CD22 is a B cell–specific transmembrane glycoprotein that acts to dampen signals generated through the B cell antigen receptor (BCR): B cells from CD22-deficient mice give increased Ca2+ fluxes on BCR ligation. Here we show that this B cell hyperresponsiveness correlates with the development of autoantibodies. After the age of eight months, CD22-deficient mice developed high titers of serum IgG directed against double-stranded DNA; these antibodies were of multiclonal origin, somatically mutated, and high affinity. Increased titers of antibodies to cardiolipin and myeloperoxidase were also noted. The results demonstrate that a single gene defect exclusive to B lymphocytes is, without additional contrivance, sufficient to trigger autoantibody development in a large proportion of aging animals. Thus, CD22 might have evolved specifically to regulate B cell triggering thresholds for the avoidance of autoimmunity.
Several autoimmune diseases are characterized by the presence in serum of high affinity antibodies to self antigens, with the B cells producing them having undergone heavy chain class switching and somatic hypermutation. This suggests that T cell help has been available, at least under these pathological conditions, to facilitate maturation of an antiself response. The defect is, however, unlikely to lie simply in the inappropriate provision of T cell help: several models reveal that multiple loci can contribute to predisposition to autoimmune disease (1–4), with intrinsic defects in the B cell lineage able to play an important role (5, 6).
Although defects in B cell apoptosis can certainly accelerate autoimmune disease (7), several lines of evidence suggest that an intrinsic hyperresponsiveness of B cells to antigen encounter could also be a contributory cause of autoimmunity. Thus, genetic dissection of the contributing loci in a mouse model of systemic lupus erythematosus reveals that one of the loci (Sle2) leads to B cell hyperactivity (8). Furthermore, the response to self antigen by B cells that express an autoreactive immunoglobulin transgene can be significantly affected by mutations that affect B cell antigen receptor (BCR)1 signaling (9, 10).
We were interested in determining whether mutations affecting B cell signaling would be sufficient to predispose autoantibody development in an otherwise normal mouse. Is a hyperresponsiveness that is restricted to the B cell compartment nevertheless sufficient to so perturb the immune system that the necessary help is recruited to allow development of high affinity antiself antibodies?
To this end, we made use of CD22-deficient mice, which exhibit a relatively mild B cell hyperresponsiveness (11–14). CD22 is a B cell–specific transmembrane glycoprotein that associates with BCR and possesses an extracellular domain that binds α-2,6-sialylated glycoconjugates (15–18). It acts as a negative regulator of antigen receptor signaling, with levels of BCR cross-linking that are too low to generate a detectable signal in B cells from control mice, nevertheless giving rise to a calcium flux with CD22-deficient B cells (11–14). Indeed, even halving the abundance of CD22 on the cell surface leads to enhanced BCR signaling (19). In our initial characterization of CD22-deficient mice, we noted a small (approximately twofold) increase in total serum IgM (but not IgG) together with a corresponding increase in total Ig anti-DNA titers in 5-mo-old animals that might be ascribable to an expanded B1 cell population (11). However, here we show that as the CD22-deficient mice age, they have a dramatically increased likelihood of producing somatically mutated, high affinity autoreactive IgG. Thus, it appears that there is tight regulation of BCR signaling and that the perturbations caused by CD22 deficiency can trigger the development of autoimmunity.
Materials And Methods
Mice were generated from chimeras established using a previously described embryonic stem cell (ES) clone (11) containing a targeted integration of a tk-neo cassette into Cd22. The chimeras (created using C57BL/6 blastocysts) were bred against both C57BL/6 and BALB/c mice, and mice from the F2 generation were maintained for up to 20 mo with tail bleeds taken every 4–6 wk. A cohort of control (129 × C57BL/6)F2 mice (that do not carry any targeted gene alteration) was established analogously. Animals were either bred in our own conventional facility or in a specific pathogen-free (barrier) unit following delivery by Caesarian section and fostering onto C57BL/6 × CBA females in isolators.
Analysis of Autoantibodies.
Serum titers of IgG anti–double-stranded (ds) DNA were measured as described elsewhere (20) using alkaline phosphatase–conjugated goat anti–mouse IgG (Sigma Chemical Co., Ltd.). Sera from four MRL/lpr mice were always titered in parallel, with one of these sera assigned a titer of 5 U/ml. The assay was calibrated using a high affinity IgG2a monoclonal anti-dsDNA antibody (S22) from mouse 9612 (see below); 1 U/ml in the ELISA was given by 24 μg/ml of S22. Titers of other IgG autoantibodies were similarly determined using plates that had been coated with either cardiolipin (Sigma Chemical Co.; 100 μg/ml in ethanol) or myeloperoxidase (Calbiochem Corp.; 250 ng/ml in sodium bicarbonate, pH 9.2). Antibody isotypes were determined using reagents from PharMingen.
Hybridomas were established from unimmunized mice by fusion with NS0 cells and autoantibodies in the supernatants monitored by ELISA developing with biotinylated goat anti–mouse κ (Southern Biotech) and alkaline phosphatase–conjugated streptavidin (Dako). The binding of monoclonal anti-DNA antibodies at 20°C to a 5′-biotinylated ds 48mer oligonucleotide that had been immobilized on a streptavidin-coated chip (SA-Biacore chip; Pharmacia) was monitored by surface plasmon resonance as previously described (21).
Sequencing of Expressed VH Segments.
Oligo-dT–primed cDNA prepared from RNA extracted from the hybridomas was PCR amplified using a consensus VH oligonucleotide for forward priming (5′-CGGGATCCTGAGGTGCAGCTGGAGGAGTC ) in conjunction with either 5′-CGGAATTCGGGGCCAGTGGATAGAC or 5′-CGGAATTCGGGACCAAGGGATAGAC for priming back from the CH1 domain of Cγ1, γ2a, and γb or of Cγ3, respectively. PCR products were sequenced directly as well as ligated into pUC18 with multiple DNA clones sequenced from each hybridoma.
We have previously described (11) an ES line (derived from 129 mice) that carries a targeted integration of a neomycin resistance gene into the CD22 gene; this cell line was used to establish chimeric mice by injection into C57BL/6 blastocysts and germline transmission of the targeted allele (yielding CD22+/− heterozygotes) obtained following breeding with both C57BL/6 and BALB/c females. Cohorts of animals from the F2 generations of both sets of breedings were followed with time for the development of IgG anti-dsDNA antibody. On both backgrounds, high titers of anti-DNA antibodies developed with age (particularly after 8 mo) in many of the CD22−/− animals but not in the CD22+/+ litter-matched controls. That the development of these autoantibodies was due to the targeted integration into the CD22 gene is confirmed by the fact that IgG anti-dsDNA was not detected in the sera of control (129 × C57BL/6)F2 mice (Fig. 1).
The titer of anti-DNA antibody in the CD22-deficient mice is, in many cases, of a comparable order to that found in 12-mo-old MRL/lpr mice. By 18 mo of age, over 70% of the CD22-deficient mice have at some time shown evidence of IgG anti-dsDNA antibody in their sera at concentrations >1.5 U/ml; none of the 42 control mice revealed titers of this magnitude (Fig. 2). We have also followed a limited number of CD22+/− heterozygotes and found that 3/11 had developed IgG anti-dsDNA by 12 mo of age (not shown).
Life expectancy among CD22-deficient mice was decreased (10/43 weaned CD22-deficient mice having died by 15 mo of age compared with 1/43 CD22+/+ controls), with at least 4 of the deaths due to infection. However, all but one of these deaths occurred after 7 mo of age. Furthermore, we did not detect proteinuria or antibody deposition in glomeruli in the mice harboring autoantibodies. This lack of pathology may well correlate with the fact that anti-DNA titers do not simply rise with age but, in individual animals, often rise, regress, and rise again.
dsDNA is not the sole target of autoantibody development. Mice were also monitored for the development of antibodies to cardiolipin and myeloperoxidase; a clear distinction between the CD22-deficient and control siblings was found here as well (Table I and Fig. 3 A). The largest cohort of animals was followed under conditions of conventional housing, but we also compared autoantibody development in CD22-deficient and control mice housed in a barrier unit. The results (Table I) reveal that autoantibodies also develop under these cleaner conditions.
Subclass typing of serum autoantibodies revealed that, in both the C57BL/6 and BALB/c breedings, IgG2a anti-DNA was found in ∼80% of the autoimmune animals. However, >50% of the autoimmune mice contained anti-DNA antibodies of multiple IgG subclasses. To obtain more detail about the nature of these antibodies, hybridomas were established from two unimmunized, 18-mo-old, CD22-deficient females (mouse 9612 from the BALB/c breedings and 9449 from the C57BL/6 breedings), as well as two CD22+/+ litter-matched controls. No anti-dsDNA IgG was detected in the supernatants from 198 wells from the control fusions; strong titers, however, were detected in 20/302 wells from the CD22-deficient mice (18 of these from mouse 9612; 2 from mouse 9449; Fig. 3). Similarly, whereas no cardiolipin-specific hybrids were detected in the control fusions, 19 positives were obtained from the CD22-deficient mice (10 from mouse 9612; 9 from mouse 9449). The majority of these cardiolipin-specific antibodies were IgMs, although mouse 9612 gave two IgA and mouse 9449 gave two IgG3 anticardiolipin antibodies.
The hybridomas were then expanded for further characterization. Analysis of the anti-DNA antibodies by surface plasmon resonance using a biotinylated oligonucleotide as antigen revealed that several bound DNA very tightly, with dissociation half-lives in the range of 8–500 min (Fig. 4). To ascertain whether the B cells producing these antibodies were clonally related and whether they had undergone somatic hypermutation, the VH sequences were determined from several of the anti-DNA antibodies from mouse 9612. The results demonstrate that the anti-DNA antibodies within a single CD22-deficient mouse derive from multiple, clonally expanded B cell progenitors that have undergone class switching and somatic hypermutation. Thus, for example, hybridoma S48 appears to be derived from S30, as they carry the same VH36–60/JH2 rearrangement but with S48 harboring multiple additional somatic mutations (several to arginine), which could account for its increased affinity (Fig. 5 A). Similarly, S31 (IgG1) and S35, S66, S11, and S15 (all IgG2a) all express the same (VHJ558 family member)/JH2 rearrangement with an arginine-rich CDR3 (characteristic of many anti-DNA antibodies [23–25]); the individual antibodies differ, however, in the extent of accumulated somatic mutations (Fig. 5 B). In contrast, S20 and S22 carry distinct rearrangements of the same VH7183 family member (Fig. 5 C). Thus, paralleling observations previously made with other autoimmune mice (25), multiple independent B cells appear to have seeded an ongoing anti-DNA response. Similarly, in respect of the two cardiolipin-specific IgG3s, analysis of their VH sequences revealed them to be clonally related (Fig. 5 D).
The development of autoantibodies in CD22-deficient mice reveals that a single gene defect exclusive to B cells is sufficient to trigger autoimmunity in a large proportion of mice. This presumably means that the restriction of the provision of T cell help to foreign antigens is intrinsically imperfect: T cell help for an autoantibody response can be elicited by a hyperreactive B cell compartment.
The CD22-deficient animals do not, however, go on to develop autoimmune disease. This is consistent with genetic analyses of predisposition to systemic autoimmune disease in lupus-prone mouse strains, which reveal a role for multiple genetic loci (1–4, 26). Indeed, one of the loci contributing to autoimmunity in NZM mice (Sle3) has been mapped to a region of chromosome 7 in the vicinity of Cd22 and, when bred into C57BL/6 mice, causes production of IgG anti-dsDNA antibodies as well as lupus nephritis (27). It will be interesting to ascertain whether this, at least in part, reflects a functionally relevant Cd22 polymorphism. By analogy with studies in the MRL mouse (28), it will also be interesting to ascertain whether mutations in Fas or its ligand exacerbate autoimmunity in CD22-deficient mice.
The precise mechanism by which CD22 deficiency predisposes to autoimmunity remains to be definitively identified, but we believe the hyperresponsiveness of CD22-deficient B cells to BCR ligation is likely to be of central importance. Phosphorylation of CD22 on its cytoplasmic tyrosines following BCR ligation is mediated by the Lyn kinase and leads to the recruitment of the phosphatase SHP1 (29–34). It is therefore notable that deficiencies in either Lyn or SHP1 both lead to autoimmunity (35–38). However, this autoimmunity is more severe than that in CD22-deficient animals and is most unlikely to simply reflect defects in CD22-mediated regulation of BCR. Indeed, the increased severity probably correlates with both Lyn and SHP-1 being implicated in signal transduction through multiple cell-surface receptors, with their functions not being limited to the B cell lineage.
Thus, the significance of the autoantibody development in CD22-deficient mice lies in the fact that these autoantibodies arise as a consequence of a relatively mild perturbation that is exclusive to B lymphocytes and that affects the BCR signaling threshold. Experiments performed using transgenic mice that have been engineered to express high affinity autoreactive specificities on a substantial proportion of their B cells have revealed that the fate of such B cells is sensitive to modifications in CD22, Lyn, and SHP-1 as well as other genes that affect BCR signaling (9, 10). Our findings are entirely consistent with these earlier studies but reveal that CD22 deficiency alone, without additional contrivance, is sufficient to predispose autoimmunity in normal animals.
It is attractive to speculate from our results that the major physiological function served by CD22 in normal mice is to mediate the avoidance of autoimmunity. In light of the diminished level of CD22 expression in immature B cells (39), we previously suggested (11) that CD22 plays a role in raising the threshold of sensitivity to antigen that accompanies differentiation of an immature B cell (sensitive to tolerization/deletion by low affinity antigen) into a mature B cell that awaits triggering by exogenous antigen (40). Such a proposal could well explain the autoimmunity in CD22-deficient mice. However, a role for CD22 should also take into account the specificity of its extracellular domain for α-2,6-sialoglycoconjugates (18). Intriguingly, the sialylated moieties present on eukaryotic membranes enhance the interaction between complement components C3b and factor H, thereby leading to inhibition of the alternative complement pathway; this serves to bias activation of the innate immune system toward microbial infection and away from autoreactivity (41, 42). Maybe CD22 recognition of the sialoglycoconjugates expressed on mammalian cells serves an analogous role in the adaptive immune system, dampening the BCR signaling that might otherwise be triggered by low affinity autoantigens. It will be interesting to ascertain whether making mutations in the CD22 extracellular domain that abolish recognition of sialoglycoconjugates will be sufficient to predispose autoimmunity.
We thank Michael Ehrenstein for provision of (129 × C57BL/6)F2 control mice and Angela Middleton and Theresa Langford for animal husbandry.
Abbreviations used in this paper
T.L. O'Keefe was supported by an Oliver Bird Fund fellowship and an International Research Scholar's award from the Howard Hughes Medical Institute (to M.S. Neuberger). F.D. Batista was supported by fellowships from the European Molecular Biology Organization and the Arthritis Research Campaign.
Theresa L. O'Keefe's present address is LeukoSite Inc., 215 First Street, Cambridge, MA 02144.
Address correspondence to Michael Neuberger, MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, United Kingdom. Phone: 44-1223-402245; Fax: 44-1223-412178; E-mail: firstname.lastname@example.org