Highly sensitive and unbiased approach for elucidating antibody repertoires

Significance

Antibodies are generated by B cells of the adaptive immune system to eliminate various pathogens. A somatic gene rearrangement process, termed V(D)J recombination, assembles antibody gene segments to form sequences encoding the antigen-binding regions of antibodies. Each of the multitude of newly generated B cells produces a different antibody with a unique antigen-binding sequence, which collectively form the primary antibody repertoire of an individual. Given the utility of specific antibodies for treating various human diseases, approaches to elucidate primary antibody repertoires are of great importance. Here, we describe a new method for high-coverage analysis of antibody repertoires termed high-throughput genome-wide translocation sequencing-adapted repertoire sequencing (HTGTS-Rep-seq). We discuss the potential merits of this approach, which is both unbiased and highly sensitive.

Abstract

Developing B lymphocytes undergo V(D)J recombination to assemble germ-line V, D, and J gene segments into exons that encode the antigen-binding variable region of Ig heavy (H) and light (L) chains. IgH and IgL chains associate to form the B-cell receptor (BCR), which, upon antigen binding, activates B cells to secrete BCR as an antibody. Each of the huge number of clonally independent B cells expresses a unique set of IgH and IgL variable regions. The ability of V(D)J recombination to generate vast primary B-cell repertoires results from a combinatorial assortment of large numbers of different V, D, and J segments, coupled with diversification of the junctions between them to generate the complementary determining region 3 (CDR3) for antigen contact. Approaches to evaluate in depth the content of primary antibody repertoires and, ultimately, to study how they are further molded by secondary mutation and affinity maturation processes are of great importance to the B-cell development, vaccine, and antibody fields. We now describe an unbiased, sensitive, and readily accessible assay, referred to as high-throughput genome-wide translocation sequencing-adapted repertoire sequencing (HTGTS-Rep-seq), to quantify antibody repertoires. HTGTS-Rep-seq quantitatively identifies the vast majority of IgH and IgL V(D)J exons, including their unique CDR3 sequences, from progenitor and mature mouse B lineage cells via the use of specific J primers. HTGTS-Rep-seq also accurately quantifies DJ_H intermediates and V(D)J exons in either productive or nonproductive configurations. HTGTS-Rep-seq should be useful for studies of human samples, including clonal B-cell expansions, and also for following antibody affinity maturation processes.

The B-lymphocyte antigen receptor (BCR) comprises identical Ig heavy (IgH) and Ig light (IgL) chains. Antibodies are the secreted form of the BCR. The V(D)J recombination process assembles germ-line V, D, and J gene segments into exons that encode the antigen-binding variable region exons of the BCR. The RAG 1 and 2 endonuclease (RAG) initiates V(D)J recombination by generating DNA double-stranded breaks (DSBs) between V, D, and J gene segments and their flanking recombination signal sequences (RSSs) (1). In this process, the V, D, and J coding ends are generated as covalent hairpins that must be opened and that are often further processed, before being joined by classical nonhomologous end joining (2). Processing of V, D, J coding ends can involve generation of deletions or insertions of nucleotides at the junction regions (2), including the frequent de novo addition of nucleotides by the terminal deoxynucleotidyl transferase component of the V(D)J recombination process (3). Notably the V(D)J junctional region encodes a major antigen contact region of the antibody variable region, known as complementarity determining region 3 (CDR3), and thus these junctional diversification processes make a huge contribution to antibody diversity.

The mouse IgH locus spans 2.7 megabases (Mb). There are 100s of V_Hs in the several megabase distal portion of the IgH, with the number varying substantially in certain mouse strains (4). The V_Hs lie ∼100 kb upstream from a 50-kb region containing 13 D_Hs, which is followed several kilobases downstream by a 2-kb region containing four J_Hs. The IgH constant region (C_H) exons lie downstream of the J_Hs. After assembly of a V_HDJ_H exon, transcription initiates upstream of the V_H and terminates downstream of the C_H exons, with V(D)J and C_H portions being fused into the ultimate IgH messenger RNA (mRNA) via splicing of the primary transcript. Due to the random junctional diversification mechanisms, only about 1/3 of assembled IgH V(D)J exons are able to generate in-frame splicing events that place the V(D)J and C_H exons in the same reading frame to generate productive (in-frame with functional V_H) rearrangements that encode an IgH polypeptide, with the remainder being nonproductive (out-of-frame, in-frame with a stop codon, or using a pseudo-V_H) (5). IgL chain variable region exons are assembled from just V and J segments but otherwise follow similar basic principles to those of IgH. The mouse Igκ light chain locus spans 3.2 Mb with 100s of Vκs in a 3.1-Mb region separated by 20 kb from five Jκs downstream whereas the Igλ light chain locus is smaller and less complex (6). RNA splicing again joins assembled VJ_L exons to corresponding C_L exons.

During B-cell development, V(D)J recombination is regulated to ensure specific repertoires and prevent undesired rearrangements. IgH V(D)J recombination occurs stage-specifically in progenitor B (pro-B) cells before that of IgL loci, which occur in precursor B (pre-B) cells. IgH V(D)J recombination is ordered, with D-to-J_H joining occurring, usually on both alleles, before appendage of a V_H to a DJ_H complex (Fig. S1A) (2). In addition, the V_H-to-DJ_H step of IgH V(D)J recombination is feedback-regulated with a productive rearrangement leading to cessation of V(D)J recombination on the other allele if it is still in the DJ_H configuration (2). In contrast, initial nonproductive IgH V(D)J rearrangements do not prevent V_H-to-DJ_H rearrangements from occurring on the other allele. Such feedback regulation generally leads to the typical 40/60 ratio of mature B cells, with two IgH V(D)J rearrangements (one productive) versus one IgH V(D)J plus a DJ_H rearrangement (7). V_H-to-DJ_H rearrangement is also regulated to generate diverse utilization of the 100s of upstream V_Hs. Although proximal V_Hs, notably the most proximal V_H (V_H81X), are somewhat overused in pro-B V(D)J rearrangements, the sequestering of the D_Hs and J_Hs in a separate chromosomal domain from that of the V_Hs (8, 9), coupled with the phenomenon of locus contraction (10, 11), allows even the most distal V_Hs to be used. Subsequently, the somewhat biased primary V_H repertoire in pro-B cells is subjected to cellular selection mechanisms to generate a more normalized primary repertoire in newly generated B cells (12).

Fig. S1.

Schematic for HTGTS-Rep-seq. (A) Schematic of the generation of DJ and VDJ rearrangements via V(D)J recombination showing Vs (green), Ds (purple), and Js (orange). Representative DJ and VDJ joining events are shown. (B) HTGTS-Rep-seq method overview. Briefly, genomic DNA from B-cell populations are sonicated and linearly amplified with a biotinylated primer that anneals downstream of one specific J segment. The biotin-labeled single-stranded DNA products are enriched with streptavidin beads, and 3′ ends are ligated in an unbiased manner with a bridge adaptor containing a 6-nucleotide random nucleotide (highlighted in the rectangular box). Products were then prepared for 2 × 300-bp sequencing on an Illumina MiSeq. Generated reads were analyzed with the Ig/TCR-Repertoire analysis pipeline described in Materials and Methods.

Each B cell expresses a unique BCR, and each individual mouse or human has the capacity to generate up to 10¹³ or more distinct BCRs in the primary repertoire (13), with a large fraction of these being generated by junctional diversification of IgH and IgL CDR3s (14). In this regard, the ability to quantitatively identify the IgH and IgL variable region exons that contribute to the primary antibody repertoire is of great interest in elucidating contributions of this repertoire to immune responses and to immune diseases (15). Several important repertoire sequencing assays that use next-generation sequencing have been developed. These approaches involve the generation of repertoire libraries from either genomic DNA or mRNA (15). Most prior DNA-based approaches rely on use of upstream degenerate V primers, each designed to identify members of particular V_H families, and a downstream degenerate J primer, an approach that covers many, but not necessarily all, V(D)J exons and likely not all equally. RNA-based approaches generally require only one downstream primer (from the J or constant region) and thus obviate biases in prior DNA-based assays, but these approaches can severely underestimate nonproductive rearrangements due to decreased transcript levels (15). In addition, the long length of the 5′ RACE-derived complementary DNAs can also pose a challenge because sequencing technologies cannot always cover the entire length of the V(D)J exons.

We developed linear amplification-mediated high-throughput genome-wide translocation sequencing (LAM-HTGTS) to identify unknown “prey” sequences that join to fixed DSB-associated “bait” sequences (16). LAM-HTGTS, like its predecessor HTGTS (17), employs a single primer for a DSB-associated bait sequence to perform linear amplification across bait–prey junctions to identify all prey sequences joined to the bait DSBs in an unbiased manner (16, 18). We have used various types of DSBs as bait for LAM-HTGTS, including those generated by engineered nucleases and endogenous DSBs (17–22). Because V(D)J recombination generates rearrangements with junctions at borders of V, D, and J segments, we can use primers for any of these gene segments as LAM-HTGTS bait to identify sites of RAG-generated DSBs, both in progenitor or precursor lymphocytes undergoing V(D)J recombination, as well as in mature lymphocytes to retrospectively identify V(D)J recombination events that occurred earlier in development. Notably, LAM-HTGTS using endogenous RAG-generated DSBs identified RAG-generated DJ_H joins, RSS joins in excision circles, and off-target junctions in developing B-lineage cells that were not detected by prior assays (22), illustrating the high sensitivity of the assay. Based on these earlier studies, we now describe an adaptation of LAM-HTGTS as a robust repertoire-sequencing assay that we term “HTGTS-adapted repertoire sequencing” (HTGTS-Rep-seq).

Results

Overview of LAM-HTGTS Adapted Repertoire Sequencing.

For HTGTS-Rep-seq libraries, we used bait coding ends of J segments to identify, in unbiased fashion, mouse IgH DJ_H repertoires, along with both productive and nonproductive IgH V(D)J repertoires from both pro-B and peripheral B cells. Similarly, we also identified mouse productive and nonproductive Igκ repertoires from peripheral B cells. For all samples analyzed, genomic DNA isolated from a pool of the given type of B cells was sonicated to generate fragments with an average size of ∼1 kb and that thus would be expected to harbor IgH V(D)J or DJ rearrangements, Igκ VJ rearrangements, or unrearranged J_Hs or Jκs (Fig. S1B). Biotinylated primers that anneal to sequences downstream of the coding end of a particular J_H or Jκ segment will allow linear amplification of any fragments containing the bait J segment(s). Subsequent streptavidin purification, adapter ligation, and library construction steps were carried out as previously described (16) (Fig. S1B). To generate longer sequencing reads for more accurate alignment of Vs and Ds, we positioned bait primers closer to the coding ends of bait Js and used MiSeq 2 × 300-bp paired-end sequencing to capture full-length V(D)J sequences in recovered junctions. For bioinformatic analysis, we combined our LAM-HTGTS pipeline with IgBLAST (23) to generate an analysis pipeline that provides comprehensive information on productive or nonproductive junctions and CDR3 sequences (see Materials and Methods for details).

For the HTGTS-Rep-seq, we generally kept for analysis all recovered junctions, including all duplicates for reasons described previously (22). To control for experimental variations, we generated three technical repeat HTGTS-Rep-seq libraries from the same splenic B-cell DNA samples, which yielded highly reproducible repertoires with correlation coefficient (r) values of 0.99 (Table S1). Even for biological repeat IgH or IgL HTGTS-Rep-seq libraries from pro-B or splenic B cells of three different mice, correlation analyses revealed highly reproducible repertoires with r values greater than 0.9 in most of the datasets (Tables S1 and S2). However, as described below, detailed analyses of certain aspects of such libraries, such as the fraction of unique CDR3s in the total repertoire, revealed expected biological variations (Table S1).

Table S1.

Summary of VDJ_H joins analysis from HTGTS-Rep-seq libraries of C57BL/6 mice

Cell type	Locus	Input DNA	Coding end primer	Exp.*	No. of V(D)J junctions, including duplicates	Correlation between % productive V(D)J subsets†	No. of unique CDR3 junctions	Correlation between % productive V(D)J subsets†
Pro-B	IgH	2 μg	JH4	Mouse 1	8,299	vs. M2: 0.88	1,639	vs. total: 0.97
				Mouse 2	25,957	vs. M3: 0.98	9,133	vs. total: 0.99
				Mouse 3	22,613	vs. M1: 0.85	8,894	vs. total: 0.99
Splenic B	IgH	2 μg	JH4	Mouse 1	45,857	vs. M2: 0.99	20,583	vs. total: 0.97
				Mouse 2	57,477	vs. M3: 0.97	38,091	vs. total: 0.99
				Mouse 3	16,504	vs. M1: 0.96	7,053	vs. total: 0.97
Splenic B	IgH	2 μg	JH1	Mouse 1	54,172	vs. M2: 0.98	N.D.	N.D.
				Mouse 2	77,670	vs. M3: 0.98	N.D.	N.D.
				Mouse 3	37,648	vs. M1: 0.98	N.D.	N.D.
Splenic B	IgH	2 μg	JH2	Mouse 1	66,589	vs. M2: 0.98	N.D.	N.D.
				Mouse 2	81,547	vs. M3: 0.98	N.D.	N.D.
				Mouse 3	74,857	vs. M1: 0.99	N.D.	N.D.
Splenic B	IgH	2 μg	JH3	Mouse 1	26,586	vs. M2: 0.99	N.D.	N.D.
				Mouse 2	29,619	vs. M3: 0.98	N.D.	N.D.
				Mouse 3	23,097	vs. M1: 0.99	N.D.	N.D.
Splenic B	IgH	2 μg	JH4	Mouse 1	21,532	vs. M2: 0.97	N.D.	N.D.
				Mouse 2	26,838	vs. M3: 0.95	N.D.	N.D.
				Mouse 3	12,669	vs. M1: 0.95	N.D.	N.D.
Splenic B	IgH	2 μg	JH4	Repeat 1	20,703	vs. R2: 0.99	14,461	vs. total: 1.00
				Repeat 2	18,959	vs. R3: 0.99	13,045	vs. total: 1.00
				Repeat 3	22,118	vs. R1: 0.99	14,799	vs. total: 1.00
Splenic B	IgH	500 ng	JH4	Repeat 1	20,605	vs. R2: 0.98	5,093	vs. total: 0.99
				Repeat 2	18,897	vs. R3: 0.97	5,374	vs. total: 0.99
				Repeat 3	20,105	vs. R1: 0.97	5,570	vs. total: 0.99
Splenic B	IgH	100 ng	JH4	Repeat 1	12,007	vs. R2: 0.77	1,163	vs. total: 0.96
				Repeat 2	6,968	vs. R3: 0.79	649	vs. total: 0.92
				Repeat 3	8,106	vs. R1: 0.86	896	vs. total: 0.96
Splenic B	Igκ	1 μg	Jκ1	Mouse 1	46,554	vs. C1: 0.99	N.D.	N.D.
			Jκ2	Mouse 1	26,117	vs. C1: 0.98	N.D.	N.D.
			Jκ4	Mouse 1	16,047	vs. C1: 0.98	N.D.	N.D.
			Jκ5	Mouse 1	9,782	vs. C1: 0.98	N.D.	N.D.
Splenic B	Igκ	1 μg	Jκ1	Combined	10,988	vs. M1: 0.99	N.D.	N.D.
			Jκ2	Combined	10,159	vs. M1: 0.98	N.D.	N.D.
			Jκ4	Combined	8,613	vs. M1: 0.98	N.D.	N.D.
			Jκ5	Combined	17,750	vs. M1: 0.98	N.D.	N.D.

N.D., not determined.

^*Mouse 1, 2, 3, the experiments were performed from three different mice; Repeat 1, 2, 3, the experiments were performed using DNA from the same mouse.

^†The correlation coefficient values (r) were derived from two sets of productive V(D)J exons (%) via the CORREL function in Excel. The bigger the value, the more similar the two sets of data.

Table S2.

Summary of HTGTS-Rep-seq libraries from 129SVE mice

Cell type	Locus	Input DNA	Coding end primer	Experiment	No. of junctions, including duplicates	Correlation between productive V(D)J subsets
Pro-B	IgH	2 μg	JH4	Mouse 1	20,081	vs. M2: 0.95
				Mouse 2	14,950	vs. M3: 0.93
				Mouse 3	21,701	vs. M1: 0.96
Splenic B	IgH	2 μg	JH4	Mouse 1	52,140	vs. M2: 0.98
				Mouse 2	67,885	vs. M3: 0.95
				Mouse 3	68,337	vs. M1: 0.96
Splenic B	IgH	2 μg	JH1	Mouse 1	165,224	vs. M2: 0.94
				Mouse 2	102,858	vs. M3: 0.95
				Mouse 3	97,125	vs. M1: 0.96
Splenic B	IgH	2 μg	JH2	Mouse 1	191,016	vs. M2: 0.98
				Mouse 2	123,362	vs. M3: 0.96
				Mouse 3	95,336	vs. M1: 0.96
Splenic B	IgH	2 μg	JH3	Mouse 1	77,966	vs. M2: 0.97
				Mouse 2	50,202	vs. M3: 0.97
				Mouse 3	43,512	vs. M1: 0.97
Splenic B	IgH	2 μg	JH4	Mouse 1	85,649	vs. M2: 0.96
				Mouse 2	37,247	vs. M3: 0.96
				Mouse 3	40,422	vs. M1: 0.97
Splenic B	IgH	2 μg	JH1	Combined	3,652	N.D.
			JH2	Combined	27,701	N.D.
			JH3	Combined	11,104	N.D.
			JH4	Combined	12,713	N.D.

N.D., not determined.

HTGTS-Rep-Seq Reveals IgH V_HDJ_H and DJ_H Repertoires in Developing and Mature B Cells.

To test the ability of HTGTS-Rep-seq to detect differences between primary pro–B-cell IgH repertoires versus those of peripheral B lymphocytes, we purified primary B220⁺CD43⁺IgM⁻ pro-B cells from the bone marrow and B220⁺IgM⁺ B cells from the spleen of wild-type (WT) C57BL/6 mice. We first used 2 μg of genomic DNA isolated from these cell populations to perform HTGTS-Rep-seq with a J_H4 coding end bait primer to capture V_HDJ_H4 and DJ_H4 rearrangements (Fig. 1A and Table S1). Libraries from both cell types showed broad use of V_Hs in V_HDJ_H4 rearrangements throughout the IgH variable region locus, with some V_Hs used more frequently (e.g., V_H5-2, V_H2-2, V_H3-6, V_H1-26, V_H1-64, V_H1-72, and V_H1-81) (Fig. 1B). The C57BL/6 IgH locus has ∼110 potentially functional V_Hs and 74 pseudo V_Hs categorized into 16 families (24). In the IgH repertoire libraries generated with a J_H4 coding end bait, we detected in V_HDJ_H exons 107 functional V_Hs from all 16 families, as well as 21 pseudo V_Hs with relatively conserved RSSs (Fig. 1C). Notably, the three “functional” V_Hs (V_H1-62-1, V_H2-6-8, and V_H7-2) not detected by HTGTS-Rep-seq also were not found by another high-throughput repertoire sequencing method (25), suggesting that they may actually be nonfunctional with respect to the ability to undergo V(D)J recombination.

Fig. 1.

HTGTS-Rep-seq of V_HDJ_H and DJ_H repertoire in pro-B cells and splenic B cells of C57BL/6 mice. (A) Schematic of the murine IgH locus showing V_Hs (green, functional; black, pseudo), D_Hs (purple), J_Hs (orange), and C_H region (black). The red arrow indicates the J_H4 coding end bait primer. (B) V_H repertoire with productive and nonproductive information from V_HDJ_H joins in pro-B cells (Upper) and IgM⁺ splenic B cells (Lower). Some of the most frequently used V_Hs are highlighted with arrows as indicated. (C) Utilization numbers of functional V_Hs and pseudo V_Hs across 16 families in HTGTS-Rep-seq libraries described in B. (D) Pie chart showing the average overall percentage of productive and nonproductive V_HDJ_H joins from libraries described in B. (E) D use in V_HDJ_H and DJ_H joins in pro-B cells and IgM⁺ splenic B cells as indicated. (F) DJ_H:V_HDJ_H ratios in pro-B cells and IgM⁺ splenic B cells as indicated. All of the data are showed by mean ± SEM, n = 3.

V_H-to-DJ_H rearrangements occur at the pro-B stage, with only one in three expected to be in-frame (5). In the V_HDJ_H4 exons we identified by HTGTS-Rep-seq, on average 65% were productive, and, correspondingly, 35% were nonproductive (Fig. 1D). This ratio likely reflects a dynamic differentiation process in which pro-B cells with two nonproductive rearrangements are negatively selected and those with a productive rearrangement on one allele are positively selected (12). Due in large part to feedback mechanisms from productive V(D)J_H rearrangements during pro–B-cell development, ∼40% of splenic B cells displayed V_HDJ_H rearrangements on both alleles (one productive and one nonproductive) and the remaining 60% had one productive V_HDJ_H and one DJ_H rearrangement (5). Thus, a population of splenic B cells theoretically would be expected to have about 71% productive V_HDJ_H exons and 29% nonproductive V_HDJ_H exons. Indeed, we observed a very similar ratio of productive/nonproductive V_HDJ_H4 exons (73:27) in the HTGTS-Rep-seq libraries from splenic B-cell DNA (Fig. 1D). In the DJ_H joins revealed by HTGTS-Rep-seq, D_H1-1 (also known as DFL16.1) was used most frequently in libraries from both pro-B and splenic mature B cells (Fig. 1E). Moreover, we observed a much higher percentage of DJ_H exons in pro-B cells compared with that of splenic B cells (45% vs. 25%) (Fig. 1 E and F), in line with D-to-J_H rearrangement on both alleles preceding V_H-to-DJ_H rearrangement in developing pro-B cells (5, 26, 27).

Biased Proximal V_H Use in 129SVE Mice Revealed by HTGTS-Rep-Seq.

The 129SVE mouse strain IgH locus contains more V_Hs than the C57BL/6 IgH locus with a somewhat different organization (24). Given that 129SVE mice and cell lines have frequently been used in V(D)J recombination studies, we used the same J_H4 bait primers to also generate HTGTS-Rep-seq libraries from 129SVE bone marrow pro-B cells and splenic B cells (Table S2). The 129SVE IgH locus V_H sequences are annotated up to ∼1 Mb into the variable V_H region, but V_H sequences lying within the relatively large more distal region of the locus are not completely annotated. Thus, to generate an approximate 129SVE V_HDJ_H repertoire, we ran IgBLAST analyses against a combination of all of the known 129SVE V_H sequences and the annotated distal V_H sequences from the C57BL/6 background starting from V_H8-2 (Fig. S2 A and B). As with the C57BL/6 libraries, the V_Hs were widely used, and we detected 128 functional V_Hs out of 133 distinct members of the 15 V_H families, plus 34 pseudo V_Hs (Fig. S2C).

Fig. S2.

HTGTS-Rep-seq of V_HDJ_H and DJ_H repertoire in pro-B cells and IgM⁺ splenic B cells of 129SVE mice. (A) Schematic of the murine IgH locus showing V_Hs (green, functional; black, pseudo), D_Hs (purple), and J_Hs (orange). The red arrow indicates the J_H4 coding end bait primer. (B) V_H repertoire with productive and nonproductive information from V_HDJ_H joins in pro-B cells (Upper) and IgM⁺ splenic B cells (Lower). Some of the most frequently used V_Hs are highlighted with arrows as indicated. Mean ± SEM, n = 3. (C) Utilization numbers of functional or pseudo V_Hs across 16 families in the HTGTS-Rep-seq libraries described in B. (D) Pie charts showing the average overall percentage ± SEM of productive and nonproductive V_HDJ_H joins in pro-B cells (Upper) and IgM⁺ splenic B cells (Lower). (E) D use in DJ_H joins in pro-B cells and IgM⁺ splenic B cells as indicated. Mean ± SEM, n = 3. (F) Comparison of DJ_H:V_HDJ_H ratios in pro-B cells and IgM⁺ splenic B cells as indicated. Mean ± SEM, n = 3. Details of the analysis are as described for Fig. 1.

In contrast to the IgH V_HDJ_H4 repertoire in C57BL/6 mice, we found a highly biased use of proximal V_Hs, especially V_H5-2 (also known as V_H81X) and V_H2-2, in 129SVE mice (Fig. 1B and Fig. S2B). The D-proximal V_H5-2 was used in 9.5% (1.7% productive; 7.7% nonproductive) of all V_HDJ_H4 exons in pro-B cells and about 4% (0.3% productive; 3.5% nonproductive) of all V_HDJ_H4 exons in splenic B cells of 129SVE mice (Fig. S2B). In contrast, V_H5-2 appeared in only about 3.5% (0.7% productive; 2.8% nonproductive) and about 1.8% (0.15% productive; 1.6% nonproductive) of the V_HDJ_H4 exons in C57BL/6 pro-B and splenic B cells, respectively (Fig. 1B). The majority of V_H5-2–containing V_HDJ_H4 joins in splenic B cells were nonproductive in both mouse strains, in contrast to other highly used V_Hs throughout both alleles (V_H2-2, V_H5-4, V_H3-6, V_H1-26, V_H1-55, V_H8-8, V_H1-64, V_H1-72, and V_H1-81), consistent with previous reports that most V_H5-2–containing productive rearrangements are selected against due to their autoreactive properties or inability to properly pair with IgL or surrogate IgL chains (28–30). Because the V_H5-2 gene body, associated RSS, and downstream region are conserved in C57BL/6 versus 129SVE mouse strains, the basis for greatly increased V_H5-2 utilization in primary repertoires of the 129SVE strain remains to be determined.

A comparison of V_HDJ_H and DJ_H rearrangements in 129SVE pro–B-cell libraries also revealed a relatively lower ratio of productive/nonproductive V_HDJ_H exons (39:61 in 129SVE vs. 65:35 in C57BL/6), as well as a lower ratio of V_HDJ_H/DJ_H rearrangements (about 45:55 in 129SVE vs. about 55:45 in C57BL/6) (Fig. 1 D–F and Fig. S2 D–F). V_H5-2 rearrangements did not substantially contribute to these differences. Both pro–B-cell libraries were generated in 4-week-old mice, suggesting that the lower relative proportion of productive V_HDJ_H exons in 129SVE compared with C57BL/6 pro-B cells might be attributed to differential timing of B-cell checkpoint selection in these two mouse strains. For both mouse strains, the splenic B-cell libraries showed comparable productive/nonproductive and VDJ/DJ ratios (Figs. 1 D–F and Fig. S2 D–F).

IgM⁺ Splenic B-Cell V_HDJ_H Exons Display Similar V_H Use Profiles Across Different J_Hs.

We also designed bait primers to the other three J_Hs in the IgH locus and made libraries from splenic B cells of both C57BL/6 and 129SVE mice to compare V_H and D utilization among the different J_Hs. These assays revealed similar V_H and D utilization repertoires for the four different J_Hs, indicating that selection for a particular V_H or D in a V_HDJ_H join did not vary substantially between the J_Hs in both C57BL/6 and 129SVE mice (Fig. 2A and Fig. S3A). However, we did find higher proportions of nonproductive V_HDJ_H rearrangements using the J_H2 and J_H3 baits, compared with the J_H1 and J_H4 bait libraries (Fig. 2A and Fig. S3A). In this regard, the stretch of sequence from the J_H coding ends to the highly conserved WGXG-motif that is crucial for a stable antibody structure (24) is shorter in the J_H2 and J_H3 segments relative to the J_H1 and J_H4 segments (Fig. S4A). Thus, some V_HDJ_H2 and V_HDJ_H3 join sites could lie too close to the WGXG-encoded sequences and be selected against due to unstable antibody structure (Fig. S4B). Moreover, we observed moderate differences in the D_H use profiles among the four J_Hs and a larger ratio of V_HDJ_H:DJ_H joins for the J_H4 bait libraries, which potentially could reflect the relative positions of these J_Hs in the recombination center that initiates V(D)J recombination (31) (Fig. 2 B and C and Fig. S3 B and C). Finally, we prepared HTGTS-Rep-seq libraries from 129SVE splenic B cells with four sets of J_H HTGTS-Rep-seq primers combined (Fig. S5A and Table S2). This approach, which allowed us to detect all V_HDJ_H1-4 exons in one HTGTS-Rep-seq library, revealed general V(D)J repertoires similar to those detected with individual J_H primers (Fig. S5 vs. Fig. S3).

Fig. 2.

V_HDJ_H and DJ_H repertoires in IgM⁺ splenic B cells across four J_H baits. (A) V_H repertoire with productive and nonproductive information from V_HDJ_H joins (Left) and pie charts showing the average overall percentage of productive and nonproductive V_HDJ_H joins (Right) in IgM⁺ splenic B cells using each of the J_H coding end bait primers as indicated. (B) Comparison of D use in DJ_H joins in IgM⁺ splenic B cells using each of the J_H coding end bait primers. (C) Comparison of DJ_H:V_HDJ_H ratios in IgM⁺ splenic B cells using each of the J_H coding end bait primers. Mean ± SEM, n = 3 for all of the data. Other analysis details are as described for Fig. 1.

Fig. S3.

Comparison of V_HDJ_H and DJ_H repertoire in IgM⁺ splenic B cells of 129SVE mice using four different J_H baits. (A) V_H repertoire with productive and nonproductive information from V_HDJ_H joins (Left) and pie charts showing the average overall percentage ± SEM of productive and nonproductive V_HDJ_H joins (Right) in IgM⁺ splenic B cells using individual J_H coding end bait primers. (B) Comparison of D use in DJ_H joins in IgM⁺ splenic B cells using each of the J_H coding end bait primers. (C) Comparison of DJ_H:V_HDJ_H ratios in IgM⁺ splenic B cells using each of the J_H coding end bait primers. Mean ± SEM, n = 3 for all of the panels. Other analysis details are as described for Fig. 1.

Fig. S4.

In-frame V_HDJ_H proportions across J_H coding end lengths for J_H1-4. (A) Alignment of the germ-line sequences of J_H1-4. The sequences were extracted from the mm9 genome and are highly conserved between 129SVE and C57BL/6. The WGXG-encoding sequences are in red. J_H length is marked with blue arrowheads, with 1 indicating the nucleotide most proximal to the bait primer. (B) Red line plots show the number per 10,000 total V(D)J joins that retained the indicated J_H length for each J_H bait. Blue bar graphs show the percentage of in-frame V(D)J exons at each retained J_H length. Mean ± SEM, n = 3.

Fig. S5.

IgM⁺ splenic B-cell V_HDJ_H use profiles in a 129SVE mouse using four J_H baits combined. (A) Schematic of IgH locus as in Fig. 1. Red arrows indicate mixed primers that bind downstream of each J_H. (B) V_H use profiles separated by J_H segment baits. One representative profile is shown here from two repeats of combined primer HTGTS-Rep-seq libraries. (C) D use in DJ_H joins in IgM⁺ splenic B cells using each of the J_H coding end bait primers.

HTGTS-Rep-Seq Detects Diverse Igκ VJ Rearrangements.

In mice, the Igκ locus generates the majority of IgL-expressing B cells (32). The Vκ locus organization is distinct from that of the V_H locus. Besides not having D segments and, therefore, undergoing direct Vκ-to-Jκ rearrangements, the Vκ locus contains V segments organized in both direct and inverted orientation relative to the Jκ segments (6) (Fig. 3A). Thus, for some Vκs, joining to Jκ occurs deletionally like V_H-to-DJ_H joining, but, for others, it occurs via inversion of the intervening sequence. Direct and inverted Vκs generally occur in distinct clusters but also can be individually interspersed (Fig. 3A). To first assess the Igκ repertoire, we performed HTGTS-Rep-seq on 1 μg of genomic DNA from C57BL/6 splenic B cells using a Jκ5 coding end bait primer. Similar to the IgH locus, we also observed widespread use of Vκs across the entire locus to the Jκs (Fig. 3 A and B). All of the 100 functional Vκs across 20 Vκ families were detected by HTGTS-Rep-seq, and 11 out of 62 pseudo Vκs were also detected (Fig. 3C). We saw productive/nonproductive VJκ joins at a 63:37 ratio in splenic B cells (Fig. 3B), which is slightly lower than the predicted 67:33 ratio (33). This small deviation might reflect the presence of nonproductive VJκ joins in Igλ-positive cells (32).

Fig. 3.

HTGTS-Rep-seq of VJκ repertoire in IgM⁺ splenic B cells of C57BL/6 mice using Jκ5 bait primer. (A) Schematic of the murine Igκ locus showing Vκs and Jκs. Green and orange bars indicate functional Vκs with convergent and tandem transcriptional orientations, respectively, to the downstream Jκs. Black bars indicate pseudo Vκs. The red arrow indicates the Jκ5 coding end bait primer. (B, Left) Vκ repertoire with productive and nonproductive information from VJκ joins in IgM⁺ splenic B cells with Jκ5 bait primer either individually (Upper) or from combined Jκ bait primers (Lower). Some differentially used Vκs among four different Jκs are highlighted with arrows as indicated (see also Fig. S6). (Right) Pie chart showing the overall percentage of productive and nonproductive VJκ joins. Representative results from two repeats are shown. (C) Utilization numbers of functional and pseudo Vκs across 20 families in libraries described in B.

We also generated HTGTS-Rep-seq libraries from splenic B-cell DNAs to capture VJκ joins from the three other functional Jκ segments separately or in a combination of all four Jκ primers. In contrast to IgH repertoires with different J_H primers, the Igκ repertoires showed apparently different utilization of some Vκs (e.g., Vκ6-15, Vκ6-23, Vκ19-93, Vκ10-96, and Vκ1-135) between different Jκ baits. Moreover, the productive/nonproductive ratios from the other Jκ primer libraries were slightly lower than those observed with the Jκ5 primer (Jκ1, 53:47; Jκ2, 60:40; Jκ4, 53:47; vs. Jκ5, 63:37) (Fig. S6). These differences in utilization and ratios likely reflect the occurrence of sequential VJκ recombination events (34). In this context, alleles containing nonproductive VJκ joins with the three Jκs upstream of Jκ5 have the ability for an unrearranged Vκ upstream of the nonproductive VJκ to join to a remaining Jκ (34). If this secondary rearrangement is inversional, the nonproductive VJκ joins would be retained in the genome and add to the nonproductive fraction of VJκ1, VJκ2, or VJκ4 joins that are detected by HTGTS-Rep-seq. Given this scenario, VJκ5 rearrangements, which are terminal rearrangement events, would be expected to reflect the theoretical productive/nonproductive ratios, as we have found.

Fig. S6.

Igκ repertoire in IgM⁺ splenic B cells of C57BL/6 mice using different Jκ baits. (A) Schematic of Igκ locus, as in Fig. 3. Red arrows indicate the position of used Jκ bait primers. (B) Vκ use profiles and overall productive/nonproductive ratios of VJκ separated by Jκ baits in IgM⁺ splenic B cells. In each panel, representative Vκ repertoires with productive and nonproductive information from VJκ joins with each Jκ bait primer either individually (Upper) or from combined Jκ primers (Lower) are shown. Some differentially used Vκs among four different Jκs are highlighted with arrows as indicated (see also Fig. 3). Representative results from two repeats are shown.

HTGTS-Rep-Seq Revealed Characteristic CDR3 Properties.

We analyzed the CDR3 sequences from productive V_HDJ_H and VJκ rearrangements in pro-B and splenic B cells. The CDR3 of productive V_HDJ_H exons in pro-B and splenic B cells showed a diverse range of lengths from 3 to 24 amino acids (aa) with a peak at 11–15 aa (Fig. S7 A and B). The consensus CDR3 motifs of these V_HDJ_H exons, made from the unique subset, from unimmunized pro-B and splenic B cells, shared the same V_H contributed and J_H4 contributed amino acid sequences as anticipated (Fig. S7 A and B). Given that the gene bodies of J_H2 and J_H3 are shorter than those of J_H1 and J_H4, the average lengths of V_HDJ_H2 and V_HDJ_H3 exons were shorter than those of V_HDJ_H1 and V_HDJ_H4 (median length 11 aa vs. 13 aa) (Fig. S7C). In contrast to productive V_HDJ_H exons, ∼85% of productive VJκ exons from splenic B cells showed a CDR3 length of 9 aa. The VJκ CDR3 motif also showed the expected flanking cysteine and phenylalanine (Fig. S7D). Thus, HTGTS-Rep-seq produces sequences with CDR3 characteristics expected from the various bait loci.

Fig. S7.

CDR3 length distribution and consensus motif of productive V_HDJ_H and VJκ exons. (A) CDR3 length distribution of productive V_HDJ_H exons in C57BL/6 pro-B libraries made with J_H4 bait primer. Consensus CDR3 motif plots were made for the subset of 11- to 13-aa-length CDR3 sequences, flanked on either end by the consensus cysteine and tryptophan. (B) As in A, for C57BL/6 splenic B libraries made with J_H4 bait primer. (C) As in A, for C57BL/6 splenic B libraries made with the four J_H bait primers. Mean ± SEM, n = 3 for A–C. (D) As in A, for C57BL/6 splenic B libraries made with Jκ5 primer. Note that we noticed some errors in our CDR3 sequence analyses due to the basal levels of sequencing errors of current high-throughput sequencing methods, including Illumina MiSeq, and the read length (maximum 600 bp) that are not sufficient to cover entire sequences of longer DNA fragments containing V(D)J exons. However, we eliminated such potential ambiguities by including in our analyses only overlapping joined reads and/or by increasing thresholds for read quality.

HTGTS-Rep-Seq Can Be Used with Low Amounts of Starting Material.

We generated libraries from J_H4 coding end baits with starting DNA amounts of 2 μg, 500 ng, and 100 ng, each purified from the splenic B cells of the same C57BL/6 mouse. Libraries generated from 2 μg and 500 ng of genomic DNA were almost identical (r > 0.97) in V_H use and productive/nonproductive rearrangement ratios (Fig. 4 and Table S1). Even though we saw a slight decrease in the number of detected V_Hs from the libraries generated from 100 ng of genomic DNA, they still displayed a similar repertoire profile (r ≈ 0.8) and productive/nonproductive ratio (Fig. 4), suggesting that HTGTS-Rep-seq can be used to generate a quite representative V_HDJ_H repertoire library from as little as 20,000 B cells.

Fig. 4.

Representative V_HDJ_H repertoire can be generated from small amounts of starting genomic DNA. (A) V_H repertoire with productive and nonproductive information from V_HDJ_H joins (Left) and pie charts showing the average overall percentage of productive and nonproductive V_HDJ_H joins (Right) in IgM⁺ splenic B cells cloned from indicated amounts of genomic DNA using J_H4 coding end bait primer. Mean ± SEM, n = 3. (B) V_H utilization numbers separated by family, organized as in Fig.1C.

We further evaluated V(D)J_H junctional diversities in these titrated libraries by comparing the percentages of unique CDR3 sequences (35). We found that the proportion of V(D)J exons containing unique CDR3 sequences substantially decreased with reduced amounts of starting material (Fig. S8A), indicating that higher amounts of DNA starting material allow us to detect a greater fraction of the highly diverse IgH CDR3 repertoire. Although sequencing errors might in theory lead to minor overestimation of CDR3 diversity, the biological diversity of CDR3 in these samples was so high that we observed only a very small overlap portion in detected V(D)J_H CDR3 sequences (<1%) between the three technical repeats of 2-μg DNA libraries and even less between 500-ng or 100-ng DNA library repeat subsets (Fig. S8B). Thus, 100 ng of DNA is enough to generate a representative V(D)J_H library with respect to V_H use, but even 2 μg of DNA reveals only a very small fraction of the immense diversity of IgH CDR3s.

Fig. S8.

Characterization of unique CDR3 reads. (A) Proportion of unique CDR3 sequences for each technical repeat library from Fig. 4. Mean ± SEM, n = 3. (B) The number of identical CDR3 sequences between technical repeat libraries at varying amounts of starting material.

Discussion

HTGTS-Rep-seq is a DNA-based method that requires only a single bait PCR primer, reads out both deletional and inversional V(D)J joins, and can readily be adapted to identify low frequency recombination events invisible to prior repertoire sequencing assays (22). In addition, HTGTS-Rep-seq can be used to comprehensively study productive and nonproductive V exon use. We also can use HTGTS-Rep-seq to developmentally assess the frequency of V(D)J intermediates, most notably by quantitatively identifying the frequency of particular DJ_H rearrangements (22) (Fig. 1 E and F). HTGTS-Rep-seq also could be adapted for revealing joining patterns of individual Ds or Vs by using them as baits. Thus, this assay, or adaptations of it, could be useful for detecting changes in repertoires that occur during development, or during an immune response. However, use of HTGTS-Rep-seq for assaying certain antigen receptor repertoires, most notably TCRα repertoires, would currently be more limited given the very large number of different Jαs (24).

HTGTS-Rep-seq requires as little as 100 ng of genomic DNA (and potentially less) from mouse splenic B cells to capture a representative profile of V_H use. Thus, this technique can be applied to relatively small numbers of cells and yield accurate repertoire profiles. However, we find that much larger amounts of starting material would be required to capture the full extent of the immense complexity of the CDR3s that we demonstrate to exist in a given population of splenic B cells. Moreover, potential inaccuracies that do arise in quantifying certain rearrangements via HTGTS-Rep-seq, such as productive/nonproductive ratios for the Igκ repertoire, are due to inherent biological events that would be detected in other DNA-based repertoire-sequencing methods, such as nonproductive VJκ rearrangements in the genome in Igλ-expressing cells or sequential rearrangements involving inversional VJκ joining (34) (Fig. S6). This ambiguity in the assay for the Igκ locus could be minimized if desired by adding an initial step to enrich for sonicated DNA fragments containing sequences just downstream of the whole Jκ region.

The ability to use linear amplification with only a single J primer or set of J primers by HTGTS-Rep-seq avoids the necessity of using sets of degenerate V primers (along with J primers) required by prior DNA-based repertoire-sequencing methods, which could lead to variable amplification efficiencies of different V families or Vs within a family (15). Being DNA-based, HTGTS-Rep-seq also bypasses a major limitation of RNA-based methods for certain applications by quantitatively capturing the frequency of Ig rearrangements in a population regardless of their expression level or whether they are productive or nonproductive. Current means to address biases due to multiplex PCR or varying expression levels between cells include the use of universal identifiers (25, 36, 37) or single cell methods (38), but HTGTS-Rep-seq can accurately identify a population repertoire profile without the additional cost or steps of synthesizing primers with random barcodes, or sorting for single cells.

It is striking that, in experiments where we sequenced about 15,000 unique V(D)J rearrangements from each of three technical repeats, we found less than 1% overlap of unique CDR3 sequences, emphasizing the great sensitivity of the approach. This highly sensitive HTGTS-Rep-seq approach should easily be adapted for application to human samples. In that regard, the sensitivity of HTGTS-Rep-seq should provide a low cost and rapid method for identifying clonal rearrangements (even DJ_H rearrangements) that would be diagnostic of clonal B- or T-lymphocyte expansions that occur in the context of certain immune system diseases, including cancers. Finally, in our libraries, approximately one-third of our joined sequences cover the entire length of the ∼370-bp V(D)J exons, making HTGTS-Rep-seq applicable to tracking dominant populations of particular V(D)J exons, including particular CDRs, that appear in the B-cell repertoire during antibody affinity maturation in an immune response. This application may be enhanced as high throughput sequencing technologies are advanced to achieve greater lengths and accuracy.

Materials and Methods

Mice.

WT 129SVE and C57BL/6 mice were purchased from Charles River Laboratories International. All animal experiments were performed under protocols approved by the Institutional Animal Care and Use Committee of Boston Children’s Hospital.

B-Cell Isolation from Bone Marrow and Spleen.

Bone marrow-derived pro-B (B220⁺IgM⁻CD43⁺) cells were purified from 129SVE or C57BL/6 mice by sorting and after the depletion of erythrocytes. Single cell suspensions were stained with B220-APC, CD43-PE, and IgM-FITC antibodies. Splenic resting B cells were purified using biotin/streptavidin bead methods (B220-positive selection) (130-049-501; Miltenyi) or EasySep negative B-cell selection (19754; Stem Cell Technologies).

HTGTS-Rep-Seq.

HTGTS-Rep-seq was performed as described (16). Primers are listed in Table S3. For the DJ_H joins analysis, we used the standard LAM-HTGTS bioinformatic pipeline (16). For the V_HDJ_H and VJκ identification, we demultiplexed MiSeq reads using the fastq-multx tool in the ea-utils suite (https://code.google.com/archive/p/ea-utils) and trimmed adaptors with cutadapt software (https://cutadapt.readthedocs.io/en/stable/). The paired reads were then joined using the fastq-join tool from the ea-utils suite (overlap region ≥10 bp and mismatch rate ≤8%). Reads were then grouped as joined reads and unjoined and were analyzed separately in the following analysis. We used IgBLAST (23) using joined reads and unjoined reads against V(D)J gene databases using default parameters. The V(D)J gene sequences were obtained from IMGT (24), manually curated, and used to generate IgBLAST sequence databases. Various stringencies were applied to filter reads that can align to V, D, and J genes (IgBLAST score >150, total alignment length >100, overall mismatch ratio <0.1). In unjoined reads, the top V gene identified in R1 and R2 reads must match. The use of V genes can be computed based on the processed IgBLAST results. A pipeline named “HTGTSrep” was developed to conduct the above-mentioned processing and analyzing and can be downloaded at Bitbucket (https://bitbucket.org/adugduzhou/htgtsrep).

Table S3.

Oligos used for library construction

Name	Sequence	Purpose
JH1-bio	/5BiosG/CTGCAGCATGCAGAGTGTG	HTGTS bio primer for JH1 coding end
JH1- red	TGACATGGGGAGATCTGAGA	HTGTS red primer for JH1 coding end
JH2- bio	/5BiosG/ACCCTTTCTGACTCCCAAGG	HTGTS bio primer for JH2 coding end
JH2- red	CCCCAACAAATGCAGTAAAATCT	HTGTS red primer for JH2 coding end
JH3- bio	/5BiosG/GGGACAAAGGGGTTGAATCT	HTGTS bio primer for JH3 coding end
JH3- red	CCCGTTTGCAGAGAATCTT	HTGTS red primer for JH3 coding end
JH4_bio	/5BiosG/CCCTCAGGGACAAATATCCA	HTGTS bio primer for JH4 coding end
JH4 _red	CTGCAATGCTCAGAAAACTCC	HTGTS red primer for JH4 coding end
Jκ1_bio	/5Biosg/TTCCCAGCTTTGCTTACGGAG	HTGTS bio primer for Jκ1 coding end
Jκ1_ red	AGTGCCAGAATCTGGTTTCAGAG	HTGTS red primer for Jκ1 coding end
Jκ2_bio	/5Biosg/ATTCCAACCTCTTGTGGGACAG	HTGTS bio primer for Jκ2 coding end
Jκ2_ red	TCCCTCCTTAACACCTGATCTGAG	HTGTS red primer for Jκ2 coding end
Jκ4_bio	/5BiosG/CGCTCAGCTTTCACACTGACTC	HTGTS bio primer for Jκ4 coding end
Jκ4_ red	CAGGTTGCCAGGAATGGCTC	HTGTS red primer for Jκ4 coding end
Jκ5_bio	/5Biosg/GCCCCTAATCTCACTAGCTTGA	HTGTS bio primer for Jκ5 coding end
Jκ5_ red	GTCAACTGATAATGAGCCCTCTCC	HTGTS red primer for Jκ5 coding end