Skewed primary Igκ repertoire and V-J joining in C57BL/6 mice: implications for recombination accessibility and receptor editing

Abstract

Previous estimates of the diversity of the mouse antibody repertoire have been based on fragmentary data owing to many technical limitations, in particular the many samples necessary to provide adequate coverage. Here, we used 5' coding end amplification of Igκ mRNAs from bone marrow, splenic and lymph node B cells of C57BL/6 mice combined with amplicon pyrosequencing to assess the functional and non-functional Vκ repertoire. To evaluate the potential effects of receptor editing, we also compared V/J associations and usage in bone marrows of mouse mutants under constitutive negative selection or an altered ability to undergo secondary recombination. In order to focus on preimmune B cells, our cell sorting strategy excluded memory B cells and plasma cells. Analysis of approximately 90 Mbp, representing >250,000 individual transcripts from 59 mice, revealed that 101 distinct functionalVκ genes are used, but at frequencies ranging from ~.001% to ~10%.Usage of sevenVκgenes made up over 40% of the repertoire. A small class of transcripts from apparently nonfunctional Vκ genes was found, as were occasional transcripts from several apparently functional genes that carry aberrant recombination signals. Of 404 potential V-J combinations (101 Vκs X 4 Jκs), 398 (98.5%) were found at least once in our sample. For most Vκ transcripts, all Jκs were used, but V-J association biases were common.Usage patterns were remarkably stable in different selective conditions.Overall, the primary κ repertoire is highly skewedby preferred rearrangements, limiting antibody diversity, but potentially facilitating receptor editing.

Introduction

Immunoglobulin genes encode antibodies vital to adaptive immunity. In B cell development, antibody heavy and light (L) chain genes are assembled independently at sequential developmental stages by recombination of the Ighfollowed byIgκ (or Igλ) gene loci, respectively. Igκ and Igλ loci are encoded on different chromosomes in vertebrates and recombine in cis to assemble variable (V) and joining (J) gene segments. In the C57BL/6 strain mouse, the genomic sequences of the antibody loci have been assembled and analyzed(1–13).The Igκlocus occupies >3 megabase of DNA, carrying over 100 potentially functional V elements and 4 functional J elements (designated throughout as Vκ and Jκ, respectively) (1–7, 9, 14–17).

Although it is widely assumed that V-J association is relatively random, consistent with a role in diversification of the antibody repertoire (18), the overall biases in Vκ association with various Jκs are not known. Moreover, the κ locus plays an important role in receptor selection. Vκ usage and joining to Jκ elements are influenced by a feedback process that is regulated by the nature of the L-chain generated. As a consequence, rearrangements resulting in out-of-frame, underexpressed, or autoreactive receptors fail to terminate recombination, frequently leading to secondary rearrangements or receptor editing(reviewed in 19). The murineIgκ locus favors primary rearrangements to upstream Jκs (Jκ1 and Jκ2)and secondary rearrangements to downstream Jκ4 and Jκ5 (Jκ3 is a pseudogene)(14, 20–23). Adding to the complexity is the fact that over half of Vκs are in the opposite transcriptional orientation to the Jκ-Cκ cluster(9, here we find 64%).As a result, some primary V-J rearrangements lead to deletion of intervening DNA, preventing the subsequent use of potentially functional elements in between, and limiting the repertoire available for secondary recombination, while other rearrangements lead to inversion, retaining the Vκ repertoire for secondary use. Thus, the pattern of Vκ-Jκ associations may provide clues as to the rules and regulation of receptor editing.

The Igκ locus also carries an element called the kappa deleting elementRS (recombining sequence) in mice, located ~25kb downstream of Cκ, which can undergo recombinationin cis with Vκsor heptamer sites in the Jκ-Cκintron, leading to silencing of previously functional Vκ-Jκ rearrangements (24–28). RS recombination occurs in most cells that go on to λ L-chain expression (29) and on the silentIgκallele in about 10% of κ B cells (30). Mutation of the recombination signal of RS in the mouse germline suppresses receptor editing in a subset of cells and indirectly reduces the production of Igλ-expressing cells (31), consistent with a role in rescuing autoreactive B cells that have “run out” of editing possibilities on a particularIgκ allele(26).

Expression of the Igκrepertoire has been analyzed over the years, first with plasmacytoma panels and protein sequencing, subsequently by cDNA cloning and sequencing, then later by direct sequencing and reverse transcriptase (RT)/polymerase chain reaction (PCR) approaches, often using hybridoma panels(32–34). However, the sheer size, diversity, and polymorphism of the Igκ repertoire have hindered a comprehensive view of antibody L-chain usage. Analysis of publicly available sequence compendia,such as the NCBI (http://www.ncbi.nlm.nih.gov/), Kabat (35) and IMGT (17) databases have been important tools, but they combine data collected in different ways for varying purposes and from multiple mouse strains. With high throughput sequencing, it is now possible to obtain sufficient data from 5' RACE amplicons to obtain a better picture of the overall baseline usage. In addition to an analysis of functional transcripts from normal mice, we were able to obtain a large number of non-functional VJ sequences from bone marrow B cells of normal mice and functional sequences from bone marrow B cells of mice under uniform negative selection, thus providing information on rearrangement preferences.

Materials and Methods

Mice

All mice were bred and maintained inThe Scripps Research Institute Animal Resources facility accordingto the Institutional Animal Care and Use guidelines. C57BL/6J mice were purchased from Jackson Laboratories and bred in the TSRI custom breeding colony. pUliκ(36), JκCκ deleted (JCκ−/−) (37), and RS−/− (31) mice have been described. All mice used in this study had been extensively backcrossed to a C57BL/6 (B6) background.

B cell isolation from spleen and lymph nodes

B6 B cells were isolated from four 3-month old mice, which included two females and two males. Spleen and lymph node cells were harvested, then depleted of erythrocytes using ammonium chloride lysis prior to B cell isolation. For B cell isolation, 50 million spleen cells or 40 million lymph node cells from each gender were combined and B cells were isolated using the “no touch” B cell isolation kit (Miltenyi Biotec).

Bone marrow B cell isolation

CD4/8⁻B220⁺IgD⁻CD138⁻ bone marrow B cells were isolated from 2–3 month old C57BL/6J male mice (n=13), JCκ+/− mice (n=9), RS−/− (n=11), or pUliκ mice (n=11) as follows. Bone marrow cells were harvested from femurs, pelvis and humerus, then depleted of erythrocytes. Prior to FACS sorting, B220⁺ B cells were enriched by using CD45R (B220) MicroBeads (Miltenyi Biotec). Briefly 400 to 500 million cells were incubated with 400 to 500 μl of B220 MicroBeads and B220⁺ cells were magnetically separated. Then cell surface marker stains were performedusing a standard protocol. All of the following antibodies wereused at 1:200 dilution in FACS buffer (HBSS containing 1% BSA,2 mM EDTA, and 0.1% NaN₃), with 3 × 10⁶ cells/stain: FITC anti-B220 (RA3-6B2, BD), PE anti-CD138 (281-2, BD), APC anti-IgD, PerCPCy5.5 anti-CD4 (GK1.5, BD) and PerCP-Cy5.5 anti-CD8 (53-6.7, BD). After staining, CD4/8⁻B220⁺IgD⁻CD138⁻ bone marrow B cells were isolated by sorting using a FACAria (BD) machine.

454 sequencing

Total RNA was obtained from purified B cells using TRIZOL^@ reagents (Invitrogen). κ chain cDNAs from each sample were synthesized using a 5' RACE kit(Ambion) according to the manufacturer's protocol. 0.5 μg of total RNA per sample was used. For RT or PCR, Transcriptor High Fidelity (Roche) and Phusion Hot Start (NEB) was used, respectively. For spleen and lymph nodes B cells, a bar code strategy was used to distinguish samples in later analysis. Thefollowing oligonucleotides were used to amplify κ chain sequences.

K-R1 (Male) 5'-TTGACTGCTCACTGGATGGTGGGAAGATGG-3', K-R2 (Female) 5'-TTATCTGCTCACTGGATGGTGGGAAGATGG-3', RACE-1 (Male) 5'-TTGACGCGGATCCGAACACTGCGTTTGCTGGCTTTGATG-3', RACE-2 (Female) 5'-TTGTCGCGGATCCGAACACTGCGTTTGCTGGCTTTGATG.

PCR products were purified on agarose gels and 2.5μg each of female and male-derived amplicons were combined and used for 454 sequence analysis. For bone marrow B cells, the “A-key” sequence was appended to the 3' primer and the B-key was added to the 5' primer to allow selective sequencing from the 3' end which included the V-J junctions. Sequencing was performed using GS FLX Titanium emPCR Kits (Lib-L, Roche). Briefly, the beads included in the kit capture only nucleotides carrying the B-key and yield sequences that start from the A-key primer site. A barcode strategy with multiple identifier (MID) sequences was also used to distinguish each strain. Thefollowing oligonucleotides were used to amplify κ chain sequences. A-key+MID1+KR (B6) 5'-CCATCTCATCCCTGCGTGTCTCCGACTCAGACGAGTGCGTCTGCTCACTGGATGGTGGGAAGATGG-3', A-key+MID3+KR (pUliκ) 5'-CCATCTCATCCCTGCGTGTCTCCGACTCAGAGACGCACTCCTGCTCACTGGATGGTG GGAAGATGG-3', A-key+MID4+KR (pAlb-κ) 5'-CCATCTCATCCCTGCGTGTCTCCGACTCAGAGCACTGTAGCTGCTCACTGGATGGTGGGAAGATGG-3', B-key+RACE (For all) 5'-CCTATCCCCTGTGTGCCTTGGCAGTCTCAGCGCGGATCCGAACACTGCGTTTGCTGG CTTTGATG-3'.

Each PCR product was purified on agarose gel and 100 ng of each amplicon library was combined and used for 454 sequencing analysis. GS FLX Titanium sequencing kit XLR70 (Roche) was used for sample preparation and a 4-region gasket set up was used for processing. Data were collected at the TSRI Next Generation Sequencing core and at the UCLA Sequencing & Genotyping Core.Data were analyzed using stand-alone BLAST program (38) downloaded from NCBI. Reference Vκ and Jκ sequences were obtained from the IMGT website (www.imgt.org) (17).We analyzed clones without unread bases using the MEGABLAST algorithm as previously described (39).Sequencing data has been assigned NCBI accession numberSRX113094 (http://www.ncbi.nlm.nih.gov/sra/?term=SRX113094).

Results

Identification of 101 functional Vκ genes and nearly 400 Vκ/Jκ combinations

To evaluate the Igκ repertoire of C57BL/6 (B6) mice, we analyzedIgκ transcripts in bone marrow (BM), spleen (SP) and lymph nodes (LN). BM samples were sorted on B220⁺IgDCD138⁻ cells to exclude from analysis mature B cells and plasma cells. Splenic and LN samples were sorted similarly. As described below, we also analyzed BM B cells of selected B6 congenic mutants of interest, including heterozygous JκCκ deleted (JCκ+/−), murine kappa deleting element knockouts (RS−/−), and mice transgenic for a ubiquitously expressed κ superantigen (pUIiκ).Amplicons were generated from κmRNAusing a 5'-RACE PCR approach with two invariant primers, one in Cκ and the second in a universal 5' linker;we then carried out 454 sequencing and analysis as described(39).These methods were designed to sample the preimmune repertoire and to avoid biases caused by preferential PCR amplification. Table I summarizes the data set. We could obtain for each sample at least 11,000 reads with identifiable Vκs (mean 17,222 ± 5,850). Most of these carried identifiable VJ combinations (mean 14,067 ± 4,833). 101 functional Vκ genes were found (Supplemental Table I³), including 10defined as “open reading frames” (ORF) of unclear functionality by IMGT (17), and the expected 4 Jκs. Theoretically, there should be as many as 404 VJ combinations represented. We found from 309–369 combinations in individual samples of the various tissues of B6 mice, and somewhat higher numbers of combinations in B6 congenic BM with defined mutations. Combining data from all samples, ~98.5% of possible VJ combinations were seen at least once. In addition to functional sequences, in our Vκ usage analysis we also captured non-productive sequences that provide information on the underlying probabilities of rearrangement.

Table I.

Summary of 5'-RACE analysis of Igκ gene usage by 454 sequencing

Sample genotype:	B6	B6	B6	JCk+/−	RS−/−	pUliK
Sample tissue:	Bone marrow	Spleen	Lymph nodes	Bone marrow	Bone marrow	Bone marrow
Run number 1	20,099 (n=5, M)	11,870 (n=2, F)	12,403 (n=2, F)	24,377 (n=3, M)	22,467 (n=4, F)	19,939 (n=5, Mix*)
2	15,711 (n=4, M)	15,430 (n=4, a+b)*** a. 2,761	33,638 (n=4, a+b)*** a. 6,331	25,392 (n=3, F)	19,828 (n=3, M)	19,619 (n=3, F)
3	14,018 (n=4, F)	(n=2, M) b. 1,278 (n=2, F)	(n=2, M) b. 4,125 (n=2, F)	25,211 (n=3, F)	20,312 (n=3, M)	24,394 (n=3, F)
Total reads:	49,828	27,300	46,041	74,980	62,602	63,952
Identifiable V**: 1	17,306	11,322	11,404	23,318	19,150	13,274
2	12,194	14,675	31,025	24,459	17,712	14,709
3	12,247			23,387	19,061	21,511
Productive VJ: 1	14,596	10,201	10,155	21,149	16,241	9,847
2	9,736	9,201	17,765	22,261	14,237	10,477
3	10,026			20,669	15,927	18,028
Non-product. VJ: 1	2,682	1,122	1,249	2,145	2,884	3,428
2	2,420	2,309	3,059	2,167	3,450	4,237
3	2,197			2,688	3,134	3,457
VJ combinations:	382	327	356	382	378	386

n: mouse number used in a sample, M: male, F: female.

a, b: the reads with male tag or female tag for the statistics.

^*2 male and 3 female are mixed.

^**including pseudogenes.

^***In SP and LN samples sample 2 included bar codes to distinguish transcripts from two subsets of mice, allowing analysis of independent samples among a smaller subset of reads in which the bar code information was captured.

Skewed Igκ rearrangements in immature B cells in BM

To evaluate the Igκ repertoire in the context of minimal selection, we analyzed Vκ usage in B220⁺IgD⁻CD138⁻CD4⁻CD8⁻ B cells of the BM, which include preB and newly formed sIgM⁺and sIgM⁻B cells (the latter cells have often downregulated their receptors (39)). These samples will be called simply “BM” B cells.Previous studies suggested that chromosome contraction might equalizethe chancesof rearrangement of distaland proximal Igh and IgκVgene segments(40, 41). Our results revealed a highlyuneven distribution ofVκ usage in the BM B cells, with many individual genes used rarely and a few used remarkably frequently, ranging from .001% to nearly 10% of total sequences (Fig. 1 A, Fig. 2 A).Two potentially functional genes (Vκ1–131 and Vκ8–26) were not represented in our samples from wild type B6 mice, but were sporadically found in BM samples from B6 congenic strains to be described later.Fig. 1A plots the frequency of usage in BM B cells of individual Vκ genes, arranged distal-to-proximal in their order along the chromosome.Seven IgκVgenes 1–135,9–120, 10–96, 19–93, 6–23, 6–17 and 6–15 were each found to be used at a frequency of 5% to 7%in BM(Fig. 1A), which is much higher than expected if they were used randomly (p<0.0001, single value test of a proportion).

FIGURE 1.

Distribution of IgκL-chain gene segment usage in B cells isolated from lymphoid tissues of C57BL/6 (B6) mice.IgκcDNA derived from 5'-RACE amplicons were sequenced by 454 technology and analyzed as described (39). A-C, Percent usage of each Vκ gene is displayed according to chromosomal order relative to the JκCκ cluster. A, Bone marrow (BM, solid bar); B,spleen (SP, open bar); C, lymph nodes (LN, lined bar). Arrows indicate 7 genesused in more than 4% of total in BM, which is 4 times above the theoretical average and indicated with dotted lines.Vκare listed according to the IMGT nomenclature (17). D,Comparison of Vκ family usage in B6 B cells.n=3 with each sequencing sample representing a pool of from 4 or 5 mice per group in BM and 2 mice per group in SP and LN samples. Mean for each group indicated by a horizontal bar. Asterisks show p values calculated by Chi-square test. ***; p<0.001. E, Frequency of Jκ usage in the indicated lymphoid organs is shown. p values calculated by Chi-squared test. **; p<0.01, ***; p<0.001.

FIGURE 2.

Analysis of functional and non-functional Vκ rearrangements in BM. A,Shown are frequencies of functional (filled diamonds) and non-functional (open squares) rearrangements in B6 BM as a percentage of the total. Each point represents the mean value obtained for a Vκ. The order follows that on the chromosome from Jκ distal (left) to proximal (right).B, Comparison of relative Vκusage frequency of non-functional B6 BM transcripts (open squares) to functional pUliκtranscripts(filled diamonds). n=3 with each sample representing a pool of from 4 or 5 mice in B6 and 3 to 5 mice in pUliκ BM samples. Mean for each group is indicated by a horizontal bar. C,Shown are percentage usagesof functional sequences in the major genes in B6 BM. n=3 with each sample representing a pool of from 4 or 5 mice per group with a mean for each group indicated by a bar.***; p<0.0001, Chi-squaredtest.D, Jκ usage frequencies among functional (“F”) and non functional (“NF”) B6 BM samples. E,Jκ usages among the seven most commonly used Vκs. In D and E Jκs are as indicated by the fills of the square symbols: Jκ1, black; Jκ2, white; Jκ4, gray; Jκ5, striped.

We conclude that 101 distinctVκ genes are used, but with wide disparities in frequency.

Minor overall Vκ usage change in peripheral B cells after BM development

We next compared Vκrepertoire usage in SP and LN to that of BM. Despite their differences in maturity, the major Vκs used in BM B cells were also dominant in SP and LN B cells, and the overall usage patterns weresimilar (Fig.1,B and C).There were exceptions, however, notably1–135, whose usage was significantly lower in the periphery (6.4% in BM versus 3.3% in SP and 2.3% in LN), and 10–96 (6.4% vs 9.2% and 10.2%) and 6–15 (5.5% vs 8.8% and 9.3%) whose usages were higher (Fig. 1D). These changes compared to BM were larger in LN than in SP. Usage in the periphery suggested positive selection of 10–96 and 6–15 and negative selection of 1–135, which presumably corresponds to changes in frequency of B cells expressing these Vκs. We address the features of the frequently-used Vκs in a later section.

Subset of Vκs not found in peripheral B cells

Certain genes found at low frequency in BM were not represented at all in samples from the LN and SP from B6 mice from this study (<1/70,000 reads) or our previous study involving 3H9 anti-DNA Ig heavy-chain transgenic mice lacking or carrying BAFF transgenes (39) (<1/100,000 reads). These Vκ genes included 1–131, 4–62, 4–54, 1–35 and 8–26. According to the IMGT database, none of these genes has been seen previously as transcripts or even as DNA rearrangements and all except 4–54 have significant apparent defects in their recombination signals (http://www.imgt.org/IMGTrepertoire/LocusGenes/index.php?repertoire=genetable&species=M us_musculus&group=IgkV). Our data suggest that they can indeed rearrange, albeit at low frequency. It is unclear why they are not represented to some extent in peripheral immune tissue samples. Although its recombination signal appears normal, 4–54 lacks a highly conserved W in the second framework region and so may be counter-selected owing to defective protein function.

Vκ family usage patterns

When the BM repertoire was analyzed with respect to Vκ family, we similarly found wide ranges of usage (Fig. 1D), with Vκ19/28 and Vκ9/10 used often and single-member Vκ11, Vκ22, VκRF and Vκdv36 families used rarely, as previously reported(3, 34, 42). 20-101-2also appears to be a functional, butrarely used, single-member family. The centrally-located Vκ4 gene family is the largest, consisting of 28 functional genes including open reading frame (ORF) genes. This family carriespredicted promoter elements different from other Vκs(9).Although, the Vκ4 family is preferentially expressed in fetal and vAbl-transformed B cells (43–45), we found in our adult sample that Vκ4 family usage hada lower probability per gene (11.1% Vκ4 family÷ 28 members = 0.4% per gene) compared to other large families such as Vκ9/10 (22.3 ÷ 13 = 1.7 % per gene), Vκ8 (8.4 ÷ 11 = 0.8 % per gene) and Vκ19/28 (20.8 ÷ 9 = 2.3 % per gene) (Fig.1D). Moreover, in the periphery a further reduction of Vκ4 family usage was observed (Fig. 1D, BM vs SP or LN, p<0.0001, chi-square test).Thus, disparities in gene usage extend to entire Vκ gene families.

Jκ usage in B6 mice

Jκusage in BM, SP and LN was analyzed among productive sequences, revealinga stable hierarchy (Fig.1E) with Jκ1 used most frequently (34%, 35% and 33% in BM, SP and LN, respectively, p<0.0001 compared to other Jκs), followed by Jκ5 (29%, 29%, 31%), Jκ2 (20%, 22%, 21%), and Jκ4, which was used significantly less than other Jκs (16%, 15%, 14%, p<0.0001 compared to other Jκs). In LN, Jκ5 usage was significantly elevatedcompared to BM (p<0.0001, chi-square test).Jκ2 has a nonamer and a heptamer that are non-canonical:AGTTTTTGT instead of GGTTTTTGT and CAGTGTG instead of CACTGTG, respectively. Heptamers and nonamers with such sequences are rare among Vκ and Jκ gene segments (Supplemental TablesI, II). Jκ4 has a non-canonical spacer, which is one base longer than normal (24 bases instead of 23, Supplemental Table II). A comparable increase in spacer length of recombination substrates leads to 15–40% decrease in recombination efficiency in transient transfection assays (46). These differences may explain in part the relative under-representation of Jκ2 and Jκ4.

BM Vκ usage in non-functional transcripts and upon uniform negative selection

To assess the underlying probability of rearrangements to particular Vκs, the frequencies of nonfunctional (NF) and functional (F) joins in B6 BM samples were analyzed (Fig. 2A). In general, individual Vκ usages among the NF set tracked the F set. The overall frequency of NF rearrangements among all transcripts carrying identifiable VJ combinations was 18.3% ±1.6% (81.1% ± 1.6% were F joins).This is an underestimate of the actual frequency of NF joins because of the lower stability of such messages (47), but allows analysis of the relative frequencies of recombination of different Vκs. Calculating the F frequencies for joins carrying the frequently used Vκs revealed a similar range of frequencies (81.7 to 85.5%), however 1–135 had a lower frequency (70.6% ± 1.9%, p<0.0001, χ²test) (Fig. 2C).Because of its high level of usage and distal location, 1–135 likely represents a special case (see Discussion). The patterns of Jκ usage among F and NF sequences were similar overall and for individual highly used genes (Fig. 2, D and E).

We then compared Vκ usage of non-functional B6 sequences with functional sequences found in BM of B6 congenic mice carrying a ubiquitously-expressed κ superantigen transgene (pUIiκ). The superantigen negatively selects allκ B cells and stimulates receptor editing, leading to a massive increase in λ B cell production (36). Remarkably, the pattern of Vκ usage was nearly identical with that of non-functional B6 BM (Fig. 2B). Again, Vκ genes dominantly seen among functional B6 BM samples were most frequent.Thus, functional Vκ usage in immature BM B cells under conditions of uniform negative selection was similar to Vκ usage among nonfunctional transcripts. We conclude that the dominance of certain Vκ gene usage is a result of intrinsic rearrangement (and perhaps expression) preferences rather than subsequent selection.

Analysis of the features of frequently used Vκs

We sought explanations for the preferred usage of the 7most heavily used Vκs (presumptively called the “munificent seven”). These were also most abundantly represented among nonfunctional sequences (Fig. 2A), arguing against simple somatic selection. As can be seen in Fig. 1A, there was some clustering, but no obvious trend in Vκusage based on chromosomal position, consistent with previous conclusions (48, 49). Nor did we identify any clear correlation between Vκ usage and recombination signal sequence, either when assessed bythe extent of identity with consensus heptamer/nonamer(Supplemental Table I) or by predicted bioactivity as calculated using a recombination signal information content (RIC) algorithm (50) (Fig. 3A).RICexplicitly takes into account the contribution to recombination ofthe spacer in addition to heptamer/nonamer sequences. We also found no clear correlation with Vκcoding end sequence, which has been shown to impact recombination efficiency (51, 52). Abundantly used genes lacked obvious differences from closely related family members in their promoter regions with respect to predicted transcription factor binding sites, as described by Brekke and Garrard (9).We examined mouse IgκV loci on the NCBIgenome browser to look forlonger-range features peculiar to these frequently usedVκ genes (Supplemental Fig.1).Interestingly,6–23, 6–17 and 6–15are adjacent to, and in opposite transcriptional orientation to,pseudogenes homologous toOdc1, ornithine decarboxylase, structural 1. Preliminary studies indicate that these pseudogenes are expressed in B220⁺CD138⁻IgD⁻bone marrow B cells (not shown). Homologues 6–25, 6–14 and 6–13are in a similar context, but are used less frequently.6–13has a non-canonical nonamer, ACATAAGCC instead of ACAAAAACC or ACAAAAATA, in which the G is considered “anti-consensus” (46, 53), perhaps explaining its less frequent representation in our samples.Overall, the peculiarly high level of usage of particular Vκ genes appears to berelated primarily to preferential rearrangement or expression that was unrelated to proximity to the Jκ cluster, and could not be ascribedto cellular outgrowth or selection.

FIGURE 3.

Analysis of the correlations between Vκ gene usage, recombination signal efficiency, chromosomal location and orientation, and Jκ usage.A,Correlation between individual Vκ gene usage and predicted recombination signal efficiency (RIC value).B,comparison of Vκ associations with Jκ1 and Jκ5 in B6 BM. Percent usages ofJκ1 (filled portion of bars) and Jκ5 (white portion of bars)among functional rearrangements are shown (stacked bars show combined frequency). Vκ genes are ordered by their chromosomal position from distal to proximal to Jκs.Arrows indicate the 7 major genes. Solid diamonds below the x-axis indicate genes that are in the same initial orientation as the JκCκ cluster (hence “deletional” when associated with Jκ1). n=3 with each sample representing a pool of from 4 or 5 mice per group with a mean for each group indicated.

Secondary rearrangements and Jκ usage bias among individual Vκ genes

Data were analyzed to assess the patterns of V-J association in B6 bone marrow B cells. We wondered if Vκ genes located distally from the Jκ cluster might preferentially use downstream Jκs, while proximal Vκs would more often use upstream Jκs. Overall, there was no such correlation, with several distal genes (1–135, 10–96, 19–93) preferentially rearranged to Jκ1, while many proximal and intermediately positioned Vκs (6–23, 6–15, and multiple members of the Vκ4 family) preferentially rearranged to Jκ5. These usage data are shown in Fig. 3B, which for clarity only shows V-J association data for Jκ1 and Jκ5. Preferential association of Vκ4 members with Jκ5 has been noted previously(45, 54).We also asked if genes that were frequently used might be associated preferentially with Jκ1, while rarely used Vκs would be limited to downstream J usage. Among the seven most frequentlyused Vκs, five of seven were associated more often with Jκ1 than with Jκ5, however, 6–15 and 6–23 were preferentially associated with Jκ5 (Fig. 3B).The twenty most frequently used Vκs made up 72% of the sample and had a Jκ usage distribution similar to the overall averages (Jκ1, 35%; Jκ2, 21%; Jκ4, 16%, and Jκ5, 27%). On the other hand, among the twenty least-used Vκs, which accounted for 0.62% of the sample,49% were joined to Jκ5while only14% used Jκ1, indicating that secondary recombinations were indeed important for theirrepresentation.

Analysis of BM Vκ repertoire in B6 congenic mice carrying mutations that affect secondary recombination

To evaluate how the expressed repertoire changes under conditions predicted to alterreceptor editing, we analyzed Vκand Jκusage in BM samples of mice carrying one of three types of genetic modification (Fig. 4): (i) a heterozygous JκCκ deletion (JCκ+/−)(37), which causes peripheral B cells to have enhanced usage of downstream Jκs and Igλ(21, 37), (ii) a ubiquitously-expressed κ superantigen (pUIiκ),which negatively selects allκ B cellsand stimulates receptor editing, leading to a massive increase in λ B cell production (36), or (iii) homozygous mutation of the murine κ deleting element RS (RS−/−), which hinders destructive editing of functional rearrangements (particularly Jκ5) and indirectly suppresses λ rearrangements(31). Here we analyzed sequences from BM samples in which both Vκ and Jκ were identified in order to assess J usage.In JCκ+/−BM, significant changes in Vκ usage were seen,compared to B6 BM, including a significant increase in the representation of1–135(Fig. 4,A and B, 9.8% ± 0.6 versus5.5% ± 0.6%p<0.0001), along witha decreasein usage of other frequently used Vκs (p<0.0001), except for 19–93.In particular,6–23, 6–17 and 6–15 genes,which are located proximally, decreasedmuchmore than other genes located on thedistal site (Fig. 4, compare B to A).

FIGURE 4.

Igκ repertoire in BM from B6 and genetically modified mouse strains. Shown are usage frequencies of Vκ genesamong functional rearrangements in B cells from the following strains.A,B6;B,JCκ+/−;C,pUliκ; and D,RS−/−. Arrows indicate 7 major geneswith open bar that were found in >4% of sequences from B6 BM. Vκ genes are ordered by their chromosomal position from distal to proximal to Jκs.n=3 with each sample representing a pool of from3 to 5 mice per group in each strain with mean for each group indicated by a horizontal bar.

In the analysis of JCκ+/− Vκ usage, it was also of interest to assess specifically the Jκ1 usage pattern as it provides a useful window into initial rearrangement preferences. (Because rearrangement often happens on both Igκ loci in wild type cells, Jκ1 rearrangements are not unequivocally the first in such cells, whereas in JCκ+/− cells they are.) Jκ associated Vκ usage values for JCκ+/− BM are given in Supplemental Fig. 2. They clearly show a significant bias to distal Vκ usage among Jκ1 rearrangements. These values are helpful in mathematically modeling different scenarios of secondary gene rearrangement.

TheVκ usage pattern in pUliκ BM (Fig. 4C) wassimilarto that seenJCκ+/− BM (Fig. 4B) except that1–135 wasusedeven more frequently (12.6% ±0.5% in pUliκ; 9.8% ±0.6 % in JCκ+/−, p<0.0001, chi-square test).4–71 and 4–59 were also overrepresented in our pUliκsample, however,there was unusually large variance for these particular genes in this strain.1–135's mode of rearrangement is obligatorily deletional, and itis the penultimate Vκ on the distal end of the locus (Fig. 3), with only the infrequently used2–137beyond (2–137 usage: B6, 0.28%, B6;JCκ+/−,0.41%; RS−/−, 0.16%; pUIiκ, 0.37%) (Fig. 4).Moreover, as mentioned above, 1–135was found associated preferentially to Jκ1 (Fig. 3). Therefore, our data indicate that 1–135/Jκ1 rearrangement is often the first and only rearrangement to occur on individual Igκ alleles.

Changes in Vκ usage in RS−/−,compared to B6, BM were more subtle, with a modest increasein usage of 1–135 (Fig. 4D, 6.7% ± 0.8, p=0.0004). In addition, several genes normally represented at moderate frequency and showing Jκ5 skewing were consistently more abundant than usual. These included14–111 (p<0.0001), 4–55 (p<0.0001), 8–28 (p<0.0001), 6–20 (p<0.0001) and 8–19(p=0.0008)(Fig. 5A). The results indicated thatunder normal conditions RS recombination sometimessilenced these genes, presumably because theirproperties promoted editing under some circumstances.

FIGURE 5.

Comparative usage analysis in BM of B6 and genetically-modified strains showing Vκswith elevated usage in RS−/− and overall Jκ usage distributions. A, Frequencies of 6 genes with increased representation in RS−/− samples is shown. B,BM Jκ usages of the indicated strains are shown. B6, solid bar;JCκ+/−, open bar; RS−/−, grey bar; pUliκ, lined bar. n=3 with each sample representing a pool of from3 to 5 mice per group in each strain.Asterisks show significant differences, with p values calculated by Chi-square test.*;p<0.05, **; p<0.01, ***; p<0.001.

Jκ usages in BMs of genetically modified strains

Although overall Jκ usage preference was not radically changed in mutant strains in that Jκ1 and Jκ5 continued to be used mostfrequently, significant differences were seen (Fig. 5B). JCκ+/− and pUliκBM hadsimilarreductions inusage of Jκ1 (B6 vs JCκ+/−; p<0.0001, B6 vs pUliκ; p<0.0001) and increases inusage of Jκ5 (B6 vs JCκ+/−; p<0.0001, B6 vs pUliκ; p<0.0001).In addition, RS−/− BM B cells had significantly higher Jκ5 usage compared to all other strains (p<0.0001).Surprisingly, however, when VJ rearrangements in the 7 most frequently used Vκs were compared between B6 and the genetically modified mice,no major changes in their Jκ associations were observed. For each Vκ, either Jκ1 or Jκ5 usage predominated, with1–135, 9–120, 10–96, 19–93 and 6–17 genes preferentiallyrearrangedto Jκ1 (Fig. 6, A–D,F) and 6–23, 6–15 preferentially rearranged to Jκ5 (Fig.6, E and G).

FIGURE 6.

Jκ usages associated with the major Vκ genesin BM of B6, JCκ+/−, RS−/−, and pUIiκ mice. B6, solid bar;JCκ+/−, open bar; RS−/−, grey bar; pUliκ, lined bar. n=3 with each sample representing a pool of from3 to 5 mice per group in each strain with a mean for each group indicated by a horizontal bar.

CDR3 analysis

Our VJ analysis suggests that Igκ rearrangements converge on certain major usages in many situations with predicted limitations on diversity. However, this analysis does not take into account junctional diversity. To address this, we analyzedpredicted CDR3 amino acid diversity generated byindividualVκgene association with particular Jκs. In the case of1–135/J1, there were 38 different sequences generated.Together, the 7 major Vκs from B6 BM, with a combinatorial diversity of 28(=7 Vs× 4 Js), generated 283distinct CDR3 amino acidsequences, a 10-fold difference (data not shown).Overall, 1000 different L-chains (including CDR3 amino acid diversity) accounted for91% of all sequences in our sample.

Discussion

We find that the Igκlocus in C57BL/6 strain micecontains 101 potentially productive Vκ genes along with the well known 4 Jκ genes,henceit is theoretically capable of generating 404combinations.Light chains contribute to antibody diversity by further combination with different H chains. However, Igκrearrangements also play an important role in receptor editing. These dual roles perhaps explain whyIgκrecombination clearly fails to approach an equal rearrangement frequency for each V gene that might be considered optimal for diversity. Nor was V-J association random. In B6 BM we found that43% of B cells used one of 7 V genes. Moreover, these and other Vκs showed biased association with particular J elements, further restricting diversity.Junctional diversity did augment overall diversity, but only by about 10-fold.What are the implications of this striking skewing?

At least two non-mutually exclusive explanations may be offered for the evolution of biased Vκ usage andV-J association. One is simply that particular combinations generate useful specificities favored by natural selection for combating microbes. Our analysis did not focus on specialized B cell subsets, such as those in the fetus, gut, and pleural tissues, which might have quite different developmental precursors (55). Vκ genes that were poorly used in the tissues that we tested might be better expressed in alternative sites. (It should also be emphasized in this regard that ourB6 Igκ repertoire usagedata need to be compared to other mouse strains and to other species.)An alternative explanation is that rearrangement bias facilitates secondary rearrangements and editing by favoring a sequenced order of gene recombination. For example, an initial bias to inversional rearrangements to Jκ1, or deletional rearrangements to Jκ1 involving proximal Vκs, would retain the maximum numbers of Vκs available for subsequent rearrangements. Such rearrangements are indeed preferred, accounting for about 2/3 of Jκ1 sequences.In theory, inversionprovides the maximum flexibility for cells to subsequently correct Igκrearrangements that are nonfunctional, under-expressed or autoreactive, which collectively constitute the vast majority of rearrangements. Hence, the heavy usage and Jκ1 skewing of the distal, inversionally-rearranging10–96 and 19–93genes might be important in initial scrambling of the locus to allow a more efficient and wider range of secondary recombinations than would otherwise occur. Without such effects, there is predicted to be strong bias for usage of distal Vκs with proximal Jκs and vice versa, which we failed to see.

What mechanism accounts for the dominance of certain Vκ genes? This skewing could not be readily attributed to predicted recombination signal strength, proximity to the Jκ elements, or somatic cell selection. For example, 10–96 carries a recombination signal identical to those of adjacent10–94 and 10–95 genes, but is used much more frequently. In BALB/c mice, 10–96 is also vastly overrepresented compared to 10–95, despite carrying functionally indistinguishable promoter elements and recombination signals(56). Our conclusion that many of the dominant genes have promoter elements that are indistinguishable from their less well expressed counterparts comes from the detailed bioinformatic analysis of Brekke and Garrard, who annotated the Igκ locus based on data largely collected by Zachau`s group and the genome project (9). No real differences were seen in this analysis between highly-expressed and poorly-expressed family members in any promoter feature characterized (9). Surprisingly, major Vκusage was also not appreciably altered in the BM ofthe gene-modified strains studied here, including in pUIiκ mice, which are subject to strong negative selection. Collectively, these results suggest that rearrangement of the major Vκgenes is strongly preferred owing to enhanced accessibility to rearrangement in preB cells.

Ig gene accessibility to rearrangement is complex and multifactorial, as it is affected by chromatin structure, histone acetylation, DNA methylation, and conventional enhancer and promoter activity, in addition to the effects of antisense and sense germline transcription (57–59). One possibility for how differential accessibility is controlled genetically is that cis-acting elements adjacent tofrequently used genes are preferentially recognized by transcription factors.Some exceptional features in Vκ4 family genes have been noted, including clusters of potential binding sites for Lef-1, Ebf, and Oct1/2 in the upstream, intron and downstream regions of the RSS with functional significance in targeting for V-J recombination(9). In some studies, Vκ4 family genes hadelevated usage and skewed Jκ association, particularly in fetal cells(34, 43–45, 54).However, in this study of young adult mice, we found asomewhat reducedaverage usage per Vκ4 family gene.We couldnot find significant differencesbetweensix of the seven frequently used genes compared to other genes in their families in terms of predicted transcription factor binding sites or recombination signals. The seventh gene,19–93, is unique in that it is a member of a single gene family. 19–93 hasan extra octamer (Oct-binding) site upstream of the conserved one close to the TATA box, possibly explaining its elevated usage. Other notable features of 19–93wereits preferred association with Jκ1, and that its usage was not diminished in abundance or skewed much in its Jκ association in cells predicted to undergo more editing, such as in pUIiκ BM B cells. The explanation for this is not clear, but could come about froman idiosyncrasy of the locus organization: 19–93 rearranges to Jκ1 by inversion, which then puts the under-used Vκ4 family proximal to the JκCκ cluster and frequently-used Vκ6 genes distal, potentially lowering the chances of secondary recombinationsediting away the 19–93/Jκ1 join.

The frequently-used Vκ19/28 family genes 6–23, 6–17,and 6–15are adjacent to Odc1-likepseudogenes in an opposite transcriptional orientation.Nearby 6–14 and 6–13genes have the same features, but are used less frequently. Other non-Ig pseudogenes are also near 1–135, and several other less well used genes.These features raise the possibility of a regulatory function of non-coding RNAs originating in adjacent pseudogenes, be they Vκ or non-Ig.Although sterile transcription of germline V coding gene segments was shown many years ago to correlate with accessibility to V(D)J recombination (60), non-coding RNA has recently been implicated in actively promoting accessibility. Bolland et al. identified in early B cells genic and intergenic antisense transcripts throughout the IGHV region that were strictly upregulated at the transition from D_H-to-J_H to V_H-to-DJ_H recombination (61). Krangel and colleagues showed that transcriptional elongation regulates the programmed accessibility of TCR Jα genes (62). Such transcripts may promote chromatin accessibility and allow RAG binding.

An additional possibility for how preferential Vκ usage is regulated relates to chromatin remodeling mediated by longer range looping interactions by CTCF, cohesins, and associated proteins(41, 59, 63–66).Locus contraction is believed to facilitate Ig gene rearrangements(40, 67). Locus looping might promote accessibility or bring separated gene segments closer together, facilitating rearrangements. Alternatively, loop tethers may serve as insulators that delimit Vκ groups, whichmight compete with one another for rearrangement. In the Igh locus, CTCF binding has been implicated both in limiting and facilitating rearrangements to V_H, through down-regulation of transcriptional read-through, retention of DNA associate with heterochromatin, and by facilitating locus contraction and element colocalization(63–65). CTCF binding sites within the Igh and Igκ loci clearly correlate with proximity to V_Helements,but not to Vκ elements (41, 59, 63). Nor are CTCF sites particularly close to frequently-used Vκs (Supplemental Fig. 1). One CTCF binding site is associated with the recombination silencer Sis (68) located in the Vκ-Jκ intron (41, 66). Regulation involving CTCF might involve insulator function, as proposed by Hendriks and colleagues, who showed dramatic changes in Vκ usage in B cells that were conditionally-deficient in CTCF, leading to an increase in proximal Vκ usage at the expense of distal usage (41). For example, use of 1–135might beenhanced because it is literally insulated from competition for recombination by looping anchored by a CTCF binding site ~35 kb downstream (Fig. 8), which in turn might be tethered near the intronic CTCF. (Alternatively or in addition, the distal end of the locus may carry an enhancer or LCR element providing an additional tether.) A similar argument could potentially explain the usage pattern of other genes. Two strong CTCF binding sites flank a total of 30 Vκ elements that includes all of the Vκ4 family members. If the two sites form the base of a large loop with access to the JCκ cluster, Vκ4 family genes may compete with each other within the loop, possibly explaining the relatively low frequency of Vκ4 family usage per gene (Fig. 8). However, there appear to be many additional lower occupancy CTCF sites among Vκ4 family genes as well (41).

FIGURE 8.

Potential looping of Igκ locus based on the reported CTCF binding sites. Cartoon depictsIgκ locus assuming clustering of major CTCF sites, highlighting distribution of major Vκs and Vκ4 family on loops. CTCF sites, blue circles; major Vκ genes, yellow arrows; Vκ genes used at 1–3.9 % of total, medium sized green arrows; infrequently used Vκ genes (<1% of total), small red arrows. Jκs and Cκ are indicated by pink symbols and blue arrow, respectively. RS; blue solid diamond.

The 1–135 gene had many unexpected features, including high frequency usage of 1–135/Jk1 in BM, increased frequency in the BM B cells of JCκ+/− and pUliκmice, counter-selection from BM to SP, and elevated proportion of nonfunctional sequences in BM.In particular, out-of-frame rearrangements to Jκ1 should be rapidly followed by secondary rearrangements. We ascribe these special featuresof 1–135to its chromosomal position rather than to its protein sequence.Because1–135is locatedin the penultimate position on the distal end of the locus and undergoes deletional rearrangement, subsequent rearrangement or editing on that allele is limited just to the rarely usedVκ gene 2–137.Apoptosis of autoreactive B cells is a slow process preceded by a time window in which editing takes place, so autoreactive 1–135-expressing “dead-end” cells unable to edit further should tend to build up in BM, but be barred from release to the peripheral immune system.

A striking finding from this study was the overall similarities in κ gene segment usage in BM B cells from pUIiκ, RS−/−, and JCκ+/− mice compared to wild type BM B cells. Although some significant differences were documented, they were relatively subtle. We had expected the overall Jκ usage and V-J associations to be more strikingly skewed, particularly in pUIiκ mice, in which editing is massively induced (36). One explanation is that secondary rearrangements are already extensive in wild type cells owing to the high proportion of nonfunctional rearrangements (>2/3) and additional editing of functional light chains owing to autoreactivity, non-association with H-chain, and suboptimal expression. Hence, additional editing may have a relatively subtle overall impact over this high background. When a B cell`s Ig H-chain is fixed by transgenesis, L-chain selection is strikingly revealed. For example, in the anti-DNA 3H9 H-chain transgenic mice, several of the normally dominant Vκs are strongly underrepresented in peripheral B cells and major V-J associations are also skewed(23, 39, 69).In mice with a diverse repertoire of H-chains, these skewings may balance out, ultimately reflecting underlying rearrangement preferences. A second explanation is that autoreactive B cells in BM may be actually overrepresented because they transit more slowly through that compartment (36) and are especially enriched among BM B cells that have downregulated sIgM (70).

FIGURE 7.

Analysis of the diversity of the expressedκ-chain repertoire in B6 BM. A,The integrated frequency of Vκ usage is plotted in declining order of usage frequency. Asterisk indicates the contributions of the 7 major genes, yielding a 43% integrated frequency of these genes. B, CDR3 diversity produced by all VJ associations is shown. Asterisk indicates total CDR3 amino acid variety contributed by the major 7 genes.