Rheumatoid factor isotype switch and somatic mutation variants within rheumatoid arthritis synovium

Abstract

The presence of clonally-related B-lymphocyte aggregates within synovial lining tisue of rheumatoid arthritis (RA) patients suggests a germinal centre-like reaction, which may hold implications for disease pathogenesis and the causes of chronic inflammation. We studied 250 rheumatoid factor (RF) heavy-chain sequences cloned from the synovium of three patients with RA, to determine whether they undergo both somatic mutation and isotype switching consistent with this hypothesis. Size analysis of immunoglobulin heavy-chain cDNAs from synovial RF⁺ B cells revealed oligoclonal RF⁺ populations and identically-sized V_H-D-J_H transcripts of different immunoglobulin isotypes. Sequencing of individual inserts selected from cloned immunoglobulin heavy-chain cDNAs demonstrated a clonal relationship between immunoglobulin M (IgM) RF and IgA RF, suggesting that this isotype switch occurred in synovium. Furthermore, most somatic mutations were found to have occurred after this isotype switch. This finding suggests that the RA synovial microenvironment sustains somatic mutation and isotype switching in RF-specific B lymphocytes akin to secondary lymphoid organs.

INTRODUCTION

Inflamed synovial tissue of rheumatoid arthritis (RA) patients contains B lymphocytes and plasma cells, producing immunoglobulins, some with specificity for self-immunoglobulin G [IgG; rheumatoid factors (RF)].¹ These synovial B cells form aggregates which resemble germinal centres of secondary lymphoid organs. These structures appear to support clonal expansion and somatic hypermutation in the synovium, indicating that abnormal proliferation may occur continually in the synovial tissue. Moreover, serum RF, which is a diagnostic criterion of the disease, is suggested to represent a flow-over of RF synthesized in the synovium.² Whilst the exact way in which RF contributes to RA pathology remains unclear, patients with high serum levels of IgM RF (seropositive patients) have more aggressive disease and poor prognosis. Furthermore, elevated titers of IgA RF and IgG RF are associated with bone erosion and extra-articular disease, respectively.^3–5 Synovial infiltrate of seropositive patients produces RF, but its isotype distribution is unclear. During progression of RA, there is a gradual reduction in the expression of public idiotypes on RF,⁶ and an increase in RF affinity.⁷ Recent studies have suggested that some B lymphocytes in the synovial aggregates of RA patients are clonally related, suggestive of a germinal centre-like reaction.^8,9 Furthermore, Randen et al. showed that clonal expansion and somatic mutation were associated with increased binding affinity in potentially pathogenic monoclonal RFs derived from hybridomas which were generated from B lymphocytes in the synovial membrane of patients with RA.¹⁰ However, these important aspects of RF production remain poorly understood. Intriguingly, RFs from the peripheral blood of immunized healthy individuals also have evidence for somatic mutation in their immunoglobulin variable region genes but display relatively low binding affinity to IgG.¹¹ Furthermore, there are no adequate studies of the clonal relationship between non-manipulated RF sequences isolated from the diseased tissue. In the current study we have examined the nature of the RF response maturation in the synovial tissue. Clonally related synovial RF were identified by immunoglobulin cDNA spectrotyping, which is based on size variation within the CDR3 of the rearranged V_H-D-J_H cDNA.¹²

MATERIALS AND METHODS

Patients

Three female patients with long-standing RA¹³ aged 80 (patient A), 68 (patient B) and 63 (patient E) years, attending Royal Birmingham Orthopaedic Hospital (UK) and being treated with non-steroidal anti-inflammatory drugs at time of synovectomy were investigated

Cell isolation

Synovial mononuclear cells (MNCs) were separated by centrifugation on Ficoll–Paque of 70-μm nylon mesh (Beckton-Dickinson, Oxford, UK) -filtered cells from collagenase (5 mg/ml) -digested diced synovium. Monocytes and natural killer (NK) cells were eliminated by 35-min incubation with l-leucine methyl ester (5 mm) (Sigma, Poole, UK). T cells were rosetted with neuraminidase-treated sheep erythrocytes. B-cell-enriched MNCs (0·5–1 ml) were separated into RF⁺ and RF⁻ B cells by binding (4° for 30 min) to M-450 Dynabeads (0·4×10⁷; Dynal, Wirral, UK) coated with heat-aggregated (63°,15 min) human IgG4

B-cell culture and RF detection by ELISA

B-cell-enriched synovial MNCs were cultured according to Banchereau et al.¹⁴ with anti-CD40 monoclonal antibody (mAb) (clone G28-5: 0·5 μg/ml), interleukin-4 (IL-4) and IL-10 (5 ng/ml) on mouse Fcγ RII-transfected L cells. Total immunoglobulin and RFs were quantified in 10-day culture supernatants by enzyme-linked immunosorbent assay (ELISA) using 96-well plates coated with either mAbs specific for immunoglobulin isotypes (clones: AF6 anti-IgM; 8a4 anti-IgG; 2D7 anti-IgA; all kindly provided by Professor R. Jefferis, University of Birmingham). Bound immunoglobulins were revealed with peroxidase-conjugated goat F(ab′)₂ anti-human μ-, α-, or γ-chain antibodies. IgM RF and IgA RF were detected using ELISA plates sensitized with human IgG4 or IgG1 paraprotein (25 μg/ml; 37°, 2 hr), and bound antibodies were detected with peroxidase-conjugated goat F(ab′)₂ anti-human μ or α light chains. IgG RF was detected using ELISA plates sensitized with a human IgG Fc paraprotein (Fc-Per) and bound IgG RF revealed with the mouse monoclonal antibody ZB8 which has specificity for the Fab region of human IgG isotypes (both reagents provided courtesy of Professor R. Jefferis). In supernatants from lines with no detectable RF activity, as detected by the above methods, peroxidase-conjugated anti-light chain antibody was used to ensure the detection of all RF in the culture supernatants

Extraction of mRNA and cDNA synthesis

One microgram of mRNA from fractionated B cells (Microfast mRNA kit; Invitrogen BV, Leek, the Netherlands) was reverse-transcribed with SuperScript II RT (Life Technologies, Paisley, UK) (200 U; 1 hr at 37°), then elongated with a 3′ dG tail using TdT (Pharmacia, St Albans, Herts, UK)

Anchor polymerase chain reaction (PCR) amplification

Poly(G)-tailed cDNA (5 μl), anch2pc primer (Oswel, Southampton, UK), dNTP (0·175 mm), in PCR buffer (50 μl) were incubated (96°, 5 min; 60°, 5 min, then 70° for 15 min) with 0·2 U of pfu-polymerase (Promega, Southampton, UK), then 3′ primer (100 ng) [one of six C_H primers: CM1w (IgM), CA1w (IgA), CG1w, CG2w, CG3w, CG4w (IgG subclasses 1–4)¹²] was added and cycled: 92° for 1 min; 60° for 1 min; 72° for 1 min; for 30 cycles, with final extension at 71° for 10 min.

1° nested PCR for cloning and 2° nested PCR for cDNA spectrotyping

These were performed as previously described.¹² The anchor-primed PCR product (0·5 μl) was amplified using pfu-polymerase (0·2 U per 50 μl reaction), V_H and C_H primers (0·25 μm), dNTPs (0·2 mm), in dimethylsulphoxide (DMSO; 10% v/v), with 35 cycles of: 92° for 15 seconds, 52° for 50 seconds, 72° for 90 seconds to generate V_H-D-J_H-C_H1 product. The 2° nested PCR (30 cycles) was done with the same V_H primers and a consensus J_H primer, to generate V_H-D-J_H products. The V_H primers^12,15 were designed to amplify most of the V_H repertoire; primers CM1w, CA1w and CG(1–4)w were in the hinge of IgM, IgA and IgG(1–4), respectively. The 5′–3′ sequences of the primers were: anch2pc, 5′-ACGAATTCTAGAGTCGACCCCCCCCCCCCCC-3′; CA1 w, 5′-AGGTGGAGTTGAACTAGTTGGGCAGGGCACAGT CAC-3′; CM1w, 5′-GAAGACGCTCACACTAGTAGGCA GCTCAGCAATCAC-3′; CG1w, 5′-CGGTGGGCATGTACTA GTTTTGTCACAAGATTTCCC-3′; CG2w, 5′-TGGGCACTC GACACTAGTTTTGCGCTCAACTGTCTT-3′; CG3w, 5′-TG GGCATGT GTGACTAGTGTCACCAAGTGGGGTTTT-3′; CG4w, 5′-TGCTGGGCATGAACTAGTTGGGGGACCATAT TTGGA-3′; VH1a, 5′-CAGGTGCAGCTCGAGCAGTCTG GG-3′; VH1f, 5′-CAGGTGCAGCTGCTCGAGTCTGGG-3′; VH2f, 5′-CAGGTGCAGCTACTCGAGTCGGG-3′; VH3a, 5′- GAGGTGCAGCTCGAGGAGTCTGGG-3′; VH3f, 5′-GAGG TGCAGCTGCTCGAGTCTGGG-3′; VH4f, 5′-CAGGTGCAG CTGCTCGAGTCGGG-3′; VH4g, 5′-CAGGTGCAGCTACT CGAGTGGGG-3′; VH6f, 5′-CAGGTACAGCTGCTCG AGTCAGGTCCA-3′; JH, 5′-ACCTGAGGAGACGGTGACC AGGGT-3′.

Complementary DNA spectrotyping

This has been described previously.¹² The 2° PCR product in formamide (90%v/v)/ethylenediamine tetraacetic acid (EDTA) 1 mm/NaOH 10 mm (20–200 μl), was electrophoresed in Tris, borate, EDTA (TBE)/urea polyacrylamide gels (4% w/v, 0·4 mm thick), at 70 watts for 200 min, and was silver-stained (Promega). For calibration, mixture (Y) of the 2° nested PCR products of clones with V_H-D-J_H lengths: 115, 116, 118, 119, 120, 121, 122, 123, 124, 125, 127, 129, or 131 codons or mixture (X) of the 2° nested PCR products of clones with V_H-D-J_H lengths: 118, 119, 120, 121, 122, 123, 124, 125, or 127 codons were applied to each gel. The length indicated is the number of codons within the complete V_H-D-J_H region.

Cloning and sequencing of cDNA

The V_H-D-J_H-C_H1, 1° PCR product was ligated (T4 DNA ligase: Life Technologies) into pT7-Blue(R) vector (Novagen, Madison, NJ) and transformed Escherichia coli XL1 blue. The insert size of cDNA clones was determined by 2° PCR and the DNA sequence was determined using manual T7 DNA Polymerase (Sequenase: USB, Cleveland, OH) or automated Taq-FS DNA polymerase (Perkin Elmer Ltd, Warrington, UK), with primers: ‘U-19mer’ and ‘T7 promoter’ (Novagen). Sequence analysis used Lasergene software (DNAstar, Madison, NJ)

RESULTS

Efficiency of separating RF⁺ B cells using human IgG4-coated magnetic beads

Magnetic beads were coated with an IgG4 paraprotein, which is known to have a high affinity for RF but has no affinity for the Fc receptor (FcR-II) present on B cells (see the Materials and Methods). Uncoated beads did not bind B cells. B-cell-enriched MNCs, and the RF⁺ and RF⁻ B cells from the separation using IgG4-coated beads, were cultured for 10 days with anti-CD40 and IL-4 stimulation (see the Materials and Methods¹⁴) in 96-well plates. Supernatants were assayed for total immunoglobulin, and RF production by ELISA. Forty-eight per cent of immunoglobulin-producing lines from MNCs, in three experiments, contained RF. In contrast, 4% and 9% of fractionated RF⁻ B cells (two experiments) and 81% of RF⁺ B cells had RF activity. This indicated that the RF⁺ B-cell fraction was significantly enriched in B cells which could secrete RF in culture.

IgH cDNA spectrotypes of RF⁺ cells from three RA patients

Poly(dG)-tailed cDNAs from RF⁺ B cells of patients A, B and E were anchor-PCR-amplified using anch2pc primer and a C_H primer. Subsequent 1° PCR amplification, with a set of eight V_H primers and six C_H primers yielded 48 V_H-D-J_H-C_H1 cDNAs which were used for 2° PCR and cloning.

The PCR product size was determined by electrophoresis of V_H-D-J_H cDNAs from a 2° PCR amplification using V_H primers and a J_H primer. Electrophoretograms of 48 V_H-D-J_H PCR products from patient B (Fig. 1) showed IgM spectrotypes were clonally most diverse, containing three or four spectrotopes, compared with one or two spectrotopes in each IgG subclass (except for amplification with the V_H4g primer; Fig. 1, section 4G), but no IgA products (Fig. 1, lane A). The small number of spectrotopes indicated that the RF⁺ B-cell population was oligoclonal. Size identity of the V_H3f-primed, 123-codon IgG1 and IgM products (Fig. 1, lanes B and M: section 3F) suggests a clonal relationship. Similar 123-codon products were primed by V_H1a, V_H1f and V_H3a, which differ by only two and five of the 24 nucleotides; because the last eight 3′ nucleotides are identical these primers may amplify the same template

Figure 1.

IgA, IgG and IgM cDNA spectrotypes of RF⁺ cells from Patient B. Silver-stained DNA polyacrylamide (4%) gel electrophoretogram of V_H-D-J_H products of 2° PCR primed with eight V_H primers (Figure sections labelled V_H:1A; 1F; 2F; 3A; 3F; 4F; 4G; 6F) together with the consensus J_H primer. The templates were derived from the RF⁺ synovial cells from Patient B by previous PCR amplification using the same V_H primers together with six C_H primers CA1w (lane A), CG1w (lane B), CG2w (lane C), CG3w (lane D), CG4w (lane E) and CM1w (lane M). The template-negative PCR control using mixed primers (lane O) and the set of calibration markers of known codon lengths (lane X), together with arrows identifying 118 and 127 codon-length bands are also shown.

Each IgG spectrotype from patient E was oligoclonal (Fig. 2, lanes B, C, D and E) containing one or two bands compared with the more polyclonal 11 bands for IgM (Fig. 2, lane M). No IgA products were detected (Fig. 2, lane A). Interestingly, a much larger 140-codon spectrotope was resolved, in the IgG4 products only, from PCR amplifications using V_H1f, V_H2f, or V_H4g (Fig. 2, lane E: sections 1F, 2F, 4G).

Figure 2.

IgA, IgG and IgM cDNA spectrotypes of RF⁺ cells from Patient E. Silver-stained DNA polyacrylamide (4%) gel electrophoretogram of V_H-D-J_H products of 2° PCR primed with eight V_H primers (figure sections labelled V_H:1A; 1F; 2F; 3A; 3F; 4F; 4G; 6F) together with the consensus J_H primer. The templates were derived from the RF⁺ synovial cell population from Patient E as described for Patient B in Fig. 1.

The cDNA spectrotypes from patient A were oligoclonal (Fig. 3) containing at least 12 different spectrotopes, sized between 116 and 128 codons. The V_H2f-primed 125-codon IgA and IgG4 spectrotopes (Fig. 3, lanes A and E: section 2F), suggested a clonal relationship. Similarly the V_H3a-primed 124-codon IgG3 and IgM spectrotopes (Fig. 3, lanes D and M: section 3A) and the V_H4g-primed 120-codon IgA and IgM spectrotopes (Fig. 3, lanes A and M: section 4G) could be clonally-related.

Figure 3.

IgA, IgG and IgM cDNA spectrotypes of RF⁺ cells from Patient A. Silver-stained DNA polyacrylamide (4%) gel electrophoretogram of V_H-D-J_H products of 2° PCR primed with eight V_H primers (Figure sections labelled V_H:1A; 1F; 2F; 3A; 3F; 4F; 4G; 6F) together with the consensus J_H primer. The templates were derived from the RF⁺ synovial cell population from Patient A as described in Fig. 1. The template-negative PCR control using mixed primers (lane O) and the sets of calibration markers, X and Y, of known codon lengths (lanes X and Y), together with arrows identifying 118 and 127 codon length bands are also shown.

IgH cDNA spectrotypes of RF⁺ and RF⁻ B cells from patient A

The cDNA spectrotypes from RF⁺ and RF⁻ B cells from patient A were compared to analyse their resolution. Each RF⁻ B-cell population contained more spectrotopes than the RF⁺ population and their V_H-D-J_H lengths were different (Fig. 4), indicating our cell-affinity separation strategy isolated genetically distinct B-cell populations

Figure 4.

IgA, IgG and IgM cDNA spectrotypes of RF⁺ and RF⁻ cells from Patient A. Silver-stained DNA polyacrylamide (4%) gel electrophoretogram of V_H-D-J_H products of 2° PCR primed with six V_H primers (figure sections labelled V_H:1A; 3A; 3F; 4F; 4G; 6F) together with the consensus J_H primer. The templates were derived from the RF⁺ (RF⁺ section) and RF⁻ (RF⁻ section) synovial cell populations from Patient A. The template-negative PCR control using mixed primers (lane O) and the set of calibration markers, Y, of known codon lengths (lane Y), together with arrows identifying 118 and 127 codon length bands are also shown.

Isolation of clonally-related IgH cDNA clones

PCR products from different immunoglobulin isotypes containing same sized spectrotypes were cloned and re-analysed by cDNA spectrotyping (data not shown). The V_H-D-J_H length distribution of clones from patient A reflected the uncloned PCR products (Fig. 3). Five IgA, one IgG1 and 12 IgG4125-codon V_H1f-primed clones and seven IgG1, five IgG3 and 10 IgM 124-codon clones were isolated. Twenty-six IgA and nine IgM 120-codon V_H4g-primed clones were also isolated. Twelve IgG1, one IgG2 and nine IgM V_H1a-primed 122-codon clones were isolated from patient B. No clones with identical V_H-D-J_H lengths were isolated from patient E using different constant region primers.

Sequences of RF⁺ immunoglobulin clones

Ten 125-codon V_H1f and CG4w-primed clones were identical (Table 1) and confirmed as IgG4. Their V_H was 92·6% similar to germ-line gene DP-79 V_H4. These IgG4 clones were unrelated to four 125-codon IgA clones (primed with V_H1f and CA1w). These IgA sequences were related, and 94·0% similar to germ-line gene DP-58 V_H3 (Table 2). The IgA and IgG4 sequences were encoded by J_H4b gene. The R:S ratios of these IgA sequences were 4:1 in the complementarity-determining regions (CDRs), taking into account shared mutations only, but only 2·4:1 overall (Table 3). This suggests that the parental RF⁺ B cell containing the shared mutations was selected through antigen-binding, but that mutations in the progeny were accumulated without further selection.

Table 1.

IgG4 RF sequences of 125-codon length clones, from Patient A. Plasmid mini preps of V_H-D-J_H-C_HI cDNA clones amplified from RF⁺ synovial cells of Patient A using VH1f and CG4w primers (Figs 3 and 5) were sequenced. The sequences are compared and aligned with the germ-line V_H gene sequence DP-79 and J_H4b. Positions of identity are indicated by a dot. Numbering is according to Kabat et al.⁴⁰

Table 2.

IgA RF sequences of 125-codon+clones from Patient A. Plasmid mini preps of V_H-D-J_H-C_HI rearrangements, amplified from RF⁺ synovial cells of Patient A using VH1f and CA1w primers (Figs 3 and 5) were sequenced. The sequences are compared and aligned with the germ-line V_H gene sequence DP-58, DK5-A/DK-5B and J_H5b

Table 3.

R:S ratios for the V_H gene CDR and FR in each set of RF IgH cDNA clones

The 124-codon IgG1 clones (Table 4) differed only in the CDR2 region at positions 52a, 60, or 61, and were 97% similar to germ-line gene DP-77. The R:S ratio of the CDRs, taking into account only the shared mutations was 6:2, but 9:2, overall (Table 3). The 124-codon IgM clones were unrelated to these IgG1 clones, and were 98·9% similar to germ-line gene DP-67 (Table 5). The IgM and IgG1 124-codon clones utilized the D21–9 D_H gene. The IgM R:S ratios were 1:0 in the CDRs and 1:1 in the framework region.

Table 4.

IgG1-RF sequences of 124-codon clones from Patient A. Plasmid mini preps of V_H-D-J_H-C_HI rearrangements, amplified from RF⁺ synovial cells of Patient A using VH1f and CG1w primers (Fig 3), were sequenced. The sequences are compared and aligned with the germ-line V_H gene sequence DP-77, D21-9 and J_H5b.

Table 5.

IgM RF sequences of 124-codon clones from Patient A. V_H-D-J_H-C_HI rearrangements, amplified from RF⁺ synovial cells of Patient A using VH1f and CM1w primers (Fig 3). The sequences are compared and aligned with the germ-line V_H gene sequence DP-67, D219 and J_H5b.

The 120-codon IgA and IgM clones from Patient A (Table 6) were related and encoded by the DP-63, DXP′1 and J_H4b genes. The distribution of mutations was consistent with accumulation of mutations in the CDRs and framework regions (FRs) in several steps. Four replacement mutations and one silent mutation were common to this family of clones (codons 31, 35, 82a and 100). The overall R:S ratios for the IgM CDRs was 5·5:1 compared with 2:1 or 5:0 for related IgA clones (Table 3). Compared with the R:S ratios of 2:0 (CDRs) and 0:1 (FR) in the hypothetical parental rearranged immunoglobulin gene, the later mutations appear to be selected by antigen.

Table 6.

DNA sequences of 11 120-codon IgA and three IgM clones (901, 902, 1376) from RA Patient A using DP-63 (V_H4.21), DXP’1 and J_H4B. Plasmid mini preps of V_H-D-J_H-C_HI rearrangements, amplified from RF⁺ synovial cells of Patient A using VH1f and CM1w or CA1w primers (Fig 3). The three IgM sequences are clones 901, 902 ad 1376. The sequences are compared and aligned with the germ-line V_H gene sequence DP-6 3, DXP’1 and J_H4b

DISCUSSION

Clonal maturation and isotype switch from IgM RF to IgA RF occurs in RA synovium

It has been suggested that transient RFs occurring in healthy individuals, especially after infection, or immunization, are normal components of the immune system, probably with regulatory functions.^16,17 In RA, chronic RF production, isotype switching from IgM RF to IgG RF and IgA RF, predominant synthesis by B2 lymphocytes, somatic mutation and loss of germ-line gene configuration are hallmarks of an altered RF response.^18,19 RFs are synthesized in RA synovium, with potential alterations in participating accessory cells and regulation, compared to the non-RA individual.

The synovial IgG RF clones from patients A and B lacked a related IgM-RF counterpart. It is unlikely that lower RF affinity led to the inability to isolate the IgG-related IgM RF-producing B cells because our RF⁺ B-cell isolation strategy depends on polyvalent interaction, thereby increasing avidity of interaction with low-affinity RF-expressing B cells. Furthermore, the IgG RF spectrotypes were not diverse, so that masking of IgM-related IgG spectrotopes by abundant unrelated clones was unlikely. It is more likely that the lack of demonstrated clonal relationship between synovial IgG RF and IgM RF cDNAs, arises because class switch occurred outside synovium

In contrast, the isolation of clonally-related IgA and IgM RF cDNAs suggests that this class switch occurred in synovium. Association between serum IgA RF and bone erosion in RA could be related to this class switch, and the consequent IgA RF synthesis within the diseased joint. Mucosal lymphoid tissue is the conventional site of IgA switching and a source of synovial lymphocytes.²⁰ IgA secretion depends on IL-10, transforming growth factor-β (TGF-β) and dendritic cells.^21,22 TGF-β₁ is synthesized by synovial fibroblasts²³ and synovial macrophages,²⁴ whilst synovial fluid is rich in TGF-β₁ and TGF-β₂.²⁵ IL-10 is synthesized by synovial macrophages.²⁶

Somatic mutation and antigen-selection in isotype-switched synovial RF

In our preliminary experiments using Taq polymerase with a cloned template, the two rounds of PCR generated an error rate of 0·46% nucleotide substitutions. However, using pfu-polymerase, which has proof-reading ability, we have isolated 10 IgG4 clones (Table 1) and six IgM clones (Table 5) which are internally identical. These data indicate an error rate of less than 0·027% and suggest that the nucleotide substitutions we observed are not artefacts of the PCR technique. Our 120-codon IgA and IgM clones appear to have undergone several rounds of somatic mutation (Table 6). Of 68 nucleotide changes, only four (codons 31, 35, 82a and 100) occurred before the isotype switch and so are present in both the IgA and IgM clones. These four mutations could alternatively be allelic variants because comparisons were made with the V-Base directory, not with the patient’s genome. Co-localization of mutated IgA and IgM sister immunoglobulin mRNAs in synovium is evidence for local mutation and isotype switch. Only CDRs of these IgM-RF clones had an elevated R:S ratio (observed 5·5, predicted 4·2), suggesting clonal selection by antigen (Table 3). Contribution of mutated CDR residues to antigen contacts has been observed in the RF-AN light chain.²⁷ On the other hand, a lack of selective pressure on the other RF CDRs is suggested from their low R:S ratios (Table 3). Low R:S ratios in peripheral blood IgM RF sequences of 2·4 (CDR) and 0·9 (FR) are intermediate between RFs from healthy donors (selective pressure against replacement mutations) and anti-rhesus antibodies (affinity-matured, antigen-selected).²⁸ Although R:S ratios of three IgG RF hybridomas from rheumatoid synovial fluid were high,²⁹ this result may be influenced by bias during hybridoma selection. In our study, a single step of antigen-specific cell isolation, with multivalent attachment, was used, facilitating the isolation of low-affinity RF B lymphocytes.

Accumulation of mutations in the 120-codon long clone family is partially antigen-selected

A high level of CDR replacement mutations in the 120-codon IgM-RF clones, was not found in the related IgA-RF clones, although in every case the level of silent mutations in the CDRs was low (Table 3). Since most of these mutations were not observed in the IgA sister clones, it would appear that the mutations are not due to an intrinsic hot-spot corresponding to an antigen-independent mutational bias. Additional mutations in IgM clone 1376, and in IgA clone 892, were restricted to the FR (Table 6), whereas in IgA clone 1341, they were more abundant in the CDR. Thus, preferential accumulation of amino acid replacements in the CDR were mostly found in the IgM clones and in IgA clone 1341, suggesting that partial clonal selection may have occurred on the basis of antigen binding.

Of the six families of IgH clones examined, only the 120-codon long IgM family exhibited a high CDR R:S ratio. The rest had low R:S values consistent with low, or no, selective pressure by antigen (Table 3). Each RF clone family [excepting the 125-codon IgG4 clones (Table 1) and the 124-codon IgM clones, from patient A (Table 5)], had accumulated significant numbers of mutations, suggesting that the original RF⁺ B cells had undergone a germinal-centre-type reaction.^2,30 Progressive loss of cross-reactive idiotype on serum RF,⁶ increase in affinity⁷ and the association of mutation with affinity change¹⁰ suggest that RF immunoglobulin somatic mutation occurs in RA, but affinity maturation, and loss of lower affinity RF, is probably only partial. The survival of these low-affinity RF B cells may be due to the favourable environment within RA synovium provided by synoviocytes.³¹ Normal selection against replacement mutations in RF in healthy individuals²⁸ is by-passed in RA perhaps via this unique environment.

In our sample of clonally expanded synovial RF, there were three V_H4 and two V_H3 genes represented. This is consistent with previous observations of over-representation of V_H4 gene, in unselected immunoglobulin mRNA and hybridomas from RA synovium.^32,33 The V_H4 family contains 15 functional genes, including the DP-63/V_H4.21 gene, expressed in the 120-codon IgA and IgM RFs (Table 6). This gene, which is associated with autoreactive antibodies and RF,³⁴ is expressed by 2·8% of normal peripheral B cells and 14% of those expressing V_H4.³⁵ The DP-77/V_H3.21 gene expressed in the 124-codon IgG1 RFs (Tables 4 and 5) is the V_H3 gene most frequently expressed, mostly unmutated, by peripheral IgM-RF⁺ RA B cells.³⁶ In contrast, we have demonstrated clonal expansion of a DP-77 RF containing up to nine replacement mutations in the CDRs 1 and 2 (Table 3), consistent with clonal selection by antigen.

CDR3 motifs of synovial Rfs

In contrast to previous observations,³⁷ our RF V_H-D junctions did not contain glutamate. We found a tyrosine-rich motif (Y/F-D-x-S-G/D-Y-Y) in the CDR3 of the 124-codon IgM RFs (e.g. clone 785) and the 124-codon IgG RFs (e.g. clone 1011). This is also expressed together with D21/9 D gene in hybridoma C304,³⁸ and paraprotein WOL. This is the third observation of a common CDR3 motif in RF, the first being the tyrosine-rich Y-G-D-Y in RF-TS1 and the paraprotein KAS³⁷ and the second being the common D-J junctional amino acids in RFs encoded by the DP-54 and DP-10 V_H genes.³⁹ These findings suggest that CDR3 tyrosine contributes to RF binding to a common epitope. The D21/9 D_H gene was used in two of 15 V_H4-encoded synovial immunoglobulins, in 11% of immunoglobulins from normal peripheral blood B cells³⁵ and four of five RFs utilizing DP-54.³⁹ Common use of D21/9 in synovial RFs from RA patients, and peripheral blood immunoglobulin, may reflect common expression of RF, and its frequent utilization of D21/9 gene

Our data suggest that in the RA joint, each RF isotype is pauciclonal; there is clonal proliferation of IgG1, IgG4, IgM and IgA RF-specific B lymphocytes; and the RF isotype switch from IgM to IgA, but not to IgG, occurs locally. Confirmation of this would require demonstration of a lack of identical switched clones in a second joint sampled simultaneously. Further study of the relationship between circulating IgA-RF and IgA-RF B lymphocytes in RA synovium, and the association of local isotype-switched IgA-RF B lymphocytes with erosive RA would help to provide a mechanistic basis to the association of erosive disease with circulating IgA RF.

Abbreviations

CDR

complementarity-determining region

C_H1

first constant region domain of immunoglobulin heavy chain

FR

framework region

IgH

immunoglobulin heavy chain

R:S ratio

ratio of replacement mutations to silent mutations

RF

rheumatoid factor

VH, DH, JH, CH

variable, diversity, joining and constant gene segments of immunoglobulin heavy chain.