Characterization of a late transitional B cell population highly sensitive to BAFF-mediated homeostatic proliferation

The precise sequence of differentiation of late splenic transitional B cells remains incompletely defined. Investigation from our own and other laboratories has suggested that the T2 B cell subset (CD21^hi/intIgM^hiIgD^hiCD24^hi) as originally defined by Loder et al. (19) represents a heterogeneous population. Pillai first introduced the idea of further delineating this population based on the relative expression of CD21 into T2-follicular precursor (CD21^int) and T2-MZ precursor (T2-MZP; CD21^hi) subsets (20). Although the properties of CD21^hi T2 B cells have become relatively well defined (20–22), little experimental data exists in regard to the functional properties and potential developmental capacity of CD21^int transitional B cells. We therefore initiated studies to define the properties of this B cell subpopulation in greater detail.

As shown in Fig. 1 A, CD24^hi B cells can be subdivided by CD21 expression into CD21^lo (which are labeled T1 B cells according to the Loder classification [19]), CD21^int (henceforth labeled CD21^int T2 B cells), and CD21^hi B cells. CD21^lo T1 B cells can be further divided into CD23^lo and CD23^hi subsets (which largely correspond to the T1 and T2 subsets described by Allman et al. [23]) and are hereafter referred to as CD23^lo T1 and CD23^hi T1 B cells. CD21^int cells are largely CD23^hi, whereas CD21^hi cells contain CD23^hi T2-MZP and CD23^lo MZ B cells. Follicular mature (FM) B cells express intermediate levels of both CD24 and CD21. The CD21^int T2 cells comprise 1–2% of total splenocytes from 8–12-wk-old mice (Fig. 1 B), and using 8-color flow cytometry we determined that they express high levels of IgM, IgD, AA4.1, CD23, and CD62L and intermediate-to-low levels of LFA-1 and CD1d (Fig. 1 C and Table S1). These combined phenotypic features are consistent with classification as a late transitional population. As an alternative measure of cell maturity, we evaluated relative GFP expression in Rag2p-GFP transgenic mice (24). In this model, Rag2 expression terminates at the immature BM B cell stage and GFP expression within the spleen is restricted to newly formed immature B cells (25). Analysis of GFP expression indicated that CD21^int T2 B cells, like T1 cells, express high levels of GFP in contrast to the GFP⁻ T2-MZP, MZ, and FM B cells (Fig. 1 D). Interestingly, a small percentage of CD21^int T2 B cells were GFP⁻; the significance of this finding will be discussed in detail below.

Notch signaling has been shown to play a crucial role in the development of T2-MZP and MZ B cells (21). To further delineate CD21^int T2 B cells from CD21^hi T2-MZP and MZ B cells, we determined relative expression of the Notch target genes hes1 and deltex1. As previously shown (21), Hes1 and Deltex1 expression was increased three- to fivefold in T2-MZP and MZ relative to T1, CD21^int T2, or FM B cells (Fig. 1 E). In addition, Notch2 haplodeficient mice exhibited no reduction in CD21^int T2 B cells, whereas T2-MZP and MZ numbers were reduced by 40–60% (unpublished data), indicating that Notch signaling is not essential for CD21^int T2 B cell development.

We next determined the in vitro functional responses of CD21^int T2 B cells to Toll-like receptor (TLR) or BCR ligation (Fig. 2 A). CD21^int T2 B cells proliferated robustly in response to LPS stimulation, similar to T2-MZP or MZ B cells. This response was significantly greater than in either T1 or FM B cells. BCR engagement led to a partial proliferative response in CD21^int T2 B cells, in contrast to T1 B cells. To further characterize this response, sorted cells were labeled with CFSE and cell division and survival were evaluated. After 48 h, although only a limited number of CD21^int T2 B cells survived, this population exhibited a rate of cell division that paralleled the response of FM B cells (Fig. 2 B). In contrast, no T1 cells survived BCR cross-linking. To exclude contamination with FM as the source of proliferating cells within CD21^int T2 cells, we performed parallel assays using mixtures of 90–95% T1 and 5–10% FM B cells (to mimic the maximal level of FM potentially present within the sorted CD21^int T2 population; Fig. S1) and observed no proliferation (Fig. 2 B).

To directly evaluate proapoptotic signaling, we measured BCR-induced activation of caspase 3. Although T1 cells exhibited a rapid increase in active caspase 3, this response was markedly blunted in CD21^int T2 B cells, similar to that in FM B cells. This reduction in caspase activity correlated with increased survival relative to T1 B cells with and without BCR engagement (Fig. 2, C and D).

In summary, CD21^int T2 B cells represent a phenotypically and functionally distinct transitional population. High expression of GFP, AA4.1, and CD24 clearly characterize CD21^int T2 B cells as immature splenic B cells. They are distinct from T1 B cells based on IgD, CD62L, and LFA-1 expression; increased resistance to BCR-induced cell death; and robust proliferation to TLR signaling. Finally, in contrast to CD24^hiCD21^hi T2-MZP and MZ B cells, CD21^int T2 B cells exhibit no evidence for increased Notch signals.

The slightly lower expression of both GFP and AA4.1 in CD21^int T2 versus T1 B cells suggested that CD21^int T2 B cells were derived from T1 cells. To test this idea, we evaluated peripheral B cell reconstitution after sublethal irradiation. Adult C57BL/6 mice were irradiated with 500 cGy and splenocytes analyzed on day 11, 13, 15, 17, and 19 after irradiation (Fig. 3). Whereas T1 B cells peaked on day 13, the number of CD21^int T2 B cells was maximal on day 15–17 and decreased markedly by day 19. In contrast, the number of FM B cells progressively increased between day 13 and 17, and they comprised the majority of cells by day 17–19. Smaller numbers of CD21^hi T2-MZP and MZ B cells could be identified beginning at day 17. A similar progression in B cell subset development was seen via analysis of splenic B cell subsets in unmanipulated neonates at various ages (unpublished data).

To directly analyze the developmental capacity of CD21^int T2 B cells, we performed adoptive cell transfer experiments. To obtain sufficient numbers of this relatively rare population, CD24^hiCD21^intIgM^hi B cells were purified from sublethally irradiated mice (day 15). Postsort purities were consistently >90%, with <0.2% contaminating T2-MZP/MZ and <1.5% FM B cells. Though more contaminating T1 cells were present, these should not alter our experimental interpretation, as they comprise precursors for CD21^int T2 B cells. Sorted populations exhibited the features of CD21^int T2 cells based on phenotypic markers, relative Deltex1 and Cyclin D2 mRNA expression, and BCR and TLR responses (Fig. S2, A-E).

Purified Ly5.1⁺ CD21^int T2 B cells were labeled with CFSE and injected into nonlymphopenic, adult Ly5.2⁺ C57BL/6 mice. Similar to previously published results (22), recovery of transferred cells was 1–1.5%. Even as early as day 1, transferred CD21^int T2 B cells developed into both FM (CD24^intCD21^intIgM^loCD1d^lo) and T2-MZP/MZ (CD24^hiCD21^hiIgM^hiCD1d^hi) B cells (Fig. 4). We conclude from these data that the CD21^int T2 subset contains immediate precursors for both FM and T2-MZP/MZ B cells.

We observed a significant number of larger cells within the CD21^int T2 cell population in contrast to either T1 or FM B cells (Fig. S3 A), suggesting that they were more activated. To address this possibility, we stained splenic B cell populations immediately ex vivo using Pyronin Y and DAPI to measure RNA and DNA content, respectively, thereby allowing us to determine the percentage of cells in each phase of the cell cycle (Fig. S3 B) (26). Cell doublets were carefully excluded before cell-cycle analysis based on DAPI width and area. Although on average, >85% of T1 and FM B cells were in G0, ∼50% of CD21^int T2 cells were in G1, and ∼5% of CD21^int T2 cells had further progressed into S and G2 phases (Fig. 5 A).

These data strongly suggested that splenic CD21^int T2 cells were proliferating in vivo. To directly test this idea, we performed in vivo BrdU-labeling experiments. We tested the prediction that, at early time points, BrdU incorporation would be significantly higher in CD21^int T2 cells compared with T1 B cells. At later times, we also predicted that (after the influx of newly formed BrdU-labeled cells from the BM), a larger percentage of T1 B cells would be BrdU⁺, as this population would comprise the precursors for additional CD21^int T2 B cells. Animals were fed BrdU continuously via the drinking water, and BrdU incorporation was analyzed within each splenic B cell subset at the indicated time points. On day 1 and 2 after BrdU administration, a significantly higher percentage of CD21^int T2 cells was BrdU⁺ (Fig. 5, B and C). However, this ratio reversed on day 4 and 5, with more T1 B cells labeled with BrdU. Consistent with previous results (23), BrdU labeling was higher in CD23^lo compared with CD23^hi T1 cells (Fig. 5 D).

We compared the functional activity and phenotype of quiescent versus active (G1, S, or G2) CD21^int T2 cells. Although the majority of transgenic CD21^int T2 B cells were GFP⁺, a limited subpopulation was GFP⁻ (∼25–35%; Fig. 1 D). As GFP is no longer transcribed, cell division was anticipated to dilute out the GFP signal in this subset. Thus, we evaluated the percentage of GFP⁺ versus GFP⁻ CD21^int T2 cells that had progressed beyond G0, and we observed an approximately twofold increase within the GFP⁻ subset (Fig. 6, A and B). The surface phenotype of resting versus cycling cells, however, remained indistinguishable, indicating that these populations represented nearly identical stages in development (Fig. S4). Strikingly, GFP⁻ CD21^int T2 B cells exhibited a significant enhancement in their response to BCR engagement, mainly through increased survival, yet retained the strong proliferative response to TLR signals (Fig. 6, C and D). Although the mitogenic responses of GFP⁻ CD21^int T2 cells were nearly identical to that observed for T2-MZP B cells (Fig. 2 A) (27), this subset exhibited no evidence of increased Notch2 signaling (Fig. 6 E).

A recent study has used κ-deleting recombination excision circles (KRECs), which are generated during light chain rearrangement, to determine the in vivo replication history of B cells (28). Therefore, we used this method to independently verify that GFP⁻ CD21^int T2 B cells have undergone cell divisions in vivo. The ΔC_T between genomic coding joints and the signal joint of KRECs (equivalent to the number of cell divisions) in GFP⁻ CD21^int T2 cells was comparable to T2-MZP/MZ cells, which were previously shown to have undergone several cell divisions (Fig. 7 A) (28). In contrast, T1, FM, or GFP⁺ CD21^int T2 B cells had undergone only 1–2 cell divisions.

Although one key characteristic of CD21^int T2 B cells was their capacity to proliferate in vivo, it remained unclear which environmental factors influence proliferation. In our transfer experiments using nonlymphopenic hosts, we observed minimal proliferation for up to 7 d after transfer (unpublished data). This finding was in accordance with published data, indicating that lymphopenia is the major stimulus for B cell homeostatic proliferation (HP) (29). Thus, we used recipient B cell–deficient μMT mice to compare the capacity of different B cell subsets to respond to signals inducing HP.

Purified Ly5.1⁺ T1 cells were labeled with CFSE and transferred into unmanipulated μMT mice (Ly5.2⁺) along with “feeder” B cells to partially blunt homeostatic signals (30). Splenocytes were isolated on day 2, 4, or 7 after transfer and evaluated for surface phenotype and relative proliferation by dilution of CFSE (Fig. 8 A). Among recovered splenic B cells, 1.5–2.5% were Ly5.1⁺ (derived from transferred T1 cells). The percentage of T1 cells at each of these time points was negligible (<0.5%), and, accordingly, this subset is not shown in the data presented. Within 2 d, ∼7% of labeled B cells had proliferated, and this proportion increased progressively to 62% by day 7. Consistent with this kinetic, CD21^int T2 cells had begun to proliferate at day 2, and nearly all cells within this gate proliferated within 7 d. We also identified transferred cells with a T2-MZP/MZ or FM B cell phenotype that had proliferated, albeit at a much lower percentage. These observations indicate that CD21^int T2 B cells preferentially cycle in a lymphopenic environment.

Next, we transferred purified, CFSE-labeled, Ly5.1⁺ CD21^int T2 B cells (sorted from sublethally irradiated mice) into Ly5.2⁺ μMT mice together with “feeder” B cells to directly assess the CD21^int T2 B cell response to homeostatic signals. Transferred cells markedly proliferated, with 55% having divided by day 3, and ∼65% having divided by day 5 and 7 after transfer (Fig. 8 B). Similar to transfer experiments in B cell–sufficient hosts, CD21^int T2 B cells developed into both FM and T2-MZP/MZ B cells. Analysis of various subsets demonstrated that CD21^int T2 B cells proliferated to the greatest extent, followed by T2-MZP/MZ cells, suggesting that CD21^int T2 B cell proliferation precedes that within the MZ pool. In contrast, newly formed FM B cells exhibited minimal proliferation. Transferred CD21^int T2 B cells also proliferated at earlier time points in comparison to cells derived from T1 B cells (Fig. 8, A and B), suggesting a requirement for T1 cell differentiation before active proliferation.

We also transferred GFP⁻ FM B cells from Rag2p-GFP transgenic mice into μMT mice to determine whether FM B cells might give rise to proliferating cells with a CD21^int T2 cell phenotype. Transferred FM B cells proliferated (30–40% on day 7 after transfer), albeit significantly less extensively than transferred CD21^int T2 B cells (Fig. 8, B and C). However, although we clearly identified T2-MZP/MZ B cells (∼60–80%) derived from the input FM population, no donor cells acquired a CD21^int T2 B cell phenotype (<1%). Thus, although FM B cells can cycle in response to homeostatic signals, cycling CD21^int T2 B cells are generated at an earlier developmental stage. These findings indicate that GFP⁻ CD21^int T2 cells are not derived from FM B cells, which is consistent with our interpretation that GFP⁺ and GFP⁻ CD21^int T2 cells represent closely related precursors of mature B cells.

Collectively, these results indicate that the CD21^int T2 B cell subset is poised to respond to homeostatic signals. However, although proliferation could accompany differentiation, it is not required for generation of the mature B cell pool. These additional transfer studies also support the hypothesis that CD21^int T2 B cells develop from T1 B cells, and subsequently give rise to both FM and T2-MZP/MZ B cells.

We next sought to identify stimuli that could induce HP of CD21^int T2 B cells. Among candidate growth factors, available data strongly implicate BAFF in splenic B cell survival and homeostasis (31, 32). Recent data also indicate that B cell deficiency leads to elevated plasma levels of BAFF (18). We therefore determined whether partial blocking of BAFF in B cell–deficient mice would alter the proliferative response of CD21^int T2 B cells.

To test this idea, we used an adenovirus vector encoding for TACI-Fc (TACI-Ad), which was previously demonstrated to block BAFF activity in vivo (33). Recipient μMT mice were injected i.v. with TACI-Ad or a control GFP vector, GFP-Ad. 7 d later, we adoptively transferred CFSE-labeled splenic B cells, and proliferation of transferred cells was analyzed at day 5 after transfer (Fig. 9 A). B cells in TACI-Ad–infected recipients exhibited lower expression of CD21, which is a direct target of BAFF signaling (34), and reduced numbers of CD21^hiCD1d^hi T2-MZP/MZ B cells (Fig. S5). BAFF inhibition also significantly reduced the number of proliferating B cells (Fig. 9 A), which is consistent with unpublished observations cited in a recent review (30). Most notably, a detailed subset analysis clearly indicated the preferential loss of proliferative responses within the CD21^int T2 population.

The reduction in HP after BAFF inhibition suggested that preferential cycling of CD21^int T2 B cells in vivo might reflect a greater capacity to compete for available BAFF. To address this idea, we analyzed BAFF-R expression in CD21^int T2 versus other B cell subsets; and, specifically, within cycling versus noncycling CD21^int T2 B cells. Bulk CD21^int T2 cells had higher BAFF-R levels than either T1 or FM B cells. Strikingly, GFP⁻ CD21^int T2 cells expressed the highest BAFF-R levels among all subsets (Fig. 9 B). Further, relative BAFF-R expression was greater in CD21^int T2 B cells in G1-G2 versus those in G0. This difference was not secondary to cell size, as expression of other surface markers remained unaltered (unpublished data).

Collectively, these results indicate that CD21^int T2 B cells preferentially cycle in response to homeostatic stimuli, and imply that this differential sensitivity is controlled by an enhanced capacity to respond to BAFF.

To look specifically for evidence of antigen-driven expansion of CD21^int T2 B cells, we first determined the relative κ- versus λ-light chain usage among peripheral B cell subsets. The majority of immature B cells express a BCR containing a κ-light chain as opposed to an λ-light chain, and the relative fraction of λ⁺ B cells declines between the immature and mature stages, presumably via negative selection (35, 36). Consistent with this interpretation, previous work suggests that λ-light chain expression is enriched on antigen receptors exhibiting autoreactivity (37). As anticipated, the κ:λ ratio was increased in the FM versus transitional B cell pool. However, when we compared GFP⁺ versus GFP⁻ CD21^int T2 B cells, we observed the opposite result with a significant decrease in this ratio in the more active GFP⁻ CD21^int T2 subset (Fig. 10 A). Notably, compared with FM B cells, the κ:λ ratio was also lower within the T2-MZP/MZ compartment and consistent with previous reports of selection for self-reactive cells within the MZ B cell compartment.

To further investigate the apparent skewing of the BCR repertoire in the CD21^int T2 compartment, we used transgenic mice expressing the antiphosphorylcholine (anti-PC) M167 V_h1 heavy chain (38). In this strain, B cells expressing the PC-binding V_h1/V_κ24 heavy/light chain pair are nearly undetectable within the BM. In contrast, Id-specific cells are significantly enriched within the MZ compartment, most likely through antigen-driven expansion (14, 38). To determine the point in B cell development at which enrichment for PC-binding, V_h1/V_κ24-expressing cells is first apparent, we stained peripheral B cell subsets with anti-Id antibodies. We observed a significant increase in Id-specific, CD21^int T2 B cells compared with T1 cells (5–10% vs. <1%; Fig. 10, B and C). Id-specific cells were further enriched within the MZ B cell population (15–20%), whereas the FM subset exhibited little change relative to T1 B cells. Although Id-specific CD21^int T2 B cells were phenotypically indistinguishable from non–Id-specific cells (unpublished data), a much greater percentage of Id-specific cells had progressed beyond G0 (Fig. 10 D). Together, these data strongly suggest that low-affinity self-antigen stimulation can promote the expansion of antigen-specific CD21^int T2 B cells, and functions in concert with other environmental cues to facilitate recruitment into the MZ compartment.

Although microenvironmental-mediated selection is clearly operative within the transitional B cell compartment, the precise subsets that serve as targets for selection remain incompletely defined and are the subject of continued controversy. Herein, we define a transitional population that, in contrast to T1 B cells, exhibits augmented responses to a range of potential microenvironmental stimuli and shows evidence of antigen-driven expansion.

CD21^int T2 B cells can be distinguished both functionally and phenotypically from other splenic B cell developmental populations. High expression of GFP (in Rag2p-GFP transgenic mice) definitively establishes CD21^int T2 B cells as an immature transitional population. However, in contrast to early transitional B cells, these cells are larger, express higher levels of IgD and CD62L, and exhibit a significantly blunted apoptotic response after BCR engagement. CD21^int T2 B cells display a rapid proliferative response to LPS stimulation that, among splenic B cell subsets, is characteristic of cells committed to the MZ B cell lineage. Low expression of the Notch target genes, however, definitively distinguishes these cells from the T2-MZP developmental subset. CD21^int T2 B cells cycle in vivo, and the actively cycling pool exhibits further augmentation in the capacity to respond to BCR engagement and/or costimulatory signals. Cycling CD21^int T2 B cells express the highest levels of BAFF-R among all peripheral B cell subsets, and BAFF mediates HP of this subset in vivo. Cycling CD21^int T2 B cells are partially enriched for λ⁺ cells, and M167 BCR Id-specific B cells are first selected within the cycling CD21^int T2 population, suggesting that specific BCR ligands may promote amplification of this subpopulation.

The observation that B cells enter the periphery as immature B cells arising from the BM was first published by Allman et al. (39, 40). The term transitional B cell was subsequently introduced by Carsetti et al. describing recently migrated immature splenic B cells in transition from the immature to the mature B cell compartments and a target of BCR-mediated selection (41). Although alternative criteria have been proposed (20, 42), the Loder (19) and Allman (23) schema are now widely used by a range of investigators to characterize splenic transitional B cells in various experimental contexts anticipated to impact peripheral B cell function. However, although these schemas use a similar nomenclature, they refer to discordant late transitional B cell populations (43) and, to a large degree, have implicated alternative events in shaping the mature B cell repertoire.

Loder et al. initially defined splenic CD21^loIgM^hiIgD^loCD24^hi T1 B cells as the key target for negative selection. Cells that survived this checkpoint became CD21^int/hiIgM^hiIgD^hiCD24^hi T2 B cells, and the authors proposed that an antigen-driven selection was required for the subsequent development of this population into either mature naive FM or MZ B cells (19).

In contrast to the aforementioned model, Allman et al. described an alternative developmental schema in which three nonproliferative transitional subsets (termed T1-T3) were identified based on relative expression of the cell surface markers AA4.1, CD23, and IgM. These developmental subsets exhibited a rapid cell turnover rate (which is consistent with a transitional stage) and no capacity to enter into the cell cycle upon BCR stimulation, but they exhibited a progressive increase in their responsiveness to CD40 and TLR stimulation (23, 42, 44).

Subsequent studies have only partially resolved the differences within these developmental models. It is now clear that the CD24^hiCD21^int/hi late transitional population described by Loder et al. is both functionally and developmentally heterogeneous (19). Mice lacking expression of the NFκB p50 subunit or the Notch2 receptor exhibit a selective loss of both MZ B cells and CD24^hiCD21^hi splenic B cells (21, 45). Based on these findings, Pillai et al. suggested that the CD24^hiCD21^hi T2 population is predominantly comprised of MZ-restricted progenitors (46). Consistent with this suggestion, Srivastava et al. showed that CD21^hi T2 cells comprise a long-lived, self-renewing compartment, and that development of FM B cells in sublethally irradiated animals precedes generation of this subset (22).

In contrast to CD21^hi T2 cells, the CD21^int T2 subset, as experimentally defined here, comprises a bipotent transitional population. Kinetic analysis of the peripheral B cell compartment after sublethal irradiation or after adoptive transfer of T1 cells into adult recipients demonstrates that CD21^int T2 cells are most likely derived from less mature splenic T1 (CD24^hiCD21^lo) progenitors. However, it might be possible that some splenic CD21^int T2 cells derive from early transitional cells in the BM, as two recent publications have shown that transitional cells can also develop in the BM and exit into the periphery (36, 47). Independent of where this development occurs, analysis of GFP expression (in Rag2p-GFP Tg mice) indicates that CD21^int B cells are newly derived (within 2–4 d, based on the half-life of GFP) from BM precursor B cells, and, in contrast to CD21^hi T2-MZP subset, do not comprise a long-lived population. In addition, the Notch pathway is minimally active in CD21^int T2 cells, and it is subsequently up-regulated in T2-MZP and MZ B cells. Finally, using adoptive transfer experiments in normal and lymphopenic recipients, we clearly show that CD21^int T2 B cells include the immediate precursors for both FM and MZ B cells.

Our findings also firmly establish that environmental cues can promote the proliferation of a subpopulation of cells within the CD21^int T2 B cell compartment. The question as to whether transitional B cells cycle has been a subject of controversy. T2 B cells were originally described as being comprised of a large percentage of cycling cells (19). This view is supported by a recent study indicating that ∼8% of B220⁺AA4.1⁺CD24^hi B cells are in the S/G2 phase (48), and by studies demonstrating a proportion of Ki-67⁺ cells within the transitional B cell compartment in human peripheral blood (49). In contrast, Allman et al. have consistently reported that transitional B cells do not contain significant numbers of cycling cells (22, 23), and this conclusion has been supported by others using similar gating criteria (44). We used several alternative approaches to determine whether CD21^int T2 B cells are cycling within the spleen, including DAPI/Pyronin Y staining, measurement of KRECs in splenic B cell subsets, and short-term in vivo BrdU-labeling kinetics. Our combined data from these analyses clearly indicate that a small but consistent fraction of CD21^int T2 B cells is cycling in vivo.

Differences between our data and previously published cell cycle analyses may be caused, in part, by alternative cell gating strategies. Using the Allman transitional B cell gating schema and seven-color flow cytometry, cycling CD21^int T2 cells fall within each of the T1–T3 subsets, and because of the relatively small size of this population they are nearly undetectable (Fig. S6). In contrast, gating as shown in Fig. 1 readily identifies cycling cells. This latter observation additionally strengthens our interpretation that CD21^int T2 B cells comprise a distinct, late transitional subset.

Our observations partially define the signals that mediate cycling within the late transitional B cell compartment. When transferred into a B cell–sufficient host, CD21^int T2 B cells proliferate very little, although they develop into both FM and T2-MZP/MZ B cells, indicating that cycling is not a prerequisite for entering the mature B cell pool. In contrast, as shown here and as previously described (30, 50), both mature and transitional B cells proliferate when introduced into a lymphopenic environment, a process termed HP. Because B cell deficiency leads to increased levels of BAFF in both humans and mice (18, 51), we hypothesized that BAFF signals are involved in HP. Although B cells do not proliferate in response to BAFF in vitro, BAFF can induce entry into early G1 (52). Moreover, stimulation with BAFF leads to an increase in cell size and glucose metabolism (31), up-regulation of Ki-67, Cyclin D2, and Cdk4 expression, and phosphorylation of RB and Akt (53). These findings, combined with our experimental data, strongly suggest that in vivo, BAFF is critical for cycling of CD21^int T2 B cells. Consistent with this model, inhibition of BAFF markedly attenuated the cycling of CD21^int T2 cells in μMT hosts, and CD21^int T2 B cells in G1 and beyond expressed the highest levels of BAFF-R, findings that directly correlated with preferential proliferative responses in lymphopenic hosts.

There is mounting evidence that alterations in transitional B cell homeostasis mediated by BAFF can significantly impact the mature BCR repertoire (30, 32, 54). Under normal physiological conditions, B cells directly compete for available BAFF (18, 54), autoreactive B cells can be rescued by strong BAFF signals (17, 18), and data from both humans and mice indicate that these selective forces operate at the transitional stage (6, 54). Whether autoreactive B cells can be specifically expanded within the transitional B cell pool, however, represents a crucial unanswered question. Using the M167 BCR heavy chain transgenic mouse, we observed a marked expansion of PC-binding B cells within the CD21^int T2 compartment. Further, our data demonstrate that this Id-specific CD21^int T2 B cell population is significantly enriched for cycling cells. In addition, we also identified a clear increase in the relative percentage of λ⁺ cells specifically within the cycling CD21^int T2 B cell population. This enrichment contrasts with the progressive decline in λ⁺ cells observed across other sequential peripheral B cell subsets, as shown here and as previously described (36). Collectively, these observations strongly imply that cells expressing a self-reactive BCR can be specifically amplified with the CD21^int T2 B cell subset.

Although proliferation or antigen-driven expansion of CD21^int T2 cells is clearly not required for the generation of mature B cells, our combined data support a model in which antigen encounter (in concert with environmental cues including BAFF) can promote the competitive expansion of cells within this compartment. According to the measurement of KRECs, CD21^int T2 cells and T2-MZP/MZ cells have undergone a similar number of cell divisions. However, our BrdU studies suggest that this cell division occurs preferentially at the CD21^int T2 stage. These combined findings imply that CD21^int T2 cells, upon receiving the appropriate signals, begin to divide and are preferentially driven into a MZ B cell fate. Because BCR engagement positively modulates BAFF-R expression (16), our findings also suggest that BCR signaling coordinately amplifies BAFF-independent and -dependent signals within the cycling CD21^int T2 subpopulation. This selection program appears to be enhanced in conditions leading to high BAFF levels, as experimentally implied in our adoptive transfer and BAFF inhibition experiments. Accordingly, in BAFF transgenic mice, autoreactive cells are primarily found within the MZ (48) and high BAFF levels in humans promote autoimmune diseases, including Sjogren's syndrome (55).

Finally, in light of the developmental capacity of the CD21^int T2 B cell subset, it remains possible that antigen-expanded cells might also develop into FM B cells. Although it is not addressed in this study, such a scenario might be operative in selected autoimmune-prone genetic contexts. Identification of a transitional population where self-reactive B cells are initially expanded sets the stage for a range of future studies investigating the selective forces operative at this key point in peripheral B cell development.

BALB/c, C57BL/6 (Ly5.1 and Ly5.2), μMT, and Rag2p-GFP mice were bred and maintained in the specific pathogen–free animal facility of the Children's Hospital and Regional Medical Center (Seattle, WA) and handled according to Institutional Animal Care and Use Committee approved protocols. Notch2 haplodeficient mice were provided by T. Gridley (Jackson ImmunoResearch Laboratories). M167 heavy chain transgenic mice (M167H Tg mice, line U243-4) were provided by J. Kenny and A. Lustig (National Institute of Aging, Bethesda, MD) and established as a M167H Tg/Tg homozygous breeding colony in the animal facility of the Albert Einstein College of Medicine. Mice used in all experiments were between 6 and 16 wk of age.

Antibodies used in this study included reagents specific for the following: active-caspase 3 (C92-605), CD24 (M1/69), CD21 (7G6), B220 (RA3-6B2), and CD1d (1B1; BD Biosciences); CD23 (B3B4; Invitrogen); Ly5.1 (A20), Ly5.2 (RA3-6B2), AA4.1, and CD62L (MEL-14; eBioscience); IgD (11-26), κ-LC (187.1), λ-LC (JC5-1), and IgM (1B4B1; SouthernBiotech); CD19 (ID3; BioLegend); and BAFF-R (204406; R&D Systems). Polyclonal F(ab')₂ anti-IgM for BCR stimulation and nonstimulatory Fab anti-IgM for cell sorting were purchased from Jackson ImmunoResearch Laboratories. Antiidiotypic M167 (28-5-15) was produced from a hybridoma line provided by J. Kenny and A. Lustig, and it was purified and conjugated with Alexa Fluor 647 using standard methods. Additional reagents included the following: 7-amino actinomycin D (7-AAD), Pyronin Y, and DAPI (Invitrogen); LPS (Sigma-Aldrich); anti-CD40 (a gift from G. Cheng, University of California Los Angeles, Los Angeles, CA); and recombinant mouse BAFF (Thermo Fisher Scientific).

Single-cell suspensions from BM and spleen were incubated with fluorescently labeled antibodies for 30 min at 4°C in staining buffer (PBS with 0.5% BSA or 2.5% FCS). Data were collected on a FACSCalibur or LSR II flow cytometer (BD Biosciences) and analyzed using FlowJo software (Tree Star, Inc.). Overlays of histogram are shown using the offset mode in FlowJo. For LSR II experiments, the data were analyzed using biexponential transformation function for complete data visualization. For caspase 3 assays, cells were surface stained, fixed, and permeabilized per manufacturer's instructions, and incubated with anti–active caspase 3 for 45 min at room temperature. 7-AAD was used to differentiate between dead and live cells. For cell cycle analysis, cells were surface stained, fixed overnight with 2% PFA, washed, and incubated for 30 min with 1 μg/ml DAPI in PBS with 0.1% BSA and 0.1% Triton X-100. 1.5 μg/ml Pyronin Y was added immediately before analysis.

For cell sorting, CD43 depletion was performed using magnetic bead–conjugated anti-CD43 antibodies according to the manufacturer's instructions (Miltenyi Biotech), and enriched cells were labeled with specific antibodies in staining buffer. Sorting was performed using a FACSAria sorter with FACSDiva software (BD Biosciences).

Purified cells were cultured in complete media (RPMI 1640 with 10% FCS, 4 μM l-glutamine, 50 μM 2-ME, 10 mM Hepes, and antibiotics) at 37°C. Cells were stimulated with 10 μg/ml polyclonal goat F(ab')₂ fragment anti–mouse IgM, 1 μg/ml anti–mouse CD40, 10 μg/ml LPS, or 50 ng/ml BAFF; incubated in triplicate (5 × 10⁴ cells/well) for 48 h; pulsed with 1 μCi [³H]thymidine for 8 h before harvesting, and analyzed using a scintillation counter.

Sorted B cell populations were pelleted and frozen at −80°C. RNA was isolated using the RNeasy Micro kit (QIAGEN) and converted into cDNA by reverse transcription (Superscript II; Invitrogen) according to the manufacturer's instructions. Real-time PCR on cDNA was performed using the iCycler real-time PCR detection system with IQ SYBR Green Supermix (Bio-Rad Laboratories) according to the manufacturer's instructions. Ratios were calculated using the Pfaffl's mathematical model for relative quantification (56), with mouse β2-microglobulin as housekeeping control. All real-time PCR analysis shown includes at least three independent experiments. Primer sequences used include the following: Hes1-5′GAGAAGAGGCGAAGGGCAAGAAT; Hes1-3′ GAGGTGCTTCACAGTCA; Deltex1-5′ CGGACATTTGAGACCCACTT; Deltex1-3′ CCACTTTCAAGCAGGGAGAA; CyclinD2-5′ GCCAAGATCACCCACACT; and CyclinD2-3′ GCTGCTCTTGACGGAACT. To determine the replication history of B cells, genomic DNA was isolated from sorted populations and the ratio between the κ-deleting rearrangement (IRS1 to RS) and excision circles (KREC) was determined by TaqMan-based (Applied Biosystems) real-time PCR, as previously described (28).

Continuous in vivo BrdU labeling was performed by feeding mice BrdU via the drinking water containing 1 μg/ml BrdU (Sigma-Aldrich) and 10% sucrose. Mice were killed at indicated time points, and splenocytes were surface stained to identify splenic B cell subsets. Cells were then fixed and permeabilized, treated with DNase, and stained with anti-BrdU FITC (BrdU kit from BD Biosciences).

CD43-depleted B cells or sorted B cell subsets were incubated with 0.05 μM CFSE at 37°C for 8 min, washed three times with complete media, and then washed with PBS. In experiments using “feeder” B cells, CD43-depleted splenic B cells were prepared and mixed with sorted B cells. For transfer of sorted T1 or CD21^int T2 B cells, a total number of 15 × 10⁶ cells in 250 μl PBS were injected by tail vein into μMT or C57BL/6 mice.

TACI- or control GFP-adenovirus (provided by R. Carter, University of Alabama, Tuscaloosa, AL) were prepared as previously described (33) and injected into the tail vein at 10⁹ PFU per recipient, followed 6–7 d later by transfer of 20 × 10⁶ CD43-depleted, CFSE-labeled splenic B cells (Ly5.1⁺). Mice were killed 5 d after adoptive transfer, and splenocytes were stained for B220, CD19, CD24, CD21, CD1d, IgM, and Ly5.1.

Fig. S1 shows sorting gates and postsort purities. Fig. S2 shows the phenotypic and functional properties of CD21^int T2 cells from irradiated mice. Fig. S3 shows how cell cycle subsets are gated. Fig. S4 shows surface markers on GFP⁺ versus GFP⁻ CD21^int T2 cells. Fig. S5 compares the expression of cell surface markers on B cells from control versus TACI-Ad–treated mice. Fig. S6 compares transitional subsets, as gated by Allman et al., and CD21^int T2 B cells. Table S1 summarizes the expression of cell surface markers on peripheral B cell subsets described in this work.

We thank R. Carter for providing TACI-adenoviral vector; M. Nussenzweig (Rockefeller University) for Rag2p-GFP transgenic mice; J. Kenny and A. Lustig for M167H transgenic mice and hybridoma lines; and S. Khim, N. Blake, and A. Hui for expert technical assistance with animal studies, flow cytometry, and manuscript preparation, respectively.

Support for this work has included the Elizabeth Campbell Endowment (A. Meyer-Bahlburg), Cancer Research Institute immunology training grant (S.F. Andrews), the Medical Scientist Training Program at the Albert Einstein College of Medicine (K.O.A. Yu), and National Institutes of Health grants HD37091, CA81140, AI45889, and AI063537.

The authors have no conflicting financial interests.