Abstract
Anti-dsDNA autoantibodies in MRL mice contain a higher than average frequency of atypical complementarity-determining regions 3, including those made with D-D rearrangements. It has been reported that MRL mice have an intrinsically high frequency of creating VDDJ rearrangements; however, we show in this study that the majority of these apparent D-D rearrangements in B cell progenitors can be accounted for by a very novel germline DH gene in mice of the Ighj haplotype. This gene has the appearance of a D to D rearrangement due to the duplication of 9 bp common to most DSP2 genes. Germline DSP2 genes from Ighj mice were amplified, cloned, and sequenced, showing the presence of this novel gene as well as a new allele of a conventional DSP2 gene. Sequencing of D-J rearrangements revealed that Ighj mice also have a different allele of DFL16.1 and apparently lack DFL16.2. Despite the existence of this new DSP gene, analysis of VDJ rearrangements from adult bone marrow pre-B cells of MRL/lpr mice still revealed the presence of complementarity-determining region 3 containing apparent D-D joinings in 4.6% of the sequences. C3H pre-B cells had 4.2% of sequences with apparent VDDJ rearrangements, and BALB/c pre-B cells had ∼2%. DDJ intermediates were also observed, but at a lower frequency. However, strikingly, no VDDJ rearrangements were observed in newborn sequences, suggesting the process of assembly of VDJ rearrangements is fundamentally different in newborn mice vs adult mice.
The autoimmune strain MRL is widely used as a model of systemic lupus erythematosus and shares many features with the human disease (1). MRL mice are characterized by a high frequency of autoantibodies, including many reactive with nuclear Ags such as dsDNA, histones, nucleosomes, and small nuclear ribonucleoproteins. Anti-dsDNA Abs are the hallmark of patients and mice with lupus; these Abs are found deposited in the kidneys and are thought to be the inciting agent of the glomerulonephritis characteristic of systemic lupus erythematosus (2). Sequence analyses of autoantibodies from autoimmune mice which are directed against anti-dsDNA have shown that these Abs are enriched in arginines, which most likely aid in the binding to the acidic Ag DNA (3, 4, 5, 6, 7, 8, 9, 10). Although many of these arginines are encoded by N regions, it has also been proposed that several of the arginines in anti-dsDNA autoantibodies were encoded by atypical complementarity-determining region 3 (CDR3)3 structures such as D genes in alternative reading frames, inverted D segments, and D-D joins (3, 4, 5, 6, 7). The prevalence of such atypical CDR3 regions in these autoantibodies could either be due to antigenic selection of the rare CDR3 which encode arginines, or due to abnormalities in the control of V(D)J recombination in autoimmune mice. If the latter is true, this would be of interest, as it could document that one of the genetic traits which presumably makes MRL mice susceptible to autoimmune disease is misregulation of the V(D)J recombination mechanism. In support of this latter hypothesis, it has been shown that rearrangements from newborn liver of MRL/+ mice cells have a high frequency of atypical CDR3 regions, predominantly due to the presence of a high frequency of D-D fusions, and of inverted D regions (11). Further studies showed that these atypical CDR3 regions were also present at high frequency in adult bone marrow pre-B cells (12).
In this study we tested an alternative hypothesis to explain the high frequency with which unusual CDR3 regions are created in MRL B cell precursors, namely that many of the putative D-D segments seen in VDJ rearrangements were actually due to a new D gene in the MRL strain of mice, because many of the D-D regions from newborn MRL mice have similar identical core sequences. This putative new gene bears resemblance to the 5′ part of a DSP2 gene joined to the 3′ part of a DSP2-type gene. It is present, at least in part, in six of the 11 “atypical” rearrangements from MRL mice described by Klonowski et al. (11). The consensus which is derived for this gene from those six sequences is … . TACTATAGTTACTATAGTTACGAC, making it >24 bp long. Because all known DSP2 genes are very similar (13), and all are 17 bp, it was a priori unclear whether this was a new gene or a bona fide D to D recombination event. In the absence of N regions, junctions are often formed at the site of short sequence homologies, and this homology-directed recombination results in significant junctional homogeneity, especially in DH-JH junctions, due to the repeated presence of TAC and TAT in D and J segments near the region of joining (14, 15). It is thus possible that this particular sequence could be present in high frequency because it represents a D-D join of DSP2.X with a yet undescribed DSP2 gene, yielding a commonly occurring junction due to short stretch of sequence identity of the two D segments resulting in homology-directed recombination during D-D joining. However, we have previously uncovered the existence of two new DH genes, DSP2.10 and DST4, based on our hypothesis that if the same unidentified sequence between a VH and JH segment is observed in several independent CDR3s, it is more likely to be a new D region than a complex D-D join or inverted D (16, 17). Such new D genes are particularly noticeable if the sequences lack N regions such as is observed in fetal or newborn mice, so that the unidentified sequence must be germline encoded by V, D, or J genes.
We show here that mice of the Ighj haplotype such as MRL and C3H have two previously undescribed DSP genes, one of which is the unusual one present in many of the apparent D-D fusions. Other apparent D-D joins can be attributed to a sequence difference in the Ighj allele of DFL16.1. Ighj mice also have different VH7183 genes, one of which has a much longer portion at its 3′ end than most VH7183 genes, thus making a bigger contribution to CDR3. Although all apparent D-D fusions from newborn sequences of MRL and C3H can be accounted for by these new DH and VH alleles, we still observed that MRL and C3H mice have ∼4–5% of CDR3 regions in adult bone marrow precursors which contained apparent D-D rearrangements. Thus, D-D rearrangements are observed in adult, but not newborn, mice, suggesting major differences in the VDJ recombination mechanism at these two times in ontogeny.
Materials and Methods
Mice
MRL/lpr, C3H/HeJ, and BALB/c mice were maintained at the breeding facility at The Scripps Research Institute (La Jolla, CA). Pre-B cells were purified from bone marrow on a FACSVantage after staining with FITC-conjugated anti-B220 (Caltag Laboratories) and biotin-coupled anti-IgM followed by streptavidin-PE.
PCR assay
DNA was prepared as previously described (18). DJ rearrangements were amplified with either AF288 (5′-GCCAAAGCTTTTTGTSAAGGGATCTAC), located at the 5′ end of the recombination signal sequence (RSS) including the nonamer and part of the spacer, or with AF216 (5′-GTSAGGAAGCTTYCCAGAAACAGACC), located more 5′. The set of JH primers (JH1 + 4, JH2, JH3), located in FR4, was previously described (16). Germline D genes were amplified with the same upstream primers, and with a primer 3′ of DSP, AF289 (5′-CCCCTAGAATTCAAGCTCCTCTTGACT). VDJ rearrangements were amplified with the same set of JH primers and with previously described VH primers (VHS107, AF72 (19); VH7183, AF303; (18); and VHJ558, AF326 (5′-GGGGCTTAAGCTTAGCTGTCCTGCAAGGCT)). Amplifications were performed for 35 cycles, and PCR products were cloned and sequenced as previously described (18). For D-J and VDJ rearrangements, 10–14 PCRs were done on each DNA sample, thus minimizing duplicate sequences. Duplicate sequences were only counted once, and no more sequences were obtained from any PCR after the first duplicate sequence was encountered. For the germline DH genes, we amplified 38 sequences from MRL/lpr tail DNA from five independent PCR, 29 sequences from MRL/lpr bone marrow derived from three independent PCRs, and 16 sequences from C3H unsorted bone marrow DNA from three PCRs.
Results
Ighj mice have a different set of DSP2 genes
To determine whether many of the apparent D-D fusions in MRL/lpr VDJ rearrangements are due to a previously undescribed D gene, we amplified germline D genes with a primer, AF288, which is located in the nonamer and spacer of the 5′ RSS of all DSP and DFL16 genes, and with a primer located ∼90 bp 3′ of DSP2. Although the 5′ primer should amplify most, if not all, DSP and DFL genes, the flanking region is likely to diverge as one gets further from the coding region, so the 3′ primer may not amplify all DSP genes. Nonetheless, we were able to amplify several DSP2 genes identical with BALB/c genes with this primer set, including DSP2.2, DSP2.5, DSP2.9, DSP2.10, as well as DSP2.X, a previously described gene present in Ighb mice such as C57BL/6, but not in Igha mice such as BALB/c (Fig. 1⇓) (20). In addition, we amplified two novel DSP2 genes, including the predicted gene. This new gene, which we called DSP2.12, has a duplication of 9 bp, thus making it a 26-bp DSP segment, whereas all other DSP2 genes are 17 bp long. We also amplified DSP2.12 from C3H DNA, as would be predicted, because this strain also has the Ighj haplotype, and because this DH sequence was also observed in C3H rearrangements (11). The other new DSP gene, DSP2.13, is more conventional, having only two base pair changes from other DSP genes. Finally, we also found that Ighj mice have an alternative allele of DFL16.1 (Fig. 1⇓).
D genes of Ighj mice. Unrearranged DSP2 genes were cloned from MRL and C3H DNA. Asterisks indicate novel genes which are unique to this Ighj haplotype. The Ighj allele of DFL16 is compared with the Ighs allele. This DFL16.1j gene was only sequenced from DJ rearrangements, and not as an unrearranged gene. DST4 and DQ52 (not shown) were also found in VDJ sequences, and thus are also present in Ighj mice, as previously described (11 ). DSP2.12 and DSP2.13 have GenBank accession numbers AF428079 and AF428080.
Because our DSP2.12 sequences were derived from bone marrow DNA, it is possible, although highly unlikely due to the frequency with which we amplified it, that we amplified a D to D rearrangement which had occurred without prior rearrangement to a JH gene in vivo. Therefore, we repeated the amplifications with nonlymphoid (tail) DNA and obtained the same sequence. Hence, DSP2.12 is, in fact, a germline gene.
Usage of DSP2 genes in D-J rearrangements
We amplified D-J rearrangements for two reasons. First, we still had not observed several DSP2 genes which were present in Igha mice, and divergence of the 3′ flanking sequence in the region of the downstream primer might be the reason for our failure to amplify them. By substituting the JH primers for the downstream primers and using them with three different primers for the 5′ flanking region, including the one just described, we should detect most DSP and DFL genes. Nonetheless, we did not uncover any additional DSP genes that we had not seen with the flanking primers. Thus, with the caveat that some D genes may vary in the spacer/nonamer sequence, as well as with the more distal primers, it appears that Ighj have only these seven functional DSP genes. Also, we did not see any DFL16.2 rearrangements, thus suggesting that this gene may also be absent in Ighj mice.
Another reason to amplify DJ rearrangements was to observe the relative frequency with which the D genes rearrange in MRL/lpr vs C3H pre-B cells, because C3H mice were reported to have fewer D-D fusions in VDJ rearrangements (11, 12). Hence, they might have a different relative usage of DSP2 genes. The first primer that we used, located in the RSS, amplified DFL and DSP genes. From 80 sequences in MRL/lpr mice, and from 98 C3H sequences, we determined that 50% of MRL and 38% of C3H DFL/DSP sequences used DFL16.1. Because the DFL16 gene was used in such a large percentage of the rearrangements, we also used another primer, AF216, in other amplifications to more accurately determine the frequency of usage of individual DSP genes. In BALB mice, the AF216 primer equally amplifies DFL16 and DSP2 family genes, but in Ighj mice AF216 does not amplify DFL genes, presumably due to divergence of the germline DFL.16j gene at the site of the primer. Thus, we used DSP2 sequences from both types of amplifications to assess the relative usage of DSP2 genes (Fig. 2⇓). The frequency of rearrangement of the individual DSP genes was similar between the two strains, other than the lower frequency of DSP2.5 rearrangements in DJ rearrangements from pre-B cells from MRL/lpr mice. However, we do not see this difference in usage in sequences derived from C3H and MRL/lpr pro-B cells (data not shown), nor is it evident in the VDJ rearrangements (Fig. 3⇓). Hence, we conclude that there is not any major difference in the frequency of rearrangement of the DSP2 genes between the two strains. The novel long DSP2.12 is not one of the more frequently rearranging DSP2 genes in either C3H or MRL mice (Fig. 2⇓).
Usage of individual DSP2 genes in DJ rearrangements from adult bone marrow pre-B cells.
Individual DH usage in VDJ rearrangements from adult bone marrow pre-B cells.
Usage of D genes in VDJ rearrangements, and a novel VH7183 gene
To determine whether the frequency of individual D gene usage was altered between DJ rearrangements and VDJ rearrangements in pre-B cells, and to determine whether any D-D joinings occurred in VDJ rearrangements, we also analyzed VDJ rearrangements from adult bone marrow pre-B cell DNA (Fig. 3⇑). Primers for the VHJ558, VHS107, and VH7183 families were used. We have recently sequenced all of the VH7183 genes in Igha mice, and all but two of the 20 VH7183 genes have 2 bp in CDR3, either CA or GA (18). In our analyses of the VH7183 sequences from MRL/lpr mice, we often observed the sequence CAAGAC at the start of CDR3, and this was always preceded by a 1-bp variation in the codon after the invariant cysteine (TGT GTA AGA). The T in the middle codon is unusual for VH7183 genes. Therefore, we surmised that this was likely to be a new VH7183 gene in MRL mice. We amplified DNA from unsorted MRL/lpr bone marrow with primers located in FR1 and 3′ of the RSS and confirmed our prediction that this was indeed a new germline VH gene, with the 3′ terminal sequence of … TGT GTA AGA CAA GAC A-RSS (GenBank accession number AF428078). The RSS differs by only 2 bp in the spacer from the prototypic VH7183 group I RSS of the Igha haplotype (18). We have also begun to sequence other unrearranged VH7183 genes, all of which so far differ in coding sequence from the Igha VH7183 genes, other than 81X.
In general, the DSP2 gene segment usage was similar from D-J to VDJ (Figs. 2⇑ and 3⇑). However, DFL16.1 was lower in usage in VDJ rearrangements than in pre-B rearrangements. The primer used in the DJ rearrangements, AF288, had one degenerate position in the spacer, because DFL and DSP varied at that site, and it also had one position in the added restriction site which was identical with DFL but varied by 1 bp from DSP. Thus, this primer may have slightly favored amplification of DFL. Because the primers for the VDJ amplifications are in the VH and JH regions, the frequency of D gene usage in VDJ rearrangements is unbiased. The lower frequency of use of DFL16 in C3H pre-B cells as compared with MRL that we observed in both DJ and VDJ rearrangements is in agreement with the lower frequency of usage of DFL16.1 in VDJ rearrangements of C3H mice previously reported by Klonowski et al. (11). We found only three D segments which appeared to be inverted D genes from our 295 VDJ rearrangements. Analysis of VDJ rearrangements allowed us to determine the frequency of rearrangement of DQ52 and DST4 genes, which were not amplified by the DSP/DFL primer. Both are present at low frequency, as they are in BALB/c mice also (16, 17).
V-D-D-J rearrangements
We did observe some CDR3 regions from adult bone marrow pre-B cells which had clear-cut examples of the use of two D regions (Fig. 4⇓). For these analyses, we used the criteria that each DH gene should have five or more continuous nucleotides identical with a D region which could not have been accounted for by VH, JH, or P nucleotides. Thus, we may have underestimated some D-D joinings which contained smaller regions of the D sequences. We observed six of 130 (4.6%) CDR3 from MRL/lpr pre-B cells which contained apparent D-D rearrangements, and seven of 165 (4.2%) from C3H. Finally, we analyzed 200 adult bone marrow pre-B cell sequences from BALB/c mice and observed four examples of apparent D-D joins in the BALB pre-B sequences (2%). Thus, it appears that all strains of mice have VDDJ rearrangements in bone marrow pre-B cells.
VDDJ rearrangements in adult bone marrow pre-B cells. Sequences are shown with the 3′ end of the VH sequence and the 5′ start of the JH sequence in lowercase letters. All D sequences are underlined and identified (F, DFL16; SP, DSP2). When a D segment is too short to be definitively identified, all potential DSP2 genes are listed. P nucleotides are written adjacent to the relevant gene and are italicized. Parentheses indicate regions of homology-directed recombination.
We have also observed apparent VDDJ rearrangements in the spleen. In MRL/lpr spleen, we observed one VDDJ from 18 rearrangements, and in our previously published analysis of cDNA from LPS-stimulated spleen cells from BALB/c mice, we observed five of 397 (1.3%) sequences with D-D rearrangements (Table I⇓) (16).
Frequency of VDDJ and DDJ rearrangements in adult and newborn mice
VDDJ rearrangements could be generated by VD, DD, or D-DJ intermediates. Within our DJ sequences we observed 3 of 140 DJ rearrangements with two D segments from MRL/lpr pre-B cells, and 2 of 163 DDJ rearrangements from C3H pre-B cells (Fig. 5⇓).
D-DJ rearrangements in adult bone marrow pre-B cells. Sequences are depicted as described in Fig. 4⇑.
Lack of D-D rearrangements in newborn sequences
In contrast to the adult bone marrow pre-B cells, we did not observe any D-D joins in 30 VDJ sequences from MRL/lpr newborn liver (Table I⇑). Also, our reanalysis of the 18 published examples of D-D joins and atypical CDR3 regions from the published study of Klonowski et al. (11) revealed only one inverted D region and no examples of D-D joins from their 141 newborn MRL/+ rearrangements. Furthermore, we have previously sequenced over 300 fetal and newborn sequences from BALB/c mice (Ref. 16 and unpublished data) and have found no VDDJ rearrangements (Table I⇑).
Discussion
The analysis of CDR3 regions of autoantibodies from the autoimmune prone strain MRL has been of importance in probing the structural basis of reactivity with autoantigens, particularly anti-dsDNA. However, such studies have been hampered by the lack of information about the germline repertoire of the D and V genes involved. In this work we report that the Ighj haplotype has three novel DH genes, or alleles of genes, as compared with the BALB/c or C57BL/6 mice. One is very unique, in that it is a very long DSP2 gene. Finding alleles of DSP2 genes is not without precedence; a predominantly used DSP2 gene in C57BL/6 is DSP2.X which is not present in BALB/c mice (20). However, in this study we show that the DSP2 repertoire of Ighj mice is quite different from Igha mice, apparently missing three DSP2 genes present in BALB/c, and having instead not just the closely related new allele DSP2.13 but the highly novel long DSP.12 gene. These findings are in accord with previously published Southern blot data which showed different patterns of hybridization among various Igh haplotypes, including the ones studied here, with a DSP2.2 probe (21). In that study it was also shown that all strains analyzed contained a 4-kb EcoRI band containing DFL16.1, but that only Igha strains had the weaker hybridizing 6.7-kb band corresponding to DFL16.2. In accord with the latter observation, we did not find any DFL16.2 sequences in any DJ or VDJ rearrangements and therefore presume that it is not present in Ighj mice. However, for DFL16.1, conservation of band hybridization does not preclude the sequence variation between haplotypes which we observed in the coding region, and in at least one place in the 5′ flanking region (at the site of the AF216 primer) between the Igha and Ighj haplotypes.
We also observed one VH7183 gene with an unusually long portion in CDR3 region, and this gene was used frequently in our rearranged sequences. However, because we do not know the sequence of all of the VH7183 genes in MRL mice, we cannot be certain whether this gene rearranges more frequently, or whether our VH primer preferentially amplifies it. A search of GenBank for the new VH7183.1j gene did not show any sequences with a perfect match to this new gene.
It has been speculated by many that autoantibodies, particularly anti-dsDNA Abs, may be enriched in atypical CDR3 regions, including inverted D sequences or D-D joinings, although a detailed study of anti-dsDNA Abs in (NZB × NXW)F1 mice indicated that in this strain most of the arginines were encoded by N regions and alternative reading frames, but not inverted D regions or D-D (3, 4, 5, 6, 7, 9, 11, 12). The long DSP2.12 which we describe in this work has in fact been used in a few published sequences of anti-dsDNA and anti-histone autoantibodies and in rheumatoid factor Abs from MRL/lpr mice (3, 22, 23). However, despite the existence of the new D genes/alleles which we describe here, several of the apparent V-D-D-J rearrangements in published sequences of hybridoma autoantibodies appear to be actual D-D rearrangements, although the analysis of autoantibodies in lupus is hampered by the fact that these Abs have undergone extensive somatic hypermutation, adding much ambiguity to the prediction of the original unmutated sequence. The three new alleles described in this report do not directly encode any arginines, and even in alternative reading frames DSP2.12 and DFL16.1j will only encode one arginine each. Thus, the predominant effect of DSP2.12 and of the new VH7183 gene would be to encode a subset of Abs which could have longer CDR3s than average Abs.
Despite the reduction in the estimate of frequency of D-D rearrangements due to the discovery of DSP2.12, MRL/lpr mice still do have some apparent D-D rearrangements, with 4.6% of pre-B cell rearranged sequences displaying CDR3 regions with at least 5 bp of a second D region. Much stricter criteria for the definition of D-D joins have been proposed for human VDJ sequences by Corbett et al. (24). However, these criteria are not directly applicable to murine DH genes, because less than one third of the human DH genes are as short as the murine DSP2 genes. However, we acknowledge that we cannot be certain that these are not N regions which fortuitously match a D sequence, and often one of the two D segments is shorter than average. Nonetheless, these sequences meet most definitions of D-D rearrangements and appear to be bona fide DD joinings. D-D joining break the 12/23 spacer rule for V(D)J recombination; however, studies with recombination substrates show that 12/12 spacer RSSs, or heptamer-only RSSs, can rearrange at 1–2% of the frequency of consensus RSS (25), and D-D rearrangements have been observed in several studies of unselected human and murine Ab repertoires (15, 16, 26, 27, 28).
VDDJ rearrangements can theoretically be assembled in any of three ways. A D segment could join to a DJ rearrangement, followed by a VH joining to the DDJ. An alternative pathway would be the creation of a VD rearrangement, which would subsequently join to a DJ rearrangement. Finally, the first step could be the rearrangement of two D segments, followed by joining of a JH and then a VH. VD rearrangements and DD rearrangements have previously been described (27, 29). In this work we show evidence for DDJ rearrangements, although we do not know whether they arose by a DD to J or a D to DJ pathway. Because the D and J loci are accessible at earlier stages of B cell development than is the V locus, one might have predicted that DDJ joins might be more likely to be an intermediate than VD rearrangements, particularly because continuing DJ rearrangements are known to occur. We found DDJ rearrangements in our DJ sequences at less than half the frequency of VDDJ rearrangements, suggesting that all of these mechanisms may be operative in the formation of these VDDJ sequences.
One surprising outcome of this analysis was the total lack of VDDJ rearrangements in newborn mice. This suggests fundamental differences in the V(D)J rearrangement process during pre-B cell development in the neonatal period and the adult. It is possible that the transit time for B cell development is faster in the newborn, and so conventional VDJ rearrangements are the only ones which have time to occur. Alternatively, although the sole role of TdT has been thought to be that of adding N regions to VDJ junctions, recent data suggest that it may play a role in specific V gene use (30). Thus, it is possible that another unconventional role for TdT would be to facilitate D-D joinings. These possibilities are being explored.
Footnotes
-
↵1 This work was supported by National Institutes of Health Grant AI29672. This is manuscript number 14272-IMM from The Scripps Research Institute.
-
↵2 Address correspondence and reprint requests to Dr. Ann J. Feeney, Department of Immunology, The Scripps Research Institute, IMM-22, 10550 North Torrey Pines Road, La Jolla, CA 92037. E-mail address: feeney{at}scripps.edu
-
↵3 Abbreviations used in this paper: CDR3, complementarity-determining region 3; RSS, recombination signal sequence.
- Received July 24, 2001.
- Accepted October 22, 2001.
- Copyright © 2001 by The American Association of Immunologists
References
In this issue
