Single-cell and deep sequencing of IgG-switched macaque B cells reveal a diverse immunoglobulin repertoire following immunization

Abstract

The non-human primate (NHP) model is important for pre-clinical evaluation of prophylactic and therapeutic intervention strategies. The recent description of the rhesus macaque germline immunoglobulin loci and establishment of a database of germline gene segments offer improved opportunities to delineate Ig gene usage in the overall B cell repertoire as well as in response to vaccination. We applied 454-pyrosequencing and single-cell RT-PCR of bulk and sorted memory B cells respectively to investigate Ig heavy chain variable gene segment expression in rhesus macaques. The two methods gave remarkably concordant results and identified groups of gene segments that are frequently or rarely used. We further examined the VH repertoire of antigen-specific memory B cells induced by immunization with recombinant HIV-1 envelope glycoproteins (Env), an important vaccine component. We demonstrate that Env immunization activates a highly polyclonal response composed of most of the expressed VH gene segments illustrating the considerable genetic diversity of responding B cells following vaccination.

Introduction

The ability of the naïve B cell repertoire to recognize almost any antigen is dependent on the process of V(D)J recombination where variable (V), diversity (D), and joining (J) gene segments generate a large number of unique B cell clones. In addition, diversity is generated in the recombining D-J and V-D junctions due to trimming and addition of non-templated nucleotides. The resulting highly variable domain spanning the V-D-J junction corresponds to the complementary determining region 3 (CDR3) of the Ab heavy chains. A similar process occurs in V-J recombination of the immunoglobulin (Ig) light chain gene segments to form its CDR3. The CDR3s, together with the V-gene encoded CDR1 and 2 regions of both the heavy and light chains, usually comprise most Ab contacts with the antigen (1). Additional variation is generated through random pairing of Ig heavy and light chains in the developing pro-B cell. The resulting B cell diversity is a major component of protective immunity to pathogens following re-encounter or vaccination. Following antigen-specific BCR activation of naïve B cells, antibody (Ab) affinity maturation occurs through somatic hypermutation (SHM) of the Ig genes of B cells recruited into germinal centers (GCs) within B cell follicles, eventually resulting in T cell-dependent class switching from IgM to IgG isotype-bearing Abs (2).

In any given individual, at any given moment, the circulating B cell repertoire is comprised of naïve B cells poised to respond to new antigens and IgG-switched memory B cells generated from prior exposure to pathogens, environmental antigens or vaccine antigens (3). Antigen-specific memory B cells have the capacity to rapidly differentiate into Ab-producing cells upon antigen re-encounter (4, 5). Examination of antigen-specific memory B cell repertoires therefore comprehensively surveys the B cell clones engaged by a specific antigen following infection or immunization. Single-cell sorting of HIV-1 Env-specific memory B cells from chronically HIV-1 infected individuals indicate a limited memory B cell repertoire size of approximately 50 clonotypes, with a bias towards the use of Immunoglobulin heavy chain variable (IGHV) 1 family gene segments (6). Vaccination with tetanus toxoid on the other hand was shown to yield a repertoire size of approximately 100 clonotypes, which was not diversified further by boosting (7, 8). So far, little is known about V-gene segment usage and clonality of B cell responses elicited by other vaccine antigens. Yet, there is an increasing interest in understanding germline VH gene activation and antibody Ab maturation in response to immunization, not the least in the HIV-1 vaccine field since it is known that several highly potent, broadly neutralizing antibodies against the envelope glycoproteins (Env) from HIV-1 infected individuals utilize the same IGHV1 family gene segment (9, 10).

To establish a baseline of VH usage in rhesus macaques, we investigated the contribution of individual Ab heavy chain V-gene segments in total IgG-switched rhesus macaques B cells. Next, we similarly analyzed the antigen-specific B cells isolated from NHPs immunized with soluble HIV-1 Env trimers in adjuvant. For total IgG-switched B cells, we used two independent methods: ultradeep 454-pyrosequencing of V(D)J transcripts generated from mRNA isolated from peripheral blood mononuclear cells (PBMCs) and single-cell sorting of IgG-switched memory B cells followed by nested PCR of V(D)J sequences. We observed highly congruent results with the two methods, allowing us to identify a large number of genetically distinct VH gene segments that were frequently or less frequently used. Furthermore, when we examined the gene segment use of Env-specific IgG+ memory B cells obtained from highly specific flow cytometric sorting (11), we observed a similar broad pattern of VH usage. These data demonstrate that the polyclonal B cell response to the HIV-1 trimers used here is genetically highly diverse, providing a basis for studies aimed to activate selected VH gene segments in a more specific manner. These results represent a first comprehensive analysis of Ig VH gene usage in IgG-switched rhesus macaque memory B cells, as well as in the antigen-specific B cell response to HIV-1 Env protein immunization.

Materials and Methods

Samples and ethics statement

Immunizations and blood samples from the rhesus macaques used in the current study are described elsewhere (12). Briefly, Chinese rhesus macaques were immunized five times in a monthly interval with recombinant gp140-F trimers based on the HIV-1 YU2 isolate (13) in AbISCO-100 and CpG-C adjuvant. Peripheral bleeds were taken under ketamine sedation, 1 and 2 weeks after each immunization and the PBMCs were isolated via density-gradient centrifugation with Ficoll-Hypaque. After extensive washing in PBS, the cells were frozen in fetal calf serum (FCS) supplemented with 10% DMSO. During the study the macaques were housed at the Astrid Fagraeus Laboratory animal facility at Karolinska Institutet. All procedures were approved by the Local Ethical Committee on Animal Experiments. Some immunoglobulin sequences from macaques F125, F126, F128, F129 and F130 were described previously (11, 14), while new memory B cell isolations were performed from frozen PBMCs for macaque F124, following the fourth immunization, and macaques F125 and F128, following the second and fifth immunizations.

Cell preparation and single-cell flow cytometric sorting

Frozen PBMCs were thawed and resuspended in 10 ml RPMI 1640 media supplemented with 10% FBS and 10,000 units/ml DNase I. After washing, the pellet was resuspended in 50 μl PBS and 5 μl Aqua dead cell stain and incubated for 20 min at 4°C. The cells were then stained essentially as previously described (11) using a cocktail of Abs for human CD3 (APC-Cy7; SP34-2), CD8 (Pacific blue; RPA-T8), CD14 (Qdot 605; M5E2), CD20 (PE-Alexa Fluro700; 2H7), CD27 (PE-Cy7; M-T271), IgG (FITC; G18-145) and IgM (PE-Cy5; G20-12) in 100 μl for 1 h at 4°C. To allow sorting of B cells expressing an Env-specific BCR, gp140-F-biotin pre-conjugated to streptavidin-APC was included at 4 μg/mL. Following staining, the cells were washed in pre-chilled PBS, resuspended in 500 μl and passed through a 70 μm nylon cell mesh. Total (CD20⁺CD27⁺IgG⁺) and Env-specific memory B cells (CD20⁺CD27⁺IgG⁺gp140-F⁺) were sorted at single-cell density into 96-well PCR plates containing 20 μl lysis buffer using a 3-laser FACS Aria cell sorter. All sorted cells were negative for Aqua dead stain mix, CD14, CD3, CD8, and IgM. The lysis buffer was composed of 6.25 mM DTT, 20 U RNase inhibitor, 5 μl 5× First-strand cDNA synthesis buffer (Life Technologies), and 0.0625 μl NP-40. In sorts performed for F124, an additional 10 μg/mL carrier RNA (poly-A, Qiagen) was included in the lysis buffer.

Single-cell RT-PCR

The sorted plates were reverse transcribed and the Ab genes amplified as previously described (14). Briefly, the RNA was reverse transcribed to cDNA following addition of 150 ng random hexamers, 0.4 mM dNTP, 100 U Superscript III, and 3.5 μl water per well and incubating the plates at 42°C 10 min, 25°C 10 min, 50°C 60 min, and 94°C 5 min. Nested PCR was performed on 3 μl cDNA in 25 μl reactions with the HotStar Taq Plus kit (Qiagen) using 5’ leader sequence- and 3’IgG-specific primers. In the second round PCR, 1.5 μl PCR product was used as template. Nested PCR products were evaluated on 2% 96-well agarose gels and positive wells with a specific band of ~450 base pairs (bp) were PCR purified and sequenced. Generated sequences are available with accession numbers KF947536-KF948098 (Env-specific) and KF948099-KF948514 (Total) at GenBank (http://www.ncbi.nlm.nih.gov/genbank/).

Antibody cloning and expression

Cloning sites were introduced in the V and J regions following verification of V(D)J sequence identity and primer align. The cloning PCR was performed in a total volume of 25 μl on 2 μl nested PCR product using the Phusion® Hot Start II High-Fidelity PCR kit according to the manufacturer’s instructions (Thermo Scientific). Briefly, the PCR consisted of 5 μl 5X Phusion GC Buffer, 1 μl 10 mM dNTP, 1 μl each of 10 μM 3’ and 5’ cloning primers, 2 μl nested PCR product and water to 25 μl. The PCR had an initial denaturation at 98°C for 30 sec, followed by 35 cycles of 98°C 10 sec, 55°C (10 cycles) and 60°C (25 cycles) 30 sec, and 72°C 30 sec. There was a final elongation at 72°C for 7 minutes before evaluation on a 1 % agarose gel. Positive bands (~400 bp for the heavy chain V(D)J and ~350 bp for the kappa and lambda light chain VJ) were then gel extracted (Thermo Scientific).

Restriction enzyme digestion of PCR products and cloning into eukaryotic expression vectors containing human Igγ1 heavy, Igλ2 or Igκ1 light chain antibody expression cassettes were performed as previously described (15), with some modifications. Restriction digestions were performed using FastDigest® enzymes (Thermo Scientific) according to the manufacturer’s instructions. Following digestion, products were purified (Thermo Scientific) and ligated into linearized, SAP treated, expression vectors using T4-DNA ligase (Thermo Scientific). XL10-Gold ultracompetent bacteria were transformed according to the manufacturer’s instructions (Agilent Technologies). Colonies were screened for positive insert by PCR and plasmids containing products of the correct size were expanded and sequenced (GATC Biotech).

Antibody expression and specificity mapping

Heavy and light chain vectors with functional inserts were transfected at equal ratio into 293F cells, cultured to a density of 1.2 million cells/ml, using FreeStyle™ MAX reagent (Life Technologies) according to the manufacturer’s instructions. The cell culture supernatants were tested for the presence of secreted IgG and antigen binding via ELISA 4-5 days following transfection and harvested and purified by protein G Sepharose columns (GE Healthcare) 7 days following transfection. The cloned antibodies were tested for successful antibody production and antigen binding by ELISA. Maxisorp plates (Nunc) were coated with 1-2 μg/mL Goat-anti-IgG_Fcγ (Jackson Immunoreserach), gp140-F (the antigen used for immunizations), and Influenza HA1 included as a control protein in PBS over night at 4°C. The samples were added in a 5-fold dilution series starting from 5 μg/mL purified MAb or 1:10 diluted culture supernatant and incubated for 1.5 h at RT. HRP conjugated goat-anti-rhesus IgG (Nordic MUbio) was then added at 1:5000 dilution in wash buffer and incubated for 1 h at RT. The bound antibody was detected by adding TMB+ (Life Technologies) for 5 minutes before stopping the reaction with an equal volume 1M H₂SO₄. The optical density was measured at 450 nm. In between each step the ELISA plates were washed six times with PBS supplemented with 0.05% Tween-20.

Library preparation for 454-pyrosequencing

The 454-library preparation was largely performed as described before (16) with some modifications. Total RNA was extracted from 12 million PBMCs from NHP F128 following the fifth immunization. mRNA was isolated from the purified total RNA using the Oligotex kit (Qiagen). cDNA synthesis was performed in three reactions á 40 μl using superscript III and Oligo(dT)₁₈ primers according to the manufacturer’s instructions (Life technologies). The cDNA was then pooled and PCR purified followed by elution with 30 μl 10 mM Tris-HCl pH 8.5, making each μl correspond to roughly 0.4 million input PBMCs.

Separate PCR reactions were set up for the IGHV1, 3, and 4 families using mixes of primers covering the rhesus germline genes (Table S1). Library amplification was performed on 5 μl cDNA, corresponding to 2 million input PBMCs, in 50 μl reactions using the Phusion® High-Fidelity PCR kit with the GC buffer according to instructions. The PCR program was initiated with a 98°C incubation for 30 sec, followed by 10 cycles of 98°C 10 sec, 51°C (for VH3) or 55°C (for VH1 and 4) 30 sec, and 72°C 30 sec and 15 cycles of 98°C 10 sec, 65°C 30 sec, and 72°C 30 sec. Following cycling there was a 10 min elongation at 72°C and finally cooling to 4°C. PCR products were evaluated on a 1% agarose gel and the bands excised and gel extracted. To remove potential gel residues the PCR products were further purified by Phenol/Chloroform extraction followed by ethanol precipitation. The DNA was quantified using Qubit (LifeTechnologies) before sequencing in a GS FLX instrument (Roche 454-Life Sciences) using the manufacturer’s suggested methods and reagents. Image collection and subsequent signal processing, quality filtering, and generation of nucleotide sequence quality scores were performed as described previously (10). The sequences are available under accession number SRP033611 at the NCBI SRA database (http://www.ncbi.nlm.nih.gov/sra).

Sequence analysis

454-pyrosequencing and single-cell RT-PCR generated sequences were first filtered for read lengths of 300 to 600bp. Germline V-genes for each sequence was then assigned using a BLAST procedure with parameters set to mimic IgBLAST (17). Variable region sequences were extracted based on the ClustalW2 global alignment of each sequence to single-cell RT-PCR generated Ab sequences. Incomplete sequences and problematic sequences (sequences with internal stop codons, frame shifts or loss of any of the conserved CDRH3-flanking motifs CXR/K and WGXG) were removed from downstream analysis. The clustering program CD-HIT (18) was then used with 100% sequence coverage and identity to remove redundant reads. Finally, the divergence (percent nucleotide mutation from the assigned germline V-gene) was calculated based on ClustalW2 global alignment to the variable region of the most homologous germline V-gene segment in a germline database constructed from previously annotated rhesus macaque Ig genes (11, 14). V, D and J assignments were additionally performed by IMGT®/HighV-Quest (19) (version 1.1.3, reference directory release 201338-1) using the rhesus macaque option and further analyzed in the recently published Immunoglobulin Analysis Tool (IgAT, version 1.14) (20) program. However, individual V-gene assignments could not be performed using the IMGT® tools as their database is still limited compared to published rhesus germlines (11).

Clonality within the sorted populations (Total and Env-specific memory B cells) was determined by IgAT with the criteria that clonally related sequences i) use of the same VH and JH genes, ii) have the same CDRH3 length, and iii) have a CDRH3 nucleotide sequence homology of ≥90%.

Statistical analysis

P values for comparing Total and Env-specific V(D)J-family distributions were calculated with the chi-square test. Comparisons of specific V-gene segments were made with two-way ANOVA followed by the Bonferroni multiple comparison posttest. Correlation was determined with the non-parametric spearman correlations test with two-tailed p value calculated for significance.

Results

Antibody sequence isolation and purification

To investigate the representation of the recently described heavy chain Ig V-gene segments in the rhesus macaque B cell repertoire (11, 14), IgG-switched B cells were interrogated using two independent approaches; 454-pyrosequencing of expressed V(D)J sequences from bulk PBMCs of NHP F128 (12) (Fig. 1A) and flow cytometric sorting of single memory B cells from five different NHPs: F124, F126, F128, F129 and F130 (12) followed by RT-PCR of expressed heavy chain V(D)J sequences (Fig. 1B). The 454-pyrosequencing was performed for the major IGHV families (IGHV1, 3, and 4), comprising 53 out of 62 described germline V-genes (11). Each IGHV1, 3, and 4 library was prepared from cDNA corresponding to 2 million PBMCs and sequenced separately to reduce PCR amplification bias. For single-cell analysis, IgG-specificity was achieved by sorting memory B cells based on positive surface expression of CD20, CD27, and IgG and by using 3’-primers specific for rhesus IgG1-4. Single-cell RT-PCR for IGHV families 1-7 was performed using recently described rhesus-specific VH primers (14).

Fig. 1. IGHV sequence generation and analysis.

(A) Schematic of the single-cell sort followed by Ig V(D)J RT-PCR. IgG-switched memory B cells, defined as CD20⁺CD27⁺IgG⁺, were sorted at single-cell density into 96-well plates containing lysis buffer. Following reverse transcription with random hexamers, nested PCR was performed with a mix of IGHV family-specific primers. PCR reactions producing specific bands of approximately 450 bp were sequenced and used in subsequent analysis. (B) Schematic of the library preparation for deep sequencing using 454-pyrosequencing. mRNA was isolated from PBMCs and reverse transcribed. Individual PCR reactions were performed for the IGHV1, 3, and 4 families using high fidelity enzyme and family specific primers with 454-adaptor tags. The PCR products were purified and sequenced separately using a GS FLX 454 instrument. (C) To obtain full length functional Ig V-gene segments for subsequent analysis, raw sequences were first filtered for length (300-600 bp). They were then translated to amino acid sequence to exclude sequences containing stop codons, in addition to include the conserved WGXG and CXR/K domains. Additionally, included sequences had an aa/nt ratio ≥0.7. Functional sequences were aligned to the rhesus germline IGHV database using ClustalW2 to calculate somatic hypermutation and provide the sequence with a V-gene assignment. To determine that a diverse pool of B cells was sampled, only cells expressing unique B cell receptors, as quantified by CD-HIT clustering, were included.

To analyze the generated antibody sequences, we implemented a bioinformatics pipeline for NHP heavy chain sequences (Fig. 1C). Briefly, sequence reads were first filtered for length, where only sequences between 300-600 bp were included. The sequences were then mapped to the current rhesus germline VH database and retained only if they included the conserved amino acid CXR/K and WGXG motifs present at the beginning and end of the CDRH3 region. Sequences with premature stop codons were discarded, as were sequences with amino acid to nucleotide ratios of ≤0.7 for identity to the assigned germ line sequence due to the likelihood that insertions or deletions resulted in a change of reading frame. Additionally, to reduce potential PCR bias of specific amplicons, only non-redundant sequences were included in the analysis. After application of these filters, the 454-pyrosequencing yielded a large number of sequences for each V-gene family: IGHV1 (20,166), IGHV3 (51,320), and IGHV4 (85,645). Single-cell sorting and PCR amplification of total memory B cells generated 480 raw sequences that, following bioinformatic processing, resulted in 416 unique sequences.

Ig-gene contribution to the IgG-switched B cell repertoire

The combination of deep sequencing and single-cell RT-PCR allows for qualitative and quantitative assessment of the Ab V(D)J gene contributions. Deep sequencing provides the capacity to ascertain, with both depth and high accuracy, genes that are expressed or not, while single-cell RT-PCR provides a quantitative measurement of how often specific gene segments are used. The latter means of analysis provides non-biased PCR amplification from a single-cell in which a unique V(D)J recombination event has occurred. The V-, D-, and J-gene contribution of known annotated germline genes in IGHV families 1, 3, and 4 was analyzed for sequences generated by 454-pyrosequencing and single-cell RT-PCR (Fig. 2A-C). Due to the challenge of correctly assigning D-gene segments, the D-gene contribution was determined at the family level. Overall, the expression of the Ig genes varied greatly, with some being abundantly used while others contributed to a lesser extent to the expressed IgG-switched B cell repertoire. Additionally, for the V-gene segments, some were not detected at all in our data set, despite having high homology to the PCR primers used for library preparation or single-cell RT-PCR. Generally, there was a very high concordance between the 454-pyrosequencing and the single-cell RT-PCR, resulting in a highly correlated pattern of gene segment contribution (Spearman r=0.92, p<0.0001, Fig. 2D). This suggests that there was no major bias in the PCR protocol used for 454-pyrosequencing and that both data sets therefore can be used for quantitative assessments and qualitative observations.

Fig. 2. High concordance in Ig VDJ gene usage by single-cell RT-PCR and deep sequencing.

The frequency of heavy chain VDJ gene contribution (Y-axis) was determined for the individual V-gene segments (A) or D- and J-families (B, C) (X-axis) within the three major V-gene families, IGHV1, 3, and 4 for 454-pyrosequencing generated sequences (Open bars) and single-cell flow sorted total memory B cells (Filled bars). The number of sequences included in the V-gene analysis (A) were for 454-pyrosequencing; VH1 n=20,226; VH3 n=51,320; VH4 n=85,645 and for single-cell RT-PCR; VH1 n=41; VH3 n=152; VH4 n=188. Individual D-family and J-gene (B, C) contribution to all analyzed sequences was determined following pooling of the sequences (454-pyrosequencing n=148,105 and Single-cell RT-PCR n=379). (D) The correlation of the V-gene, D-family and J-gene contribution to the analyzed sequences (A-C), as determined by 454-pyrosequencing (Y-axis) and single-cell RT-PCR (X-axis), was calculated by determining the nonparametric Spearman correlation coefficient (r) and p value.

To further assess the germline VH gene segments contributing to the functional IgG-switched B cell repertoire, the 454-pyrosequencing-derived sequences corresponding to unique V(D)J recombination events were plotted from highest to lowest frequency of contribution within each of the IGHV1, 3, and 4 families, respectively. The percentage SHM of each contributing sequence was calculated relative to the assigned germline sequence and the average SHM levels determined (Fig. 3A-C). These analyses show that unmutated variants, or variants very close to the assigned germline sequences, were found for the majority of gene segments. However, there were some exceptions such as VH1.61 where the closest variants were approximately 5% mutated from the putative germline, suggesting that these sequences may have derived from a V-gene that deviates from the published VH1.61 germline sequence. Overall, 42 of the possible 53 annotated V-gene segments were found to contribute to the circulating IgG-switched B cell repertoire, although there was clear preferred usage of some gene segments relative to others, closely resembling the pattern of V-gene usage observed in humans (21).

Fig. 3. Differential usage of germline Ig V-gene segments.

Frequencies of the individual V-gene segment contribution to the IGHV1 (A), IGHV3 (B), and IGHV4 (C) families, as determined by 454-pyrosequencing, are plotted in descending order. The number of sequences (n) and mean SHM based on nucleotide alignment to closest germline sequence, of each V-gene segments is indicated in the table. The scatter plot (right) shows SHM for each V-gene segment analyzed with mean values indicated by the grey line.

Highly diverse V-gene usage in the HIV-1 Env-specific IgG-switched memory B cell repertoire

We next investigated the range of V(D)J gene segments used in vaccine-induced responses, using samples from rhesus macaques inoculated with well-described recombinant HIV-1 Env protein-based trimers (12, 13). Both total and Env-specific memory B cells were sorted to allow comparisons within individual animals. To isolate the Env-specific memory repertoire, B cells were sorted at single-cell density based on expression of cell-surface CD20, CD27 and IgG as well as binding to a fluorochrome-conjugated Env trimer probe as previously described (11). The Env-specific population was found to be approximately 4% of the total memory population (Fig. 4A). Next, heavy chain V(D)J sequences from sorted Env-specific memory B cells were rescued with nested RT-PCR and a total of 638 VH sequences were generated. Following processing in the bioinformatics pipeline (Figure 1) for annotation of individual V-gene segments and IMGT®/HighV-Quest followed by analysis in IgAT for V(D)J-gene family assignment, 563 and 570 unique Ab sequences were obtained respectively. Of these, 33 heavy chain V(D)J sequences from two NHP donors were cloned, along with the corresponding light chain VJ sequences from the same cell to allow expression of the complete MAbs to verify the specificity of the antigen-specific sort. The resulting MAbs were evaluated for binding to HIV-1 Env by ELISA and were all confirmed positive, indicating very high specificity of the flow cytometry-based sort (Fig. S1).

Fig. 4. The HIV-1 Env-specific memory B cell Ig V-gene segment usage is highly polyclonal.

(A) Env-specific memory B cells were sorted at single-cell density based on the positive surface expression of CD20, IgG, CD27, and Env (red gate). Gate frequencies (percent) correspond to number of cells of total input. Streptavidin-APC conjugated trimeric gp140-F-biotin was used as the Env-specific probe. (B) IGHV-, D-, and J-family distribution of RT-PCR amplified sorted cells analyzed with IMGT®/HighV-Quest and IgAT. Families are color-coded with family 1-red, 2-grey, 3-blue, 4-green, 5-yellow, 6-pink, and 7-black. The size of the colored area corresponds to the frequency (percent) out of the total number of sequences, indicated in the center of the graphs. Differences in the gene family distribution between the Total and Env-specific sequences were evaluated with the Chi-square test with *p<0.05 and **p<0.01. (C) V-gene contribution to the total number of sequences (percent of total) shown as mean+SEM for sorted Total (black, n=5 donors) and Env-specific (red, n=3 donors) memory B cells is shown for the individual V-gene segments. Statistics was evaluated with two-way ANOVA followed by the Bonferroni posttest for multiple comparisons with ***p<0.001.

V-gene segments of all IGHV families except VH6 were detected in the Env-sorted populations, although at variable frequencies, similar to the distribution observed in the total IgG repertoire (Fig. 4B). However, there was a significant overrepresentation of the VH5 family (p<0.01, Chi-square test) in the Env-specific group compared to the total IgG-switched B cells. Upon closer examination of the individual V-gene segments (Fig. 4C), VH5.46 was to found to be used significantly more frequently in the Env-specific memory B cell pool than in the total IgG memory B cell pool (p<0.001, two-way ANOVA followed by the Bonferroni posttest). D- and J-genes from all families were found to contribute to the B cell pool with similar frequencies observed in the Env-specific and total IgG-switched memory B cells, with the exception of IGHJ5 and 6, which were slightly overrepresented among the Env-specific memory B cells (p<0.05, Chi-square test). The reasons for the overrepresented gene segments will be investigated in our future studies aimed at characterizing the elicited specificities in greater detail. When investigating specific V-gene segment contribution to the IGHV families, the sorted total memory B cells were found to engage 42 out of 62 described germline genes (Fig. 4C, black bars and Fig. S2). Remarkably, the antigen-specific memory B cell repertoire elicited by HIV-1 Env immunization engaged a similar breadth of V-genes, highlighting the extensive genetic diversity in the B cell response to a complex but single protein-based vaccine antigen (Fig. 4C, red bars and Fig. S3). Furthermore, by calculating the number of participating clonal lineages in the Env-specific V(D)J sequences we found that 502 of 606 functional sequences represented unique clonotypes (Table I), indicating a high diversity of Ab specificities elicited towards the trimeric Env antigen.

Table I.

Clonal lineage analysis of Env^a-specific memory B cells

Donor	Functionald	Uniquec	Clonotypesd
F124	217	200	164
F125	185	175	156
F128	204	195	180
Total	606	570	502

^aEnv refers to CD20⁺CD27⁺IgG⁺l40-F⁺ B cells

^dNumber of functional sequences as determined by IMGT/V-Quest and IgAT

^cNumber of unique sequences as determined by IMGT/V-Quest and IgAT

^dDefined as: (i) use the same VH and JH genes, (ii) identical CDRH3 length, (iii) 90% CDRH3 nucleotide homology

Discussion

Rhesus macaques are frequently used as models of human immunology and are important for the evaluation of primate responses to infection and vaccination (22-24). Until recently, systems to investigate elicited B cell responses in NHPs at the cellular and genetic level were poorly developed. With the publishing of the rhesus macaque genome (25), the annotation of rhesus Ig germline gene segments and the description of methods to phenotype and sort rhesus macaque B cell populations (11), the ability to dissect vaccine- and infection-induced B cell responses in NHPs are now available. This capability, coupled with methods to clone and express antigen-specific human monoclonal antibodies (6, 15), recently adapted to the rhesus macaque system (14) and recent developments in deep sequencing approaches to study human Ig repertoires (16, 21, 26-39), motivated us to investigate VH gene usage in IgG-switched and antigen-specific repertoires in rhesus macaques. While isolated studies of VH gene segment use in rhesus macaque B cells were reported (40-43), this is the first comprehensive report describing the IgG-switched B cell repertoire in NHPs at a level of high resolution to more specifically elucidate the genetic diversity of rhesus macaque memory B cells.

Another unique aspect of our study is that we use both 454-pyrosequencing and single-cell sorting for genetic characterization of IgG-switched memory B cells. Despite performing the PCR under very different conditions by these two distinct means of analysis, we obtained remarkably similar results with the two methods demonstrating VH gene segments that were frequently used, rarely used or not detectably used. Most notable was the lack of representation of several gene segments of the VH3 family, the largest VH family in both humans and macaques, in agreement with results previously obtained from 454-pyrosequencing of the human Ab repertoire (21). It is possible that a greater sequencing depth, or the analysis of additional animals, would reveal the use of additional gene segments as we observed some variation in the five different macaques analyzed in this study (Fig. S2 and S3). While 454-pyrosequencing provides a large number of sequences, allowing a substantial coverage of the VH repertoire, single-cell PCR allows the isolation of matched heavy and light V(D)J sequences for subsequent cloning and expression of functional MAbs. This is particularly important for the analyses of antigen-specific repertoires where it is critical to confirm the specificity of the flow cytometry-based sort by evaluating the binding capacity and specificity of the resulting MAbs. In the current study, we observed 100% specificity of a panel of 33 cloned MAbs, strongly supporting the conclusions in regards to the Env-specific VH repertoire presented here.

In the past few years a number of studies were performed using 454-pyrosequencing to investigate the antibody repertoires in HIV-1 infected individuals with the aim to trace the evolution of broadly neutralizing Abs in these individuals (16, 27-31). These analyses provide valuable information about the antibodyome induced and shaped by chronic HIV-1 infection and highlight the extreme levels of Ab affinity maturation that can develop under these conditions (44), likely a common feature of and a consequence of chronic infection (45). Deep sequencing approaches were also recently used to investigate human Ab responses to vaccination (32, 46), but, so far, such studies have not been reported from immunization trials in NHPs. Our results obtained by deep sequencing of rhesus macaque B cell responses therefore provide timely information, which will greatly facilitate future studies of B cell/Ab repertoires in the NHP model.

In parallel, we generated over one thousand sequences from single-cell sorted total and antigen-specific memory B cells from animals inoculated with soluble HIV-1 Env trimers (12). In contrast to studies of infection-induced B cell responses, where the Env antigen is highly variable due to the continuously changing virus population and therefore not easily definable (6, 47), we used a probe for FACS sorting that was identical to the protein used for immunization, allowing isolation of all potential Env-elicited sub-specificities. Clonal lineage analysis of the isolated sequences estimated that 502 unique Ab clonotypes were present in 606 Env-specific sequences, suggesting that vaccination results in a highly polyclonal response. Additionally, the overall VDJ gene distribution of the Env-specific sequences was highly similar to that observed in the IgG-switched total memory B cell repertoire with the exception of an increase of VH5-using sequences, especially of VH5.46, in Env-immunized macaques, potentially signifying that Env-specific antibodies using this V-gene segment might have a selective advantage following immunization, a finding that is worthy of further investigation.

A similar estimation of the clonality of vaccine-induced B cell responses was reported following single-cell sorting and antibody isolation from plasma blasts of humans repeatedly immunized with the protein-based vaccine tetanus toxoid (7, 8). In that study, immunization also induced a highly polyclonal B cell response consisting of approximately 100 different Ab clonotypes utilizing variable gene segments from most of the IGHV families. Our results suggest that the vaccine-induced response against HIV-1 Env is more diverse than that induced by tetanus toxoid, but how diverse remains to be determined by analyzing a considerably larger number of B cells to reach saturation. In contrast to the broad diversity of the antigen-specific B cell repertoire following immunization, IgG repertoires in chronically HIV-1 infected humans were shown to be restricted with the dominance of a few clonal families, preferentially VH1-using antibodies over the course of the infection (6, 48, 49). This phenomenon may not be specific for HIV-1, however, as an over-representation of antibodies using the VH1-69 gene segment was also observed in antibodies generated in other chronic diseases (45), suggesting that this may be a more general consequence of persistent antigen exposure. Also, a recent study by Xiao et al. (50) demonstrate a skewing of the VH gene usage over time, with a decrease in VH3-using and an increase in VH1-using B cells at the later time point. However, this study was not based on antigen-specific cells and the contribution of Env-specific antibodies to this shift is therefore unknown. Further studies of the Env-specific repertoire in both infected and vaccinated individuals are therefore needed.

In summary, we report a comprehensive deep sequencing analysis of rhesus macaque V(D)J transcripts, which together with our analysis of a large number of antigen-specific antibody sequences obtained from single cell sorting, provide a platform for further B cell studies in the NHP model.

Non-standard abbreviations

NHP

Non-human primate

GC

Germinal center

SHM

Somatic hypermutation

Env

HIV-1 envelope glycoproteins