Massively Parallel Sequencing of Peritoneal and Splenic B Cell Repertoires Highlights Unique Properties of B-1 Cell Antibodies

The Ly-1⁺ (CD5⁺) subset of B cells, later named B-1 cells for their early appearance in ontogeny, has many unique characteristics (1, 2). In contrast to conventional B-2 cells, B-1 cells develop in the fetal liver, produce so-called “natural” Abs (NAbs) even in a germ-free environment, react to Ag independent of cognate T cell help, and their Ab production can be stimulated by non–Ag-specific signals (e.g., TLR agonists) (3–5). A phenotypically similar subset, termed B-1b cells, has been described, which shares similar surface markers with B-1a cells, but does not express CD5 (6). In contrast to B-1a cells, B-1b cells are able to expand clonally in response to Ag and can be reconstituted from a single hematopoietic stem cell from adult bone marrow, suggesting that B-1b cells develop from different stem cells than B-1a cells (7–9). B-1 cells also are the predominant B cell subset in the peritoneal cavity and B-1a cells can migrate to the spleen in response to LPS, where they differentiate and secrete Ab (10, 11). Their Abs form a first-line response against infections (e.g., Streptococcus pneumoniae), are involved in clearing apoptotic debris, and can protect from atherosclerosis by binding to oxidation-specific epitopes (OSEs) of oxidized low-density lipoprotein (OxLDL) and apoptotic cells (12, 13). We have postulated that these innate NAbs can be thought of as innate pattern recognition receptors (PRRs) that have been selected to provide homeostasis against endogenous danger-associated molecular patterns (DAMPs) such as OSEs, as well as against pathogen-associated molecular patterns (PAMPs) as found on exogenous pathogens. A combination of these functions can be mediated by only a single Ab, for example, by the phosphocholine (PC)-binding T15/E06 IgM NAb (14, 15).

Each B cell develops its BCR during development in the fetal liver or adult bone marrow by rearrangement of V, D, and J gene segments (only V and J for the L chain). During rearrangement, the junctions are modified by TdT activity, which adds random nontemplated nucleotides (N additions) to the ends of gene segments, and exonucleases, which can shorten each segment (16). Notably, TdT is active only after birth, and neonatal mouse Ig lacks N additions (17, 18). The regions constituting the Ag-binding site of the Ab are the highly variable CDRs. Of these, the most diverse is the CDR3 region of the IgH, which spans both the V-D and the D-J junction. It is the major site determining the Ab’s specificity, and a substitution in only one amino acid in the CDR3 of the H chain can alter reactivity with its Ag (19, 20).

Originally, B-1a cell Ab sequences were studied by sequencing of hybridomas derived from CD5⁺ B cells (21, 22). They were described as Abs with high analogy to germline sequence and a low number of N additions (18). A high number of these cells bound to bromelain-treated RBCs (BrRBCs), binding a hidden “autoantigen” exposed by proteolysis (21). Most of these anti-BrRBC sequences contained the gene segments V_H11 and J_H1 (23). Phosphatidylcholine competed Ab binding to BrRBCs, indicating that this might be the exposed epitope or a molecular mimic (13, 24).

Later, Kantor et al. (25) analyzed the repertoire of single cell–sorted peritoneal B-1a, B-1b, and B-2 cells by sequencing the IgH V region (IGHV) of 55–70 cells per subset, giving further insight into the properties of the B-1 cell repertoire. Vale et al. (26) showed that transcripts of peritoneal V_H5 (V_H 7183) sequences had properties and CDR3 sequences that were uncommonly found in splenic or bone marrow sequences. In a study describing functional differences of B-1a cells with different levels of surface expression of plasma cell alloantigen 1 (PC-1), Wang et al. (27) recently showed some disparities in V_H usage and N additions of PC-1^hi as opposed to PC-1^lo B-1a cells. All of these studies used traditional sequencing techniques, which only allowed the identification of a few hundred sequences at a time. With the advance of massively parallel next-generation sequencing and new amplification methods for the IGHV, it is now possible to analyze a much higher number of sequences (28, 29). In fact, during preparation of this manuscript, an article by Yang et al. (30) was published, where next-generation sequencing was used to show, among other findings, that the repertoire of B-1a cells becomes more restricted as mice age.

In this study, we analyze the repertoire of peritoneal (B-1a, B-1b, B-2) and splenic (B-1a, marginal zone [MZ], and follicular [FO]) B cell subsets by using a nonbiased 5′-RACE reverse transcription step to generate cDNAs including all potential IGHV sequences, which were then determined by massively parallel sequencing. We used RNA as starting material as a representation of the functional repertoire (i.e., actually expressed sequences). Our studies demonstrate that B-1a cells have a unique repertoire with higher usage of V_H11 and V_H12 and that these sequences have distinct properties compared with other Ab sequences. All B-1 subsets were shown to have significantly more N addition–free transcripts compared with B-2 cells. Additionally, we compare the repertoires between all the six subsets analyzed, finding notable overlaps, but also surprising differences in the Abs they express.

Ten-week-old female C57BL/6 mice were purchased from Charles River Laboratories and maintained in our vivarium for 2 wk with a 12-h light/12-h dark cycle and ad libitum access to water and food. Mice were sacrificed for analysis at the age of 12 wk. All animal experiments were performed according to the National Institutes of Health guidelines and were approved by the University of California San Diego Animal Subjects Committee.

Peritoneal lavage was performed on sacrificed mice using 10 ml of isolation buffer (PBS with 1% FCS, 10 mM EDTA). Spleens were homogenized with a syringe and cells were passed through a 70-μm cell strainer using isolation buffer to obtain a single-cell suspension. All further steps were done at 4°C or on ice. RBCs were lysed using RBC lysis buffer (BioLegend, San Diego, CA) for 3 min. Both peritoneal and splenic cells were subsequently centrifuged at 400 × g for 5 min, resuspended in sorting buffer (PBS with 5 mM EDTA, 2 mM HEPES, 1% FCS), and counted.

Two to three mice were used for each sort and their cells pooled to attain sufficient amounts of RNA for 5′-RACE and sequencing. To avoid overrepresentation of one single mouse in that pool, an equal ratio of peritoneal-to-splenic cells from each mouse was used. Cell isolations were done on 3 separate days, and sequencing libraries were prepared independently for each of the experiments and later combined, culminating in one database with sequences from eight mice.

Peritoneal and splenic cells were first incubated with anti-CD16/32 Ab to block Fc receptors and then incubated for 30 min with Abs specific for: CD19 (clone: 1D3, fluorochrome: allophycocyanin-Cy7), CD23 (B3B4, FITC), IgM (II/41, allophycocyanin), IgD (11-26c.2a, Brilliant Violet 421), and CD5 (53-7.3, PE-Cy7) for 30 min. For splenic cells, Abs to CD43 (S7, PE) and CD21 (7G6, Brilliant Violet 605) were used additionally. 7-Aminoactinomycin D was added to determine cell viability. All reagents were purchased from BD Biosciences (San Jose, CA) or BioLegend. FACS was done at the flow cytometry core of the La Jolla Institute for Allergy and Immunology, and cells were sorted into FCS. Single, live lymphocytes were gated based on scatter properties and 7-aminoactinomycin D staining. The gating strategy for peritoneal and splenic B cell subsets was modified from published methods (Fig. 1) (31, 32) and criteria are indicated as follows: peritoneal cell subsets (Fig. 1A): B-1a cells, CD19^hiCD23⁻IgM^hiIgD^loCD5⁺; B-1b cells, CD19^hiCD23⁻IgM^hi IgD^loCD5⁻; B-2 cells, CD19^midCD23⁺IgM^loIgD^hiCD5⁻; splenic cell subsets (Fig. 1B): B-1a cells, CD19⁺IgM^hiIgD^loCD23⁻CD21^−/loCD43⁺CD5⁺; MZ cells, CD19⁺IgM^hiIgD^loCD23^−/loCD21^hiCD43⁻CD5⁻; FO cells, CD19⁺IgM^loIgD^hiCD23^hiCD21^midCD43⁻CD5⁻.

Immediately after sorting, cells were centrifuged and resuspended in TRIzol and RNA was isolated by spin column centrifugation (Direct-zol RNA MiniPrep kit; Zymo Research, Irvine, CA). Reverse transcription with template switching was performed using the 5′-RACE SMARTer kit (Clontech, Mountain View, CA) and a specific primer encoding the μ-chain (5′-ATGGCCACCAGATTCTTATCAGAC-3′) (33). The resulting cDNA product of the IGHV with the Clontech oligonucleotide sequence at the 5′ end was amplified by two-step PCR with HiFi HotStart ReadyMix (Kapa Biosystems, Wilmington, MA). The first PCR reaction was designed with a primer containing both a region complementary to the C region of the μ-chain in proximity to the J_H region (underlined) and to the Illumina adaptor sequence (bold type) (5′-GACGTGTGCTCTTCCGATCTGGGAAGACATTTGGGAAGGACTG-3′) and a primer mix complementary to the oligonucleotide used for template switching (underlined) and to the Illumina adaptor sequence (bold type) (5′-ACACGACGCTCTTCCGATCTAAGCAGTGGTATCAACGCAGAGT-3′ and 5′-ACACGACGCTCTTCCGATCT-3′). PCR products were loaded onto a 1% agarose gel. After electrophoresis, the IGHV product at 450–600 bp was extracted using MinElute gel extraction kit (Qiagen, Hilden, Germany). A second PCR reaction was then used to add the remaining Illumina adaptor sequence and unique sample indexes (5′-AATGATACGGCGACCACCGAGATCTACACNNNNNNNNACACTCTTTCCCTACACGACGCTCTTCCGATCT-3′ and 5′-CAAGCAGAAGACGGCATACGAGATNNNNNNGTGACTGGAGTTCAGACGTGTGCTCTTCCGATC-3′, where NNNNNNNN is the unique index sequence).

Sequencing was performed at the Institute for Genomic Medicine at the University of California San Diego on the Illumina MiSeq using paired-end read sequencing with a length of 2 × 300 bp.

Raw sequencing data were groomed and quality trimmed at the 3′ ends of each of the paired-end reads for a Phred score >30 with Galaxy (34). Quality trimmed reads were joined in their overlapping regions using FLASH (35). The dataset of joined sequences was uploaded to IMGT/HighV-QUEST (36, 37). Data from the obtained spreadsheets were then analyzed using the Ig analysis tool (38). Specific sequence subsets were extracted based on certain characteristics (N addition, V_H usage) by custom Python scripts (available at https://github.com/maxchang/IMGT-subsets). Data from the IMGT files or Ig analysis tool output were plotted and analyzed with Prism (GraphPad Software, San Diego, CA). Only functional sequences (i.e., without a stop codon) were included in the analysis. For each of the B cell subsets, three separate pools were prepared, each of which contained B cells from two to three mice, and each pool was separately sequenced. For the main part of the analysis, we combined the data generated from these three individual experiments into one dataset for each B cell subtype. In an additional analysis, joined sequences were analyzed with the Ab Mining Toolbox to obtain a list of expressed CDR3 amino acids sequences along with their frequency in each subset (39). All raw sequence data were uploaded to the National Center for Biotechnology Information Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra/, BioProject accession number PRJNA418221).

To characterize the binding specificity of the most common CDR3 sequence observed in peritoneal and splenic B-1a cells (CMRYGNYWYFDVW, V11-D2-J1), we generated a cDNA for a single-chain variable fragment (scFv) using the original IGHV bearing this CDR3, paired with the L chain V region from the hybridoma sequences from the study originally describing this sequence (21). This was constructed by gene fragment synthesis of H and L chain variable regions and Gibson assembly (Integrated DNA Technologies, San Diego, CA) in a construct with a flexible 15-aa-long linker connecting the H and L chain domains and designated as XQ11-scFv. A His6-tag was inserted into the C terminus of the scFv, and the fusion construct cloned into a pFUSE mammalian expression vector (InvivoGen, San Diego, CA) under the control of an hEF1-HTLV promoter. This was transiently expressed in HEK293T cells cultured first in DMEM containing 4.5 g/l glucose, 10% FBS, and 15 μg/ml blasticidin, and then serum-free media. The culture supernatant was concentrated using an Amicon Ultra centrifugal filter device (Millipore, Burlington, MA) and then used in an ELISA to evaluate binding to Ags.

Chemiluminescent ELISA of XQ11 binding to indicated Ags was performed in 96-well microtiter plates as previously described (15). RBCs were obtained from C57BL/6 mice and either used for the experiment as native RBCs or incubated with 0.5% bromelain (Acros, Geel, Belgium) solution for 10 min at 37°C. Cells were then washed and stored in PBS at 4°C. For ELISA, 2 × 10⁵ cells per well were plated. Native LDL and OxLDL were prepared as described (15). BSA was from Sigma-Aldrich (St. Louis, MO), and PC-BSA was from Biosearch Technologies (Petaluma, CA). XQ11 binding to various Ags was detected using anti-His alkaline phosphatase conjugated Ab from Sigma-Aldrich.

Peritoneal B-1a, B-1b, and splenic B-1a cells, collectively referred to as B-1 cells, and peritoneal B-2, splenic MZ, and FO cells, collectively referred to as conventional B cells, were isolated from female C57BL/6 mice at 12 wk of age. B cell subsets were isolated by flow cytometry using established criteria as defined in 2Materials and Methods (Fig. 1), and the purity of the isolated pools averaged >99% in postsort analysis. Utilizing an unbiased 5′-RACE amplification strategy, we generated cDNAs of the IGHV and used these to perform massively parallel sequencing, with a yield of 711,760 total reads after quality trimming and joining of reads, as described in 2Materials and Methods. From these, we obtained 378,740 functional IGHV sequences, an average of 63,123 sequences per B cell subset (range, 48,579–81,375; Fig. 2). We analyzed the repertoire of these subsets on the level of total RNA transcripts as an estimate of the functional repertoire. The terminology used in this report is according to the IMGT nomenclature (40, 41).

Major V-D-J gene expression for IGHV sequences of each B cell subset is shown in Fig. 2, and more detailed analyses of VJ rearrangements can be found in Supplemental Figs. 1, 2, 3. Genes of the V_H1 family (the biggest V family, consisting of 53 functional genes) were the most commonly used V_H genes in all B cell subsets besides B-1a cells. They ranged from 59.6% in B-1b cells to 84.8% in FO cells, whereas they only comprised 15.8 and 46.2% of peritoneal and splenic B-1a sequences, respectively. B-1a cells expressed V_H11 genes substantially more frequently than did other B cell subsets (58.1 and 25.2% of peritoneal and splenic B-1a sequences, respectively) (Fig. 2A). All of these were of the V_H11-2 gene segment (Supplemental Figs. 1, 2). B-1b cells also expressed V_H11 transcripts, but to a relatively limited extent (3.9% of B-1b sequences). Additionally, 8.7% of peritoneal B-1a sequences contained V_H12, compared with 3.2% of B-1b and 1.0% of splenic B-1a sequences, respectively. Of note, conventional B cells showed little to no expression of V_H11 or V_H12 segments (<0.2% in all subsets). This confirms prior studies showing preferential expression of V_H11 and V_H12 in B-1 cells (23, 25, 42). In our analysis, however, we show that expression of both of these prototypic B-1 cell V_H genes was significantly higher in B-1a cells than in B-1b cells.

As shown in Fig. 2B, in most B cell populations, D1 and the D2 family were the most common D genes and were expressed to a similar extent (33.5–45.7% of transcripts). However, peritoneal B-1a cells exhibited preferential expression of genes from the D2 family (62.4%) with an accompanying decreased use of D1 (19.0%). In contrast, D3 and D4 genes were present only to a limited extent (5.9–11.8%) in all subsets. The increase in D2 family genes in peritoneal B-1a cells was mostly driven by higher expression of D2-1 and D2-5 with decreased expression of other D2 genes compared with other sequences. Less prominently, this pattern was also present in splenic B-1a cells (Supplemental Fig. 4).

As shown in Fig. 2C, conventional B cells and B-1b cells also expressed J genes in similar proportions, using mainly J_H2 (32.0–39.8%) and comparable proportions of J_H1 (13.9–24.9%), J_H3 (20.2–21.9%), and J_H4 (21.3–30.1%). However, peritoneal and, to a lesser extent, splenic B-1a cells expressed J_H1 more frequently than did other subsets (72.4 and 39.6% respectively), with an accompanying decrease in expression of other J_H genes. Interestingly, J_H1 was previously noted to be the most commonly expressed J_H gene among neonatal cells, further indicating a possible fetal origin of sequences with this gene (43).

We used the Kyte–Doolittle scale to compare average CDR3 hydrophobicity. All B-1 subsets expressed CDR3s with a lower average hydrophobicity (−0.18 ± 0.001 to −0.21 ± 0.001) compared with conventional B cell sequences, whose means also were in a relatively narrow range for all subsets, but at a higher hydrophobicity (−0.09 ± 0.002 to −0.12 ± 0.001) (Fig. 3A). Of note, Vale et al. (26) reported similar findings when analyzing V_H5 sequences in BALB/c mice. Different studies by the same group found that forcibly expressing a highly charged CDR3 led to decreases in B-1 cell populations with an increase in peritoneal B-2 and splenic MZ cells (44). In contrast to this increase in conventional B cells with expression of a highly charged CDR3, we find that in our studies these subsets actually express sequences with higher CDR3 hydrophobicity compared with B-1 cells.

CDR3s were shorter in B-1 and MZ cells (32.05 ± 0.03 to 33.59 ± 0.02 nt) compared with peritoneal B-2 and splenic FO cells (35.33 ± 0.04 and 36.27 ± 0.04 nt, respectively) (Fig. 3B). A similar trend was seen in regard to D length. Peritoneal B-1a CDR3s contained the shortest (7.85 ± 0.01 nt) and splenic FO cells the longest D segments (10.85 ± 0.02 and 11.24 ± 0.02 nt). B-1b (8.95 ± 0.02 nt), splenic B-1a (9.38 ± 0.01 nt), MZ (10.26 ± 0.02 nt), and peritoneal B-2 (10.85 ± 0.02 nt) D lengths fell in between these subsets (Fig. 3C).

We categorized all transcripts with an identifiable D segment as either N+ or N−, based on the presence of N additions at either the V-D or D-J junction. In the peritoneal cavity, 60.6% of peritoneal B-1a transcripts were N−, much more than in any other subset, suggesting a much greater contribution of fetal development to this compartment. In contrast, only 15.1 and 25.6% of B-1b and splenic B-1a transcripts, respectively, were without N additions. N− sequences were prevalent only to a very limited extent in all conventional B cell subsets (4.3–7.6%) (Fig. 3D).

We extracted N− and N+ sequences from the database and analyzed them individually for each subset. Most notably, the fraction of V_H1 family genes was relatively low in both peritoneal and splenic B-1a N− sequences, whereas it showed a higher proportion of V_H11 sequences (Fig. 4A). Furthermore, all B-1 cell N− transcripts showed substantially lower average CDR3 hydrophobicity in a tight range (−0.268 ± 0.005 to −0.292 ± 0.001) compared with N− sequences of conventional B cells and N+ sequences of all cell subsets (−0.086 ± 0.003 to −0.170 ± 0.001) (Fig. 4B). Average CDR3 lengths were similar in N− and N+ B-1a transcripts (32.50 ± 0.01 to 34.13 ± 0.04) and were also longer than those of N− sequences of other subsets (27.86 ± 0.11 to 29.36 ± 0.14). Notably, B-2 and FO N+ sequences had the longest average CDR3 length (36.27 ± 0.03 and 36.91 ± 0.04, respectively) (Fig. 4C).

We examined the frequency of the V_H and J_H rearrangements in the analyzed B cell subsets. Both peritoneal and splenic B-1a cells showed distinct rearrangements compared with conventional B cells, expressing a high number of V_H11/J_H1 recombinations, with peritoneal B-1a cells also using a considerable number of V_H12/J_H1 sequences. These findings highlight the restricted repertoire in both peritoneal and splenic B-1a cells. In contrast, peritoneal B-1b and B-2 transcripts, as well as splenic MZ and FO sequences, displayed a much more diverse set of VJ rearrangements (Supplemental Figs. 1, 2). Because V_H11 and V_H12 sequences account for >50% of the B-1a transcripts, we also present the data for peritoneal and splenic B-1a cells in a form that excludes V_H11 and V_H12 transcripts to better visualize all other clones (Supplemental Fig. 3).

V_H11 and V_H12 have been previously described as prototypical V genes of B-1a cells (45–47). Transgenic mice for either of these V genes have a marked increase of their B-1 cell population and of phosphatidylcholine-binding Abs (48). In contrast, conventional B cells showed very little to no expression of these V genes in our studies. We further investigated the properties these specific V genes have in B-1a cells, which appeared to be the most unique of all subsets. Of note, all of the V_H11/V_H12 sequences in our database contained the genes V_H11-2 and V_H12-3, respectively (Supplemental Figs. 1, 2).

Overall, D and J expression were very similar between peritoneal and splenic B-1a subsets, indicating that a substantial number of sequences expressed might be the same in both subsets. Whereas B-1a sequences used different DJ genes than other B cell subsets, sequences that contained neither V_H11 nor V_H12 showed comparable D gene rearrangement to conventional B cells or B-1b cells (Fig. 5A). The expression of D1 and D2 in non-V_H11/V_H12 sequences in B-1a cells was similar to those in conventional B cells. However, rearrangements with V_H11 and V_H12 appeared to use a very restricted set of both D and J segments. In V_H11 sequences, there was a striking predominance of rearrangement with D2 genes, and to a smaller extent also with D1, but almost never rearranged with D3 or D4 genes. Furthermore, we found that V_H12 sequences preferentially contained D2 or D3. In fact, D3 was the most common D gene in V_H12 sequences, but was found to rearrange only rarely with other V_H genes (Fig. 5A).

There was a similar restricted pattern in J usage of V_H11 or V_H12 sequences (Fig. 5B). Whereas non-V_H11/V_H12 sequences showed a slight predominance of J_H2 and almost equal usage of the remaining three J genes (again similar to conventional B cells and B-1b cells), the rearrangement of V_H11 and V_H12 with J genes was very restricted: V_H11 almost exclusively and V_H12 predominately rearranged with J_H1. Interestingly, 28.7% of V_H12 sequences rearranged with J_H4 in splenic B-1a cells, whereas peritoneal B-1a V_H12 only rarely recombined with J_H4 (0.4%), indicating different selection processes or developmental origin (Fig. 5B).

Having shown above that substantial proportions of B-1a sequences have no N additions, we next investigated whether there is a difference in N additions between sequences that have different V genes. In fact, we found that most V_H11 sequences did not contain N additions and this was true for both peritoneal and splenic B-1a cells, whereas for V_H12 sequences, the proportion of N− sequences was lower and similar to non-V_H11/V_H12 sequences (Fig. 6A). V_H11 and V_H12 sequences were also much less hydrophobic than other sequences (Fig. 6B). Interestingly, peritoneal B-1a CDR3s that did not contain V_H11 or V_H12 were the most hydrophobic subset in our analysis, even more hydrophobic than all N+/N− subsets in Fig. 4B. This increased hydrophobicity in non-V_H11/V_H12 sequences was not seen in splenic B-1a CDR3s whose hydrophobicity was in a range comparable to most conventional B cell (both N+ and N−) and B-1 N+ sequences, as shown in Fig. 4B. Furthermore, CDR3s of V_H12 sequences were longer than the CDR3s of V_H11 or non-V11/V12 sequences and in a similar range as B-2/FO cells (Fig. 6C). Interestingly, D length was the shortest in V_H11 sequences, whereas the D segment of V_H12 and non-V11/V12 sequences was ∼1 aa longer (Fig. 6D).

To determine whether the rather unique properties of V_H11 and V_H12 sequences might be caused by the high expression of very few Ab sequences, we also analyzed sequences based on clonotypes instead of total sequences. One “IMGT clonotype (AA)” is defined as using the same VDJ genes and having the same CDR3 amino acid sequence. As opposed to total sequence numbers, in this analysis, each unique clonotype was regarded as occurring only once in the analysis. We found similar rearrangements of V_H11 and V_H12 clonotypes with D and J genes, and similar CDR3 properties to those found on the analysis of total sequences, confirming that the preferential recombinations and qualities described above are unlikely to be skewed by the high expression of only a few clones of Abs (Supplemental Fig. 5).

During rearrangement, not only are nucleotides inserted, but they can also be removed from the end of each gene segment by exonucleases. We found that peritoneal B-1a sequences had less loss of J_H nucleotides than did other B cell subsets, similar to a previous report by Kantor et al. (25) in peritoneal B-1a sequences from BALB/c mice. This could be explained by the high fraction of V_H11 and V_H12 sequences that on average lost only 0.3 and 1.3 nucleotides of their J gene, respectively. These numbers were very similar in both peritoneal and splenic B-1a cells (Fig. 7). Alternatively, peritoneal and splenic B-1a sequences had on average slightly more nucleotides excised from the 5′ D end. This is an attribute of their V_H11 sequences, which had a higher 5′ D loss than any other sequence subset analyzed. Remarkably, however, almost all V_H11 gene segments were completely preserved with an average loss of only 0.04 and 0.07 (peritoneal and splenic, respectively) nucleotides. Alternatively, V_H12 sequences lost an average of six nucleotides, considerably more than any other group analyzed (Fig. 7).

B-1 cells have been thought to have a unique repertoire of IgM Abs. We thus investigated whether B-1 cells, and B-1a in particular, would contain distinct CDR3 amino acid sequences and, in general, to what extent the expressed sequences overlap between B cell populations. The Ab Mining Toolbox was used to extract CDR3 sequences from the database. For this analysis, we included only CDR3 amino acid sequences that occurred at least five times in the whole dataset across all subsets to reduce the probability that a given sequence might be a result of amplification or sequencing error. By this definition, there were 8820 unique CDR3s that accounted for 71% of the total functional reads (267,840 of 378,740 total reads).

Fig. 8 presents a series of Venn diagrams that examine the shared repertoire of unique CDR3 sequences in the different B cell subsets. In these diagrams, the unique and shared repertoire of CDR3 sequences for each B cell subset in both the peritoneum and spleen are compared. In each horizontal line, the pink color identifies the unique CDR3 sequences of the indicated cell type on the left margin of that line, whereas the blue color identifies unique sequences of the cell type in the header of that column. The green field indicates the shared CDR3s of these two subpopulations. A shared sequence was defined as one expressed at least once in each of the subsets being compared.

It has been assumed that the B-1a population of the peritoneum represents the reservoir of unique B-1a cells, which then migrate to the spleen to transform to plasma cells and secrete IgM as needed. Indeed, 59.6% of the peritoneal B-1a sequences were found in splenic B-1a cells as well. However, our analysis shows that the repertoire of the splenic B-1a compartment exceeded that of peritoneal B-1a cells, such that splenic B-1a cells had a 5-fold larger repertoire of unique CDR3 transcripts compared with peritoneal B-1a cells (1882 versus 376 CDR3s, respectively), with 554 additional shared CDR3s between the two subsets. In a separate analysis not shown, even when one looked at the number of unique CDR3s that occurred only one to three times, the splenic B-1a transcripts were still substantially more numerous. This suggests that the spleen could be a reservoir of unique B-1a cells, perhaps being seeded not only from the various tissue compartments that B-1a cells are thought to reside in, such as peritoneal and pleural cavities and the omentum, but also the bone marrow (2, 49, 50). Additionally, peritoneal B-1a cells also shared 31.9% of their CDR3s with B-1b cells and 26.5% with MZ cells, but only small proportions with B-2 and FO cells (8.3 and 13.7%, respectively).

The repertoire of splenic B-1a cells, alternatively, overlapped to a modest degree with MZ cells (27.8%) and to a smaller amount with B-1b (18.1%) and FO cells (14.0%). However, again the overlap with B-2 cells was much lower (8.4%), similar to the small overlap between peritoneal B-1a and B-2 cells.

The B-1b CDR3 repertoire, alternatively, shared sequences with B-1a cells (both peritoneal and splenic) and MZ cells (13.3, 19.8, and 20.0%, respectively). However, there was little overlap of B-1b cells with B-2 (8.3%) or FO cells (9.1%). Another notable finding was that the CDR3 repertoire of peritoneal B-2 cells was the most unique compared with the other B cell subsets studied. The B-2 cell repertoire shared 20.2% of its sequences with MZ, but only 5–11% with B-1 cell subsets and, interestingly, only 12.1% with those of FO cells.

The analyses above were performed using all sequences. To better visualize clonotypes that are developmentally related among the B cell subsets, we performed a similar type analysis using N− and N+ transcripts separately (Fig. 9). In these analyses, most B-1a N− sequences were shared between peritoneal and splenic B-1a cells, but the splenic B-1a subset contained many more N+ sequences than did peritoneal B-1a cells, consistent with more postnatal development of splenic B-1a cells. Interestingly, there is relatively high overlap of N− sequences between B-1 and MZ cells (30.6% of peritoneal B-1a, 30.6% of B-1b, and 33.1% of splenic B-1a, respectively). However, there was little overlap between B-1 cells and conventional B cells; that is, only 6–13% of the B-1 N− sequences were found in the B-2 or FO repertoires. This could indicate prenatal selection processes that are similar for B-1 and MZ cells, but do not include B-2 and FO cells. Complicating this picture, a substantial number of conventional B cell N− sequences were shared with MZ cells (39.2% of B-2 and 55.0% of FO, respectively), thereby making the N− repertoire of MZ cells a hybrid of B-1 and conventional B cell repertoires (Fig. 9A). On the N+ level, besides B-1a cells in the peritoneal cavity and spleen, there was also notable overlap in the repertoires of MZ cells with splenic B-1a cells and FO cells (23.1 and 25.8% of MZ sequences, respectively). MZ cells shared few N+ sequences with the peritoneal subsets (7–14% of MZ sequences), suggesting a postnatal separation of the peritoneal and splenic B cell repertoires (Fig. 9B).

In Table I, we show a list of the most common peptide CDR3s in each subset. Remarkably, we found that two specific peptide CDR3 sequences, which differ only in 1 aa (CMRYGNYWYFVW, CMRYSNYWYFDVW), made up a large proportion of total B-1a CDR3 sequences. Collectively, these sequences accounted for 43.4% of the total number of peritoneal B-1a and 15.9% of splenic B-1a peptide sequences, and thus they are a major contributor of the V_H11 pool in B-1a cells. They were the most common B-1a transcripts in each of the three independent B cell sorts, as well as in a pilot experiment on peritoneal B-1a cells performed on 10 female C57BL/6 mice bred in our vivarium, as opposed to the purchased mice used in the present studies. Both sequences are known V11-D2-J1 sequences that have been previously described and studied as prototypic B-1 cell Abs binding to BrRBCs and phosphatidylcholine (21, 42). These same sequences were also recently reported by Yang et al. (30) as consistently among the top 10 recurring CDR3 sequences in splenic B-1 samples in both specific pathogen-free and germ-free mice.

To characterize the binding specificity of the most common CDR3 sequence, CMRYGNYWYFVW, as described in 2Materials and Methods, we generated a cDNA for an scFv Ab using the original IGHV bearing this CDR3, paired with the L chain V region from the original hybridoma as described by Reininger et al. (21), which we termed XQ11-scFv (Fig. 10A). Its cDNA was transfected into HEK293 cells and the binding characteristics of XQ11-scFv secreted into the culture medium were characterized by chemiluminescent assay. In earlier studies, the IgM Ab had been shown to bind to BrRBCs and it was postulated that this was to a cryptic epitope that might be represented by binding to phosphatidylcholine present on liposomes (21, 24). As shown in Fig. 10B, XQ11-scFv binds to murine BrRBCs as originally reported, but we now show that it binds specifically to the PC epitope present on BSA, that is, to the headgroup of phosphatidylcholine. In contrast, XQ11 does not bind to unmodified BSA. Importantly, XQ11 also binds prominently to OxLDL, a rich source of oxidized phospholipid (OxPL), which prominently displays such PC. The binding of XQ11 to OxLDL and OxPL is analogous to the binding of the prototypic IgM E06 (or IgA T15) NAb, which binds the PC of OxLDL or the PC present on the cell wall of S. pneumoniae (14, 15). We have previously described PC in this context as an OSE and shown that IgM NAbs to PC attenuate atherosclerosis development (15, 51). Of interest, XQ11-scFv also appears to bind to a limited extent to the starting preparation of murine RBCs not treated with bromelain, perhaps consistent with the concept that RBCs steadily accumulate OSEs with aging (52).

In this study, we used massively parallel sequencing to define the complete IGHV repertoire of peritoneal (B-1a, B-1b, and B-2) and splenic (B-1a, MZ, and FO) B cell subsets from female C57BL/6 mice 3 mo of age. B-1 cells in particular are a unique subset of lymphocytes whose repertoire is thought to have developed through natural selection and whose Abs have important homeostatic and housekeeping functions. We have suggested that in particular a substantial subset of these IgM NAbs are directed to OSEs and not only provide homeostasis to OSEs found on OxLDL but also on apoptotic cells and microvesicles, which otherwise would be both immunogenic and proinflammatory (reviewed in Ref. 13). We have also suggested that because such innate IgM represent soluble PRRs, their selection has been additionally influenced to provide homeostasis against PAMPs of pathogens. A prototypic example of such an IgM NAb is the B-1 cell derived T15/E06 idiotype Ab that was first identified for its binding to PC on the cell wall of S. pneumoniae, and which provides optimal protection to mice against lethal infection with S. pneumoniae infection (20, 53). Additionally, we have shown that E06 provides homeostasis by neutralizing inflammatory properties of microvesicles and apoptotic cells bearing PC containing OxPL (12, 54), and it restricts atherosclerosis by both inhibiting uptake of OxLDL by macrophages and by preventing inflammatory properties of OxPL (14, 15, 55). In a similar manner, we have shown that an even greater number of both murine and human cord blood IgM NAbs bind to other OSEs, and in particular malondialdehyde-type adducts (12, 13, 51). Of course, it has been long known that B-1 cell Abs provide the first line of protection against many bacterial and viral pathogens (7, 56, 57). Furthermore, it has been reported that the titers of such innate IgM NAbs decline with age, and could thus contribute to a general weakening of innate immune responses with aging (58, 59). Thus, knowing the baseline repertoire and understanding how the B-1 cell Ab repertoire changes with aging and with disease can give insight into beneficial functions of these Abs that could eventually be used in humans by passive or active immunization strategies.

In a recent elegant publication, Yang et al. (30) traced the early fate of B-1a cells and demonstrated that B-1a represent a B cell lineage whose IGHV recombinations represent fetal cells, as their IGHV rearrangements are primarily without N additions typical of postnatal TdT activity. They found a paucity of N additions in splenic B-1a cells up to 6 d of life, which slowly increases until weaning at ∼3 wk. Thereafter, the proportion of splenic B-1a cells with N additions appears to stabilize with ∼80% showing multiple insertions. Our data are consistent with these observations, as 74% of the splenic B-1a sequences contained N additions. Interestingly, however, in contrast to the splenic B-1a population, we found that ∼60% of all peritoneal B-1a rearrangements still lacked N additions at 3 mo of age, and it was particularly striking that among the V_H11 sequences that dominated the B-1a peritoneal repertoire at that age, ∼90% of the sequences lack N additions, consistent with a fetal origin of these cells. It might thus be surmised that the peritoneal B-1a population is more isolated developmentally than the splenic B-1a pool. Furthermore, Vale et al. (26) previously demonstrated that very few V_H5 (V_H7183) sequences from the neonatal liver have N additions, whereas almost all sequences obtained from adult bone marrow contained N insertions. Additionally, Holodick et al. (60) showed recently that B-1a cells derived from adult bone marrow progenitors in fact had a higher number of N additions compared with fetal liver progenitor cells. Also note that the total unique expressed repertoire of the peritoneal B-1a cells was only 930 peptide sequences in this analysis, which is the smallest number of unique sequences among all the B cell subsets studied. This is even more striking when it is realized that two sequences alone accounted for 43% of all expressed sequences, as noted below. This restricted repertoire, which is nearly 3-fold smaller than that of the splenic B-1a population, plus the largely prenatal origin of these cells, is consistent with a very protected and isolated set of “innate” soluble PRRs that are the products of natural selection mediated by both innate DAMPs and exogenous PAMPs and are tasked with maintenance of homeostatic functions against “conserved” Ags.

As noted above, it is also of considerable interest that B-1a cells in the spleen have a larger repertoire than peritoneal B-1a sequences, and ∼75% of these sequences contain N additions, quite different from what is seen with peritoneal B-1a cells. This suggests that at least by 3 mo of age, most of the splenic B-1a cell population is postnatal in origin, consistent with findings of Yang et al. (30). One might thus speculate that the spleen could act as a “melting pot” of prenatal B-1a cells residing in the peritoneal cavity and postnatal B-1a cells, the latter of which might be greatly enriched from other sites such as the bone marrow (50, 60). B-1a cells from the peritoneum migrate to the spleen when activated (11), whereas the vascular architecture of the spleen could enhance the possibility of bone marrow–derived B-1a cells to take up residence in the splenic pulp. Further studies tracking B-1a cells from the fetal liver and bone marrow are needed to confirm these hypotheses.

Yang et al. (30) used a deep sequencing strategy to report on the IGHV repertoire of various B cell subsets of varying ages in C57BL/6 mice, including B-1a cells, but their studies did not include B-1b cells. Thus, to our knowledge, the current results are the first to include a full IGHV characterization of all peritoneal and splenic B cell subsets including the B-1b subset. It has long been discussed how closely B-1a and B-1b cells are related, as their surface markers only differ in expression of CD5. In striking contrast to the peritoneal B-1a population, we found in the present study that ∼85% of the IGHV sequences from peritoneal B-1b cells from 3-mo-old mice had N additions, and thus are likely to have developed postnatally. Additionally, the V_H11 sequences that were expressed so prominently in peritoneal B-1a cells were very rarely found in B-1b sequences, further distinguishing B-1a from B-1b populations. Indeed, a recent study by Ghosn et al. (9) showed that a single adult hematopoietic stem cell could replenish the B-1b, but not the B-1a, cell population. Further evidence of a distinct origin of B-1a and B-1b may also be seen in the Venn diagrams of Fig. 8, which compare unique sequences in different B cell subsets. Note that the shared repertoire of peritoneal B-1a and B-1b CDR3 peptide sequences is very small. From among 2229 unique B-1b sequences, there are only 297 (13.3%) shared sequences, further distinguishing the unique origins of these two seemingly closely related B-1 cell subsets. Despite this difference in the overall functional repertoire, it appears that both B-1a and B-1b cell compartments secrete IgM that are anti-inflammatory and prevent atherosclerosis development and progression. Indeed, we and others have shown that B-1a cells provide atheroprotective IgM (13) and, more recently, that B-1b Abs also decrease atherosclerosis and protect against obesity-associated inflammation (61, 62). Interestingly, the overlap of the B-1b repertoire with MZ sequences was larger than with peritoneal B-1a and at about the same level as with splenic B-1a sequences. Splenic B-1a cells, alternatively, also shared a significant number of sequences with MZ cells. However, although MZ cells also share a significant portion of their repertoire with splenic B-1a cells, interestingly, this was less so with peritoneal B-1a cells. On a more detailed analysis, MZ cells share substantial parts of their N− repertoire with all other subsets whereas there was little overlap in N− sequences between B-1 and B-2/FO cells. Therefore, MZ cells could be regarded as the cell type that “bridges the gap” between B-1 and conventional cells on a repertoire level. However, this interaction did not appear as strong on the N+ level, suggesting that these cells have more similarities in repertoire development prenatally, whereas the repertoires diverge postnatally.

Other B cell subsets thought to be closely related are peritoneal B-2 cells and splenic FO cells. Interestingly, as shown in the Venn diagrams of Fig. 8, the shared repertoire between peritoneal B-2 and splenic FO cells is very small and B-2 cells seem to be very isolated in their repertoire. In a prior study, B-2 cells transplanted into peritoneal cavities of Rag^−/− mice led them to acquire a B-1b like phenotype, suggesting that there might be some form of relationship or interchange between them (63). Notably, this could not be confirmed on a repertoire base in our mice, as we found very little overlap between peritoneal B-1b and B-2 cells. Of note, the highest numbers of sequences B-2 cells share with any other subset are MZ cells, both on the level of N− and N+ sequences. Thus, one might speculate that B-2 cells represent MZ cells in a different environment, rather than FO cells. However, even this amount was low (∼20% of B-2 sequences), and B-2 cells notably lack the overlap of their repertoire with B-1 cells that MZ cells show, as noted above.

As suggested by the literature and confirmed by our studies, V_H11 and V_H12 are V genes commonly expressed by B-1a cells, but not conventional B cells. They showed a remarkable specificity for rearrangement with certain D and J genes that were used to a lesser extent in all other sequences, indicating that these are conserved sequences that need a very specific rearrangement to lead to the formation of a mature B cell. The high number of transcripts without N additions in V_H11 sequences (both on the level of total sequences and clonotypic analysis) further suggests that prenatal development of V_H11 is predominant in B-1a cells. Interestingly, ∼75% of V_H12 transcripts contained N additions, indicating that this V gene is also selected postnatally to a much greater extent. In accordance to these findings, Gu et al. (43) showed that J_H1 (which is the predominant J gene of V_H11 sequences) is the most common J gene in neonatal pre-B cells, but not later in life. In line with these observations, we found that most N− sequences in B-1a cells (∼85–95%) contained this J gene, whereas B-1a N+ sequences contained this V gene at a much lower frequency (∼25–40%).

In our analysis, B-1 cell CDR3s had a lower hydrophobicity compared with conventional B cells, mainly driven by their N− sequences, consistent with a previous report by Vale et al. (26) of peritoneal V_H5 sequences in BALB/c mice. Interestingly, the same group showed that forced expression of a highly charged CDR3 led to significantly decreased B-1a cell numbers and a slight increase in peritoneal B-2 and splenic MZ cells (44). This is in apparent contrast to our study in which we found lower hydrophobicity of B-1 cell CDR3s compared with conventional B cells. However, it has been shown that C57BL/6 mice (used in our study) and BALB/c mice (used in the cited reports) regulate highly hydrophobic or highly charged CDR3s differently, which might at least partially account for these results (64). Of the B-1a sequences, V_H11/V_H12 sequences had a substantially lower average CDR3 hydrophobicity compared with other sequences. Non-V_H11/V_H12 sequences of B-1a cells in fact had an average CDR3 hydrophobicity in a range much closer to N+ sequences or the N− sequences of conventional cells, indicating that the V_H11 and V_H12 sequences contribute strongly in giving B-1a cells the unique repertoire properties found in this study.

Notably, many of the most highly expressed sequences in the report of Yang et al. (30) contained short CDR3s of five or less amino acids, almost exclusively using J_H2 genes. However, we did not make the same observation in our data where only 581 (0.15%) of all transcripts had a CDR3 length of ≤5 aa (11 J_H1, 215 J_H2, 319 J_H3, and 36 J_H4 sequences, respectively). We used a 5′-RACE amplification method of RNA as opposed to multiplex PCR used by Yang et al. but in theory this should not have any influence on these findings. Similar to Yang et al., we used the same strain of mice, but only studied a narrow age range. Also, as noted below, unlike the findings of Yang et al., there was only infrequent expression of the E06/T15 idiotype in our mice. Thus, one might speculate that these differences and others in Ab expression between our data and those of others is in part due to differences in strains and ages of mice studied. Additionally, different Ag exposures of the mouse colonies could have caused further disparities in the Ab repertoire.

One of the most striking and unique qualities of B-1a cells in both the peritoneal cavity and the spleen is the expression of substantial numbers of V_H11 sequences, which were very rarely found in other B cell subsets, including B-1b. This is consistent with previous reports that have found high numbers of these sequences in hybridomas generated from B-1a cells (23). The proportion of V_H11 sequences was higher than in studies using BALB/c mice for analysis. However, this is in line with Seidl et al. (42), who found much higher proportions of V_H11 in phosphatidylcholine-binding B-1 cells of C57BL/6 mice compared with BALB/c mice. This V gene has also been reported to primarily rearrange with J_H1, a finding that we can also confirm, and is commonly found in Abs that bind BrRBCs and phosphatidylcholine (24, 65). Yang et al. (30) in fact reported that the expression of V_H11 in splenic B-1a cell sequences increases substantially at 2–6 mo of age compared with younger mice. Strikingly, in our 3-mo-old mice, we identified two CDR3 sequences that were the most dominant sequences in B-1a cells both in the peritoneal cavity and in the spleen, accounting for 43.4 and 15.9% of all B-1a CDR3s, respectively. These exact CDR3 amino acid sequences have also been previously reported to bind BrRBCs and phosphatidylcholine present on liposomes, making them prototypical B-1a sequences (21). They were also highly expressed in all of the B-1a populations reported by Yang et al. (30) as well, including in mice reared in a germ-free environment. These sequences were the most common B-1a transcripts not only in the mice used in this study, but also in a pilot experiment performed only on peritoneal cells, pooled from 10 female C57BL/6 mice that were bred in our vivarium. Their overwhelming dominance in B-1a cells warrants further studies about their target epitopes and their possible important physiological functions in vivo.

Our laboratory is especially interested in the role of B-1 cells in generating IgM Abs that target what we have termed OSEs. We discovered that up to 20–30% of all IgM in noninfected mice, and in human umbilical cord blood, bind to various OSEs, such as PC, OxPL, malondialdehyde, malondialdehyde acetaldehyde adducts, 4-hydroxynonenal, and others (51). We have observed that all of these bind not only to a variety of oxidatively modified structures, but to apoptotic cells as well. We have thus suggested that they are part of the innate immune response providing homeostasis against oxidative stress. A major purpose in undertaking this study was to provide a basal library of the B-1 cell IGHV sequences at steady-state. In further studies, we are exploring the impact on these sequences to perturbations of oxidative stress, such as high-fat high-cholesterol feeding, to determine which clones expand, and to annotate the epitopes to which such clones bind.

Surprisingly, the classic E06/T15 idiotype occurred only infrequently in our analysis, most commonly in MZ cells (45 transcripts) and rarely in splenic B-1a and peritoneal B-1b cells (1 and 4 transcripts, respectively). However, we show that the most frequently observed CDR3 amino acid sequence CMRYGNYWYFVW, which together with a sequence that differs by only one amino acid, accounted for an astounding 43% of all unique B-1 peritoneal sequences, encodes for an OSE binding Ab. As noted above, this Ab previously was noted to bind to a “cryptic site” on BrRBCs and to also bind phosphatidylcholine present on liposomes (21, 24). We have prepared XQ11-scFV containing the H and L chain segments described in the original hybridomas, and while confirming that it does bind BrRBCs, we found it is an example of an Ab targeting an OSE, that is, the PC of OxLDL and PC-BSA, which is also bound by T15/E06 (15). We also have preliminary data that it binds to lipoprotein(a), a lipoprotein that is known to be greatly enriched in OxPL (data not shown) (66). It is of interest that it also bound modestly to “native” RBCs. RBCs are known to be under increased oxidative stress as they age, and it is likely that they accumulate OSEs as they age (52). The binding of IgM targeting OSEs, such as OxPL, may in time mediate clearance of the aged RBCs. Furthermore, this Ab, similar to other Abs that bind OSEs, may play generalized roles in maintaining homeostasis against oxidative stress and in providing protection against inflammation and atherosclerosis (15, 51, 67). Further studies are needed to test these suggestions.

In summary, we used an unbiased 5′-RACE amplification strategy with massively parallel sequencing to define the IGHV functional repertoire of all murine peritoneal and splenic B cells. We could analyze a total of ∼379,000 productive transcripts from six B cell subsets including peritoneal B-1a, B-1b, and B-2 cells, and splenic B-1a, MZ, and FO cells. An early analysis has revealed that B-1a cell sequences differed remarkably from other subsets, adding further evidence for a distinct selection process. Furthermore, we found that the peritoneal B-1a compartment also appeared to be a unique compartment derived predominantly prenatally, and quite distinct from B-1b cells, which appear to be separate and derived postnatally. Interestingly, peritoneal B-2 cells appear to be the most isolated subset based on their CDR3 repertoire. Massively parallel sequencing is a powerful tool to analyze the B cell repertoire and can be used in further studies to investigate how the repertoire changes during development of environmental influences or disease states, and with aging, giving further insight into these exceptional cells and their roles in innate and acquired immunity.

X.Q. and J.L.W. have patents held by the University of California San Diego related to use of oxidation-specific epitopes. J.L.W. consults for several pharmaceutical houses unrelated to the topic of the manuscript. X.Q. and J.L.W. are coinventors and receive royalties from patents owned by the University of California San Diego on oxidation-specific Abs. J.L.W. is a consultant for Ionis Pharmaceuticals. The other authors have no financial conflicts of interest.