Systemic lupus erythematosus (SLE) is characterized by polyclonal B-lymphocyte activation, which leads to production of autoantibodies with various specificities, principally against nuclear antigens including double stranded (ds)DNA, ribonucleoprotein particles, histones and nonhistone chromatin proteins. Other antibodies bind cell surface structures and cytoplasmic antigens. Of these, serum high-affinity IgG antibodies that are specific for native dsDNA are believed to be the principal pathogenic agents and are used as a diagnostic indicator [1]. These autoantibodies differ from anti-DNA antibodies found in the sera of healthy individuals in that they bind to dsDNA with high affinity, they are often cationic in charge and they do not usually cross-react with unrelated antigens [2]. Despite intensive study, the factors that lead to the production of such autoantibodies remains in dispute, although a number of hypotheses have been suggested. Previous studies, using serum antibodies, hybridomas generated from peripheral blood lymphocytes (PBLs) and mouse models, concluded that autoantibodies produced in SLE are associated with particular properties. These include expression of characteristic idiotypes, clonal restriction of anti-DNA and anti-Sm antibodies, somatic hypermutation, V gene bias, and the presence of positively charged complementarity determining region (CDR) residues or sequence motifs in anti-dsDNA antibodies [3].

V(D)J rearrangement of immunoglobulin genes has the capacity to generate an immense repertoire of immune receptors that are able to recognize virtually any foreign substance via somatic hypermutation, but because of the nature of this process a number of immune receptors with specificity for self molecules are also generated. These self-reactive B cells are normally eliminated in the bone marrow, but self-reactivity can also be generated in the periphery by somatic mutation. For example, mutation of a single amino acid at position 35 on the heavy chain culminates in a switch from anti-phosphoryl choline (a bacterial hapten) to anti-dsDNA [4]. This supports the hypothesis that the aetiological stimulant of the autoimmune response observed in SLE may be of bacterial or viral origin, and this is further supported by the observation that the anti-DNA response is clonally restricted in both mouse models and SLE patients [5, 6]. This hypothesis suggests that self-reactive B cells may arise from B cells that express low-affinity V genes, which are cross-reactive for self and microbial antigens, by somatic hypermutation to generate high-affinity autoantibodies.

There have been only a limited number of studies on the immunoglobulin V gene repertoire and somatic hypermutation in SLE, with the majority of those investigations performed in PBLs. PBLs comprise a population of recirculating memory cells that have encountered a vast range of antigens, including many environmental antigens, over a prolonged period of time, whereas germinal centres (GCs) in the spleen or lymph nodes provide a profile of B cells that respond to antigen at a given time point. In an earlier investigation, Ravirajan and coworkers [7, 8] demonstrated the presence of autoantibody-producing B cells in the spleen of an SLE patient by analysis of hybridomas generated from splenic B cells. The aims of the present study were to identify the immunoglobulin V genes used by proliferating B cell clones in GCs of a SLE spleen, and to determine whether there are abnormalities in the pattern of somatic hypermutation and antigen selection. To our knowledge, this is the first detailed study of the repertoire of the splenic GC response in SLE.

The immunohistology of the spleen from the SLE patient produced a picture similar to the cellular architecture of healthy spleens in mice [15] and humans [16], which is known to be interspersed with GCs. In the present study, however, a mantle surrounding the GC was not identified. Each mature GC is generally derived from one to three B-cell clones, which manage to survive a significant reduction in clonal diversity and then go on to endure V(D)J hypermutation. The GC is of most interest because it is the site of antigen driven V(D)J hypermutation and selection [15] where antigen-specific B cells acquire point mutations in the V regions of transcriptionally active rearranged immunoglobulin genes. These mutations accumulate steadily during expansion of B-lymphocyte clones in the dark zone of the GC. This clonal evolution occurs independently in each GC, because little trafficking of B cells between GCs has been observed [17].

The cellular components necessary for a GC response were present in both GC A and GC B (i.e. B cells, FDCs capable of presenting antigen and T cells). However, none of the GCs and B-cell clusters had a discernible mantle zone. No plasma cells were located within the GCs themselves but they were loosely distributed in the surrounding tissue within close proximity to the GCs (data not shown). The immunohistochemistry also demonstrates that both GCs are of approximately equal size, which validates comparison of the data produced from each.

A recent study in Science [18] identified autoreactive B cells in MRL.Fas^lpr mice proliferating in the T-cell zone of lymphoid tissues. This was thought to be due to their deficiency of the Fas receptor, because these cells would normally be deleted through the Fas receptor/Fas ligand-mediated pathway of apoptosis. A similar explanation may account for the GCs identified in this study not exhibiting a traditional mantle, and explain the lack of negative selection illustrated by the low number of nonfunctional genes.

Of the total number of rearranged V_H genes, 19% were found to contain stop codons and out-of-frame rearrangements, and were therefore deemed to be nonfunctional, the majority being found in cluster D. This correlates with a similar study conducted by Jacobi et al. [19], who found 13% of PBL V_H gene sequences amplified from an SLE patient to be nonfunctional, as compared with 53% of genes amplified from PBLs of a healthy individual. Those investigators observed a similar R/S ratio in the productive rearrangements to that seen in the nonproductive rearrangements. It was therefore suggested that there may be some abnormality in selection in SLE related to an intrinsic failure of B-cell apoptosis or enhanced B-cell activation by T cells, which overwhelms protective mechanisms that are effective in normal individuals.

The most abundantly expressed V_H gene family was V_H3, which is not surprising because it is the largest family and has been found to be the most dominant in the normal repertoire [20]. What is perhaps more interesting is the large number of V_H5 sequences (16.6%) as compared with the 3.9% expected to be produced randomly in the normal human repertoire. V_H5–51 has also been isolated from breast tumours (Nzula et al., unpublished data) and thymic GCs from a patient with myasthenia gravis [9] previously in our laboratory, but not from GCs from the salivary glands of two patients with Sjögren's syndrome, using the same primers [21]. It is therefore unlikely that the family V_H5 primers used here preferentially amplify V_H5–51. V_H5 sequences have been consistently associated with IgE antibodies, and it has been suggested that these antibodies may be associated with an unidentified superantigen [22]. Two studies analyzing anti-DNA antibodies in SLE both identified a heavy chain clone comparable to V_H5–51 [23, 24]. Comparison of our V-D-J rearrangements with the two anti-DNA specific antibodies revealed very few somatic mutations in common, however. One hotspot highlighted at position 77 (in the tip of the FR3 loop) [25] was found to be present in the anti-DNA specific antibodies as well as in most of the analyzed V_H5–51 sequences from this study, but this was not often the same amino acid substitution.

V_H5–51 was also used by two human IgG monoclonal antibodies that bind phospholipid, derived from the PBL of a SLE patient. In this instance the J segment used was J_H6b, which was not found in combination with V_H5–51 in the present study. It may be significant that patient M produced anticardiolipin antibodies, but not anti-dsDNA, although there is no direct evidence that the cardiolipin antibodies used V_H5–51 because no information about the specificity of these antibodies was obtained.

A complete absence of V_H1 family genes was observed in the present study of splenic GC, which contrasts with the theoretically expected value of 20–30%. This was also found to be the case in another study of SLE patients conducted by Hansen et al. [26]. They found that 13% of functional PBL V genes from healthy control individuals were from the V_H1 family, but only about 1% of the V_H genes expressed by PBLs of a SLE patient used this family. de Wildt et al. [27], on the other hand, found very little difference between expression of V_H1 in healthy control individuals and SLE patients. The PBL B-cell repertoire may not reflect the repertoire of splenic B cells.

Using high-fidelity Taq polymerase and the same primers as used here, the PCR error rate for V_H genes was less than one base per four V_H genes [9]. The average numbers of single base mutations in the V_H genes from GC A and GC B were 7.6 and 15.8, respectively, and for clusters C and D they were 15 and 14.9, respectively. This is significantly higher than the Taq polymerase error rate, demonstrating somatic hypermutation in vivo. The difference in mutation rates between the two GCs may indicate that they are in different states of maturity or that GC B might have been founded by a memory B cell that was already mutated (e.g. the founder cell of clone B-b; Fig. 5). Hypermutation within a GC is closely linked to antigen-induced B-cell proliferation; thus, from the data presented here, this appears to be the case in both GCs and the B-cell clusters. The second phase of this process, during an immune response against a xenoantigen, is selection of B cells that express high-affinity antigen receptors resulting from rare mutations, by competition for binding to antigen on the surface of FDCs. This process is generally believed to result in an increase in the ratio of replacement to silent mutations, especially within CDRs, and often in selection against replacement mutations in the FRs. The evidence presented here is supportive of antigen selection in GC B and clusters C and D, in which the R/S ratio is higher than random in the CDRs. Selection for replacement mutations in the FRs has previously been observed, for example in two anti-hen egg lysozyme antibodies (HyHEL-5 and HyHEL-10) [28, 29], in which both had contact residues in the FRs. It is also possible that the high R/S ratio in the FR region of the sequences in this study are the result of antigen selection, but this was not confirmed because of the absence of data on specificity.

A serine codon bias was also observed (not shown), with 70% of the serines in CDR1 and CDR2 represented by AGC or AGT, both of which are recognized targets of the hypermutation machinery (for review [30]). Only AGC and AGT serine codons produced replacement mutations in the CDRs. Of these, the mutation from serine to asparagine was the most prevalent, which is in accordance with mutational analysis performed on patients with myasthaenia gravis [9].

The clonal genealogies show that the groups of rearranged V_H genes included sequences that could be assigned to parental and daughter cells on the basis of shared mutations and junctional sequences. By far the most dominant sequences were V_H3–30*01 and V_H5–51*01 (approximately 16% each). In studies of PBLs in SLE, the most common V_H segment observed was V_H3–23 (12%) [27], which we have seen in similar numbers (11%).

The present study highlights a possible bias toward expression of V_H5 immunoglobulin V genes by splenic GC B cells in SLE, as well as normal high levels of V_H3 and an under-representation of V_H1. It also confirms that there is a defect in the negative selection process in SLE patients.