1. Cell Biology
  2. Microbiology and Infectious Disease
Download icon

Distinct clonal evolution of B-cells in HIV controllers with neutralizing antibody breadth

  1. Deniz Cizmeci
  2. Giuseppe Lofano
  3. Evan Rossignol
  4. Anne-Sophie Dugast
  5. Dongkyoon Kim
  6. Guy Cavet
  7. Ngan Nguyen
  8. Yann Chong Tan
  9. Michael S Seaman
  10. Galit Alter
  11. Boris Julg  Is a corresponding author
  1. Ragon Institute of MGH, MIT and Harvard, United States
  2. Atreca Inc, United States
  3. Center for Virology and Vaccine Research, Beth Israel Deaconess Medical Center, United States
Research Article
  • Cited 0
  • Views 385
  • Annotations
Cite this article as: eLife 2021;10:e62648 doi: 10.7554/eLife.62648
Voice your concerns about research culture and research communication: Have your say in our 7th annual survey.

Abstract

A minor subset of individuals infected with HIV-1 develop antibody neutralization breadth during the natural course of the infection, often linked to chronic, high-level viremia. Despite significant efforts, vaccination strategies have been unable to induce similar neutralization breadth and the mechanisms underlying neutralizing antibody induction remain largely elusive. Broadly neutralizing antibody responses can also be found in individuals who control HIV to low and even undetectable plasma levels in the absence of antiretroviral therapy, suggesting that high antigen exposure is not a strict requirement for neutralization breadth. We therefore performed an analysis of paired heavy and light chain B-cell receptor (BCR) repertoires in 12,591 HIV-1 envelope-specific single memory B-cells to determine alterations in the BCR immunoglobulin gene repertoire and B-cell clonal expansions that associate with neutralizing antibody breadth in 22 HIV controllers. We found that the frequency of genomic mutations in IGHV and IGLV was directly correlated with serum neutralization breadth. The repertoire of the most mutated antibodies was dominated by a small number of large clones with evolutionary signatures suggesting that these clones had reached peak affinity maturation. These data demonstrate that even in the setting of low plasma HIV antigenemia, similar to what a vaccine can potentially achieve, BCR selection for extended somatic hypermutation and clonal evolution can occur in some individuals suggesting that host-specific factors might be involved that could be targeted with future vaccine strategies.

Introduction

The induction of cross-reactive broadly neutralizing antibodies (bNAbs) that can effectively neutralize a majority of circulating HIV-1 strains is a major goal of current HIV vaccine development, but no vaccine candidate to date has achieved this goal (Burton and Hangartner, 2016). During natural HIV-1 infection, development of neutralizing antibody responses is seen in about 10–30% of HIV-1-infected individuals but often requires years of infection (Pilgrim et al., 1997). Specifically, high viremia (Mikell et al., 2011; CAPRISA002 Study Team et al., 2011) with rapid viral diversification (NISC Comparative Sequencing Program et al., 2014; NISC Comparative Sequencing Program et al., 2013) along with infection-associated immune activation (Sather et al., 2009) have been associated with the evolution of antibody neutralization breadth. One specific feature that characterizes many bNAbs is the high mutation frequency in their variable domains, indicative of many cycles of somatic hypermutation (SHM) in immunoglobulin (Ig) variable-diversity-joining [V(D)J] gene segments, ultimately resulting in superior B-cell receptor-antigen affinity maturation (Victora and Nussenzweig, 2012). The requirement for repeated cycling through the germinal centers (GCs) to accumulate affinity and specificity enhancing mutations is a lengthy process and is reflected by the time required for bNAbs to be elicited in the presence of a constantly evolving virus. However, bNAbs with high potency and exquisite breadth against cross-clade viral isolates have also been isolated from HIV-infected individuals who spontaneously control HIV to low plasma levels in the absence of antiretroviral therapy (‘HIV controllers’) (Walker et al., 2009; Simek et al., 2009; González et al., 2018; Schoofs et al., 2019; Freund et al., 2017), thus suggesting that neutralizing antibody breadth may be achievable in the absence of high levels of viremia and rapidly evolving viral diversity, a scenario that more likely can be accomplished with vaccination strategies. In these HIV controller individuals, evolution of neutralization breadth was linked to low but persistent HIV viral antigenemia in the setting of a unique inflammatory profile, while antigenic diversity was not a significant contributor (Dugast et al., 2017). Thus, neutralizing antibody breadth may be achievable in the absence of high levels of viremia and rapidly evolving viral diversity, therefore suggesting a scenario that more likely can be accomplished with vaccination strategies. In this study, we selected 22 controllers who exhibited different levels of neutralization breadth (ranging from 0/11 to 11/11 of tier 2/3 viruses neutralized) (Freund et al., 2017; Dugast et al., 2017) to elucidate features of the HIV-specific antibody repertoire that are associated with neutralization breadth. We performed a natively paired heavy and light chain B-cell receptor repertoire analysis of HIV-1 envelope (Env)-specific memory B-cells (MBCs) using high-throughput single-cell sequencing to determine alterations in the BCR Ig gene repertoire and B-cell clonal expansions that track with neutralization breadth.

Results

Study cohort characteristics

Twenty-two HIV-infected individuals, 4 female and 18 male, were included in this study. 11 participants were African American and 11 were of European descent. All individuals were US-based and therefore likely infected with clade B viruses. Plasma samples were tested for neutralizing activity using a standard reference panel of tier 2 and 3 clade B Env pseudoviruses (Li et al., 2005). Six individuals neutralized >90% of viruses at 50% inhibitory dose (ID50) titers above background, and were classified as top-neutralizer (TN), while seven individuals neutralized <10% of the panel viruses and were classified as non-neutralizer (NN). The remaining individuals were considered intermediate neutralizer, exhibiting neutralization of 36–82% of the virus panel (Supplementary file 1). There was no statistically significant difference in age (mean 53 years; min 44, max 63), CD4 count (mean 726 cells/μl [SD 168]), HIV RNA levels (mean 961 copies/ml [SD 1645]), and time-off-ART (mean 17 years; min 4.5, max 30) between the TN and NN groups (Figure 1—figure supplement 1). Single HIV-1 Env-specific CD19+CD20+IgM-IgA-MBCs were FACS-sorted (Figure 1—figure supplement 2) and we obtained repertoires of natively paired, full variable region IGH and IGL by Immune Repertoire Capture analysis; 12,591 total sequence pairs were generated for all study participants, with 5771 sequences from TN and 2707 sequences from NN (Figure 1—figure supplement 2). To exclude potential bias caused by the number of input cells or sequences, we normalized the proportion of each repertoire signature across all sequences in each individual (see 'Materials and methods'). TN MBCs showed an Ig subclass distribution similar to serum IgG antibodies, with average levels of IgG1 (78.6%) > IgG2 (12.5%) > IgG3 (8.8%) > IgG4 (0.1%) while Ig subclasses from NNs differed in hierarchy with average levels of IgG1 (82.5%) > IgG3 (11.1%) > IgG2 (6.2%) > IgG4 (0.2%) (Figure 1A).

Figure 1 with 3 supplements see all
Non-neutralizer (NN) and top-neutralizer (TN) memory B-cells (MBCs) B-cell receptor (BCR) repertoires.

(A) IgG subclass distribution in NNs and TNs. (B) The proportion of IGHV-IGLV gene combination in NN and TN individuals' usage was averaged and plotted in heatmaps representing the intensities of heavy and light chain paired V gene usage. (C) CDRH3 and CDRL3 length distribution (in aa) in NN and TN.

IGH-VDJ and IGL-VJ gene segment usage

The MBC repertoires were found to use 40 IGHV genes representing all seven human VH gene families. Of the seven IGHV gene families, IGHV1 was most commonly used in both NN and TN repertoires, consistent with prior reports that anti-HIV antibodies frequently use IGHV1 genes (Breden et al., 2011; Gorny et al., 2009Figure 1—figure supplement 3A). The top two most frequently used IGHV genes were IGHV1-69 and IGHV4-34 in both the TN and NN repertoires, followed by IGHV4-4 in the TN and IGHV1-24 and IGHV1-2 in the NN repertoire (Figure 1—figure supplement 3B). For IGHD and IGHJ genes, no statistical significant differences were observed. The most frequently used heavy chain VDJ genes were IGHV1-69, IGHD3-22, IGHJ4 in NN and IGHV1-69, IGHD4-17, IGHJ4 in TN. Among IGLV and IGLJ genes, IGLKV3 and IGLKV3-20 dominated in both TN and NN repertoires (Figure 1—figure supplement 3B). This is consistent with previously reported data showing the dominance of IGLKV3-20 (Jackson et al., 2012).

We next compared the proportions of specific IGH/IGL pairs between TNs and NNs. IGHV1-69/IGKV3-20 combination followed by IGHV1-69/IGKV1-5 dominated in NN while IGHV4-4/IGLV6-57 followed by V1-69-2/LV1-51 were most frequently found in TNs (Figure 1B). Specifically the most frequent IGHV-IGHJ-IGLV-IGLJ combination in NN was IGHV1-69/IGHJ6/IGKV3-20/IGKJ5 and IGHV4-4/IGHJ6/IGLV6-57/IGLJ2 in TNs (Figure 1—figure supplement 3C). Overall, the frequencies of the top 3 IGHV/HJ-IGLV/LJ pairs differed significantly between NNs and TNs (Mann Whitney U-test, p<0.05). We next were interested to determine if we could find sequences of known bNAbs as listed in the CATNAP database (Yoon et al., 2015) and as summarized by Sok and Burton, 2018, using the same IGHV and IGLV gene segment usage and equal CDRH3 length as minimum criteria. Overall, very few sequences in our data matched known bNAb IGHV and IGLV gene pairs and of those none showed high levels of CDRH3 sequence similarity. No difference in these frequencies between TNs and NNs was observed. Only the V3 glycan antibody BG18 that had been identified previously in TN 622800 (Freund et al., 2017) was again detected in the BCR repertoire of this individual (Supplementary file 3). Taken together, these results suggest that V(D)J gene usage was similar overall between TNs and NNs, but the frequency and combination of IGH/IGL pairs differed between groups.

Characteristics of CDRs

Complementarity determining regions (CDRs) play critical roles in the binding of antibodies to antigens and unusually long CDRH3s have been described for some bNAbs (Sok and Burton, 2018). The CDRH3/L3 length in TNs was between 3 and 37 aa and 5 and 17 aa, respectively, and 4 to 31 aa and 5 to 14 aa in the NNs, respectively. CDRH3 length distribution in TN sequences was significantly different to CDRH3 length distribution in NN sequences (p<2.2e-16 Kolmogorov-Smirnov test; Figure 1C). Furthermore, significant differences in length were observed for CDRH2, CDRL1 (data not shown), and CDRL3 (p<2.2e-16 Kolmogorov-Smirnov test; Figure 1C). There were no significant differences for CDRH1 and CDRL2. The mean CDRH3 lengths observed here are more consistent with what has been described in the CD4 binding site-specific bNAbs like VRC01 and 3BNC117, which have CDRH3 lengths of 14 and 12, respectively, while V2 and V3 glycan-specific bNAbs like PGDM1400 and PGT121 have substantially longer CDRH3 length with 34 and 26 aa, respectively (Sok and Burton, 2018). Nevertheless, CDRH3 length by itself is not a determining feature for neutralization potency and breadth but rather CDRH3-specific sequence characteristics that contribute to the superior binding activity of certain bNAbs. Other than for BG18, however, no CDRH3 sequences, that matched known bNAbs exactly, were indentified in our dataset.

Rates of SHM in IGHV and IGLV

When we calculated the number of SHMs in IGHV, we noticed that the TN repertoires had significantly higher frequencies of nucleotide mutations compared to NN (mean ± SD: 54 ± 17 mutations for TN and 29 ± 8 mutations for NN, unpaired t-test, p<0.05) with more than 65% of sequences showing mutation frequencies greater than 15% (Figure 2A). When comparing the frequency of mutations in IGHV (TNs, NNs, and intermediate neutralizers) with the degree of serum neutralization breadth across the entire cohort, a direct correlation was observed (Spearman’s rho = 0.60, p<0.01) (Figure 2B and C). Similarly, SHM in IGLV (Figure 2D) was more frequent in TN than NN (mean ± SD: 35 ± 18 mutations for TN and 14 ± 5 mutations for NN, unpaired t-test, p<0.05) with more than 50% of sequences showing mutation rates greater than 10%. Here as well, the frequencies of mutations in IGLV correlated with neutralization breadth (Spearman’s rho = 0.63, p<0.01; Figure 2E and F). We were next interested to determine whether the IGH and IGL V-J gene usage would differ between the more (with higher than 15% mutations in IGH, 10% mutations in IGL) and less (≤15% mutations in IGH, 10% mutations in IGL) mutated sequences. While including all sequences, independent of the mutation rates, there was no apparent difference in IGHV-J (Figure 2G) or the IGLV-J combinations between NNs and TNs (Figure 2H). For the sequences with higher than 15% mutations in IGHV, however, which occurred nearly exclusively in the TNs, the pattern was distinct, demonstrating a selection for fewer dominant V-J combinations (Figure 2G). In particular HJ4/HV1-69-2, HJ6/HV4-4, HJ6/HV3-30.33, and HJ6/HV4-34 were significantly enriched in TN sequences with high mutation rates (p<0.01, one-sided Fisher’s exact test followed by correction for multiple comparison using Benjamini-Hochberg method). The enrichment of HV4-34 usage (p<0.001, one-sided Fisher’s exact test followed by correction for multiple comparison using Benjamini-Hochberg method), a feature of autoimmunity (Wardemann et al., 2003), in the TN sequences with high mutation rates, is consistent with a recent report that suggested that HIV infection induces a permissive state in which potentially autoreactive clones that would otherwise be eliminated can persist (Roskin et al., 2020). A similar contraction to fewer dominant V-J combinations was also observed for sequences with higher than 10% mutations in IGLV (Figure 2H). The frequencies of the paired IGHV and IGLV genes (Figure 2I) demonstrated a dominant pairing of the HV1-69-2 and HV3-30.33 with LV1-51, and HV4-4 with LV6-57 in TNs with >15% heavy chain mutations. Assuming that the somatic IGHV mutation rates in B-cells correlate with the total exposure time in the GC environment, and in light of the fact that the TN and NN did not differ in age, time-off-ART, and viral loads, our data suggests that TNs were more permissive for extensive SHM of B-cell clones.

Mutation characteristics of IGHV and IGLV genes and associations with neutralization breadth.

(A) The overall mutation frequencies in IGHV in top-neutralizers (TNs) versus non-neutralizers (NNs) are significantly different with mean number of mutations of 54 in TNs and 29 in NNs (unpaired t-test, p<0.05). (B) The number of mutations in IGHV for each of the 22 study participants is plotted and the serum neutralization breadth is color-coded ranging from blue (0% neutralization) to red (100% neutralization). (C) The mean frequency of mutations in IGHV correlates with the serum neutralization breadth in % (Spearman’s rho = 0.60, p<0.01). Similarly, for IGLV, the mean number of mutations differ (35 in TNs and 14 in NNs) (D) and correlates with the serum neutralization breadth in % (r = 0.63, p<0.01) (E and F). Circos plots summarize the combinations of V and J segments used in the rearranged IGH genes (G) and IGL genes (H) and V segment combinations of IGH and IGL (I) expressed by memory B-cells (MBCs). Top circos graphs demonstrate combinations expressed by NN MBCs and second row graphs demonstrate TN MBCs independent of the number of detected mutations. Third and fourth row circos graphs demonstrate gene combinations expressed by TN MBCs with low and high mutation rates. For each (G)/(H) plot, the bottom half depicts J genes and the top half depicts V genes. For (I) top half depicts IGHV and bottom half IGLV. To exclude potential bias caused by the number of input cells, the number of sequences in each repertoire signature was weighted by the total number of sequences and total number of unique clones in each individual (see 'Materials and methods'). The arc length of each segment denotes the normalized frequency at which each gene segment was identified. Rearrangement of a J gene with a V gene segment in a clonal immunoglobulin (Ig) sequence is represented by a ribbon (ribbons carry the color of the HV or LV family of the gene participating in the pairing). The width of the ribbons corresponds to the weighted frequency at which each particular HV-HJ or LV-LJ rearrangement was used in the respective MBC repertoire.

Clonal selection

To further assess the clonal nature of the BCR repertoire in TNs, we clustered sequences into clonal groups. Each unique clone is a cluster of sequences that meet the following criteria: (1) derived from the same individual; (2) share the same V and J gene segment annotations in heavy chain and light chain; (3) have equal CDRH3 length; and (4) junction regions share similarity by a defined sequence distance cutoff (see 'Materials and methods'). This clonal assignment identified 770 clones (mean number of sequences 7.4) and showed that larger sized clusters (dominant clones) consist of highly mutated sequences (Figure 3A). To quantify the greater clonality of the highly mutated sequences, we used an unevenness measure, the Gini index. We selected individuals (subject IDs: 330183, 280008, 622800) for whom a total of more than 600 sequences were available. Sequences were divided into two sets, highly mutated (>15% mutations) and less mutated (≤15% mutations) sequences. Each set of sequences was clustered into clones and the Gini index was calculated based on the clone size distributions. For all three subjects, the Gini index was greater in the highly mutated set (paired t-test, p<0.04), thus supporting the observation that the BCR repertoire becomes more narrow with increasing mutation rates, suggesting the dominance of certain clones (Figure 3B).

Clonality analysis.

A total number of 770 clones were obtained for top-neutralizers (TNs). (A) Each clone is represented as a circle. The sizes of the circles are scaled to the number of sequences in each clone (mean number of sequences 7.4); the colors represent mean number of mutations per clone. Dominant clones consist of highly mutated sequences. (B) Gini index (0 represents perfect equality and one perfect inequality) was calculated as an estimate of clonality for TNs for which a total of more than 600 sequences were obtained (subject IDs: 330183, 280008, 622800). Per each subject, sequences were divided into two sets as high mutated sequences (>60 mutations) and low mutated sequences (≤60 mutations). Each set of sequences were clustered into clones and Gini index was calculated. Gini Index is higher (paired t-test, p=0.04) in highly mutated sequence sets supporting the presence of fewer dominant clones. (C) Maximum likelihood point estimates of ω with 95% confidence intervals. (D) Estimates of mean tree length (total substitutions per codon within a lineage, averaged across all lineages within a repertoire) were compared between TN and non-neutralizer (NN) using Wilcoxon rank sum test (left graph). The correlation between ω and mean tree length calculated using Spearman’s correlation. Largest lineage trees in NNs (E) and TNs (F).

Next, we estimated the evolutionary process of mutation and selection in B-cells using phylogenetic models. Selective dynamics have been estimated using ω: dN/dS, the ratio of nonsynonymous (amino acid replacement) and synonymous (silent) mutation rates (Nielsen and Yang, 1998; Hoehn et al., 2019). Low ωCDR values indicate fewer amino acid changes in CDRs than expected suggesting negative BCR selection to remove affinity-decreasing variants, while positive selection to introduce new affinity-increasing amino acid variants is associated with higher ωCDR values. When comparing TNs to NNs we found an overall trend toward fewer amino acid changes in the CDRs of TNs (maximum likelihood estimate of ωCDR in Figure 3C) indicating stronger purifying selection. Furthermore, the average lineage tree lengths (the total expected substitutions per codon site within an individual lineage phylogeny) was significantly higher in TNs compared to NNs (p<0.05, Wilcoxon rank sum test) and correlated negatively with ω (Figure 3D). The largest lineage trees constructed in TNs and NNs are visualized in Figure 3E and Figure 3F, respectively. These results are therefore consistent with prior observations that B-cell lineages shift toward negative selection over time as a general feature of affinity maturation (Hoehn et al., 2019) and suggest that the dominant clones in the TNs have reached a level, where the continuous accumulation of additional random amino acid changes would rather be detrimental and decrease affinity (Clarke et al., 1985).

Discussion

Overall, when comparing frequencies of either IGH or IGL genes between those individuals who developed antibody neutralization breadth compared to those who did not, the repertoire analysis did not reveal substantial differences. This is consistent with the findings from Scheepers et al. who analyzed the IGHV repertoires of bulk peripheral blood B-cells for 28 HIV-1 clade C infected South African women with different ranges of serum neutralization breadth (Scheepers et al., 2015) and observed a wide range in the number of IGHV alleles in each individual but did not find significant differences in germline IGHV repertoires between individuals with or without bNAbs. When we however compared the repertoire of paired IGH/IGL sequences between both groups, clear differences in the hierarchy of IGHV/IGHJ/IGLV/IGLJ combinations were observed. Despite these distinguishing pattern, we did not observe an enrichment of known bNAb sequences in the TNs. In fact, we found very few sequences in our cohort that matched IGHV-IGLV gene combinations and/or CDRH3 -sequences of published Nabs (Sok and Burton, 2018) with BG18 being the exception. Indeed, additional bNAbs, such as the CD4 binding site antibody 3BNC117, have been isolated from HIV controllers (Scheid et al., 2011). The general absence of these or similar bNAb sequences in our data therefore suggests that the neutralization breadth observed in the TNs was either from single bNAbs with non-classical epitope specificities or due to multiple neutralizing antibodies targeting multiple different epitopes. This distinction necessitates the generation of antibodies from a large number of BCRs in the TNs for characterization of their neutralizing activities and such studies are planned in the future.

In addition to the composition of the BCR repertoire, that separated TNs from NNs, both groups were clearly distinguishable by the significantly higher rates of mutations in the VH and VL genes in the individuals with neutralizing antibody breadth. Our data in HIV-1 specific B-cells confirms findings recently reported by Roskin et al. in a bulk B-cell repertoire analysis, demonstrating that individuals with neutralizing antibody breadth have hundreds to thousands of antibody lineages with long CDRH3s and very high SHM frequencies (Roskin et al., 2020). Moreover, the mutation frequency in our data correlated directly with the neutralization breadth consistent with the notion that high levels of SHM are generally necessary for neutralization breadth and potency (Klein et al., 2013).

Initial interclonal and then intraclonal competition between B-cells during affinity maturation is well described (Kuraoka et al., 2016; Tas et al., 2016), resulting in diverse repertoires. Indeed, we found that with increasing numbers of mutations in HV, fewer clones with less diverse HV-HJ combinations dominated, suggesting a process that allows these clones to mature and accumulate mutations. Since GCs select for affinity and not for neutralizing breadth and since evolving viruses, such as HIV-1, will continuously elicit novel immune responses from naive B-cells, strong effective selection normally occurs, so that antibodies specific to the contemporaneous virus have a sufficient advantage to outcompete any less-specific potential broad neutralizers before they take hold. Indeed, by focusing on the kinship relations among Ig gene segment recombinations within individual clones, we found evidence of natural selection in the NNs, with younger and less diverse clones continuously evolving, constantly exploring novel evolutionary space. Conversely, in TNs, this continuous evolution does not seem to occur, but rather specific evolutionary paths are pursued. One potential mechanism behind this observation is that neutralizers may generate massive bursts of clones that compete aggressively and prevent the evolution of lower affinity variants. Alternatively, the effects of immunodominance where B-cell specific to an epitopic site dominate B-cells that target other sites, as has been described in influenza infection (Angeletti et al., 2017), might not occur to the same degree in TNs as it might do in NNs. Indeed, a generalized deficiency of normal selection against B-cells expressing IgG with long CDRH3 regions and high SHM frequencies has been associated with the development of HIV neutralizing breadth (Roskin et al., 2020).

While the selection process might be suppressed in the TNs early on, to allow extended SHM with ongoing clonal evolution and increasing affinity, this process seems to revert during the later stages of antibody evolution when the benefit of additional amino acid changes decreases, and purifying selection becomes dominant and removes most non-synonymous mutations. Indeed, Yaari et al. demonstrated that early mutations in the trunk of B-cell clonal lineages from healthy subjects were consistent with positive selection pressure than more recent mutations in the canopy (Yaari et al., 2015) and a negative relationship between both signs of increased negative selection (low ωCDR values) and mean repertoire tree length have been reported across subjects of different age and sex (Hoehn et al., 2019). Interestingly, Wu and Sheng et al. observed that the evolutionary rates of HIV-1 bNAb antibody lineages derived from different donors consistently decreased over time (NISC Comparative Sequencing Program et al., 2015; Sheng et al., 2016), also suggesting that SHM occurs at a faster rate during the early phase of lineage development while later on, a decrease in mutation rates may be a mechanism to protect immune memory.

Several potential mechanisms for differences in BCR evolution and selection can be envisioned in TNs versus NNs. B-cells with the highest affinity receptors tend to acquire the most antigen from the follicular dendritic cell network and present the highest density of cognate peptides to CD4+ T-follicular helper (Tfh) cells, which respond with survival signals to the B-cell. While some studies have suggested that the quality and quantity of GC Env-specific Tfh cells are associated with the expansion of Env-specific B-cells and broader antibody neutralization activity (Yamamoto et al., 2015; Moody et al., 2016), others have failed to find a correlation (Roskin et al., 2020). Interestingly, Cirelli et al. recently demonstrated that slow delivery of antigen resulted in enhanced autologous tier 2 nAb development in non-human primates by inducing higher frequencies of total and Env-specific GC-Tfh cells accompanied by larger and more diverse Env-specific B-cell lineages. In this model, evolving B-cells with BCRs able to bind to a large array of Env variants, are likely able to capture more antigen, present more effectively to Tfh cells, and thus gain higher survival signals. Conversely, B-cells able to bind tightly, but to only a narrow range of Env variants, will only capture a fraction of arrayed antigens and receive less help. Thus, high affinity, narrow specificities may gain less support compared to lower -affinity, broad -binders (De Boer and Perelson, 2017). Thus, while antigenemia is lower in controllers, the low-level titration of antigen into the system may create a higher bar of selection for evolving B-cell clones. Along with this high degree of competition, B-cell lineages in HIV controllers able to bind many variants may capture larger amounts of antigen and therefore gain enhanced signals from Tfh cells (Fukazawa et al., 2015). Additionally, HIV controllers with low-level viremia also exhibit unique inflammatory profiles, and sustained GC activity, collectively also potentially providing an optimal environment for B- cell development.

In summary, development of B-cell lineages with superior neutralization breadth in HIV controllers is linked to selected clonal evolution, therefore allowing B-cells to accumulate mutations and to diversify without the constant risk of clonal interference. Immunization strategies that facilitate this process by providing continuous, diverse, and subtherapeutically dosed antigen, in the setting of target inflammatory signals, could potentially accelerate the development of broad nAb responses in vaccinees.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional information
Chemical compound, drugPropidium iodideLife TechnologiesCat# P1304MP
AntibodyAnti-CD3 (FITC, clone UCHT1)BiolegendCat# 300406, RRID:AB_31406010 μg/ml
AntibodyAnti-CD14 (FITC, clone HCD14)BiolegendCat# 325604, RRID:AB_830677Dilution (1:100)
AntibodyAnti-IgM (FITC, clone MHM-88)BiolegendCat# 314506, RRID:AB_493009Dilution (1:100)
AntibodyAnti-CD20 (PE-cy7, clone 2H7)BD BiosciencesCat# 560735, RRID:AB_1727450Dilution (1:100)
AntibodyAnti-CD19 (BV421, clone HIB19)BiolegendCat# 302233, RRID:AB_10897802Dilution (1:100)
AntibodyAnti-IgA (FITC, clone IS11-8E10)Miltenyi BiotecCat# 130-093-071, RRID:AB_1036156Dilution (1:100)
AntibodyhCD40L His tagBPS BiosciencesCat# 71191Dilution (1:100)
Chemical compound, drugStreptavidin conjugated to PE or AlexaFluor 647Life Technologies
Recombinant proteinClade B JR-CSF gp120, JR-FL gp140, 92BR020 gp120, clade A BG505 SOSIP, clade C IAVI C22 gp120Duke Human Vaccine Institute, protein production facility
Cell line
(Homo sapiens)
TZM-bl cellsNIH AIDS
Reagent Program
Cat# 8129–442, RRID:CVCL_B478
Recombinant DNA reagentAC10.0.29, RHPA4259.7, THRO4156.18, REJO4541.67, WITO4160.33, TRO.11, SC422661.8, QH0692.42, CAAN5342.A2, PVO.4 TRJO4551.58Seaman labPlasmids
Commercial assay or kitAmpure XP beadsBeckman CoulterCat# A63881
Commercial assay or kitQuant-iT PicoGreen dsDNA Assay KitInvitrogenCat# P7589
Software, algorithmFlowJohttps://www.flowjo.com/solutions/flowjoRRID:SCR_008520
Software, algorithmIgBLASThttps://www.ncbi.nlm.nih.gov/igblast/RRID:SCR_002873
Software, algorithm454 GS FLX data analysis softwareRocheRRID:SCR_018028
Software, algorithmR version 4.0.0https://cloud.r-project.orgRRID:SCR_001905
Software, algorithmIgphyml (version 1.1.0)https://igphyml.readthedocs.io/en/latest/
Software, algorithmAlakazam (version 1.0.0)https://alakazam.readthedocs.io/en/1.0.1/news/

Human subjects

Request a detailed protocol

A total of 22 HIV-infected individuals were recruited for this study. All participants underwent leukapheresis and routine blood draws as per study protocol. All subjects signed informed consent, and the study was approved by the MGH/Partners Institutional Review Board (Protocol # 2003P001894).

Neutralization assay

Request a detailed protocol

HIV-1 neutralization breadth was assessed using the Tzm-bl cell-based pseudovirus neutralization assay, as described (Sarzotti-Kelsoe et al., 2014), against a standard panel of Env-pseudoviruses derived from nine clade B tier 2 viruses: AC10.0.29, RHPA4259.7, THRO4156.18, REJO4541.67, WITO4160.33, TRO.11, SC422661.8, QH0692.42, CAAN5342.A2, and two tier 3 viruses: PVO.4 and TRJO4551.58 (Li et al., 2005). Murine leukemia virus (MuLV) was included in all assays as a negative control. Neutralization titers (50% inhibitory dose [ID50]) were defined as the reciprocal of the plasma sample dilution that caused a 50% reduction in relative luminescence units (RLUs) compared to virus control wells after subtraction of background RLUs. Neutralization breadth was determined as the proportion of pseudoviruses with an ID50 score threefold above background titers observed against MuLV negative control virus (3× ID50 of MuLV) (Sarzotti-Kelsoe et al., 2014).

Cell lines

Request a detailed protocol

TZM-bl cells (also called JC53BL-13) were obtained from the NIH AIDS Reagent Program and authenticated by (1) morphology and growth characteristics, (2) cell surface expression of CD4 and co-receptors was assessed by flow cytometry, and (3) susceptibility in vitro to HIV, with readout being firefly Luc activity in infected cell lysates. All cell lines were tested to confirm absence of mycoplasma contamination.

Single-cell flow cytometry sorting

Request a detailed protocol

Isolated cells were stained with fluorochrome-antibody conjugates and reagents to identify antigen-specific MBCs. The panel consisted of propidium iodide (Life Technologies); CD3 (FITC, clone UCHT1), CD14 (FITC, clone HCD14), IgM (FITC, clone MHM-88), CD20 (PE-cy7, clone 2H7), CD19 (BV421, clone HIB19) (All Biolegend); IgA (FITC, clone IS11-8E10) (Miltenyi Biotec). Preformed conjugates for antigen-specific B-cell sorting were made as described (Sok et al., 2014) using streptavidin conjugated to PE or AlexaFluor 647 (Life Technologies). B-cell probes were made using clade B JR-CSF gp120, JR-FL gp140, 92BR020 gp120, clade A BG505 SOSIP, and clade C IAVI C22 gp120 tags (Duke Human Vaccine Institute, protein production facility) and a cocktail of all antigens was used for cell sorting. IgG+ B-cells were defined as CD3/14−, CD19+, CD20+, and IgA/IgM-; antigen-specific B-cells positive for probes in either PE or APC color were sorted into 96-well U-bottom plates containing 200 µl of B-cell culture medium (IMDM supplemented with FBS, Normocin, hIL-2, hIL-21, hCD40L His tag, and anti-His antibody). After 4-day culture, the B-cells were sorted into microtiter plates at one cell per well. Sorted plates were frozen immediately and maintained at −80°C before reverse transcription (RT)/PCR.

BCR sequencing

Request a detailed protocol

Natively paired variable region sequences from individual cells were generated by RT, cDNA barcoding, amplification, and sequencing as described previously (Tan et al., 2014; DeFalco et al., 2018). cDNA sequences were determined by 454 Titanium sequencing. A minimum of 10 reads for each chain (heavy and light) was required, and a contig was kept only if it included at least 90% of the reads for that chain from that well. V(D)J assignment and mutation identification was performed using IMGT reference sequences and IgBLAST through Immcantation analysis framework (Supplementary file 2). Percentage SHM was calculated by calling observed mutations of the entire input sequence compared to the germline sequence.

Analysis

Statistical analyses and visualizations were performed in R version 4.0.0 (April 24, 2020). Subclass distribution, gene family, and gene usage analysis: to exclude potential bias caused by the number of input cells or sequences, we divided the number of occurrences of each repertoire signature by the total number of all sequences for each individual. Gene recombination profiles: to avoid skewing of mean frequencies when averaged across all individuals by individuals with high or low number of available sequences (in particular for sequences with higher than 15% heavy chain mutations), normalized frequencies were multiplied by a factor (number of unique clones per individual). For IGHV-IGLV recombinations, the highly mutated sequences were defined using only heavy chain mutation cutoff. CDRH3 length distributions consist of CDRH3 length of each unique clone. Kolmogorov-Smirnov test was used to compare the distributions between TN and NN repertoires.

Mean number of mutations between TN and NN were compared using unpaired two-samples t-test. All correlations were performed using Spearman’s correlation. Visualizations were performed using R package ggplot2 (version 3.3.0). The flow graph of IGHV/IGHJ/IGLV/IGLJ combinations were plotted using ggalluvial extension of ggplot2. Circos plots were generated using R package circlize (version 0.4.9) chordDiagram function. The layout for clonal network visualization was arranged using R package packcircles (version 0.3.3).

Clones were defined using DefineClones tool of the Immcantation analysis framework (50). Clones consist of sequences that meet the following criteria: (1) come from the same individual; (2) share the same IGHV and IGHJ gene segment annotations; (3) have equal CDRH3 length; and (4) share sequence similarity calculated by nucleotide Hamming distance threshold of 0.16. Clonal groups were then split that have differing IGLV and IGLJ gene segment annotations.

The clonality measure was estimated using Gini index and calculated using R package bcRep (version 1.3.6) clones.giniIndex function.

Lineage trees were constructed using IgPhyML (version 1.1.0) (Stern et al., 2014). The HLP19 model was used to estimate maximum likelihood tree topologies. Clonal selection was estimated using ω, also called dN/dS, or the ratio of nonsynonymous (amino acid replacement) and synonymous (silent) mutation rates. A value of ω≈1 indicates totally neutral amino acid evolution, ω<1 negative selection, and ω>1 diversifying selection. The topology analyses were carried out using Alakazam (version 1.0.0) (Gupta et al., 2015) and the lineage trees were visualized using R ggtree (version 2.2.1).

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. A source data file (Supplementary file 2) has been provided for Figures 1-3.

References

Decision letter

  1. Pamela J Bjorkman
    Reviewing Editor; California Institute of Technology, United States
  2. Satyajit Rath
    Senior Editor; Indian Institute of Science Education and Research (IISER), India

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This study clearly shows that the donor antibodies' levels of somatic hypermutation correlate with serum neutralization breadth. The authors suggest that B cells expressing BCRs capable of binding a large array of Env variants gain more T help than B cells expressing BCRs that target a narrower range of Env variants, which is an interesting hypothesis for interpreting immune responses in the context of a continuously-evolving virus.

Decision letter after peer review:

Thank you for submitting your article "Distinct clonal evolution of B-cells in HIV controllers with neutralizing antibody breadth" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Satyajit Rath as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data.

Our expectation is that the authors will eventually carry out any additional experiments, if necessary, and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Summary:

Cizmeci and colleagues determined the serum neutralizing antibody activities of 22 HIV-infected individuals that spontaneously controlled HIV replication to low plasma levels in the absence of therapeutic interventions. They identified six individuals whose plasmas displayed broad neutralizing activities (top neutralizers, TN) and seven with narrow/no neutralizing activities (non-neutralizers, NN). They then isolated Env-specific class-switched memory B cells from these individuals and sequenced their IGH and IGL genes. This allowed them to determine the frequencies of specific IGH and IGL gene segments (including germline genes of known broadly neutralizing antibodies) and to compare these frequencies between TNs and NNs. Their analysis clearly shows that the donor antibodies' levels of somatic hypermutation correlate with serum neutralization breadth. The authors propose that in the context of prolonged, but relatively low HIV replication, B cells expressing BCRs capable of binding a large array of Env variants may gain more T help than B cells expressing BCRs that target a narrower range of Env variants. The former BCRs could be those from which broadly neutralizing antibodies are derived from, while the later ones could be those that give rise to neutralizing antibodies with narrow breadth. In the context of a continuously evolving virus, this proposal makes sense. The data presented in this manuscript does provide a valuable view into the natural development of neutralization breadth. However, the reviewers have substantial concerns, as outlined below, that they invite you to address in a revised manuscript. While they appreciate the valuable dataset presented by the authors, they believe that the analysis could be more thorough, and that further exploration of neutralizing Abs from the TN would strengthen the study.

Essential revisions:

1. The authors report that the composition of the BCR repertoire alone was insufficient to identify the TNs, although higher rates of somatic mutations were observed in the TNs than NNs. It is questionable whether they can draw this conclusion from the way the sequence analysis was performed. Although it is mentioned (first line in the Discussion) that 'paired' IGH and IGL sequences were compared, what is discussed throughout the manuscript are frequencies of either IGH or IGL but not frequencies of specific IGH/IGL pairs. For example, they mention that 'certain genes that are distinctive for known bNAbs, like VH1-2 (and they discuss VRC01-class antibodies)…..'. However, VRC01-class antibodies specifically utilize the *02 VH1-2 allele. Also, VRC01-class antibodies are formed by the pairing of VH1-2*02 with LCs expressing 5 amino acid long CDRL3 domains. We expect that most human VH1-2*02 HCs are paired with LCs expressing >5 amino acid long CDRL3s. They also mention that the CDRH3 lengths were not different between the IGHs found in the TN and NN (and they discuss that certain bnAbs that bind the apex of the trimer on the N332 site). On its own, the length of the CDRH3 loop is unlikely to be a good predictor of antibody neutralizing potency or breadth, because specific amino acids sequences CDRH3 loops of particular lengths are linked with the mechanism by which some anti-apex or anti-N332 antibodies act. The authors could determine the relative frequencies of VRC01-class BCRs from the existing data. Similarly, they could mine their data to identify particular CDRH3 sequences. This would be the minimal they could do prove their point, which would actually necessitate the generation of antibodies from particular BCRs and the characterization of their neutralizing activities.

2. It is unclear whether the broad neutralizing activities in the samples examined here targeted a single or multiple envelope epitopes. The authors should explicitly discuss this point in the Discussion section.

3. Several parts of the discussion hard to read and confusing. This reflects the somewhat confusing and contradictory findings of high levels of SHM, but contracting evolution that is shown in Figure 3. The discussion of this could do with some tightening. Parts are speculative and wordy, e.g., the first paragraph of page 8 could be shortened.

4. It is unclear whether a cocktail of the five different Envs listed in the Methods section was used to sort B cells, and if the same Envs were used in every case or not.

5. The manuscript reports that the difference in CDRH3 length distribution (Figure 1F) is not statistically significant (they do not report the exact p value produced by the Wilcoxon rank sum test). However, the difference appears substantial in Figure 1F. One wonders whether a different statistical test or looking at subpopulations might give a statistically-significant result. Also, in Figure 1F, can the authors comment on the very high frequency of sequences in NN with CDRH3 lengths of 9 and those of 16 on the TN? Are these sequence artefacts, perhaps of oversampled clones in certain donors?

6. The authors state that sequence data will be deposited in GEO. The processed sequences (used to make Figure 2, etc.) should be made available, not just the raw NGS sequencing.

7. There is some inconsistency with the V gene names – "VH1-2" vs. "HV4-59".

8. The proposal in the Introduction that neutralizing breadth may be achievable in the absence of high levels of viremia is not entirely new, as broadly neutralizing antibodies have been isolated from elite controllers. For instance, Kwong and colleagues reported on the presence of broadly neutralizing antibody responses in a cohort of long-term non-progressors (Gonzalez et al. PLOS ONE 2018) and the Bjorkman group reported on the isolation of a broadly neutralizing monoclonal antibody from a 'controller' (Schoofs et al. Immunity 2019).

9. Page 4 – The authors discuss the varying Ig subclass distributions and suggest for example that the higher percentage of IgG3 in NN versus TN might suggest a higher proportion of more recently induced MBCs in the NNs. These differences are not very convincing, if anything they are driven by a single data point, and we suggest removing related discussion.

10. Figure 1F – Can the authors comment on the very high frequency of sequences in NN with CDRH3 lengths of 9 and those of 16 on the TN. Are these sequence artefacts, perhaps of oversampled clones in certain donors?

11. Page 5 – could the authors translate frequency of nucleotide mutations into an approximate % SHM for easy comparison to other studies particularly of bNAbs. How close are these mutation counts of >60 to the levels of SHM we associate with neutralization breadth?

12. Figure 2 – In both panels A (IGHV) and C (IGLV), while the NNs largely cluster towards the left (with low numbers of mutations), there are a cluster of NN sequences with higher numbers of mutations (~80 for IGHV and 50 for IGLV). Are these sequences from specific donors? What, if anything, distinguishes them from the other sequences?

13. Page 6 – the last statement that the first paragraph seems somehow circular – "Assuming that the somatic IGHV mutation rates in B cells correlate with the total exposure time in the GC environment, and in light of the fact that the TN and NN did not differ in age, time-off-ART and viral loads, our data suggests that in the TNs, B-cell clones were selected for extensive SHM." How would B cell clones be selected FOR extensive SHM?

14. Page 6 – Clones may show variation in terms of CDRH3 length – did the authors explore the effect of softening these criteria e.g. by allowing +/- 1 amino acid on clone size?

15. Figure 3 – legend refers to a mean number of sequences per clone of 7.7 that differs from the text.

16. Page 7, third line, first para of discussion – why "in contrast"?

"Moreover, the mutation frequency in our data correlated directly with the neutralization breadth suggesting a skewing towards poly- or cross-reactive antibodies that are able to bind mutated forms of the original epitope" – do high mutation frequencies necessarily translate to polyreactivity?

17. In the Results section, at the end of clonal selection section (page 7), the discussion of "maximum affinity maturation" does not seem to reckon with the continuous change in the Env antigen, which generally manages to escape antibody pressure.

18. Page 8 – correct spelling of "alternatively"

19. Page 2 and page 20 – Supplemental Table 1. Why have the authors used a background of 3x MuLV? This is different from most studies of neutralization breadth. Why do some donors have activity against MuLV? What contributes to this non-specific background? Also, the legend contains a typo – please correct.

https://doi.org/10.7554/eLife.62648.sa1

Author response

Essential revisions:

1. The authors report that the composition of the BCR repertoire alone was insufficient to identify the TNs, although higher rates of somatic mutations were observed in the TNs than NNs. It is questionable whether they can draw this conclusion from the way the sequence analysis was performed. Although it is mentioned (first line in the Discussion) that 'paired' IGH and IGL sequences were compared, what is discussed throughout the manuscript are frequencies of either IGH or IGL but not frequencies of specific IGH/IGL pairs.

We appreciate the comment and have modified the data presentation and result section to highlight specifically IGH and IGL pairs. While we believe that reporting the frequencies of each IGH or IGL genes independently is informative (now in supplemental Figure S3B), we focus for the comparison of TN and NNs on IGH/IGL pairs depicting IGHV-IGLV pair frequencies as a heatmap (Figure 1B) but also listing the most frequent IGHV-IGHJ-IGLV-IGLJ combinations (in NN: IGHV1-69/IGHJ6/IGKV3-20/IGKJ5, and in TN: IGHV4-4/IGHJ6/IGLV6-57/IGLJ2) (Figure 1—figure supplement 3C). For the clonal analysis in figure 3, we only utilized the paired nature of the dataset by grouping sequences into unique clones only if they have the same IGHV-IGHJ-IGLV-IGLJ genes.

For example, they mention that 'certain genes that are distinctive for known bNAbs, like VH1-2 (and they discuss VRC01-class antibodies)…..'. However, VRC01-class antibodies specifically utilize the *02 VH1-2 allele. Also, VRC01-class antibodies are formed by the pairing of VH1-2*02 with LCs expressing 5 amino acid long CDRL3 domains. We expect that most human VH1-2*02 HCs are paired with LCs expressing >5 amino acid long CDRL3s. They also mention that the CDRH3 lengths were not different between the IGHs found in the TN and NN (and they discuss that certain bnAbs that bind the apex of the trimer on the N332 site). On its own, the length of the CDRH3 loop is unlikely to be a good predictor of antibody neutralizing potency or breadth, because specific amino acids sequences CDRH3 loops of particular lengths are linked with the mechanism by which some anti-apex or anti-N332 antibodies act. The authors could determine the relative frequencies of VRC01-class BCRs from the existing data. Similarly, they could mine their data to identify particular CDRH3 sequences. This would be the minimal they could do prove their point, which would actually necessitate the generation of antibodies from particular BCRs and the characterization of their neutralizing activities.

As suggested by the reviewer, we mined the data set for the frequencies of known bNAbs, including the VRC01-class antibodies. For the latter, we identified 3 unique clones utilizing IGHV1-2/IGKV3-20 genes, however these antibodies had 8-9 aa long CDRL3 and no sequence similarity for CDRH3. We screened for all neutralizing antibodies listed in the comprehensive review by Sok and Burton, Nat Immunology 2018 and also compared our sequences to Nabs that are listed in the CATNAP database.

We found very few sequences that match IGHV-IGLV genes with known bNAbs in either TN or NNs and overall there was no difference (Supplementary file 3). Beside the V3 glycan antibody BG18, that had previously been identified in one of our TNs (622800), we did not find any sequences that matched the exact CDR3-sequences of known bNAbs. However, these results are not surprising given that few of the TN subjects exhibited classic epitope-specific reactivity. Specifically, 22 controllers were profiled against a panel of epitope specific knock out variants in collaboration with Lynn Morris (Author response image 1) (Approach described in PMID: 24204277, PMID: 29950423). 5 TN, 6 MN, and 2 NN were profiled. Among the TN, none possessed class CD4-bs neutralizing antibodies. 622800 possessed PGT128-like response (likely from BG18). Conversely, 3 of the MN possessed CD4bs antibodies. Thus, not unexpectedly, we were not able to identify an enrichment of known bNAb sequences in our cohort.

Author response image 1

Conversely, it is important to note that BG18 and 3BNC117 were isolated from controllers, supporting our findings that extended SHM and affinity maturation occurs in these unusual individuals, and interestingly leads to the evolution of unique clonal lineages that give rise to potent broadly neutralizing antibodies – even in the absence of substantial antigenemia.Finally, we agree 100% with the reviewer that looking at CDR3 length alone is not an adequate predictor of neutralization capacity. We have made this explicitly clear in the text. However, for simple descriptive reasons we still compared CDR length between TN and NN for completeness. However, instead of comparing length means, we deployed a Kolmogorov-Smirnov test to describe length distributions in the groups. Indeed, using this approach, and as suspected by the reviewers, the 2 groups had significantly different CDR lengths.

2. It is unclear whether the broad neutralizing activities in the samples examined here targeted a single or multiple envelope epitopes. The authors should explicitly discuss this point in the Discussion section.

We agree that this is an important point and now specifically highlighted it in the result section and the discussion.

3. Several parts of the discussion hard to read and confusing. This reflects the somewhat confusing and contradictory findings of high levels of SHM, but contracting evolution that is shown in Figure 3. The discussion of this could do with some tightening. Parts are speculative and wordy, e.g., the first paragraph of page 8 could be shortened.

We apologize if the discussion has led to confusion and have now modified and shortened the text. We hope that the findings are now better explained, specifically emphasizing the different phases during the selection process with slowing down of the continuous accumulation of additional random amino acid changes once a certain level of affinity has been reached, that explain some of the observations in Figure 3.

4. It is unclear whether a cocktail of the five different Envs listed in the Methods section was used to sort B cells, and if the same Envs were used in every case or not.

Yes, a cocktail of the five different Envs listed were used in every case and we have clarified this in the methods section.

5. The manuscript reports that the difference in CDRH3 length distribution (Figure 1F) is not statistically significant (they do not report the exact p value produced by the Wilcoxon rank sum test). However, the difference appears substantial in Figure 1F. One wonders whether a different statistical test or looking at subpopulations might give a statistically-significant result. Also, in Figure 1F, can the authors comment on the very high frequency of sequences in NN with CDRH3 lengths of 9 and those of 16 on the TN? Are these sequence artefacts, perhaps of oversampled clones in certain donors?

We now use the Kolmogorov-Smirnov test to compare CDRH3 length distributions in NN and TN rather than comparing the median CDRH3 lengths. Indeed, the CDRH3 length distribution in TN sequences was significantly different to CDRH3 length distribution in NN sequences (p < 2.2e-16 Kolmogorov-Smirnov test; Figure 1C). Furthermore, significant differences in length were observed for CDRH2, CDRL1, CDRL3 (p < 2.2e-16 Kolmogorov-Smirnov test). There were no significant differences for CDRH1 and CDRL2. As correctly assumed by the reviewer, 355 of 488 sequences in NN with CDR3 lengths of 9 belong to the VH4-34/J4 – KV2-24.2D/KJ1 clone from donor 211774, while 978 of 1355 sequences in NN with CDR3 lengths of 16 belong to the VH1-24/J4 – LV1-51/LJ1 clone from donor 330183. As now we present the CDR3 length distributions separately for each individual and test the difference between the distributions using Kolmogorov-Smirnov test rather than comparing only medians with Wilcoxon test, we believe that our results no longer represent artifacts of oversampled clones.

6. The authors state that sequence data will be deposited in GEO. The processed sequences (used to make Figure 2, etc.) should be made available, not just the raw NGS sequencing.

We agree with the reviewer and have instead attached “Supplementary file 2” with all annotated sequences.

7. There is some inconsistency with the V gene names – "VH1-2" vs. "HV4-59".

All names are corrected to be consistent in the “HV” format.

8. The proposal in the Introduction that neutralizing breadth may be achievable in the absence of high levels of viremia is not entirely new, as broadly neutralizing antibodies have been isolated from elite controllers. For instance, Kwong and colleagues reported on the presence of broadly neutralizing antibody responses in a cohort of long-term non-progressors (Gonzalez et al. PLOS ONE 2018) and the Bjorkman group reported on the isolation of a broadly neutralizing monoclonal antibody from a 'controller' (Schoofs et al. Immunity 2019).

We have modified the language in the introduction and have added the 2 references.

9. Page 4 – The authors discuss the varying Ig subclass distributions and suggest for example that the higher percentage of IgG3 in NN versus TN might suggest a higher proportion of more recently induced MBCs in the NNs. These differences are not very convincing, if anything they are driven by a single data point, and we suggest removing related discussion.

The respective discussion has been removed.

10. Figure 1F – Can the authors comment on the very high frequency of sequences in NN with CDRH3 lengths of 9 and those of 16 on the TN. Are these sequence artefacts, perhaps of oversampled clones in certain donors?

We have modified the analysis for CDR length comparisons and now use the Kolmogorov-Smirnov test to compare length distributions in NN and TN rather than comparing median lengths (Figure 1C). As correctly suggested by the reviewer, 355 of 488 sequences in NN with CDR3 lengths of 9 belong to the VH4-34/J4 – KV2-24.2D/KJ1 clone from donor 211774, while 978 of 1355 sequences in NN with CDR3 lengths of 16 belong to the VH1-24/J4 – LV1-51/LJ1 clone from donor 330183. The new analysis present CDR3 length distributions separately for each individual and test the difference between the distributions using Kolmogorov-Smirnov test rather than comparing only medians with Wilcoxon test, we believe that our results no longer represent artifacts of oversampled clones.

11. Page 5 – could the authors translate frequency of nucleotide mutations into an approximate % SHM for easy comparison to other studies particularly of bNAbs. How close are these mutation counts of >60 to the levels of SHM we associate with neutralization breadth?

Translated mutation frequencies are added along with the mutation counts. Mutation frequencies were calculated by comparing the entire input sequence to the germline sequence. Figure 2A,D,G,H,I were modified to depict the %SHM. Previously defined >60 mutation count cutoff corresponds to %15 SHM and >40 mutation count to %10 SHM. Figure 3A is replotted using %SHM. This is consistent with the spread of SHM rates in known bNAbs that can range from 10 to 30% (ie VRC01 class antibodies).

12. Figure 2 – In both panels A (IGHV) and C (IGLV), while the NNs largely cluster towards the left (with low numbers of mutations), there are a cluster of NN sequences with higher numbers of mutations (~80 for IGHV and 50 for IGLV). Are these sequences from specific donors? What, if anything, distinguishes them from the other sequences?

Reviewers suggestion of translating mutation counts to frequencies helped us resolve the cluster of NN sequences with higher numbers of mutations (~80 for IGHV and 50 for IGLV). We now present Figure 2A and 2C panels as %SHM frequency distributions for each individual separately. As NN sequences with higher mutation counts were longer sequences, translated mutation frequencies do not show these clusters of NN sequences anymore.

13. Page 6 – the last statement that the first paragraph seems somehow circular – "Assuming that the somatic IGHV mutation rates in B cells correlate with the total exposure time in the GC environment, and in light of the fact that the TN and NN did not differ in age, time-off-ART and viral loads, our data suggests that in the TNs, B-cell clones were selected for extensive SHM." How would B cell clones be selected FOR extensive SHM?

We agree with the reviewer that this is misleading and have modified the text to state that “…TNs were more permissive for extensive SHM of B-cell clones”.

14. Page 6 – Clones may show variation in terms of CDRH3 length – did the authors explore the effect of softening these criteria e.g. by allowing +/- 1 amino acid on clone size?

We appreciate the suggestion, and we repeated the clonal inference allowing +/- 1 CDRH3 amino acid length. We would expect that softening these criteria could decrease the number of unique clones identified. However, we did not see a change in clone sizes. We evaluated various options available within the Immcantation framework (the computational pipeline we chose for this study), which explores clustering thresholds and distance measures for CDRH3 sequences. We finally selected the default Hamming distance threshold as explained in the methods.

15. Figure 3 – legend refers to a mean number of sequences per clone of 7.7 that differs from the text.

These numbers have been corrected to be consistent with the recent clonal analysis. Mean number of sequences per clone was changed to 7.4.

16. Page 7, third line, first para of discussion – why "in contrast"?

"Moreover, the mutation frequency in our data correlated directly with the neutralization breadth suggesting a skewing towards poly- or cross-reactive antibodies that are able to bind mutated forms of the original epitope" – do high mutation frequencies necessarily translate to polyreactivity?

We have modified the discussion to focus more on the paired IGH/IGL analysis and now highlight the differences observed. This has changed the first paragraph of the discussion substantially. With regards to the link between high mutation rates and poly-reactivity we agree with the reviewer that high mutation frequencies rather increase affinity than polyreactivity and have changed the text accordingly.

17. In the Results section, at the end of clonal selection section (page 7), the discussion of "maximum affinity maturation" does not seem to reckon with the continuous change in the Env antigen, which generally manages to escape antibody pressure.

We agree with the reviewer that this is reads confusing and have modified the text to reflect that the process of affinity maturation has slowed down (rather than reached a maximum), and that the continuous accumulation of additional random amino acid changes would rather be detrimental and decrease affinity.

18. Page 8 – correct spelling of "alternatively"

The spelling is corrected.

19. Page 2 and page 20 – Supplemental Table 1. Why have the authors used a background of 3x MuLV? This is different from most studies of neutralization breadth. Why do some donors have activity against MuLV? What contributes to this non-specific background? Also, the legend contains a typo – please correct.

The non-specific background activity that was observed when using negative control virus MuLV was most likely due to anticoagulants that are present in the patient plasma samples. All anticoagulants can cause some level of cell stress/toxicity of TZM.bl cells at low plasma dilutions. While serum is always preferred for measuring neutralizing antibodies in the TZM.bl assay to avoid this issue, plasma samples are also acceptable when proper controls are in place to account for any non-specific activity, which may also vary between individual samples. The use of a 3-fold cut-off for determining positive responses is actually a standard practice in the field. The validation criteria for the TZM.bl assay outline that titers within a 3-fold range can be considered equivalent (Sarzotti-Kelsoe, M. et al., 2014, J. Immunol. Methods, 409: 131-146). We have referenced this paper in the methods description. Thus, it is critical that a stringent cut-off (i.e. > 3-fold difference in ID50 titers) is utilized for establishing positivity criteria when comparing neutralization responses against HIV-1 pseudovirus strains versus non-specific background observed against non-HIV negative control viruses.

https://doi.org/10.7554/eLife.62648.sa2

Article and author information

Author details

  1. Deniz Cizmeci

    Ragon Institute of MGH, MIT and Harvard, Cambridge, United States
    Contribution
    Data curation, Formal analysis, Validation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  2. Giuseppe Lofano

    Ragon Institute of MGH, MIT and Harvard, Cambridge, United States
    Contribution
    Formal analysis, Visualization, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Evan Rossignol

    Ragon Institute of MGH, MIT and Harvard, Cambridge, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
  4. Anne-Sophie Dugast

    Ragon Institute of MGH, MIT and Harvard, Cambridge, United States
    Contribution
    Conceptualization, Resources, Data curation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Dongkyoon Kim

    Atreca Inc, Redwood City, United States
    Contribution
    Data curation, Formal analysis, Validation, Writing - review and editing
    Competing interests
    employee of Atreca Inc
  6. Guy Cavet

    Atreca Inc, Redwood City, United States
    Contribution
    Data curation, Formal analysis, Supervision, Validation, Methodology, Writing - review and editing
    Competing interests
    employee of Atreca Inc
  7. Ngan Nguyen

    Atreca Inc, Redwood City, United States
    Contribution
    Data curation, Formal analysis, Validation, Methodology, Writing - review and editing
    Competing interests
    employee of Atreca Inc
  8. Yann Chong Tan

    Atreca Inc, Redwood City, United States
    Present address
    Esco Ventures X, Singapore, Singapore
    Contribution
    Data curation, Formal analysis, Validation, Methodology, Writing - review and editing
    Competing interests
    Former employee of Atreca Inc Now employee of Esco Ventures X.
  9. Michael S Seaman

    Center for Virology and Vaccine Research, Beth Israel Deaconess Medical Center, Boston, United States
    Contribution
    Data curation, Formal analysis, Funding acquisition, Validation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  10. Galit Alter

    Ragon Institute of MGH, MIT and Harvard, Cambridge, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Investigation, Project administration, Writing - review and editing
    Competing interests
    GA is a founder of Seromyx Systems.
  11. Boris Julg

    Ragon Institute of MGH, MIT and Harvard, Cambridge, United States
    Contribution
    Conceptualization, Supervision, Investigation, Visualization, Writing - original draft, Writing - review and editing
    For correspondence
    bjulg@mgh.harvard.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4687-9626

Funding

Bill and Melinda Gates Foundation (OPP1146996)

  • Michael S Seaman

Bill and Melinda Gates Foundation (OPP1066973)

  • Galit Alter

Ragon Institute of MGH, MIT and Harvard (Internal funds)

  • Boris Juelg

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We would like to thank the participants of the Ragon HIV controller studies from whom these samples were obtained, and the staff involved in all the sample collection and processing. We thank Dr Daniel Lingwood for the careful review of the manuscript and his critical input. This work was funded by Bill and Melinda Gates Foundation Collaboration for AIDS Vaccine Discovery (CAVD) grant #OPP1066973 and #OPP1146996.

Ethics

Human subjects: All subjects signed informed consent, and the study was approved by the MGH/Partners Institutional Review Board (Protocol #: 2003P001894).

Senior Editor

  1. Satyajit Rath, Indian Institute of Science Education and Research (IISER), India

Reviewing Editor

  1. Pamela J Bjorkman, California Institute of Technology, United States

Publication history

  1. Received: September 1, 2020
  2. Accepted: April 2, 2021
  3. Version of Record published: April 12, 2021 (version 1)

Copyright

© 2021, Cizmeci et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 385
    Page views
  • 56
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    Na Li et al.
    Research Article Updated

    Adiponectin is essential for the regulation of tissue substrate utilization and systemic insulin sensitivity. Clinical studies have suggested a positive association of circulating adiponectin with healthspan and lifespan. However, the direct effects of adiponectin on promoting healthspan and lifespan remain unexplored. Here, we are using an adiponectin null mouse and a transgenic adiponectin overexpression model. We directly assessed the effects of circulating adiponectin on the aging process and found that adiponectin null mice display exacerbated age-related glucose and lipid metabolism disorders. Moreover, adiponectin null mice have a significantly shortened lifespan on both chow and high-fat diet. In contrast, a transgenic mouse model with elevated circulating adiponectin levels has a dramatically improved systemic insulin sensitivity, reduced age-related tissue inflammation and fibrosis, and a prolonged healthspan and median lifespan. These results support a role of adiponectin as an essential regulator for healthspan and lifespan.

    1. Cell Biology
    2. Structural Biology and Molecular Biophysics
    Axel F Brilot et al.
    Research Article

    Microtubule (MT) nucleation is regulated by the γ-tubulin ring complex (γTuRC), conserved from yeast to humans. In Saccharomyces cerevisiae, γTuRC is composed of seven identical γ-tubulin small complex (γTuSC) sub-assemblies, which associate helically to template MT growth. γTuRC assembly provides a key point of regulation for the MT cytoskeleton. Here, we combine crosslinking mass spectrometry, X-ray crystallography, and cryo-EM structures of both monomeric and dimeric γTuSCs, and open and closed helical γTuRC assemblies in complex with Spc110p to elucidate the mechanisms of γTuRC assembly. γTuRC assembly is substantially aided by the evolutionarily conserved CM1 motif in Spc110p spanning a pair of adjacent γTuSCs. By providing the highest resolution and most complete views of any γTuSC assembly, our structures allow phosphorylation sites to be mapped, surprisingly suggesting that they are mostly inhibitory. A comparison of our structures with the CM1 binding site in the human γTuRC structure at the interface between GCP2 and GCP6 allows for the interpretation of significant structural changes arising from CM1 helix binding to metazoan γTuRC.