Positive Selection of Natural Autoreactive B Cells

See allHide authors and affiliations

Science  02 Jul 1999:
Vol. 285, Issue 5424, pp. 113-116
DOI: 10.1126/science.285.5424.113


Lymphocyte development is critically influenced by self-antigens. T cells are subject to both positive and negative selection, depending on their degree of self-reactivity. Although B cells are subject to negative selection, it has been difficult to test whether self-antigen plays any positive role in B cell development. A murine model system of naturally generated autoreactive B cells with a germ line gene–encoded specificity for the Thy-1 (CD90) glycoprotein was developed, in which the presence of self-antigen promotes B cell accumulation and serum autoantibody secretion. Thus, B cells can be subject to positive selection, generated, and maintained on the basis of their autoreactivity.

Although it is widely accepted that B cells with self-reactivity are deleted or rendered functionally inactive (1), autoantibodies can be found in the serum of healthy animals, referred to as “natural autoantibodies,” in an apparent paradox to the clonal tolerance theory (2, 3). In contrast with disease-associated hypermutated immunoglobulin G (IgG) antibodies, these natural autoantibodies are predominantly IgM, encoded by mostly unmutated germ line variable (V) region genes, and are independent of T cell help for secretion. Natural autoantibody constitutes a large fraction of serum Ig, and the B cells that produce natural autoantibodies frequently express CD5, a phenotype rare in spleen, but more common in the peritoneal cavity of mice (4,5). The significance of such natural autoantibody is not yet clear. However, these antibodies often cross-react with antigens on bacteria or tumors, and mice deficient in serum Ig are susceptible to bacterial infection, suggesting that these Igs participate in innate immunity (3, 5, 6).

SM6C10 is an anti-thymocyte autoantibody (ATA) produced by a hybridoma derived from CD5+ B cells (B-1 cells) of SM/J mice, a strain with elevated serum ATA. SM6C10 binds a murine-specific carbohydrate epitope of the Thy-1 glycoprotein that is expressed on thymocytes, a fraction of mature T cells, and T cell tumors, as well as on the soluble form of Thy-1 (7). Expression of this epitope is Thy-1–dependent and thus is absent from Thy-1 gene-targeted mice (Thy-1) (Fig. 1). Immunoglobulin V gene sequence analysis showed a VH3609 heavy chain combined with a Vκ21C light chain in the SM6C10 hybridoma, both unmutated from the germ line (GenBank X53097 andM21522; VH3609-DHQ52-JH2, Vκ21C-Jκ2).

Figure 1

Immunofluorescence staining of T cells by SM6C10 (ATA). Thymocytes and splenic CD4+ T cells from wild-type C57/BL(B6) (thick line, Thy-1+) and B6.Thy-1−/− (thin line, Thy-1) mice were stained with FL-SM6C10. Staining was negative for all cells in the thymus, liver, bone marrow, spleen, and peritoneal cavity of Thy-1 mice.

To understand why ATA B cells develop, we generated VH3609 μa heavy chain transgenic mouse lines on the C.B17 (μb) background (8) and found that, despite introducing only the μ heavy chain transgene (Tg), these animals had high IgM ATA serum titers detectable from 1 to 2 weeks after birth onward (Fig. 2A). Most splenic B cells from transgenic mice expressed the Tgμ without endogenous μ and lacked CD5 expression (CD5Tgμ+Endoμ), and did not secrete ATA upon lipopolysaccharide (LPS) stimulation in vitro (Fig. 2B). In contrast, there was high ATA secretion from a CD5+B cell subset enriched in the peritoneal cavity (PerC) that expressed the Tgμ alone (CD5+Tgμ+Endoμ), but not from CD5+ cells expressing both Tgμ and Endoμ nor from CD5 B cells in the PerC (Fig. 2, B and C). CD5+Tgμ+Endoμ cells were also present in spleen as a minor fraction, contributing some ATA secretion from this sample. To determine the relation between ATA specificity and light chain usage, we generated hybridomas from sorted B cell fractions. More than half (8/13) of the hybridomas established from the CD5+Tgμ+ EndoμB cell fraction had ATA specificity and identical Vκ21C-Jκ2 light chains, with a V-J junction that lacked nucleotide addition and was identical to the sequence found in the SM6C10 hybridoma. In contrast, CD5B cell–derived hybridomas had a more diverse light chain usage, none expressed Vκ21C, and all lacked detectable ATA activity (Table 1). The predominance of identical Vκ21C-Jκ2 light chains by CD5+Tgμ+Endoμcells in contrast to CD5B cells was confirmed by sequencing mRNA from individual cells (9). Therefore, high ATA serum titer in the VH3609μ transgenic mice resulted from an increase of autoreactive ATA B cells bearing a canonical Vκ21C light chain, and these ATA B cells were exclusively CD5+; that is, introduction of a VH3609μ heavy chain had a positive effect on the generation of ATA B cells.

Figure 2

High serum ATA levels in VH3609μ transgenic mice and an analysis of B cell subsets. Serum and culture supernatant ATA levels were assayed by flow cytometry analysis with C.B17 or B6 thymocytes by using fluorescein (FL)-conjugated antibody to murine IgM (FL–anti-IgM) (antibody 331.12) as a detecting reagent. Serum levels are expressed as thymocyte staining intensity with 1:10 diluted serum. ATA+ serum samples from Tg+ mice were also tested with anti-μa, verifying Tgμ origin. (A) Serum ATA levels of VH3609μ Tg (ATA1 line) × C.B17 littermates. Data are from Tg+/− (circle) and Tg−/− (triangle) individuals. Bkg, FL–anti-IgM alone. (B) ATA secretion by purified B cell subsets incubated with LPS. LPS stimulation of purified B cell subsets was carried out in U-bottomed, 96-well microtiter plates with 4 × 104 cells in 0.15 ml of LPS (1 μg/ml) culture medium. Levels are staining intensity by 50 μl of 3-day culture supernatant using wild-type Thy-1+thymocytes (white bars) and mutant Thy-1thymocytes (black bars) (u.s., unseparated for CD5/μ allotype expression). PerC subsets used for purification are shown in (C) with corresponding numbers. (C) Peritoneal B cell subsets in the ATA1 mouse line. Right panel shows Tgμ(μa) and Endμ(μb) expression among CD5+ cells in the PerC. The percent of B cell fractions among total spleen cells or PerC are as follows: spleen, CD5Tgμ+Endμ (52.0%); PerC, CD5 Tgμ+Endμ(37.8%), CD5+Tgμ+Endμ(5.0%), CD5+Tgμ+Endμ+(8.2%), and CD5+TgμEndμ+(1.6%). These correspond to 98% of splenic B cells and 71, 9, 15, and 3% of PerC B cells, respectively.

Table 1

Summary of Vκ gene usage in transgenic VH3609μ-expressing hybridomas. Vκ gene family nomenclature is according to the IMGT/LIGM database and Potteret al. (26). Designations a to d are arbitrary to distinguish individual germ line gene sequences. Only sequence data from functional Vκ genes are shown. All hybridomas secreted Tgμ. Three CD5-derived hybridomas expressed λ light chain and were ATA. No hybridoma from the CD5+Tgμ+Endoμ cells were λ+. Generation of the hybridomas (25 from CD5Tgμ+Endoμ and 13 from D5+Tgμ+Endoμ) is described in (27).

View this table:

This CD5+ ATA B cell subset was not discernible at the newborn stage (<6 days old), at which time all B cells showed exclusive Tgμ expression, but was detected at 2 weeks and increased during subsequent development. This age-dependent ATA B cell accumulation suggested that their generation might be the result of cellular selection. We therefore crossed VH3609μ transgenic mice with antigen-deficient Thy-1mice. Lymphocyte development appeared normal in Thy-1mice (10). However, accumulation of the ATA B cell subset and concomitant high ATA serum titer only occurred in Tgμ+ mice that were Thy-1+ (Fig. 3). The difference of serum ATA titer held for 2 week neonates as well but became more evident as animals aged (Fig. 3A, 7 months). LPS stimulation of total B cells purified from PerC (or spleen) of Tgμ+ mice consistently yielded much higher ATA secretion from a Thy-1+ background compared with a Thy-1 background (11), consistent with an increased frequency of CD5+ ATA B cells in Thy-1+ mice.

Figure 3

Self-antigen–dependent increase of serum ATA and ATA B cells. To generate Thy-1Tg+ mice, we crossed ATA1 mice with C57BL/6 (B6) Thy-1−/− mice, and Tg+/− offspring were backcrossed with Thy-1−/− mice. Tg, Thy-1, and H-2 typing were carried out by PCR assay and peripheral blood lymphocyte flow cytometry analysis. Individual data are from analysis of six and two litters of 1- to 2-month-old and 7-month-old mice, respectively. (A) Serum ATA levels of 1- to 2-month-old littermates and 7-month-old Tg+ littermates. Serum levels are staining intensity values with 1:10 diluted serum. (B) PerC ATA B cell subset frequency in Tg+ 1- to 2-month-old littermates. Percentages of CD5+Tgμ+Endμ cells among total PerC cells (1 × 106 to 3 × 106 cells per mouse in both groups) are shown.

Therefore, in the presence of Thy-1 antigen, ATA B cells accumulate and secrete autoantibody into the serum, whereas in the absence of Thy-1 neither occurs. This absence of ATA B cell accumulation in Thy-1 transgenic mice could be rescued by providing a Thy-1+ environment, as revealed by stem cell cotransfer of newborn liver from Thy-1 transgenic mice (Thy-1Tg+) and bone marrow from Thy-1+ nontransgenic (Thy-1+Tg) mice into immunodeficient SCID (severe combined immunodeficiency disease) mice (Fig. 4). The μa+μb− fraction of B cells derived from stem cells of transgenic origin contained CD5+ cells (the ATA B cell fraction) when Thy-1+ bone marrow was cotransferred (Fig. 4C), but not when Thy-1 bone marrow was cotransferred (Fig. 4D). The same correlation held true for serum ATA. Thus, ATA B cell accumulation and serum ATA production were dependent on self-antigen.

Figure 4

ATA B cell generation from Thy-1Tgμ+ stem cells in a self-antigen–positive environment. Flow cytometry analysis of PerC cells gated for μa+μb− cells in 4-week-old (A) Thy-1+ compared with (B) Thy-1Tgμ+ littermates (H-2b/b), and (C and D) in SCID mice reconstituted with stem cells 4 weeks after transfer. Cells from a pool of newborn (d0) liver from a litter between Tgμ+ / and Tgμ−/− of Thy-1−/−H-2b/bmice (50% of newborns were Tgμ+) were divided into two groups. Cells from each of the groups were mixed with adult (3 month) bone marrow (BM) cells from either wild-type B6 (C) or B6.Thy-1−/− (D) mice, then intravenously injected into B6.scid mice (2 × 107 newborn liver cells mixed with 1 × 107 BM cells per mouse). The ATA serum level and the ATA B cell subset were analyzed 1 month after injection. Representative data of four mice per group are shown. Percentages of ATA B cells correspond to CD5+μa+μb− cell frequency (arrow marked) among total PerC cells (similar PerC recovery in all groups). Both transferred groups showed a similar predominance of μa−μb+ cell generation from BM stem cells and also from the presence of 50% Tgμ stem cells in the newborn liver used for cell transfer. CD5μa+μb− cell frequencies were comparable. Serum ATA levels are staining intensity values 1.67 ± 0.59 and 0.29 ± 0.028, in (C) and (D), respectively (n = 4).

During B cell development, immature B cells are highly sensitive to B cell receptor–mediated signaling, resulting in tolerance (12). To test whether ATA B cells might be generated because they were sequestered from antigen at an immature stage, we injected Thy-1+ thymocytes into the PerC of newborn or adult mice, assuring direct antigenic exposure of developing or established ATA B cells (13). In both cases, thymocyte treatment failed to eliminate or inactivate ATA B cells in the PerC, and the ATA serum titer was not significantly affected. Injected thymocytes were present at significant levels in the PerC for at least 2 days after each treatment. Thus, ATA B cells have not simply escaped deletion, in contrast with disease-associated autoreactive B cell models described previously (14).

For most of T cell development, some level of antigen receptor–mediated signaling by interaction with self-antigen is known to be essential for survival and maturation, a phenomenon termed positive selection. In contrast, strong self-reactivity leads to deletion or inactivation of these cells (negative selection) (15). Although negative selection has been clearly demonstrated for B cells in several transgenic model systems, it has been unclear whether there is any positive role for self-reactivity in B cell development (1, 16). Previous studies showed that functionally competent B cells can be generated in the presence of self-antigen (17), and a certain degree of signaling involving the B cell receptor may be important in B cell maintenance (18). A study of V-gene usage previously suggested that self- and environmental antigens appear to influence peripheral B cell maintenance (19). By using a germ line gene–encoded specificity, our data provide a clear demonstration that self-antigen can positively influence B cell fate, selecting B cells bearing an appropriate light chain partner and generating a B cell pool with an autoreactive specificity.

It remains unclear, however, whether all B cells require positive selection to mature and persist. Antigens such as the carbohydrate ATA determinant may signal differently from high-affinity or high-valence antigens that normally have a negative effect on newly generated B cells (20), resulting in CD5 up-regulation and continued maturation. Alternatively, it is possible that the positive selection we observed reflects a difference in the responder B cells. Tolerance susceptibility and B cell receptor signaling thresholds can be modulated by coreceptors (20, 21) whose expression on immature B cells may change with ontogeny, similar to the known difference between immature versus mature developmental stages. Autoreactive B cells (and CD5+ B cells) appear to be preferentially generated early in ontogeny (5, 22). Thus, it is possible that positive selection is a relatively common feature of B cells generated early in ontogeny, but only rarely occurs with immature B cells generated in the adult.

The presence of natural autoantibodies in serum and their increase with age have been long recognized. Although antigenic involvement has been suggested (5, 23), direct evaluation of the role of self-antigen in their development has been hampered by the difficulty of experimentally altering the expression of “natural” antigens in live animals. Such antigens are typically ubiquitous in distribution, representing important cellular components, and are often glycolipids or carbohydrate constituents shared among diverse glycoproteins and glycolipids (24). By taking advantage of the VH3609 ATA specificity restricted to a determinant on the Thy-1 glycoprotein, we demonstrate here that self-antigen can indeed promote B cell accumulation and suggest that a significant proportion of natural autoantibody in serum is the product of such a self-antigen–dependent process. Carbohydrates expressed on glycolipids or glycoproteins, some commonly up-regulated on tumors or components of pathogenic bacteria, are the frequent targets of natural autoantibodies (24). In turn, the CD5+phenotype along with autoreactivity is a common feature of chronic leukemic B cells (25). Thus, it is intriguing to speculate that self-antigens play a positive role in recruiting B cells for immunologic surveillance as a part of innate immunity, but that this process carries a risk for potential dysregulated growth. It will be important to assess whether the positive selection responsible for generating such cells demonstrated here is the property of a distinct subset of B cells.

  • * To whom correspondence should be addressed. E-mail: K_Hayakawa{at}


View Abstract

Stay Connected to Science

Navigate This Article