Evidence for a Ligand-Mediated Positive Selection Signal in Differentiation to a Mature B Cell ¹

B cell development follows a set pathway involving sequential expression of the pre-B cell receptor (pre-BCR) ³ and BCR. Pre-BII cells express the pre-BCR upon productive V_HDJ_H rearrangement and the association of H chain with surrogate L chain (reviewed in Ref.1). These cells become immature B cells upon productive V_LJ_L rearrangement and expression of an IgM BCR. Immature B cells are initially IgM^low, IgD⁻, but quickly become IgM^high and migrate to the spleen as IgM^high transitional (T) 1 B cells. In the spleen two schemes for transitional B cell maturation are proposed. In the first, described by Loder et al. (2), T1 cells (IgM^high, CD21⁻, CD23⁻) differentiate to T2 cells (IgM^high, CD21^high, CD23⁺) before entering the long-lived mature B-2 cell (IgM^low, CD21^low, CD23⁺) population. The second differentiative scheme, described by Allman et al. (3), proposes three transitional populations that are distinguished by the expression of the IgM, AA4, and CD23. In this scheme the IgM^high, AA4⁺, CD23⁻ T1 cells differentiate to IgM^high, AA4⁺, CD23⁺ T2 cells and then to IgM^low, AA4⁺, CD23⁺ T3 cells before becoming IgM^low, AA4⁻, CD23⁺ mature B-2 cells. Completing the mature B cell repertoire are two other functionally and phenotypically distinct populations: marginal zone (MZ) and B-1 B cells.

Negative selection has a well-documented role in shaping the mature B cell repertoire by eliminating autoreactive B cells at multiple points along the differentiative pathway. Some autoreactive B cells arrest in differentiation as immature B cells and undergo receptor editing (4, 5, 6). This process is a major mechanism in shaping the B cell repertoire, as ∼25% of bone marrow immature B cells are believed to undergo receptor editing (7). B cells that fail to edit successfully away from their autoreactive specificity undergo apoptosis in the bone marrow, prohibiting further differentiation (4, 8). Other autoreactive B cells exit the bone marrow and either developmentally arrest at a transitional stage in the spleen (9, 10) or become anergic B-2 cells (5, 11, 12).

Positive selection also shapes the mature B cell repertoire. MZ and B-1 cells are Ag selected (13, 14, 15, 16), but whether entry into the B-2 subset, the major population of splenic and lymph node B cells, also requires positive selection is controversial. There is indirect evidence for positive selection in the development of B-2 cells. The inability to make a pre-BCR or BCR or in their ability to signal transduce results in a block in differentiation at the bone marrow pre-B cell stage and also at the splenic transitional B cell stage (2, 17). The nature of this positive selection signal is unclear. Monroe and colleagues (18) have demonstrated that the cytoplasmic domains of Igα and Igβ in the absence of Ig are sufficient to drive B cell differentiation to a splenic B cell, arguing that positive selection at the pre-B and immature B cell stages may require only tonic signaling from the pre-BCR and BCR. However, differences in the expressed repertoires between mature and immature/transitional B cells (19, 20, 21) suggest a requirement for ligand-mediated positive selection. Indeed, BCR stimulation of T2 B cells can result in differentiation to B-2-like cells in vitro (22, 23). Ligand-mediated positive selection is also suggested by an analysis of signaling-deficient anti-HEL Tg mice. Nonautoimmune anti-HEL B cells that lack CD45 arrest at a transitional B cell stage, but become mature if HEL is available (24). Similarly, nonautoimmune anti-MHC class I B cells that lack CD19 arrest at a transitional B cell stage, but become mature in CD19-sufficient mice (25).

Our studies of V_H12 B cell differentiation have indicated unusually restricted positive selection that focuses differentiation toward specificity for phosphatidylcholine (PtC) and B-1 differentiation (15, 26, 27, 28). Positive selection at the pre-B cell stage allows for the survival of only ∼5% of V_H12 pre-B cells and enriches for cells with a V_HCDR3 of 10 aa and a Gly in the fourth position (designated 10/G4), the V_HCDR3 motif of anti-PtC B-1 cells (26). Most L chains are unable to associate (nonpermissive) with 10/G4 V_H12 H chains, and thus, most V_H12 B cells undergo multiple L chain rearrangements (27). Among splenic V_H12 B cells, there is a significant bias for the use of Vκ4/5H, the L chain used by PtC-specific V_H12 B-1 cells. Surprisingly, this bias is evident even among V_H12 B cells that cannot bind PtC and do not differentiate to B-1. We hypothesize that Vκ4/5H L chains drive V_H12 B cell differentiation to the long-lived mature stage, whereas most other V_H12-permissive L chains drive differentiation no further than the short-lived transitional stage. Our analysis of mice carrying a 10/G4 V_H12 transgene and a Vκ1A transgene (6-1/Vκ1A) supports this hypothesis; T1 B cells from these mice use the Vκ1A L chain, whereas the more mature CD23⁺ B cells often use endogenous L chains, particularly Vκ4/5H (28). We show in this study that PtC nonbinding V_H12 B cells, such as 6-1/Vκ1A B cells, terminate differentiation at a transitional stage and provide evidence that their differentiative arrest is due to a failure to receive a ligand-mediated, positive selection signal.

6-1 and 6-1/Vκ4 Tg mice have been previously described (10, 15). Vκ1A Tg mice were described previously (28) and provided by Dr. M. Weigert (Princeton University, Princeton, NJ). These mice contain a Vκ1A-Jκ1 rearrangement inserted by homologous recombination into an endogenous Jκ locus. These knock-in mice also contain a duplication of the Vκ1A-Jκ1-Cκ locus on the same locus as the knock-in. Recombinase-activating gene-1 (RAG-1)^−/− mice on a C57BL/6 background were purchased from The Jackson Laboratory (Bar Harbor, ME). All strains, with the exception of RAG-1^−/− mice, were bred and maintained in our pathogen-free facility at University of North Carolina by backcrossing to C.B17 mice. Offspring carrying transgenes were identified by tail genomic DNA as previously described (10, 15, 28).

Cells were prepared and stained as previously reported (29). mAbs specific for mouse B220, IgM, IgM^a, IgM^b, CD43, CD40, CD44, CD80, CD86, CD5, CD19, CD23, CD21/CD35, heat-stable Ag (HSA), AA4.1, and MHC class II (I-A^b) were obtained from BD PharMingen (San Diego, CA) and were either directly conjugated to FITC, R-PE, and allophycocyanin or biotinylated. Biotinylated Abs were revealed with streptavidin-conjugated PerCP (BD PharMingen). Cells were analyzed using a FACSCalibur (BD Biosciences, Mountain View, CA). Data were analyzed by WinMDI software (The Scripps Institute, La Jolla, CA).

For cell-sorting experiments, spleen cells were stained with FITC-labeled anti-B220 and sorted on a MoFlo high-speed sorter (Cytomation, Fort Collins, Co). Sorted populations were >95% pure. The cells were then used for Western blotting.

Dividing cells were labeled with BrdU in vivo as described previously (10). Briefly, BrdU (Sigma-Aldrich, St. Louis, MO) was administered in drinking water continuously for 1–6 days. At each time point, mice were sacrificed, and spleen cells were isolated for staining with anti-IgM-allophycocyanin and anti-B220-PE. Cells were then fixed, permeabilized, treated with DNase (Sigma-Aldrich), and stained with anti-BrdU-FITC (BD Biosciences). The fraction of BrdU-labeled B cells was determined by flow cytometry.

Spleen cells from non-Tg, non-Tg that had been irradiated with 500 rad 14 days earlier (Ir-d14), and 6-1/Vκ1A/RAG-1^−/− mice were purified through a B cell purification column (Accurate Chemical & Scientific Corp., Westbury, NY) according to manufacturer’s instruction. For the proliferation assay, 1 × 10⁵ cells/well in 96-well plates were cultured in triplicate in complete RPMI 1640 medium supplemented with 10% FCS. Goat F(ab′)₂ anti-mouse μ (10 μg/ml) (Southern Biotechnology Associates, Birmingham, AL), LPS (20 μg/ml) (Sigma-Aldrich), monoclonal anti-CD40 Ab (0.5 μg/ml; Southern Biotechnology Associates), and IL-4 (10 ng/ml; R&D Systems, Minneapolis, MN) were added to cultures for 4 days. On day 3, [³H]thymidine was added at 1 μCi/well and incubated for 16 h. The cells were then harvested and analyzed using a liquid scintillation counter (Beckman, Fullerton, CA). For Ab secretion assay, 2 × 10⁵ cells/well were cultured under the same conditions for 6 days. A μ-chain-specific ELISA was used to measure the secreted IgM in culture supernatant as described previously (29).

Splenic B220⁺ cells from non-Tg mice, Ir-d14, and 6-1/Vκ1A/RAG-1^−/− mice were purified by cell sorting. Three million cells were stimulated with 20 μg/ml goat F(ab′)₂ anti-mouse μ for 3 min at 37°C. The cells were then lysed in buffer containing 20 mM Tris (pH 8.0); 150 mM NaCl; 1 mM PMSF; 1 μg/ml each of aprotinin, α1-antitrypsin, and leupeptin; 10 mM tetrasodium pyrophosphate; 2 mM sodium orthovanadate; and 1% Nonidet P-40. Lysates were cleared by centrifugation and subjected to SDS-PAGE using 10% gels. Fractionated proteins were transferred to nitrocellulose membranes and blocked with 4% BSA. The blots were probed with the mAb Ab-2 recognizing the phosphotyrosine residues (Oncogene Research Products, Boston, MA), followed by an HRP-conjugated anti-rat IgG1 Ab. Immunoreactive proteins were visualized with an ECL kit (Amersham Pharmacia Biotech, Piscataway, NJ).

6-1/Vκ1A/RAG-1^−/− mice were injected i.p. with 10 or 100 μg of either anti-Igβ (HM79) or an irrelevant hamster (HM) IgG Ab (Southern Biotechnology Associates). Bone marrow and splenic cells were analyzed by flow cytometry on day 2.

For adoptive transfer experiments, C.B17 or RAG-1^−/− mice were irradiated with 500 rad 1 day before i.v. injection of 3 × 10⁷ spleen cells or 6 × 10⁶ bone marrow cells from 6-1/Vκ1A/RAG-1^−/− mice. Mice that received spleen cells were sacrificed on day 3, and the spleen cells of donor origin (IgM^a+) were analyzed by flow cytometry. Donor B cells from recipients that received bone marrow cells were analyzed at wk 3 and 4. Serum IgM levels were assayed by ELISA as previously described (29).

For cell survival assays, splenic cells from mice that had been pretreated with 10 μg of anti-Igβ or HM IgG for 3 days were cultured in RPMI 1640 medium with 10% FCS for 3 days. Cells were stained for apoptotic cells with anti-IgM-APC, annexin V-FITC (BD PharMingen) and propidium iodide at different time points and analyzed by flow cytometry.

6-1 mice carry a 10/G4 V_H12 transgene and generate large numbers of PtC-specific splenic and peritoneal B-1 cells that stain brightly with PtC-containing liposomes (PtC^bri) (15). The remaining B cells stain with liposomes with either intermediate intensity (PtC^int) or do not stain (PtC^neg), and are not B-1 (27). Previously, we considered many of the PtC^int and PtC^neg cells B-2 based on the expression of CD23, but because some transitional B cells are now known to also express CD23 (2, 3), we re-examined the phenotype of these cells. Fig. 1 and Table I show that PtC^int and PtC^neg cells include CD23⁻ T1 cells and MZ B cells, but few CD23⁺ T2 and B-2 cells, as defined by Loder et al. (2). The few cells that fall within the B-2 cell gate have atypically high IgM levels. The majority of CD23⁺ PtC^int and PtC^neg cells are CD21^low. This phenotype suggests that these cells are intermediate to T1 and T2, as defined by Loder et al. (2). These data indicate that most PtC^neg and PtC^int B cells in these mice are transitional, suggesting that they are unable to become long-lived B-2 cells.

To more precisely define the developmental arrest, we examined the B cells from 6-1/Vκ1A Tg mice. 6-1/Vκ1A B cells do not bind PtC and are not selected into the B-1 subset (27, 28). The more mature CD23⁺ cells in these mice often express endogenous L chains, whereas the less mature CD23⁻ cells predominantly use the Vκ1A L chain (28). 6-1/Vκ1A/RAG-1^−/− mice were generated to eliminate the effect of endogenous L chain expression on 6-1 B cell development. 6-1/Vκ1A/RAG-1^−/− mice have similar numbers of bone marrow IgM^a+, B220^low immature B cells as 6-1 mice (Fig. 2,A and data not shown). Strikingly, however, 6-1/Vκ1A/RAG-1^−/− mice lack recirculating IgM^a+, B220^high B cells in the bone marrow (Fig. 2,A). The pro-B and pre-B populations in 6-1/Vκ1A/RAG-1^−/− mice are small, as they are in 6-1/Vκ4 Tg mice (Fig. 2 A), presumably because the presence of H and L chain transgenes accelerates passage through the pre-B cell stage.

In the spleen B cell numbers are approximately six times lower in 6-1/Vκ1A/RAG-1^−/− mice than in non-Tg and approximately five times lower than in 6-1 mice (p < 0.0001; Table I). The surface IgM level on 6-1/Vκ1A/RAG-1^−/− splenic B cells is high, but within the normal range (Fig. 2,B), and based on CD23 and CD21 expression, both T1 and the population seen in 6-1 mice that are intermediate to T1 and T2, as defined by Loder et al. (2), are present (Fig. 2,B and Table I). The majority of the CD23⁺ and CD23⁻ cells are also AA4⁺ (Fig. 2,B). The CD23⁻, AA4⁺ cells correspond to T1 cells, and the CD23⁺, AA4⁺ cells correspond to T2 cells, as defined by Allman et al. (3). The level of AA4 on the CD23⁺ cells is lower than that on the CD23⁻ cells, in agreement with the original description of T1 and T2 (3). Thus, we conclude that the splenic 6-1/Vκ1A B cells that differentiate belong to the T1 and T2 subsets as defined by Allman (3). These cells will be referred to as T1 and T2 to conform with this nomenclature. These cells express normal levels of CD19 and are HSA^high and CD5⁻ (Fig. 3,A). Consistent with non-Tg transitional cells, 6-1/Vκ1A B cells have a rapid turnover rate, suggesting a half-life of ∼2.5 days (Fig. 3,C). These mice lack detectable splenic B-2, and MZ B cells (Fig. 2,B and Table I) and peritoneal B-1 cells (data not shown), and their lymph nodes have few B cells, with those present having either a T1 or T2 phenotype (data not shown). Thus, 6-1/Vκ1A B cells appear to differentiatively arrest as CD23⁺, AA4⁺ T2 cells soon after exiting the bone marrow. There is no detectable serum IgM in these mice (data not shown), indicating that 6-1/Vκ1A B cells do not differentiate to plasma cells.

To assess whether negative selection can explain the developmental arrest of 6-1/Vκ1A B cells, we examined activation marker expression, the functional status of the cells, and the ability of the BCR to transduce signals. Negatively selected B cells exhibit one or more of a variety of features. These include elevated expression of activation markers (30, 31, 32, 33), unresponsiveness to LPS and T cell-derived factors (32, 34, 35, 36), and deficiencies in BCR signaling (31, 34, 36, 37). However, in contrast to negatively selected cells, 6-1/Vκ1A B cells express normal levels of multiple activation markers (Fig. 3,B) and undergo proliferation (Fig. 4,A) and Ab secretion in response to LPS and anti-CD40 plus IL-4 (Fig. 4,B) similar to transitional B cells from non-Tg mice. Non-Tg transitional B cells were obtained from mice that had been given a sublethal dose of irradiation 14 days before spleen cell harvesting. Allman et al. (38) have demonstrated that these cells, designated Ir-d14, are predominantly transitional B cells. Our flow cytometric analysis confirmed that the Ir-d14 cells used in this study are predominantly T1 and T2 cells (data not shown). As expected, anti-μ stimulation does not induce proliferation or Ab secretion by 6-1/Vκ1A B cells or control Ir-d14 cells, in contrast to mature non-Tg B cells. The pattern and magnitude of protein tyrosine phosphorylation in 6-1/Vκ1A B cells are similar to those in non-Tg transitional B cells following BCR cross-linking with F(ab′)₂ anti-mouse μ Abs (Fig. 4,C). The intensity of phosphorylation of some protein bands in 6-1/Vκ1A and Ir-d14 B cells is stronger than that in non-Tg B cells (Fig. 4,C), probably due to differences in surface IgM levels (Fig. 2). Thus, although negative selection cannot be excluded, 6-1/Vκ1A B cells resemble non-Tg transitional B cells and show no evidence of negative selection.

An alternative explanation for the block in maturation of 6-1/Vκ1A B cells is that they lack a positive selection signal. To test this possibility, 6-1/Vκ1A/RAG-1^−/− mice were injected with anti-BCR Abs. Two days after injection of 10 μg of anti-Igβ, the percentage of splenic CD23⁺ cells in 6-1/Vκ1A/RAG-1^−/− spleens increased from 50 to >80%. The levels of IgM, HSA, and AA4 decreased, and CD21 increased on CD23⁺ cells compared with unstimulated 6-1/Vκ1A B cells, indicative of maturation toward B-2 (Fig. 5, A and B). The expression levels of CD40, MHC class II, and CD5 were slightly higher, whereas CD80, CD86, and CD44 were unchanged (Fig. 5,B and data not shown). In addition, anti-Igβ-treated 6-1/Vκ1A B cells survived longer in culture than untreated 6-1/Vκ1A B cells, similar to mature non-Tg B-2 cells (Fig. 6,A), and were able to secret IgM Abs in response to IL-4 and anti-CD40 (data not shown). In contrast, increasing the level of injected anti-Igβ to 100 μg induced deletion (Fig. 6 B). Thus, these data suggest that weak in vivo BCR stimulation induces maturation of 6-1/Vκ1A B cells to B-2.

As an approach to provide Ag for positive selection, 6-1/Vκ1A/RAG-1^−/− B cells were exposed to a surge of self-Ags by transfer to sublethally irradiated C.B17 mice. Three days after splenic cell transfer, ∼10% of input donor B cells were recovered from recipient spleens. Almost all the recovered cells were CD23⁺ (data not shown), and the majority of these cells expressed lower levels of IgM and HSA and higher levels of CD21, suggesting that they had differentiated to mature B-2 cells (Fig. 6 C). The expression levels of CD5 and MHC class II were also higher, suggesting Ag stimulation. Bone marrow cell transfers gave similar results. Three weeks after bone marrow transfer, ∼60% of splenic IgM^a+ B cells were IgM^low and HSA^low suggesting differentiation to a B-2 cell (data not shown). The expression levels of CD5, CD40, and CD86 were also slightly increased in bone marrow recipients, although CD44 and CD80 were unchanged (data not shown), suggesting that the cells had encountered Ag. Serum IgM^a levels remained undetectable as long as 4 wk after bone marrow transfer (data not shown), arguing that the encounter with Ag was insufficient to induce plasma cell differentiation and was more consistent with the induction of B-2 maturation.

As 6-1/Vκ1A/RAG-1^−/− mice are a mixture of C.B17 and C57BL/6 backgrounds, the observed differentiation of donor B cells could be due to the absence of a negatively selecting Ag of C57BL/6 origin in C.B17 recipient mice. To rule out this possibility, spleen cell transfers were repeated in C57BL/6/RAG-1^−/− recipients. These transfers yielded indistinguishable results from those using C.B17 recipients (data not shown). Thus, the absence of a negatively selecting Ag of C57BL/6 origin cannot explain the induced differentiation of donor B cells.

This analysis demonstrates that most V_H12 B cells not selected into the B-1 subset are transitional. As the majority of newly developing V_H12 B cells do not bind PtC and are not selected to become B-1 (27), this suggests that most newly developing V_H12 B cells are unable to become mature B-2 cells. This is confirmed by the analysis of PtC^neg B cells from 6-1/Vκ1A/RAG-1^−/− mice. Only T1 and T2 cells are present in the spleen and lymph nodes of these mice, and no B cells, B-1 or otherwise, are present in the peritoneum, indicating that developmental arrest of these PtC^neg cells occurs at T2 soon after migration to the spleen. This supports the hypothesis that the observed bias in Vκ4/5H L chain gene use by mature B cells in 6-1 mice (27) is due to an inability of non-Vκ4/5H L chains to support differentiation to the mature B cell stage. This is the third checkpoint in V_H12 B cell differentiation, after the pre-BI to pre-BII and the pre-BII to immature B checkpoints (26, 29, 39), that restricts V_H12 B cells primarily to the production of H and L chains that form an anti-PtC BCR and drives differentiation to B-1.

The developmental arrest of 6-1/Vκ1A B cells could be explained by either negative selection or a lack of positive selection. B cell negative selection induces a wide range of responses. Some negatively selected B cells, such as anti-dsDNA and anti-MHC class I B cells, undergo receptor editing as a means of altering their specificity or are deleted in the bone marrow (5, 40). Other autoreactive B cells that developmentally arrest in the spleen have lower than normal surface IgM levels (30, 41) and in some cases elevated levels of activation markers such as CD44, HSA, and MHC class II, indicating that they have encountered Ag (30). Anergic autoreactive B cells also exhibit evidence of Ag encounter. Although they are long-lived mature B cells, they are unresponsive to a variety of activating stimuli and may have low levels of surface IgM and high levels of activation markers, including CD80, CD86, and MHC class II (31, 32, 36). In addition, some anergic B cells exhibit impaired protein tyrosine phosphorylation of cytoplasmic proteins upon BCR cross-linking (34, 36). Thus, negatively selected B cells that enter the periphery show phenotypic and functional evidence of an encounter with self-Ag. 6-1/Vκ1A B cells differ significantly from the B cells in these models. They are not deleted in the bone marrow, but differentiate to a short-lived transitional B cell in the spleen (Figs. 2 and 3). They have high levels of IgM, and their activation markers are not elevated (Figs. 2 and 3). In addition, they respond normally to BCR and non-BCR activation stimuli, and protein phosphorylation upon BCR cross-linking appears normal (Fig. 4). Thus, there is no evidence that these cells have encountered Ag and no support for negative selection as an explanation for the developmental arrest of 6-1/Vκ1A B cells.

The more likely explanation for the arrested differentiation of 6-1/Vκ1A B cells is a lack of positive selection. This is suggested by the finding that weak BCR stimulation in vivo induces differentiation of 6-1/Vκ1A transitional B cells to become mature B-2 cells (Fig. 5). Not only do treated cells resemble mature B-2 cells in phenotype, they also acquire increased resistance to apoptotic cell death in vitro (Fig. 6 A). 6-1/Vκ1A B cells can also be induced to acquire a more B-2-like phenotype by in vitro stimulation with anti-Igβ or anti-IgM Abs (data not shown). One day after treatment the percentage of 6-1/Vκ1A B cells expressing CD23 increases, and the IgM level on these cells decreases. BCR signaling of already negatively selected cells would be expected to reinforce negative selection and prevent further differentiation rather than promote differentiation, further arguing against negative selection.

Tonic signaling, which derives from an intact BCR (42) and is sufficient for differentiation to a splenic B cell (18), appears to be intact in 6-1/Vκ1A B cells. We have previously shown that the 6-1 H chain and Vκ1A L chain can pair and be secreted by cells of a transfected cell line (27), and bone marrow immature and splenic transitional B cells in 6-1/Vκ1A/RAG-1^−/− mice are IgM^high (Fig. 2), indicating that this H/L chain pair can generate a high level of BCR expression. That 6-1/Vκ1A B cells differentiate to splenic transitional B cells suggests that positive tonic signaling is intact in these cells. Nevertheless, whereas it cannot be excluded that the 6-1/Vκ1A BCR is unable to provide a sufficiently strong tonic signal to drive differentiation to the mature B-2 cell stage, its inability would have to be unrelated to its BCR expression level.

The ability of weak BCR stimulation to induce maturation of these B cells in vivo (Fig. 5) suggests that it is more probable that 6-1/Vκ1A B cells lack a ligand-mediated positive selection signal. The need for a separate signal at the T2 stage is consistent with the requirement for BCR signal transduction at the transitional B cell stage (17). It is particularly noteworthy that xid B cells deficient in Btk arrest in differentiation at the T2 cell stage (3), suggesting that Btk is necessary for mediating the ligand-mediated positive selection signal. Ligand-mediated positive selection at this stage would account for the differences in V gene use between transitional and mature B cells determined by Levine et al. (19). These findings suggest that certain V gene combinations are favored for maturity, which, based on the results presented in this study, is likely to be due to differences in the ability to bind self-ligands. Moreover, it suggests, as proposed by Levine et al. (19), that some of the cell death at the transitional stage is due to a lack of positive selection, not just negative selection. Ligand-mediated positive selection would ensure that only B cells with a BCR capable of Ag binding and of initiating signal transduction can become long-lived mature B cells. This is also consistent with the model proposed by Freitas and Rocha (43) in which transitional B cells compete with other B cells for survival resources to enter and remain in the mature B cell repertoire (43).

The requirement for ligand-mediated positive selection signal can provide an explanation for the differentiation of 6-1/Vκ1A B cells after adoptive transfer to irradiated recipients. Irradiation creates space for recovery of the B cell compartment (38), and this may allow transferred B cells to differentiate to mature B-2 cells more efficiently than in the steady state condition. An alternative possibility is that irradiation provides to B cells new self-Ags or higher concentrations of already available self-Ags that can mediate positive selection. This is suggested by our observation that negative selection of B cells specific for the ribonucleoprotein Sm, a self-Ag targeted in systemic lupus erythematosus, is greatly enhanced in irradiated 2–12H Tg mice compared with nonirradiated mice (unpublished observation) because of the association of Sm with apoptotic cell blebs (46). A third possibility is that irradiation induces cytokine release, which may facilitate T2 differentiation to B-2. We cannot rule out this possibility, but B cell-activating factor of the TNF family, a critical cytokine for mature B cell differentiation and survival (44), alone is unable to induce phenotypic changes in cultured 6-1/Vκ1A/RAG-1^−/− B cells (data not shown).

The fate of cells that are not positively selected by self-ligand appears to be cell death. However, splenic CD23⁺ B cells of RAG-sufficient 6-1/Vκ1A mice have endogenous Vκ rearrangements, and these mice also have a population of splenic PtC-binding B-1 cells expressing an endogenous Vκ4/5H L chain (28). The expression levels of RAG-1 and RAG-2 in 6-1/Vκ1A immature B cells are lower than those in control immature B cells from non-Tg mice and two other H chain Tg mice (data not shown), ruling out an autoantigen-induced receptor editing mechanism. It is unclear whether endogenous rearrangement is due to secondary rearrangement induced by the absence of ligand-mediated positive selection or to a leakiness of allelic exclusion by the Vκ1A transgene. However, Vκ1A Tg mice carrying a different H chain transgene show no evidence of endogenous Vκ rearrangement (data not shown), suggesting an induction mechanism. Additional experiments will be required to clarify this.

The strength of the ligand-mediated positive selection signal is critical. Too strong a signal induces apoptosis (Fig. 6). In addition, MZ and B-1 cells require self-Ag binding for differentiation (13, 16), and it will be of interest to determine how differences in BCR engagement induces different maturation pathways. The differences may be strictly quantitative, but we have observed, contrary to expectations, that positively selected anti-Sm B-1 cells are better binders than anergic anti-Sm B-2 cells (32, 45). Thus, there may be important qualitative differences that need to be elucidated.

We gratefully acknowledge the assistance of the Flow Cytometry Facility at the University of North Carolina. We are also indebted to Dr. Martin Weigert for the Vκ1A Tg mice.