigen-specific Env-specific B cell lineages evolve following immunization is a subject of recent interest (Cirelli et al., 2019; Luo et al., 2016; Wang et al., 2016; Yacoob et al., 2018), which we address here at a comprehensive level. This aspect is important to understand to design improved vaccine regimens against pathogens for which vaccines do not yet exist.
Previously, we demonstrated that the well-ordered 16055 NFL trimers elicit autologous neutralizing Abs in rhesus macaques that target a region of the Env trimer apex (Martinez-Murillo et al., 2017). As also observed here, boosting did not significantly increase the titers of the neutralizing Ab response, suggesting restricted affinity maturation or poor persistence of the responding B cell clones over time. Here, we traced Env vaccine-induced Ab HC sequences in Rep-seq data to determine how individual Ab lineages evolved in response to trimer boosting. We demonstrate that even poorly expanded Ab lineages accumulated high levels of SHM following boosting, as we observed for the GM9_TH8 lineage. These data indicate that affinity maturation per se is not the limiting factor for the neutralizing Ab response, suggesting that it is the frequency of naive B cells capable of recognizing neutralizing Ab epitopes, and/or the level of expansion of such B cell lineages, that limits the development of effective neutralizing responses. While several clones likely contributed to the plasma neutralizing response, the other 16055-neutralzing Abs isolated here, GM3_TC2 and GM6_TG1, were markedly less potent than GM9_TH8. Furthermore, as observed in this study and in previous studies aimed at isolating Env vaccine-induced mAbs (Martinez-Murillo et al., 2017; Phad et al., 2015), typically only a small fraction of the total number of isolated mAbs are neutralizing. Here, we additionally find that several non-neutralizing mAbs, such as GM4_TD7, displayed a significantly greater level of expansion following boosting compared with the neutralizing Ab lineages traced here.
As expected, SHM accumulated primarily in the HC CDRs and in particular the CDR2; although as expected, each lineage was unique in its SHM pattern. In variants with high overall levels of SHM, mutations also accumulated in the FR3 and to a lesser extent in the FR2. High levels of SHM in Ab FRs has been described as a feature selectively detected in chronically HIV-infected individuals (Klein et al., 2013), but we show here that this can also occur in response to vaccination. The high levels of SHM (24% at the amino acid level) that can accumulate in members of given Ab lineages, affecting both CDRs and FRs, are encouraging, especially regarding the elicitation of HIV-1 neutralizing Abs following sequential Env trimer vaccination. Although we cannot determine definitively which SHM observed in the GM8_TH9 lineage are direct contacts with the cognate Env trimer antigen (some may be neutral or affect Ab stability), that such levels can occur following sequential subunit immunization and result in improved binding and neutralizing Abs is encouraging for future vaccine efforts. Together, these are important observations, since high levels of SHM are displayed by many HIV-1 broadly neutralizing Abs derived from chronic infection and long-term antigen exposure.
Our results demonstrate broad compartmental dissemination of the vaccine-elicited Ab lineages in draining LN, blood, spleen, and BM. Previously, we demonstrated active GC responses in draining LN 2 wk after inoculation with NFL Env trimers administered in Matrix-M adjuvant, also used here (Martinez-Murillo et al., 2017). The presence of GCs at this time point is consistent with the lineage-tracing analyses presented here, which identified hundreds to thousands of somatic variants for each Env-specific lineage in the draining LN and spleen, and considerably fewer somatic variants for each lineage in blood and BM. This is also consistent with the B cell ELISpot analysis reported here, which demonstrates the copresence of Env-specific memory B cells and plasma cells in the draining LN and spleen, where plasma cells likely contributed a greater total number of sequences to the Rep-seq data due to the high copy number of Ab transcripts in these cells. In contrast, we detected only memory B cells in blood and only plasma cells in the BM, which we interpret to represent the GC output. We did not detect Env-specific B cells in the ileum despite high numbers of total IgG memory B cells and plasma cells in this compartment, in agreement with a previous vaccination study (Luo et al., 2016) and a study of total B cell repertoires in humans, demonstrating limited exchange between the gut and peripheral blood B cell repertoires (Meng et al., 2017). This finding, however, does not rule out that soluble IgG induced by i.m. immunization diffuses to mucosal sites, including to the gut, to mediate protective effects in these anatomic sites.
The use of the IgDiscover germline gene inference tool (Corcoran et al., 2016) generated a robust foundation for the precise genetic analyses and lineage-tracing studies reported here. To accomplish nucleotide-level precision, an individualized database is required due to the high allelic diversity in rhesus macaques (Shen et al., 2013) and the lack of a comprehensive public database for macaque Ab germline genes. Our current understanding of the concordance between the peripheral memory B cell compartment and the BM plasma cell compartment is limited. The methodological approach described here offers possibilities to investigate this further. For example, it would be of interest to apply the analysis approach used here, combining single-cell sequencing and mAb isolation with Rep-seq, to elucidate if different modes of vaccination (Cirelli et al., 2019) alter the developmental pathways of vaccine-induced Ab lineages targeting distinct epitopes of Env.
In conclusion, the current study offers a comprehensive, in-depth analysis of the evolution of Ab responses following immunization with well-ordered HIV Env trimers. This study investigates and, to some extent, defines the process of Ab affinity maturation and dissemination of B cells to selected immune compartments. Elucidating the Env-specific polyclonal response at great depth, as performed here, improves our understanding of vaccine-induced responses and provides a basis for future efforts to design more effective immunization regimens.
Materials and methods
Four female rhesus macaques (Macaca mulatta) of Chinese origin, 4–10 yr old, were housed at the Astrid Fagraeus Laboratory at Karolinska Institutet. Housing and care procedures complied with the provisions and general guidelines of the Swedish Board of Agriculture. The facility has been assigned an Animal Welfare Assurance number by the Office of Laboratory Animal Welfare at the National Institutes of Health. The macaques were housed in pairs in 4-m3 cages, enriched to give them ability to express their physiological and behavioral needs. They were habituated to the housing conditions for >6 wk before the start of the experiment and subjected to positive reinforcement training to reduce the stress associated with experimental procedures. All immunizations and blood samplings were performed under sedation with ketamine 10–15 mg/kg i.m. (Ketaminol, 100 mg/ml; Intervet). The macaques were weighed at each sampling. All animals were confirmed negative for simian immunodeficiency virus, simian T cell lymphotropic virus, simian retrovirus type D, and simian herpes B virus.
The animal work was conducted with the approval of the regional Ethical Committee on Animal Experiments (Stockholms Norra Djurförsöksetiska Nämnd). All methods were performed in accordance with the approved guidelines.
HIV-1 Env immunogens and probes
The design, expression, and purification of clade C 16055 NFL TD CC trimers from transient expression in 293F cells were previously described (Guenaga et al., 2015). In brief, the 16055 Env coding sequence was modified as follows: the native signal sequence was replaced by the CD5 signal sequence, the furin cleavage motif at the C terminus of gp120 was replaced with two copies of the G4S flexible linker, a K334S mutation was introduced to restore the 332N-glycan supersite, and eight TD modifications were introduced to stabilize trimer formation, along with the previously described I559P, I201C, and A433C substitutions. The sequence was terminated at D664 followed by a G4S linker and His6 tag. Cell culture supernatants containing the expressed Env proteins were collected 4–5 d after transfection and purified over a Galanthus nivalis lectin-agarose (Vector Laboratories) column followed by size exclusion chromatography to isolate the trimer. For the trimer probe, the His6 tag was replaced by AviTag and biotinylated with BirA (Avidity).
Immunization and sampling
Animals were inoculated six times at weeks 0, 4, 12, 24, 36, and 45 i.m. in a total volume of 1 ml, divided equally between the left and right quadriceps (100 µg Env formulated in 75 µg Matrix-M adjuvant [Novavax] per inoculation; Sharma et al., 2015). The trimers from the first two inoculations contained 4 N-glycan deletions proximal to the CD4bs as previously described (Dubrovskaya et al., 2017). Peripheral blood samples were collected 2 wk after each immunization, and BM aspirates were taken 2 wk after the second, fourth, and sixth immunizations. Tissue samples, including iLN, spleen, and ileum, were collected at termination, i.e., 2 wk after the sixth immunization.
PBMCs and cells from BM were isolated by density-gradient centrifugation from EDTA blood by Ficoll-Paque PLUS (GE Healthcare), washed extensively in PBS, and treated with red blood cell lysis buffer. Single-cell suspensions of splenocytes and LN cells were prepared using a 70-µm cell strainer and a syringe plunger. Cell suspensions were washed once in Ca2+- and Mg2+-free PBS (Sigma-Aldrich) and treated with red blood cell lysis buffer. Cell samples collected from different immune compartments were frozen in 90% heat-inactivated FBS and 10% DMSO (Sigma-Aldrich). Intestinal lamina propria cells were isolated from duodenum, ileum, or colon tissue of rhesus macaques, as previously described for murine tissue (Parigi et al., 2018). Briefly, 5-cm-long intestinal samples were placed in ice-cold PBS. After removal of residual mesenteric fat tissue, the intestines were opened longitudinally and cut into 1-cm pieces. After extensive washing, tissues were incubated in a shaking incubator at 37°C for 30 min in 20 ml of HBSS containing 5% FCS, 5 mM EDTA, 1 mM dithiothreitol, and 15 mM Hepes. Tissue pieces were washed with 20 ml PBS with 5% FCS and 1 mM EDTA followed by PBS with 1% FCS and 15 mM Hepes. Digestion was performed in 10 ml of serum-free HBSS containing Liberase TL (0.15 mg/ml; Roche) and 0.1 mg/ml DNase I (Roche) at 37°C at 600 rpm for 45 min. Cells were washed and passed through a 100-µm cell strainer. Cells were separated at the interphase of a 44–67% Percoll (Sigma-Aldrich) gradient.
Single-cell sorting of HIV-1 Env-specific memory B cells by flow cytometry
Env-specific memory B cells were single-cell sorted from PBMCs from animal D20 sampled 2 wk after the sixth immunization. Single cells were sorted with a three-laser FACSAria cell sorter (BD Bioscience) by gating Live/Dead stain (Life Technologies)–negative, CD3−, CD14−, CD20+, IgG+, CD27+, Env+ cells into 96-well PCR plates containing 4 µl of cell lysis buffer (Sundling et al., 2012a). Plates were sealed and immediately frozen on dry ice before storage at −80°C. All fluorescently labeled Abs and the biotinylated Env-trimer conjugated to streptavidin-allophycocyanin (Invitrogen) were carefully titrated to stain rhesus macaques PBMCs.
Single B cell RT-PCR
Sorted HIV-1 Env-specific single memory B cells from animal D20 were lysed and reverse transcribed, and VDJ sequences were amplified as described previously (Sundling et al., 2012a,b). Briefly, the 96-well plates containing single B cells, were thawed on ice and reverse transcribed to cDNA by addition of random hexamers, dNTPs, and SuperScript III reverse transcription (Invitrogen). The VDJ sequences were amplified separately in 25-µl nested PCR reactions using 3 µl of cDNA in the first-round PCR and 1.5 µl PCR product in the second-round PCR. The HotStar Taq Plus Kit (Qiagen) and 5′ leader sequence-specific and 3′ IgG-specific primers were used. PCR products from the positive wells were purified, sequenced (GATC Biotech), and analyzed. The productive HC and LC sequences were reamplified in 25-µl cloning PCR reactions to add the cloning sites using 2-µl nested PCR product with Kappa Mix (Kappa Biosystems) and 5′ and 3′ custom cloning primers containing restriction sites previously described (Sundling et al., 2012b; Tiller et al., 2008).
Cloning and expression of mAbs
Cloning PCR products were evaluated on 1% agarose gel for correct size (∼450 bp for HC and ∼350 bp for κ or λ LC), and then PCR purified. Cloning of the Ab sequences into expression vectors containing human Igγ1 H, Igκ1 liter, or Igλ2 liter constant regions (Tiller et al., 2008) were performed with FastDigest restriction enzymes (Thermo Fisher Scientific) according to the manufacturer’s instructions. The digested PCR products were inserted into linearized, shrimp alkaline phosphatase–treated vectors using T4 DNA ligase (Thermo Fisher Scientific). XL10-Gold ultracompetent cells were transformed by heat shock at 42°C for 45 s according to the manufacturer’s protocol (Agilent Technologies). Positive colonies were verified for insert by PCR. Bacterial colonies containing plasmids with inserts of the correct size were then expanded, followed by plasmid purification (Qiagen) and Sanger sequencing (GATC Biotech). For Ab expression, 15 µg of each HC and LC vector DNA was transfected into FreeStyle 293-F cells and cultured in 30 ml of FreeStyle 293 expression medium (Life Technologies) at a cell density of 106 cells/ml and ≥90% viability, using 30 µl of 293fectin (Life Technologies) according to the manufacturer’s protocol. After 4–5 d, cell culture supernatants were tested for total Ab production and for binding to HIV-1 Env ligands by ELISA. Cultures containing functional Env-specific Abs were then harvested and purified 7 d after transfection using Protein G Sepharose columns (GE Healthcare). All purified recombinant mAbs were analyzed by SDS-PAGE under reducing conditions using NuPAGE Novex 4–12% Bis-Tris polyacrylamide gels and NuPAGE reducing agent (Life Technologies) according to the manufacturer’s instructions.
HIV-1 neutralization assays
Neutralization assays were performed using a single-round infectious HIV-1 Env pseudovirus assay with TZM-bl target cells (Li et al., 2005). To determine the Ab concentration and the plasma dilution that resulted in a 50% reduction in relative luciferase units, serial dilutions of the mAbs and the plasma were performed, and the neutralization dose–response curves were fitted by nonlinear regression using a five-parameter Hill slope equation using the R statistical software package. Neutralization capacities of mAbs were reported as the Ab IC50, whereas the results for plasma were reported as the plasma neutralization ID50, which is the reciprocal of the plasma dilution producing 50% virus neutralization.
ELISA for Env binding
The mAbs were tested for binding against 16055 NFL TD CC trimer, gp120, and the gp120 V2b(180–194) deletion mutant with residues replaced with GAG (Martinez-Murillo et al., 2017). MaxiSorp 96-well plates (Nalgene Nunc International) were coated at 2 µg/ml with WT gp120, gp120 V2b, or with an anti-His tag mAb (AD1.1.10; R&D Systems) to capture the His-tagged trimer in PBS at 4°C overnight. After incubation with blocking buffer (PBS containing 2% nonfat milk), the mAbs were added and incubated for 1 h at 37°C. Binding was detected by secondary HRP-conjugated anti-human Fcγ Ab (Jackson ImmunoResearch) at 1:10,000 for 1 h. The signal was developed by addition of TMB substrate (Invitrogen) for 5 min, reactions were terminated with 1 N sulfuric acid, and the OD was read at 450 nm. Between each incubation step, the plates were washed six times with PBS containing 0.05% Tween-20.
GM9_TH8 was biotinylated using EZ-Link NHS-Biotin (Pierce Biotechnology, Thermo Fisher Scientific) per the manufacturer’s protocol. For cross-competition ELISA assays, 16055 NFL TD CC trimers were captured on the ELISA plate by an anti-His tag mAb (R&D Systems) coated at 2 µg/ml in PBS at 4°C overnight. Fivefold serial dilutions of competing mAbs were preincubated with the captured trimer at RT for 30 min before addition of biotinylated GM9_TH8 at a concentration previously determined to give ∼75% of the maximum binding signal (i.e., binding to trimer with no competitor present) for 60 min at RT. The bound biotinylated mAbs were detected using HRP-conjugated streptavidin (Sigma-Aldrich). The signal was detected with tetramethylbenzidine substrate, and the reaction was stopped with 1 N sulfuric acid. Wells were blocked with 5% nonfat milk in PBS-T (0.2% Tween-20) after coating with the anti-His mAb and washed with PBS-T after each incubation step. Samples were diluted in 10% blocking buffer.
The GM9_TH8 lineage mAbs were assessed by a chaotropic ELISA wash assay using sodium thiocyanate (NaSCN). An additional wash incubation step with PBS or 1.5 M NaSCN was included after Ab binding to the His-captured 16055 NFL TD CC trimers. Binding of the mAbs was detected after incubation with the HRP-conjugated secondary Ab as described above.
GM9_TH8 lineage IgG and Fab Expression
The HCs from the selected clones were paired with the GM9_TH8 LC, expressed by transient transfection into 293F cells, and purified as described above. For Fab expression, the VDJ segments were cloned into the Fab expression vector, which contains stop codons in the HC hinge region (EPKSCDK**). All constructs were sequence verified. Similarly, each Fab HC was paired with the GM9_TH8 LC and transiently transfected into 293F cells as described above. Fabs were purified from culture supernatants using CaptureSelect IgG-CH1 affinity matrix (Life Technologies).
B cell ELISpot assay
To detect IgG-producing cells in blood, BM, spleen, LN, and gut, 96-well Multiscreen-IP filter ELISpot plates (Millipore) were activated using 70% ethanol and washed twice with sterile PBS. The plates were coated with 10 µg/ml anti-human Fcγ/Fcα Ab (Jackson ImmunoResearch) overnight at 4°C. The plates were washed with sterile PBS containing 0.05% Tween-20, and the frozen cells were thawed, plated in complete medium in dilution series, and placed in a 37°C, 5% CO2 humidified incubator overnight. To detect IgG-producing cells after stimulation, the cells were cultured for 72 h in complete medium supplemented with 5 µg/ml CpG B oligodeoxynucleotides (InvivoGen), 10 µg/ml pokeweed mitogen (Sigma-Aldrich), and 1:1,000 Staphylococcus aureus Cowan (Sigma-Aldrich) before plating. The next day, the plates were washed and incubated for 1.5 h with the biotinylated probes: goat anti-human Fcγ (0.025 µg/ml; Jackson ImmunoResearch), 16055 Env trimers (3 µg/ml), or ovalbumin (3 µg/ml) in the respective wells to detect total and antigen-specific Abs. Streptavidin-alkaline phosphatase (Mabtech AB) was added at 1:1,000 dilution in PBS and incubated for 45 min. The plates were washed to develop and visualize spots by the addition of nitro-blue tetrazolium and 5-bromo-4-chloro-3′-indolyphosphate substrate (Mabtech AB) for 5 min in the dark. The reaction was terminated by washing the plates with water. The plates were dried for counting the spots using an Immunospot analyzer (Cellular Technology), and the number of Ab secreting cells per 106 cells was calculated as previously described (Martinez et al., 2015; Sundling et al., 2010).
Rep-seq library production and sequencing
We processed all samples as single mononuclear cell suspensions and stored these at −80°C for subsequent analysis. Total RNA extraction from samples from different immune compartments (blood, BM, LN, spleen, and ileum) was performed using the RNeasy kit (Qiagen). First-strand cDNA synthesis was performed at 50°C for 45 min using 0.5 µg of total RNA, isotype-specific primers containing a 21-bp unique molecular identifier and a sequence identical to the Read 2 sequencing primer (Table S1), and Superscript III reverse transcription (Thermo Fisher Scientific). The resulting cDNA was purified using MinElute reaction cleanup kit (Qiagen). Next, 5 µl of cDNA was amplified using the 5′ multiplex primer mix (each of the forward primers targeted the leader sequence of one or more V genes, and each primer contained a tail identical to the Read 1 sequencing primer) and 3′ SR_Universal primer (Table S1). The resultant bands of ∼550 bp were gel purified using the Qiagen gel purification kit, for subsequent Rep-seq library production. 2–10 ng of each gel-purified IgM or IgG product was used in an 8–10-cycle PCR step to introduce Illumina adapters as well as sample-multiplexing indexes. Products from all samples were purified using Ampure XP beads according to the TrueSeq protocol, and the resulting libraries were validated and quantified. The various individually indexed libraries were sequenced on the Illumina MiSeq using the Illumina Version 3, (2 × 300 bp) sequencing kit. 15% PhiX174 DNA was included as a control and as a means of generating diversity within the MiSeq flow cell.
Analysis of sequencing data
The computational analysis of the libraries, including preprocessing, VDJ annotation, and generation of an individualized germline V gene database, was performed with IgDiscover software (http://docs.igdiscover.se; Corcoran et al., 2016). To trace the B cell lineages in multiple immune compartments, two new IgDiscover modules were developed. The Clonotypes module computes the set of clonotypes (lineages) that occur in a dataset. It is after the preprocessing steps of the IgDiscover pipeline, which include paired-end read merging, removal of duplicate sequences, barcode removal, and gene assignment (see IgDiscover documentation). The result is a table in which each row represents a unique merged read pair. Annotations include, among others, the V, (D), and J gene assignments and the CDR3 sequence. The Clonotypes module partitions the sequences into sets of sequences estimated to belong to the same clonotype. Two sequences are considered to belong to the same clonotype if they have (1) identical V and J gene assignments, (2) identical HCDR3 lengths, and (3) similar HCDR3 regions at either the nucleotide or amino acid level, by either percentage identity or absolute number of mismatches, but with at least the first or last n nucleotide/amino acid of the HCDR3s being exactly identical. This restriction on the CDR3 “junctions” reduces inclusion of unrelated contaminating lineages. The sequence similarity criterion is applied transitively; that is, two sequences that would otherwise be in different groups are considered to be of the same clonotype if there is a third sequence that is similar enough to both of them. For each group found by clustering the sequences in this way, a representative is chosen. The second module, Clonoquery, works with two input tables of the same type as above. The reference table contains annotations for the sequences in the dataset of interest, and the smaller query table contains sequences whose clonotypes are to be found in the reference table. Queries are looked up individually using the same criterion as described above. Unlike the first module, similarity among reference sequences is ignored; that is, only reference rows directly found to be of the same clonotype are returned.
Ab lineage sequences were first deduplicated, storing their frequencies, using preprocessing scripts implemented in the Julia language for scientific computing. To infer the position of the unmutated common ancestor sequence, the V and J were concatenated, with the junction and D gene masked with ambiguous (N) nucleotide, which are treated as missing data during phylogenetic inference. A maximum-likelihood phylogenetic tree was inferred with FastTree2 (Price et al., 2010) and visualized with FigTree (http://tree.bio.ed.ac.uk/software/figtree/).
GM9_TH8seq732127 Fab production, crystallization, and structure determination
The GM9_TH8seq732127 IgG protein was expressed in HEK293F cells after transfection of HC and LC plasmids. The protein was purified from the supernatant using protein A resin. The purified IgG was cleaved with LysC, and the GM9_TH8seq732127 Fab was further purified by size-exclusion chromatography HiLoad 16/600 Superdex 200 (GE) column using buffer containing 5 mM Hepes, pH 7.5, and 150 mM NaCl. The purified GM9_TH8seq732127 Fab was screened for crystallization using an NT8 (Formulatrix)-dispensing robot. Screening was done with Wizard Precipitant Synergy block no. 2 (Rigaku), Proplex screen HT (Molecular Dimensions), and Crystal Screen HT (Hampton Research) using the sitting drop vapor diffusion method in 96-well crystallization plates (MRC plates, Hampton Research). The protein drop was mixed with reservoir solution in a 1:1 ratio, equilibrated against 40 µl of reservoir solution, and incubated at 20°C. High-quality morphology crystals of the GM9_TH8seq732127 Fab were obtained in screening condition of Crystal Screen HT (Hampton Research) with 4.0 M sodium formate. The screening condition was expanded using the hanging drop vapor diffusion method to obtain diffractable quality crystals. The crystals were cryoprotected in 15% ethylene glycol and 4 M sodium formate. The cryoprotected crystals were flash-cooled in liquid nitrogen, and data collection was done at ALS 5.0.2. The crystals diffracted to 2.26-Å resolution. The collected data were processed using HKL2000 (Otwinowski and Minor, 1997) and imported using Aimless in CCP4 (Collaborative Computational Project, Number 4, 1994). The structures were solved by molecular replacement using Phaser in CCP4 and unpublished solved Fab structure as a search model. The structure building and refinement were performed in Phenix (Adams et al., 2010) and COOT (Emsley and Cowtan, 2004). The refinement statistics are summarized in Fig. S5 B.
Rep-seq data from this study were deposited in the European Nucleotide Archive under accession number PRJEB29348. Accession numbers for the individual Rep-seq library are given in Fig. S2. V and J alleles identified in animal D20 by IgDiscover are from GenBank, available under the accession numbers MK111753–MK111853 (V alleles) and MK111854–MK111861 (J alleles). The Env-specific single-cell HC and LC sequences are available from GenBank under accession numbers MK104588–MK105259 and MK105260–MK105280, respectively. Structure factors and coordinates have been deposited in the Protein Data Bank under PDB ID 6U9U. The authors declare that all data supporting the findings of this study are available within the article and its supplemental files or can be obtained from the corresponding author upon request.
The Clonotypes and Clonoquery modules are the part of the IgDiscover software, which is available as open source software under the Massachusetts Institute of Technology license from http://docs.igdiscover.se.
Online supplemental material
Fig. S1 shows Env-specific memory B cell sorting FACS gating and epitope-specificity of the isolated mAbs. Fig. S2 shows Ab repertoire sequencing metadata. Fig. S3 shows B cell ELISpot analysis of the different immune compartments 2 wk after the sixth immunization and intraclonal SHM analysis of Ab lineages traced in the BM compartment. Fig. S4 shows alignments of variants displaying increasing SHM for a subset of Ab lineages found in blood. Fig. S5 shows autologous tier 2 neutralizing Ab titers and data collection and refinement statistics for the structural analysis of GM9_TH8seq732127. Table S1 lists primers used for Next Generation Sequencing library production.
We thank Dr. Mats Spångberg and Dr. Bengt Eriksson and all personnel at the Astrid Fagraeus Laboratory for expert assistance with rhesus macaques and Novavax, AB, Uppsala, Sweden, for generously making the Matrix-M adjuvant available to this study. We also thank Fondation Dormeur, Vaduz, and the J.B. Pendleton Charitable Trust for generous support of equipment.
X-ray diffraction data were collected at the Berkeley Center for Structural Biology beamline 5.0.2, which is supported in part by the National Institute of General Medical Sciences. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the US Department of Energy under contract number DE-AC02-05CH11231. This work was funded by an National Institutes of Health P01 HIV Research and Design grant (AI104722) to G.B. Karlsson Hedestam, R.T. Wyatt, M. Pancera, and L. Shapiro; a Swedish Research Council Distinguished Professor grant (2017-00968) to G.B. Karlsson Hedestam; the intramural research program of the Vaccine Research Center, National Institute of Allergy and Infectious Diseases, National Institutes of Health; and a grant from Karolinska Institutet for doctoral education to G.E. Phad.
The authors declare no competing financial interests.
Author contributions: G.E. Phad, P. Pushparaj, R.T. Wyatt, and G.B. Karlsson Hedestam conceived the study. K. Tran and V. Dubrovskaya produced the 16055 NFL TD CC trimers and analyzed mAbs and Fabs. G.E. Phad, P. Pushparaj, M. Àdori, N. Vázquez Bernat, P. Martinez-Murillo, C. Sundling, C. Sorini, V. Dubrovskaya, and E.J. Villablanca processed samples. G.E. Phad, P. Pushparaj, M. Àdori, V. Dubrovskaya, and K. Bhullar isolated and analyzed mAbs. S. O’Dell and J.R. Mascola evaluated Ab neutralizing activity. P. Pushparaj, P. Martinez-Murillo, and M. Àdori performed ELISpot analysis. G.E. Phad, S. Narang, and M. Corcoran performed Rep-seq and analyzed the data. B. Murrell performed phylogenetic analysis. S. Singh, M. Pancera, G. Dionne, and L. Shapiro performed structural analysis. M. Martin wrote software. All authors contributed to the interpretation of data. G.E. Phad and G.B. Karlsson Hedestam wrote the paper with input from the co-authors.