Natural IgM Prevents Autoimmunity by Enforcing B Cell Central Tolerance Induction

Immunoglobulin M is produced by all jawed vertebrates. It is the first isotype produced in ontogeny and the first Ig produced in response to an insult. Its pentameric structure is also unique among the other Ig isotypes, indicating its unique contributions to immunity and the host’s interactions with its environment (1). Spontaneous “natural” IgM secretion occurs without external microbial stimulation (2, 3). Major sources of natural IgM in mice are B-1 cells situated in spleen and bone marrow, producing at least 80% of the circulating IgM (4, 5). Natural IgM-producing B-1 cells appear to be selected on self-antigens (6, 7) and exhibit dual reactivity to both self and common microbial Ags (1, 8, 9). This selection process might ensure the generation of evolutionary “useful” specificities (8). Indeed, natural Abs appear to bind particularly to “altered” self-antigens, such as Ags expressed on dead and dying cells, which is thought to allow the efficient removal of tissue debris and thereby the removal of potential autoantigens (1, 9–12).

Rapid T-independent IgM responses to systemic application of microbial components, such as LPS of Gram-negative bacteria, or polysaccharide Ags are induced by both B-1 (13, 14) and by marginal zone (MZ) B cells (15), which have a high propensity for rapid differentiation to IgM-secreting cells. Finally, most conventional B cell responses result in the initial production of IgM by early activated B cells, prior to class-switch recombination to IgG, IgA, or IgE (16). Early low-affinity IgM may facilitate Ag deposition in the developing germinal centers (17).

Selective IgM deficiency is a little studied, relatively rare primary immunodeficiency of humans, reported to occur at a prevalence rate of 0.03% (18). Selective IgM deficiency is often associated with recurrent infections (18), consistent with findings in sIgM-deficient mice (μs^−/−), which showed increased morbidity and mortality from various bacterial and viral infections (19–22). The data highlight the importance of both natural and Ag-induced IgM in immune protection from pathogen encounter.

Mechanistically less well understood is the observed development of autoantibodies against dsDNA (12, 23) and the increased risk of autoimmune diseases such as arthritis and systemic lupus erythematosus in a subset of humans with selective IgM deficiency and in μs^−/− mice (11, 12, 18). It has been argued that this is attributable to a break of peripheral B cell tolerance due to ineffective removal of cell debris in the absence of natural Abs (1, 11, 12). This is consistent with the repertoire of self-specificities that preferentially bind to dead and dying self and other components of the altered self (24, 25). Yet, no studies to date have demonstrated such lack of self-antigen removal. Moreover, various BCR transgenic and knock-in mice have been generated during the last two decades, which express highly restricted oligoclonal or even monoclonal B cells, and often lack B-1 cells and/or B-1 cell–derived IgM (26–29). These mice do not appear to suffer from autoimmune disease, indicating that autoantibody production in IgM deficiency may have other underlying causes.

Positive and negative selection events during B cell development are critical for the elimination of self-reactive B cells. The fate of the developing B cells is strongly dependent on the strength of BCR interaction with self-antigens (30, 31). Autoreactive immature B cells may either 1) undergo L chain rerearrangement, that is, change their Ag specificity, 2) become anergic, that is, unresponsive, and express the BCR inhibitory surface molecule CD5, or 3) die via apoptosis (31, 32). Overall strengths of the selecting signals appear to determine also B cell subset selection. Relatively strong signals seem to favor development of B-1 and follicular (FO) B cells, whereas weaker signals drive MZ B cell development (33, 34).

Lack of sIgM may affect B cell development, possibly via expression of the recently identified FcμR (35–38). However, reported alterations appeared not only subtle but also difficult to reconcile: two independently generated strains of μs^−/− mice were reported to have increased numbers of MZ B and B-1 cells, but a normal FO B cell compartment (39, 40). The increased MZ B cell development in μs^−/− mice may indicate that sIgM may affect B cell subset selection during development, and that in its absence overall strengths of the selecting signals is reduced, resulting in increases in the MZ B cell compartment. However, this is not consistent with the reported expansion of their B-1 cell compartment (40), which one would expect to be reduced and with a lack of effects on FO B cells.

To determine the causes of autoantibody production in selective IgM deficiency we re-examined the previously generated μs^−/− mice (40). Our findings demonstrate that the accumulation of autoantibodies is due to nonredundant effects of sIgM on B cell development and repertoire selection. Escape from central tolerance induction in the absence of sIgM explains the accumulation of anergic CD5⁺CD43⁻ B-2 cells and the generation of autoantibodies and identifies sIgM as a nonredundant regulator of B cell selection.

Eight- to 12-wk-old female and male mice, C57BL/6J (wild-type [WT]; CD45.2, Igh-b), B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1), B6.129S7-Rag1^tm1Mom/J, and B6.Cg-Igh^a Thy1^a Gpi1^a/J (Igh-a) mice were obtained from The Jackson Laboratory. Breeding pairs of B6.129S-sIgM^−/− (μs^−/− CD45.2, Igh-a) mice were a gift from Dr. Frances Lund (University of Alabama at Birmingham). Heterozygous μs^+/− mice were created by intercrossing μs^−/− and C57BL/6J mice. Heterozygous CD45.1 × CD45.2 F₁ mice were created by intercrossing B6.SJL-Ptprc^a Pepc^b/BoyJ with C57BL/6J mice. Igh-a × Igh-b mice were created by intercrossing B6.Cg-Igh^a Thy1^a Gpi1^a/J (Igh-a) with C57BL/6J mice. All mice were kept in specific pathogen-free housing, placed in filtered-air ventilated racks, screened for the absence of 17 common mouse pathogens (list available upon request). Mice were euthanized by overexposure to carbon dioxide.

To generate mixed bone marrow irradiation chimeras, single-cell suspensions of bone marrow from the femur and tibia of WT (C57BL/6J) and μs^−/− mice were injected i.v. at a 1:1 ratio into CD45.1 × CD45.2 F₁ mice lethally irradiated by exposure to a gamma irradiation source (800 rad). Similarly, WT and FcμR^−/− (CD45.2; provided by Dr. Hiromi Kubagawa [University of Alabama at Birmingham]) mixed bone marrow chimeras were generated by injecting WT or FcμR^−/− bone marrow into lethally irradiated CD45.1 mice. Chimeras were allowed to reconstitute for at least 7 wk before use in experiments. All procedures and experiments involving animals were approved by the Animal Use and Care Committee of the University of California, Davis.

Groups of four WT and μs^−/− mice were sublethally irradiated by whole-body exposure to a gamma irradiation source (350 rad). The mice were allowed to reconstitute peripheral B cell compartments for 12 d. Frequencies of transitional B cells (CD93⁺) in bone marrow and spleen were determined by flow cytometry.

Groups of three or four μs^−/− mice were injected i.p. with either 50 μl WT serum, containing ∼200 μg IgM as assessed by ELISA, μs^−/− serum, or 200 μg monoclonal IgM (Sigma-Aldrich, clone MOPC-104E, specific for α-1,3-glucose) per mouse three times per week for 3 wk to reconstitute sIgM. Control mice received PBS only.

Spleen and lymph node cells were obtained by grinding the organ between frosted ends of glass slides. Bone marrow was obtained by flushing fibula and tibia with medium using a 23-gauge needle. Single-cell suspensions were stained as previously described (41). Briefly, after Fc receptor blocking (5 μg/ml anti-CD16/32) for 20 min on ice, cells were stained with the following fluorochome conjugates: streptavidin (SA)–Qdot 605, SA-allophycocyanin, CD93-PE, CD80-PE, CD86-allophycocyanin, CD45.1–(PerCP, Cy5.5-PE), CD45R–Cy7-allophycocyanin, CD3e–Cy55-allophycocyanin, BrdU-FITC (all BD Pharmingen), CD45.2–(Cy7-allophycocyanin, Cy7-PE) (BioLegend), CD5–(biotin, Cy5-PE), CD45R–(FITC, Cy7-allophycocyanin), FITC-labeled, phophatidylcholine (PtC)-containing liposomes–FITC, CD21-(FITC, biotin), CD23-(FITC, allophycocyanin), CD43–(PE, Cy7-allophycocyanin), CD19–(Cy5-PE, allophycocyanin), CD21–Cy5.5-PE, CD24–Cy5.5-PE, BP-1–allophycocyanin, IgD–Cy7-PE, and IgM–(Cy7-allophycocyanin, Cy55-allophycocyanin) (all in-house generated). Dead cells were excluded by Live/Dead Pacific Blue staining (Invitrogen). Staining for BrdU was done according to the manufacturer’s protocols using a BD Pharmigen BrdU flow kit. To accurately set gates to identify CD5- and CD43-expressing cells, we used “fluorescence minus one” controls in which cells were stained with all reagents except anti-CD5 and anti-CD43, respectively.

For in vitro sIgM binding, sIgM (SouthernBiotech, clone 11E10, specific for LPS) was conjugated to biotin and added to bone marrow and spleen cells for 1 h at 4°C. Staining was revealed with SA-allophycocyanin. Dead cells were excluded by Live/Dead Pacific Blue staining.

For cell cycle analysis, surface-stained splenocytes were fixed with 2.5% paraformaldehyde, washed in PBS plus 1% BSA, and stained with anti–Ki67-FITC and 7-aminoactinomycin D (7-AAD) (BD Pharmingen) overnight at 4°C. FACS analysis was done using a 15-parameter FACSAria (41) or an LSRFortessa (BD Biosciences) equipped with four lasers and optics for 22-paramenter analysis. Analysis was done using FlowJo (gift of Adam Treister).

FACS sorting of live “non-dump,” non–autofluorescent CD19⁺ B cell subsets was done using a FACSAria (BD Biosciences). Sort gates were defined as follows: FO B cells, CD21^intCD23⁺; MZ B cells, CD21^hiCD23⁻; anergic B cells, CD21^intCD23⁻. Sort gates for bone marrow CD45R⁺CD43⁻ B cell precursors were defined as follows: fraction D, IgD⁻IgM⁻; fraction E, IgD⁻IgM⁺; fraction F, IgD⁺IgM⁺. Sorting purities were >95%.

Splenic B cells were enriched using a mixture of depleting Abs (CD90.2-biotin, Dx5-biotin, Gr-1–biotin) and anti-biotin microbeads (Miltenyi Biotec) after Fc blocking. CD23⁻ (non-FO) B cells in the peritoneal cavity were enriched by adding anti–CD23-biotin to the above Ab mixture. Nylon-filtered splenocytes were separated using autoMACS (Miltenyi Biotec). Purities of enriched B cells were >90% as determined by subsequent FACS analysis.

FACS-purified MZ, FO, and anergic CD21^intCD23⁻ B cells from WT and μs^−/− mice were labeled with 0.5 μM CFSE in PBS at a concentration of 10⁷ cells/ml for 20 min at 37°C, washed twice with PBS, and cultured at 2.5 × 10⁵ cells/well in the presence or absence of 20 μg/ml anti-IgM medium (RPMI 1640, 292 μg/ml l-glutamine, 100 μg/ml penicillin/streptomycin, 10% heat-inactivated FCS, 0.03 M 2-ME) in 96-well round-bottom plates for 3 d at 37°C in 5% CO₂. FACS analysis was done to identify the number of dividing cells.

The method of anti-nuclear Ab (ANA) staining was adapted from previous publications (42). 3T3 cells (A31, provided by Dr. Peter Barry, University of California, Davis) were grown on glass slides overnight and fixed with acetone for 10 min. The slides were blocked with Fc block in 10% FCS/PBS buffer for 1 h, washed twice with PBS and once with washing buffers (PBS plus 0.1% BSA plus 1% normal calf serum), and then incubated with mouse serum in PBS for 2 h (1:100 dilution for 2.5-mo-old mice and 1:1000 dilution for 8-mo-old mice). Staining was revealed using goat anti-mouse IgG-biotin and SA–Alexa Fluor 594, each incubated for 1 h at room temperature. The 3T3-stained slides were washed with PBS for 5 min at least three times and mounted with Fluoromount-G (SouthernBiotech). Images were collected with an Olympus BX61 microscope with an Olympus DP72 color camera and were processed by using MetaMorph (Molecular Devices) and ImageJ (National Institutes of Health) software. Differential interference contrast images were taken for outlining the cells. Compared images were collected on the same day and analyzed in the same way. The scale bars were from two images (original magnification ×20 and ×40) and applied to others.

DNA-specific IgG was measured by ELISAs. Ninety-six–well plates (Maxisorb; Thermo Fisher Scientific) were coated with 10 μg/ml dsDNA from calf thymus or 10 μg/ml ssDNA (degraded dsDNA after heating at 94°C for 14 min in 1 N NaOH) in PBS overnight. After blocking with PBS/1% heat-inactivated calf serum/0.1% milk powder/0.05% Tween 20 for 1 h, 2-fold serially diluted serum in PBS was added for 2 h. Plates were washed and Ab binding was revealed using biotin-conjugated anti-IgG (SouthernBiotech) followed by a SA-HRP (Vector Laboratories, Burlingame, CA) incubation for 1 h and then substrate (10 mg/ml 3,3′,5,5′-tetramethylbenzidine in 0.05 mM citric acid, 3% hydrogen peroxide) for 20 min. Reactions were stopped with 1 N sulfuric acid, and absorbance was read at 450 nm and reference wavelength of 595 nm using a SpectraMax M5 (Molecular Devices). Relative unit Ab levels were calculated by comparison with NZB mouse serum.

DNA was isolated from sorted cells by a DNeasy kit (Qiagen) and stored in the DNA storage buffer at −20°C until processed. IgHV regions were amplified by quantitative PCR as described previously (43) in SYBR Green buffer (Applied Biosystems). For quantitative RT-PCR analysis of VH11 mRNA expression, total RNA from peritoneal cavity B cells was isolated with RNeasy (Qiagen) and stored in the RNA storage buffer (Ambion) at −80°C until processed. cDNA was prepared by using random hexamers (Promega) with SuperScript II reverse transcriptase (Invitrogen). Primers and probe for IGVH11 were designed using IDT software: forward primer, 5′-GAAGTGCAGCTGTTGGAGA-3′; reverse primer, 5′-GTACAGGGTGCTCTTGTCATT-3′. Relative expression was normalized to GAPDH (IDT).

Statistical analysis was done using a two-tailed Student t test. A p value <0.05 was considered statistically significant.

Measurements of autoantibodies by ELISA in 10-wk-old μs^−/− and age-matched WT controls confirmed earlier reports on the generation of high titers of IgG autoantibodies against dsDNA and ssDNA in μs^−/− mice (Fig. 1A, 1B) (12, 23). We also found increased anti-nuclear IgG Abs, which further increased with age (Fig. 1C, 1D). Thus, selective IgM deficiency causes the development of autoantibodies against multiple self-antigens, which increase with age. Comparison of total serum IgG levels in WT and μs^−/− mice as assessed by ELISA revealed no significant differences between the mouse strains at 3 mo of age and only very modest increases at 8 mo (Supplemental Fig. 1A). Similarly, the frequencies of IgG-secreting cells in the spleen as measured by ELISPOT and the frequencies of spleen CD19⁺CD138⁺ plasma cells were not significantly different between these two mouse strains (Supplemental Fig. 1B, 1C). Thus, the difference in autoantibody production in μs^−/− mice cannot be explained by differences in total IgG levels and instead suggest B cell repertoire differences.

Modest increases in MZ (39) and peritoneal cavity B-1 cells in μs^−/− mice (40) were previously reported. We confirmed the increased frequencies of MZ B cells and a concomitant decrease in FO B cells in μs^−/− mice (Fig. 2A, 2B), but we identified multiple additional defects. First, μs^−/− mice had significantly reduced numbers of CD19⁺ B cells, especially in lymph nodes (Fig. 2A, 2B). Moreover, we noted an unusual population of CD21^intCD23⁻CD19⁺ B cells in spleens and lymph nodes of μs^−/− mice that was not present in WT mice and blurred the differentiation of MZ and FO B cells somewhat (Fig. 2A). Further analysis identified them as CD5^+/−CD21^loCD23⁻CD43⁻CD45R^hiCD93⁻ (Fig. 2C). Despite expression of CD5 on some of these cells, they were not classical B-1a cells, as they were mostly CD43⁻, and CD5 expression levels of those expressing CD5 were lower and their expression of CD45R was higher compared with those of B-1a cells (Fig. 2C). They were not transitional B cells either, as they lacked expression of CD93. Based on their expression of CD80 and CD86 they appeared to be Ag experienced. In WT mice, very few B cells fell into the CD21^intCD23⁻ B cell gate. Those that did were either B-1 (CD5⁺CD21^loCD23⁻CD43⁺CD45R^lo) or transitional B cells (CD5⁻CD21^loCD23⁻CD43⁻CD45R^hiCD93⁺) (Fig. 2C, 2D). FO cells also showed phenotypic alterations in μs^−/− mice. They expressed reduced levels of IgD and some seemed to express very low levels of CD5 based on comparison with control stains that did not include CD5 (Fig. 2D and data not shown). These dramatic changes of the peripheral B cell compartments suggested that sIgM is a nonredundant component required for normal B cell development and B cell subset selection and/or maintenance. We therefore explored the relationship between the effects of IgM on the B cell compartment and the regulation of autoantibody production.

First, we determined whether the B cell developmental changes in μs^−/− mice could be normalized by the administration of sIgM. For that we injected polyclonal IgM purified from WT serum into μs^−/− mice three times per week for 3 wk. Transfer of sIgM resulted in significant increases in the frequencies of FO B cells and reduced frequencies of CD21^intCD23⁻ B cells in spleen and lymph nodes compared with mice receiving serum from μs^−/− mice, a decrease that reached statistical significance in the lymph nodes (Fig. 3A). Owing to the very short half-live of IgM (2 d) (44), the levels of circulating serum sIgM could not be normalized with this approach (Fig. 3B), which may explain why a more complete rescue of B cells was not achieved.

Therefore, to test for the development of B cells from precursors of μs^−/− mice in the presence of normal levels of serum IgM we generated mixed bone marrow chimeras, transferring WT (CD45.1) and μs^−/− (CD45.2) bone marrow at a 1:1 ratio into lethally irradiated CD45.1 × CD45.2 F₁ recipients. The CD45.1/CD45.2 double-positive radio-resistant recipient cells were easily distinguished from the single-positive WT and μs^−/− donor B cells in the host (Fig. 3C, second panel). Recipients of WT or mixed μs^−/− WT bone marrow had similar serum IgM levels, whereas recipients of μs^−/− bone marrow lacked sIgM (Fig. 3E). Precursors from μs^−/− mice reconstituted roughly half of each B cell compartment in the mixed bone marrow chimeras (Fig. 3C, 3D). Frequencies and phenotypes of all major B cell subsets were normal (Fig. 3C, 3D, and not shown). Furthermore, CD21^intCD23⁻ B cells did not develop in the mixed chimeras, but they did develop in controls that received μs^−/− bone marrow only and thus lacked IgM (Fig. 3C, 3E). Importantly, ANA development was greatly reduced in mixed bone marrow chimeras, compared with mice receiving μs^−/− bone marrow only. The latter generated measurable levels of ANA within 10 wk after adoptive transfer (Fig. 3F). Thus, precursors of μs^−/− mice have no inherent developmental defect in B cell development when raised in the presence of sIgM.

Increases in peritoneal cavity CD5⁺CD45R⁺ B-1 cells have been reported for μs^−/− mice (40). Although our results confirmed increases in frequencies of CD5⁺ B cells (Fig. 4A), these cells were mostly CD43⁻CD45R^hi and CD19^int, that is, they did not represent classical CD43⁺CD45R^loCD19^hi B-1a cells of WT mice (Fig. 4A, 4B). The difference in CD43 staining shown (Fig. 4A, 4B) is highly significant (mean fluorescence intensity of 2131 ± 556 on WT CD19⁺IgM⁺ cells versus 195 ± 34 for μs^−/− cells, p < 0.005). Also, the difference for CD45R expression was greatly significant (1,758 ± 259 versus 3,727 ± 115, p < 0.0001), as was the reduction in expression of CD19 (12,720 ± 1,036 versus 8,920 ± 320, p < 0.005). Therefore, CD19^hiCD45R^loCD43⁺ cells, that is, classical B-1a (CD5⁺) and B-1b (CD5⁻) cells, were greatly reduced in the peritoneal cavity of μs^−/− mice (Fig. 4C). Further confirmation of a reduction in B-1 cells came from the analysis of PtC-specific cells, a B-1a cell–restricted specificity (25, 45). PtC-binding CD5⁺ B cells made <1.0% of CD19⁺ peritoneal cavity B cells in μs^−/− mice compared with nearly 10% in C57BL/6 mice and 15% in 129 WT controls (Fig. 4D and data not shown). Because PtC-binding B-1 cells in C57BL/6 and 129 mice preferentially use VH11 (46), we determined whether the lack of B-1 cells might be the outcome of altered B cell selection. Indeed, CD23⁻ peritoneal cavity B cells of μs^−/− mice showed a nearly complete absence of VH11 mRNA compared with the high expression seen in WT mice (Fig. 4E). The strong reduction of B cells in the μs^−/− peritoneal cavity was not due to increases in B-1 cell death (Supplemental Fig. 2). It also did not appear to be due to an increase in B-1 cell migration from the cavities to the spleen, and subsequent differentiation to Ab secretion cells, as is often seen after activation of peritoneal cavity B-1 cells, because frequencies of ASC in the spleen were comparable between WT and μs^−/− mice (Supplemental Fig. 1). Thus, lack of B-1 cells in the peritoneal cavity may suggest differences in B cell selection. In contrast, B-1 cell numbers in the spleen and lymph nodes were unaffected (Fig. 4F).

To determine whether peritoneal cavity B-1 cell development can be rescued by the presence of sIgM, we created heterozygous μs^−/− (Igh-a) × C57BL/6 (Igh-b) mice. As B cells undergo allelic exclusion, each B cell expresses IgM either from the WT or μs^−/− allele. Differences in IgM allotype expression identified their origins. The sIgM levels in μs^+/− sera were comparable to those of WT mice (data not shown), and heterozygous mice had normal frequencies of total and PtC-binding B-1 cells in the peritoneal cavity (Fig. 4G). Furthermore, μs^−/−-derived (Igh-a) and WT-derived (Igh-b) B cells were present at similar frequencies in bone marrow, spleen, and peritoneal cavity (Fig. 4H). The CD5^+/−CD21^intCD23⁻ B cell population observed in μs^−/− mice was absent from peritoneal cavity and spleen (data not shown), and IgG autoantibody levels against dsDNA (Fig. 4I), ssDNA (Fig. 4J), and ANA-specific autoantibodies (Fig. 4K) were strongly reduced in sera from 10-wk-old μs^+/− mice compared with age-matched μs^−/− controls. The slight but not statistically significant increases of autoantibodies in the heterozygous compared with the WT mice might be due to local concentrations of sIgM in the immediate vicinity of B cells in the bone marrow. In summary, sIgM is a nonredundant factor required for the normal development of B-2 cells and peritoneal cavity, but not splenic B-1 cells.

CD5 is a negative regulator of BCR signaling, and its expression on conventional B cells has been previously linked to anergy (32). The expression of CD5 on some FO and CD21^intCD23⁻ B cells suggested the emergence of anergic B cells in μs^−/− mice. Anergic B cells are relatively short-lived and are unresponsive to BCR-mediated stimulation (47). Indeed, B cell subsets in μs^−/− mice, including MZ and FO B cells, showed enhanced turnover compared with their WT counterparts as measured by enhanced BrDU incorporation (Fig. 5A), as noted by others (48). The turnover rate of CD21^intCD23⁻ B cells was similar to that of FO B cells in μs^−/− mice, but it was enhanced compared with those in WT mice (Fig. 5A), and their rate of cell death, as measured by 7-AAD-incorporation, was highest among all B cell populations (Fig. 5B).

To determine whether the higher rate of BrDU incorporation was due to reduced B cell survival in μs^−/− mice, we transferred equal numbers of CFSE-labeled splenic B cells from WT (CD45.1) and μs^−/− (CD45.2) mice into WT mice. All B cell subsets from μs^−/− mice, including the CD5^+/−CD21^intCD23⁻ cells, survived poorly compared with WT B cells at both 2 and 25 d after transfer (Fig. 5C, 5D). Additionally, μs^−/− mice had increased frequencies of B cells expressing the proliferation marker Ki67, as well as increased frequencies of CD19⁺ cells in the G₁ phase of the cell cycle, compared with WT cells (Fig. 5E). Transfer of either sIgM containing WT serum or polyclonal IgM could reverse these increases, whereas monoclonal IgM could not (Fig. 5E, 5F).

We tested whether the reduced ability of μs^−/− B cells to survive in vivo could be due to a lack of BCR-mediated signaling, as unresponsiveness to BCR signaling is a hallmark of anergic B cells, and continuous “tonic” signaling through the BCR is required for B cell survival. Indeed, all B cell subsets from μs^−/− mice showed greatly reduced proliferation in response to anti-IgM stimulation compared with their WT counterparts. The CD21^intCD23⁻CD5^+/− B cells appeared least responsive (Fig. 5G). Thus, B cells from μs^−/− mice have a shortened lifespan and are hyporesponsive to BCR stimulation, suggesting that a lack of sIgM leads to the development of anergic B cells.

Given that the transfer of mature μs^−/− B cells into a sIgM-complete environment could not rescue μs^−/− B cell survival (Fig. 5D), sIgM might affect very early stages of B cell development. Analysis of bone marrow precursors showed that early pro–B cells (Hardy scheme, fraction A) and pro–B cells (fraction B) were increased in μs^−/− compared with WT mice, but they were reduced after the pre–B cell stage (Fig. 6A, Supplemental Fig. 3A). Staining for CD93 expression on Hardy fraction E further demonstrated a strong reduction of immature B cells in the bone marrow (Fig. 6A). Because reduced BCR expression might affect identification of immature B cells (usually done via IgM/IgD staining), we added staining for CD93 to the analysis, which confirmed the decrease of immature B cells in μs^−/− mice (Fig. 6A).

The rate of proliferation among pre-pro– and late pre–B cells was similar between μs^−/− and WT mice (Fig. 6B). Thus, the increases of fractions A and B were not due to their increased expansion, but were likely attributable to increased accumulation due to reduced progression through the pre–B cell stage. Consistent with the presence of reduced B cell precursors in the bone marrow, frequencies of splenic transitional B cells were reduced at transitional stage T1, but not at T2, in μs^−/− compared with WT mice (Fig. 6C, Supplemental Fig. 3B). Finally, to directly compare bone marrow output, we assessed peripheral B cell reconstitution on day 12 after sublethal irradiation of μs^−/− and WT mice. The results showed significant reductions in the frequencies (Fig. 6D) and total numbers (not shown) of immature/transitional CD19⁺CD93⁺ B cells in the bone marrow and spleen of μs^−/− mice compared with their WT controls. We conclude that the presence of secreted IgM is required for normal bone marrow B cell development.

The data thus suggest that bone marrow B cell development is dependent on the presence of sIgM. To determine whether sIgM can bind to bone marrow B cells, and thus to assess whether sIgM/B cell precursor direct interaction may be responsible for the effect of sIgM on B cell development, we measured the binding of sIgM to B cells in bone marrow and periphery.

Staining for total surface IgM, which stains for bound sIgM as well as for membrane-bound BCR, showed that immature B cells from μs^−/− mice had reduced IgM staining compared with WT controls (Fig. 6E). To distinguish surface IgM expression from binding of sIgM to B cells via Fc or other receptors, which would not occur in μs^−/− mice, we created heterozygous Igh-a × Igh-b mice, in which the expressed IgM allele and surface-bound IgM from the nonexpressed allele were differentiated by surface staining with allotype-specific anti-IgMa and anti-IgMb. All B cells stained significantly for the nonexpressed Igh allotype, with stronger surface binding on peripheral B cells compared with bone marrow precursors (Fig. 6F), which explained the reduced staining for total surface IgM in μs^−/− compared with WT mice (Fig. 6E). Also, in vitro exposure of spleen and bone marrow cells to sIgM showed strong sIgM binding to B cells, particularly B cells from μs^−/− mice (Fig. 6G), including immature and mature B cells, but little to other cell types (Fig. 6G, Supplemental Fig. 4). We conclude that sIgM binds to B cells and their precursors in vivo.

Of the known surface IgM receptors, only the FcμR is expressed by early B cell precursors (35, 37), and we aimed to assess its effects on B cell development. First, we tested for its contribution to sIgM binding to B cells and non–B cells. For that we exposed in vitro IgM labeled with biotin to bone marrow and spleen cells from bone marrow irradiation chimeras generated by reconstituting WT C57BL/6 (CD45.1) mice with bone marrow from either WT C57BL/6 mice or FcμR^−/−mice. WT and FcμR^−/− B cells were able to bind sIgM on their cell surface. The lack of the FcμR led to a measurable but only very modest reduction in sIgM binding to B cells of bone marrow and spleen, whereas non–B cells showed no discernable IgM binding (Fig. 7A, 7B).

To determine whether direct binding of IgM via the FcμR might regulate B cell development, we studied the effects of B cell development in these bone marrow irradiation chimeras. Bone marrow B cell development appeared normal (Fig. 7C), and peripheral B cell subset composition in the spleen was similar to that of WT but not μs^−/− mice. An exception was the significant reduction in MZ B cells (Fig. 7D, 7E), whereas total serum IgM levels were increased in chimeras reconstituted with FcμR^−/− bone marrow compared with those of WT-reconstituted mice (Fig. 7F). These findings were similar to the original reports with this strain of FcμR^−/− mice (49).

In vitro responses to anti-IgM stimulation were comparable between FcμR^−/− B cells and WT B cells, whereas the μs^−/− B cells showed significant reduced B cell proliferation (Fig. 7G, 7H). Thus, sIgM direct binding to B cells via the FcμR cannot explain the significant effects of sIgM on bone marrow B cell development. Because the lack of FcμR expression reduced, but did not abrogate, sIgM binding to B cells (Fig. 7A), the presence of other IgM-binding proteins on the surface of the developing B cells may regulate B cell development.

The observed binding of sIgM onto B cell precursors may alter B cell selection at the immature B cell stage and with it the B cell repertoire. Comparison of the BCR repertoire of μs^−/− and WT mice using quantitative PCR on FACS-purified B cell fractions taken at different checkpoints during bone marrow B cell development showed significant differences of VH usage among bone marrow fractions D and E from μs^−/− and WT mice (Fig. 8A, 8B). Mature B cells in the bone marrow and FO B cells in the periphery showed similar repertoire differences (Fig. 8C). Comparison of IgHV usage among FO B cells between WT and μs^−/− mice and the CD21^intCD23⁻ B cells in μs^−/− mice also revealed repertoire differences demonstrating the strong effects of sIgM on development of all B cell compartments (Fig. 8D). Combined with the findings of increased autoantibody generation in μs^−/− mice, as well as reduced VH11 gene usage among B-1 cells, the data demonstrate that the lack of sIgM causes repertoire changes in both the B-1 and the B-2 cell compartments.

The present study revealed profound effects of sIgM on the development of most B cells and B cell subsets, thereby providing a mechanism through which sIgM prevents autoantibody formation, namely by facilitating normal B cell development and enforcing negative selection of autoreactive B cells. Although the a lack of autoantigen clearance may contribute to autoantibody development in μs^−/− mice (1, 9–12), the observed effects of sIgM on B cell development in bone marrow and periphery, as well as the reversal of autoantibody production following the re-establishment of normal B cell development, suggest that these effects are the major underlying cause for the development of autoreactive B cells in IgM deficiency.

The direct binding of sIgM to B cells, which we observed at all stages of B cell development from the pre–B cell stage onward (Fig. 6E, 6F, and data not shown), may provide the nonredundant signal for B cell development and selection. Among the three known B cell sIgM receptors, neither the Fcα/μR nor the complement receptors CR1/2 are expressed on pro–B cells, pre–B cells, and immature B cells (50–52), and thus they cannot be responsible for the observed effects of IgM deficiency. FcμR is expressed early in B cell development and was a potential candidate receptor for IgM (37, 53). Additionally, mice lacking the FcμR^−/− develop autoantibodies despite harboring normal levels serum sIgM (37, 53), further suggesting a central role for sIgM in regulation of B cell development rather than removal of autoantigens for the enforcement of central tolerance induction. However, studies with bone marrow FcμR^−/− chimeras (Fig. 7) failed to reveal B cell developmental defects similar to those observed in the μs^−/− mice. The phenotype of the bone marrow irradiation chimera is consistent with the original description of the FcμR^−/− mice from which we obtained the bone marrow (49). In those mice, overall B cell development was found to be unaffected, except for the reduction in MZ B cells and an increase in splenic B-1 cells, as well as increases in serum IgM. A second report with independently generated FcμR^−/− mice reported no B cell abnormalities (50). However, a third group who generated FcμR^−/− mice reported reduction in all B cell precursors in the bone marrow, except mature fraction F B cells cells. In contrast to our findings with the sIgM^−/− mice, they found increased numbers of peritoneal cavity B-1 cells in the FcμR^−/− mice (38). Thus, based on published work with three different FcμR^−/− mice, the lack of this receptor appears to have some effects on B cell development. Those reported changes are more subtle and/or different from those shown in the present study with the sIgM^−/− mice. Most prominently, the accumulation of CD5^+/−CD43⁻CD23⁻CD21^int anergic B cell population in the μs^−/− mice is not noted in any reports of the FcμR^−/− mice, nor was it found in the FcμR^−/− bone marrow chimeras we created (Fig. 7). Given the somewhat contradictory reports on the phenotype of the three independently generated FcμR^−/− mice, compensatory mechanisms may make any differences on B cell development less obvious than a complete lack of sIgM. Alternatively, the various approaches that were taken in targeting the FcμR might have led to the differences in the effects. A direct comparison between the different gene-targeted mice could help to resolve some of these issues.

Because natural polyclonal, but not monoclonal, IgM rescued B cell development (Fig. 5F), our data indicate that natural IgM either directly or indirectly enhances BCR signaling strengths and thereby alters B cell central tolerance induction at the pre–B and immature B cell stage, possibly by facilitating autoantigen presentation, as natural IgM-secreting cells appear selected on and directed against self-antigens (6). However, a BCR ligand requirement for pre–B cell selection has not been documented (54). Alternatively, sIgM may provide a costimulatory signal more akin to that provided by pattern recognition receptors, increasing responsiveness of the developing B cell to self-antigens. In support of a model in which sIgM affects BCR signaling strengths, the lack of sIgM altered the peripheral B cell VH repertoire (Fig. 8), and thus B cell selection, and it led to the accumulation of autoantibody-secreting cells as well as to the presence of CD5^+/− anergic and autoreactive B cells in the periphery, possibly in an attempt to silence these autoreactive cells. Taken together, the data provide evidence that B cell signaling and negative B cell selection at the immature B cell stage are defective in the absence of sIgM. The increases in MZ and decreases in FO and body cavity B-1 cells are also all consistent with reduced BCR-mediated signaling in μs^−/− mice (33, 34).

Our findings of reduced peritoneal cavity B-1 cell frequencies in the μs^−/− mice are opposite to conclusions drawn previously by others (40), but these differences can be explained by the presence of CD5⁺B220⁺ anergic B cells in the peritoneal cavity and spleen, which were likely misidentified as B-1a cells. Their lack of CD43 expression, high expression of B220, intermediate expression of CD19, lack of VH11 expression, and of PtC binding are not consistent with emergence from the B-1 cell lineage. Furthermore, their turnover was among the highest of any B cells studied (Fig. 5), whereas B-1 cell turnover in the peritoneal cavity is very low (55, 56). Multiple mouse models have demonstrated that reduction in BCR signaling strength reduces peritoneal cavity B-1 cell development (7). Thus, the demonstration of curtailed B-1 cell development in the μs^−/− mice further supports a model in which sIgM acts as a positive costimulatory signal during B cell development.

A “chicken and egg” problem emerges from the fact that sIgM is required for normal B cell development but that it is not usually transmitted from the maternal circulation to the fetus. Thus, the development of fetal IgM-secreting cells must not require IgM. Because B-1 cell numbers in the spleen were largely unaffected in μs^−/− mice (Fig. 4), they might be the earliest source of natural IgM supporting further sIgM-dependent B-1 and B-2 cell development. This is consistent with studies showing profound reductions in peritoneal cavity B-1 cells after splenectomy (57) and with studies indicating that B-1 cell development arises in waves (58). Splenic B-1 cells may arise as a first wave of “IgM-independent” B-1 cells during fetal development, whereas body cavity B-1 cells emerge later and develop in an sIgM-dependent manner. The data indicate a new, nonredundant role for splenic B-1 cells in the regulation of B-2 cell development.

In summary, the present study demonstrates that B-1 cells produce natural IgM to regulate B cell development and selection, thereby suppressing the development of autoreactive B cells. Identification of natural IgM as a critical enforcer of central selection suggests use of these proteins as potential therapeutics against Ab-mediated autoimmune diseases.

We thank Abigail Spinner and Frank Ventimiglia (California National Primate Research Center, University of California, Davis) for help with flow cytometry and assistance with cell imaging, respectively, and Adam Treister for FlowJo software. Special thanks to Dr. Frances Lund and Hiromi Kubagawa (University of Alabama at Birmingham) for sharing the sIgM^−/− mice and FcμR^−/− bone marrow, respectively, for these studies.

The authors have no financial conflicts of interest.