Foxo3 Promotes Apoptosis of B Cell Receptor–Stimulated Immature B Cells, Thus Limiting the Window for Receptor Editing

The development of a diverse B cell repertoire is crucial for normal humoral immune responses. However, this diversity comes at a price, as many of the B cells generated in the bone marrow express BCRs that recognize self-antigens. Failure of tolerance checkpoints that eliminate or inactivate these autoreactive B cells can lead to autoimmune diseases such as systemic lupus erythematosus (SLE), in which autoantibodies are produced and form immune complexes that induce inflammation and tissue damage. At the immature B cell stage of development, the BCR is first fully assembled and tested for functionality. A basal or tonic signal through an unligated, innocuous (nonautoreactive) BCR is necessary for continued cell survival and maturation (1–3). This is mediated by PI3K signaling (2, 4). Disruption of this tonic signal, inhibition of the PI3K pathway, or strong engagement of the BCR by self-antigen result in receptor editing, in which B cells continue L chain rearrangements in an attempt to change their specificity. Cells remaining autoreactive after a few rounds of editing are eliminated by clonal deletion (2–6).

Foxo transcription factors are downstream targets of PI3K that have proapoptotic and antimitogenic effects in numerous cell types (7, 8). Two Foxo family members, Foxo1 and Foxo3, have each been shown to play unique roles at several stages of B cell development (9–14). Upon activation of mature B cells via the BCR, PI3K signaling is activated and downregulates Foxo function at two levels: 1) by reducing their expression at the mRNA level (10, 14) and 2) by inducing their phosphorylation by Akt and their subsequent exclusion from the nucleus (7, 9). In contrast, BCR cross-linking blocks activation of PI3K in immature B cells (2), resulting in nuclear localization of both Foxo1 and Foxo3 (11, 15). The activation of Foxo family transcription factors in Ag-engaged immature B cells suggests that they might play a role in central B cell tolerance. Indeed, Foxo1 is known to promote Rag expression in immature B cells and thus receptor editing, whereas the role of Foxo3 in these processes is poorly understood (11–14). We previously demonstrated that although Foxo3^−/− mice have reduced numbers of pre–B cells (for unknown reasons), they have normal numbers of immature B cells (14). We hypothesized that this relative increase from the pre–B cell to the immature B stage could be indicative of increased immature B cell survival in the absence of Foxo3 due to a role for Foxo3 in immature B cell apoptosis.

In this study, we show that Foxo3 plays a unique role in promoting apoptosis of BCR-stimulated immature B cells. Our results suggest that receptor editing is unimpaired and in fact enhanced in Foxo3^−/− mice, as measured by both Igλ expression and recombining sequence (RS) recombination. This is likely a result of a longer editing window due to reduced apoptosis, as germline Igλ expression was not significantly elevated in Foxo3^−/− pre–B cells. These results support a model in which Foxo1 and Foxo3 promote receptor editing and apoptosis, respectively, in immature B cells expressing a nonfunctional or autoreactive BCR. Although Foxo3^−/− mice do not develop autoantibodies, reduced expression of Foxo3 mRNA was observed in B cells from SLE patients with anti-dsDNA Abs and high disease activity compared with patients without these characteristics. Although this is likely due in part to increased B cell activation in these SLE patients, it is also possible that low-affinity B cells that remain autoreactive after editing may survive inappropriately in the absence of Foxo3 and become activated to secrete autoantibodies in the context of other SLE-associated defects.

Animal studies were approved by the University of Texas Southwestern Institutional Animal Care and Use Committee. Foxo3^−/− mice (16) and wild type controls were on the FVB background. MD4 (anti–hen egg lysozyme [HEL] Ig) (JAX 002595) (17) and KLK (membrane-bound HEL [mHEL]) (JAX 002598) (18) mice on the C57BL/6 background were obtained from The Jackson Laboratory and crossed to Foxo3^−/− mice, which had been backcrossed three generations onto the C57BL/6 background. Foxo3^−/− mice and littermate controls backcrossed to C57BL/6 (B6) for four or five generations were also used in some experiments. Experimental and control mice were age and gender matched, and littermate controls were used whenever possible.

Apoptosis was measured by staining with Annexin V FITC (BD Biosciences). To assay Igλ usage, splenocytes were depleted of RBCs and stained with anti–CD21-FITC, anti–CD23-PE, anti–B220-PerCP, and anti–Igλ_1,2,3-biotin (BD Biosciences or Tonbo Biosciences). The biotinylated Ab was detected with streptavidin-allophycocyanin (BD Biosciences). Bone marrow cells were depleted of RBCs and stained with anti–B220-FITC, anti–Igλ_1,2,3-biotin, anti–B220-PerCP, and anti–CD93/AA4.1-allophycocyanin (BD Biosciences or Tonbo Biosciences). The biotinylated Ab was detected with streptavidin-PE (BD Biosciences). T3 cells were identified by staining splenocytes with anti–B220-FITC, anti–CD23-PE, anti–IgM-PerCP, and anti–AA4.1-allophycocyanin (BD Biosciences or Tonbo Biosciences). In studies with the MD4 anti-HEL Ig transgene system, splenocytes were depleted of RBCs and stained with combinations of anti–IgMb-FITC, anti–IgMa-PE, anti–B220-PerCP, anti–CD19-allophycocyanin (antibodies from BD Biosciences or Tonbo Biosciences), and HEL (Sigma-Aldrich) labeled with Alexa 488 using an Alexa Fluro 488 Microscale Protein Labeling Kit (Molecular Probes) according to the manufacturer’s instructions.

Samples were run on a FACS Calibur (Becton Dickinson). Data were analyzed with CellQuest (BD Biosciences) or Flowjo (Tree Star).

Bone marrow was first depleted of RBCs, and B cells were then purified with anti-B220 magnetic beads using the IMag system (BD Pharmingen). B cells were subsequently stained with anti–IgM-PE, anti–B220-PerCP or anti–B220-FITC, and anti–CD93/AA4.1-biotin or anti–CD93/AA4.1-allophycocyanin (BD Biosciences or Tonbo Biosciences). The biotinylated Ab was detected with streptavidin-allophycocyanin (Caltag or BD Biosciences). Immature (B220⁺, IgM⁺, AA4.1⁺) and pre–(B220⁺, IgM⁻, AA4.1⁺) B cells were sorted on a FACS Aria (Becton Dickinson) or a MoFlo (Cytomation) cell sorter. Samples were kept at 4°C at all times prior to and during the sort and until stimulation to avoid activating the BCR with the sorting Abs (11). Purified cells were either harvested immediately or stimulated with 10 μg/ml goat anti-mouse IgM F(ab′)₂ (Jackson ImmunoResearch Laboratories) for the indicated times at 10⁶ cells/ml in RPMI 1640 plus 10% FBS. Alternatively, bone marrow B lineage cells were expanded by culturing for 4–6 d in 10 ng/ml IL-7 (R&D Systems) at 2 × 10⁶ cells/ml and pre– and immature B cells sorted as above.

Total splenic B cells were purified from RBC-depleted splenocytes by either negative selection with anti-CD43 magnetic beads (Miltenyi) or by positive selection with anti-B220 beads (BD Biosciences) according to the manufacturer’s instructions. To purify Igλ⁺ and Igλ⁻ (and thus Igκ⁺) cells, Igλ⁺ cells were purified from splenocytes using anti–Igλ_1,2,3-biotin and streptavidin magnetic beads (BD Biosciences). B220⁺ cells were then purified from the Igλ-depleted cells by positive selection with magnetic beads (BD Biosciences). To purify Igκ⁺ and Igκ⁻ (and thus Igλ⁺) cells, Igκ⁺ cells were purified from splenocytes using anti–Igκ-biotin and streptavidin magnetic beads (BD Biosciences), and B220⁺ cells were purified from the Igκ-depleted cells by positive selection with magnetic beads (BD Biosciences). To purify marginal zone (B220⁺CD21⁺CD23⁻) and transitional (B220⁺CD21⁻CD23⁻) B cells, splenocytes were depleted of CD23⁺ cells using anti-CD23 magnetic beads (Miltenyi Biotec) according to the manufacturer’s instructions, stained with anti–CD21-FITC, anti–CD23-PE, and anti–B220-PerCP or -allophycocyanin (BD Biosciences or Tonbo Biosciences), and sorted on a FACS Aria (Becton Dickinson) or a MoFlo (Cytomation) cell sorter.

Total RNA was prepared using the RNeasy kit (Qiagen). cDNA was generated with a cDNA Archive Kit (Applied Biosystems). Real-time PCR was performed using a Bio-Rad CFX96 Real-Time System using TaqMan reagents for mouse Bcl2l11 (Bim) and the internal control GAPDH (Applied Biosystems). Data were normalized to GAPDH using the δ comparative threshold cycle method.

Total cell lysates were generated by boiling cells in 2× Laemmli sample buffer. Samples were run on a 4–15% gradient SDS–PAGE gel (Bio-Rad) and transferred to nitrocellulose (GE Healthcare, Amersham, U.K.). Blots were blocked in 5% milk and probed with a 1:1000 dilution of anti-Bim (clone C34C5; Cell Signaling Technology) or anti–β-actin (clone D6A8; Cell Signaling Technology) rabbit mAbs. Blots were washed three times in TBST, probed with a 1:2500 dilution of goat-anti-rabbit HRP (Bio-Rad), washed three times in TBST, and developed with Clarity ECL substrate (Bio-Rad). Bands were detected using a Chemidoc System (Bio-Rad) and quantified with Image Lab Software (Bio-Rad).

Total RNA was prepared from pre–B cells using the RNeasy kit (Qiagen). cDNA was generated with a cDNA Archive Kit (Applied Biosystems). Serial dilutions of cDNA were subjected to PCR for germline transcripts for Igλ1, Igλ2, and Igλ3, as well as β-actin as a loading control [as described in (19, 20)], using the following primers: Igλ1: Stlambda1for2 5′-CTTGAGAATAAAATGCATGCAAGG-3′, Clambda1R3 5′-TGATGGCGAAGACTTGGGCTGG-3′; Igλ2: Stlambda2for 5′-GAGAACAGGACCAGGTGCTG-3′, Clambda2rev 5′-ACACGGTGAGAGTGGGAGTG-3′; Igλ3: Stlambda3for 5′-CCCAGGTGCTTGCCCCAC-3′, Clambda3R3 5′-TGTTTTCCTGGAGCTCCTCAGG-3′; and β-actin: betaActfor 5′-CTTTTCCAGCCTTCCTTCTT-3′, betaActrev 5′-ACGCAGCTCAGTAACAGTCC-3′. PCR reactions were run on an agarose gel and band intensities were quantified using VisionWorks LS software (UVP).

Genomic DNA was isolated using the DNeasy Blood and Tissue kit (Qiagen) according to the manufacturer’s instructions. Serial dilutions of DNA were subjected to PCR to detect RS recombination with the following primers [as described in (21–23)]: Vk degenerate primer, 5′-GGCTGCAGSTTCAGTGGCAGTGGRTCWGGRAC-3′(S = G or C, R = A or G, W = T or A) and RS-101, 5′-ACATGGAAGTTTTCCCGGGAGAATATG-3′. Primers from the Ets1 gene (5′-AAAGCTGACCCATGTGCTCT-3′ and 5′-TGCAGTGGATTCAGGCATTA-3′) were used as a loading control. PCR reactions were run on an agarose gel, and band intensities were quantified using VisionWorks LS software (UVP).

Autoantibody reactivity against a panel of 124 autoantigens was measured using an autoantigen microarray platform developed by the University of Texas Southwestern Medical Center (https://microarray.swmed.edu/products/category/protein-array/) (24). Briefly, serum samples were pretreated with DNAse-I and then diluted 1:50 in PBS with Tween 20 buffer for autoantibody profiling. The autoantigen array bearing 124 autoantigens and 4 control proteins was printed in duplicates onto Nitrocellulose film slides (Grace Bio-Labs). The diluted serum samples were incubated with the autoantigen arrays, and autoantibodies were detected with cy3-labeled anti-mouse IgG and cy5-labeled anti-mouse IgM using a Genepix 4200A scanner (Molecular Device) with laser wavelengths of 532 and 635 nm. The resulting images were analyzed using Genepix Pro 6.0 software (Molecular Devices). The median of the signal intensity for each spot was calculated and subtracted from the local background around the spot, and data obtained from duplicate spots were averaged. The background-subtracted signal intensity of each Ag was normalized to the normalization factors generated based on the overall background signal of each array. Finally, the net fluorescence intensity for each Ag was calculated by subtracting a PBS control, which was included for each experiment as negative control. Signal-to-noise ratio was used as a quantitative measurement of the true signal above background noise. Signal-to-noise ratio values equal to or greater than three were considered significantly higher than background and were therefore true signals. The net fluorescence intensity of each autoantibody was used to generate heatmaps using Cluster (http://bonsai.hgc.jp/~mdehoon/software/cluster/) and TreeView (http://sourceforge.net/projects/jtreeview/files/jtreeview/) software. Each row in the heatmap represents an autoantibody, and each column represents a sample. Red color represents the signal intensity higher than the mean value of the raw, and green color means signal intensity is lower than the mean value of the raw. Gray or black color represents the signal is close or equal to the mean value of the raw.

Total splenic B cells, marginal zone B cells, transitional B cells, or Igκ-depleted B cells were cultured at 10⁶ cells/ml in complete RPMI 1640 media with or without 5 mg/ml LPS (Sigma-Aldrich). Supernatants were harvested after 5 d of culture.

Anti-dsDNA and/or anti-ssDNA ELISAs were performed on serial dilutions of serum (1:100, 1:400, 1:1600) or culture supernatant (neat, 1:5, 1:25, 1:125) as described in (25). Anti-IgM, IgG, anti-Igκ, or anti-Igλ alkaline phosphatase–conjugated detection Abs (Southern Biotech) were used as indicated in the figure legends.

Publicly available gene expression data from GSE10325 (26) were analyzed for correlation of Foxo3 (204131_s_at) expression in B, T, and myeloid cells with the SLE disease activity index (SLEDAI) [as reported in (26)] and for the level of Foxo3 (204131_s_at) expression in B, T, and myeloid cells in patients with versus without anti-dsDNA Abs [as defined in (26)]. Publicly available gene expression data from GSE49454 (27) were analyzed for the level of Foxo3 (ILMN_1844692) expression in patient samples with <50 U versus >60 U anti-dsDNA Abs. Statistical analysis was performed using Graph Pad.

We previously demonstrated that although Foxo3^−/− mice have reduced numbers of pre–B cells (for unknown reasons), they have normal numbers of immature B cells (14). We confirmed this in an additional cohort of mice (Fig. 1A, Supplemental Fig. 1A). Given the known proapoptotic role of Foxo3 in other cell types (7, 8), we hypothesized that this relative recovery of the immature B cell population in Foxo3^−/− mice may be due to a reduction in apoptosis during central tolerance mechanisms. Indeed, BCR-stimulated primary immature B cells from Foxo3^−/− mice showed increased survival compared with their wild type counterparts (Fig. 1B, 1C). The proapoptotic BH3 family member Bim is a known Foxo3 target in other cell types (7) and has been shown to promote immature B cell apoptosis (28, 29). However, although Bim mRNA expression was decreased in BCR-stimulated Foxo3^−/− immature B cells in some experiments, this was not consistent (Fig. 1D) nor was there a significant difference in Bim protein expression (Fig. 1E, 1F). Thus, Foxo3 promotes apoptosis in BCR-stimulated immature B cells via mechanisms other than control of Bim expression.

In addition to undergoing apoptosis, immature B cells can undergo receptor editing in response to Ag encounter. This has been shown to be mediated by Foxo1, whereas the role of Foxo3 in this process is controversial (11, 12). To determine whether Foxo3^−/− mice exhibit signs of altered receptor editing, we first assessed their usage of Igλ. Rearrangement of the κ L chain locus occurs first, and production of a functional κ L chain inhibits rearrangement of the λ locus. Thus, the λ locus will usually only undergo recombination if an in-frame κ L chain cannot be produced or if receptor editing occurs because of an autoreactive κ L chain (30). We thus examined bone marrow and splenic B cell subpopulations for Igλ expression by flow cytometry. We found that a significantly larger proportion of immature B cells in the bone marrow and spleen of Foxo3^−/− expressed Igλ than was the case in wild type mice (Fig. 2, Supplemental Figs. 1A, 1B, 2A, 2B). The frequency of Igλ⁺ cells in more mature B cell populations (follicular and marginal zone B cells in the spleen and mature recirculating cells in the bone marrow) was more similar between wild type and Foxo3^−/− mice, although it was still slightly increased in some animals in the absence of Foxo3 (Fig. 2).

To determine whether the increased Igλ usage could result in part from enhanced accessibility of the Igλ locus, we examined expression of Igλ germline transcripts in pre–B cells from wild type and Foxo3^−/− mice (Fig. 3). No significant difference was observed. This suggests that receptor editing not only occurs independently of Foxo3 but appears to be increased in its absence. As a more direct measure of receptor editing, we performed a PCR assay for Vκ-RS recombination, which occurs when B cells have exhaustively rearranged an Igκ allele (21, 22). Although RS recombination levels in total, Igλ⁻ (and thus Igκ⁺), and Igκ⁺ splenic B cells were normal in Foxo3^−/− mice, both Igλ⁺ and Igκ⁻ (and thus Igλ⁺) B cells from Foxo3^−/− mice demonstrated increased Vκ-RS recombination (Fig. 4). This suggests that developing autoreactive Foxo3^−/− B cells are more likely to edit their Igκ alleles to exhaustion prior to using Igλ because of a longer editing window.

Reduced apoptosis of immature B cells in the absence of Foxo3 may result in the inappropriate release of B cells that remain autoreactive even after editing. We thus sought to determine whether Foxo3^−/− mice demonstrate signs of humoral autoimmunity. We first asked whether Foxo3^−/− mice accumulated anti-DNA Abs with age. There was no difference between anti-ssDNA or anti-dsDNA IgM levels between wild type and Foxo3^−/− mice at 9–11 mo of age (Fig. 5A). To obtain a broader view of potential autoreactivity, we analyzed serum from aged wild type and Foxo3^−/− mice using an autoantigen array that allows the simultaneous interrogation of autoreactivity against ∼100 self-antigens commonly targeted in autoimmune disease (24). We did not observe an increase in either IgM or IgG autoantibodies in Foxo3^−/− mice using this approach (Fig. 5B, Supplemental Tables I, II).

To test whether autoreactive B cells may be present at increased frequencies in Foxo3^−/− mice but not activated in vivo, we purified B cells from wild type and Foxo3^−/− mice by negative selection with anti-CD43 magnetic beads, stimulated them with LPS to induce differentiation and Ab secretion, and measured anti-dsDNA Abs by ELISA. There was no difference in the level of anti-dsDNA Igκ between genotypes, and anti-dsDNA Igλ was not detected (Fig. 6A). This was not due to reduced response of Foxo3^−/− B cells to LPS (14). We also purified B cells with anti-B220 magnetic beads in case a CD43⁺ B cell subset was enriched for autoreactivity in Foxo3^−/− mice. Still, there was no increase, and in fact there was a slight decrease, in the secretion of anti-dsDNA Abs in response to LPS by Foxo3^−/− B220⁺ cells (Fig. 6B). We also did not observe enhanced production of anti-dsDNA Abs by marginal zone, transitional, or Igκ-depleted B cells (thus enriched for Igλ⁺ cells) from Foxo3^−/− mice (Fig. 6C–E). We considered the possibility that there was an increase in autoreactive cells that were not activated by LPS because of their anergic state, and we thus enumerated splenic T3 cells (B220⁺ AA4.1⁺ CD23⁺ IgM^lo), which are known to be enriched in anergic, autoreactive B cells (31, 32). However, there were normal numbers of these cells in young Foxo3^−/− mice (wild type 1.36 ± 0.72 × 10⁶, Foxo3^−/− 1.13 ± 0.75 × 10⁶, n = 8), (Supplemental Fig. 2E), and this population was in fact reduced, rather than increased, in aged Foxo3^−/− mice (wild type 8.32 ± 1.75 × 10⁵, Foxo3^−/− 1.93 ± 0.95 × 10⁵, n = 16). Taken together, these results suggest that the increased window of receptor editing afforded by reduced apoptosis of Foxo3^−/− bone marrow immature B cells allows for cells to redeem themselves away from autoreactivity.

To determine whether autoreactive Foxo3^−/− cells that should be deleted survive in the periphery in situations in which receptor editing does not efficiently eliminate autoreactivity, we used the anti-HEL Ig (MD4) × mHEL (KLK) transgenic system. The anti-HEL Ig transgene, which confers high-affinity reactivity with the Ag HEL, is not in the endogenous Ig locus and thus is not edited in the presence of self-antigen (17, 18, 33). Instead, exposure to mHEL results in deletion of HEL-reactive cells (17, 18, 33). We generated anti-HEL Ig × mHEL × Foxo3^−/− mice and asked whether anti-HEL B cells were present in the periphery in greater numbers than in anti-HEL Ig × mHEL mice. Surprisingly, Foxo3^−/− anti-HEL Ig B cells were deleted normally in the presence of mHEL (Fig. 7). This suggests that the high-affinity BCR/Ag interaction in the anti-HEL/mHEL system provides a strong enough signal to overcome a requirement for Foxo3 in BCR-induced apoptosis.

Reduced Foxo3 levels have been observed in B cells from polygenic mouse models of lupus (34). We thus asked whether alterations in Foxo3 expression were observed in B cells from SLE patients by analyzing previously published (26) gene expression profiling data. Although Foxo3 levels did not differ overall between healthy controls and SLE patients (data not shown), SLE patients with high anti-dsDNA levels (>60 IU) had lower levels of Foxo3 mRNA expression in B cells, but not T or myeloid cells, compared with patients with normal anti-dsDNA levels (<50 IU) (Fig. 8). Furthermore, Foxo3 levels in CD19⁺ B cells, but not CD4⁺ T cells or CD33⁺ myeloid cells, correlated inversely with disease activity as measured by SLEDAI (Fig. 8). We also analyzed another publicly available study in which whole blood gene expression analysis was performed on SLE patients (27). We divided these patients into groups based on anti-DNA titers (<50 and >60 IU) using the same criteria as in the first study (26) and found that whole blood Foxo3 mRNA levels were also reduced in patients with high anti-DNA titers (Fig. 8).

Elimination of autoreactive B cells during development is critical to limit the frequency of potentially pathogenic B cells in the periphery and prevent autoimmune disease. This central tolerance checkpoint involves receptor editing, followed by apoptosis of cells that remain autoreactive after several rounds of editing. In this study, we define the role of the transcription factor Foxo3 in these processes.

Upon Ag encounter or inhibition of the tonic BCR signal, immature B cells express Rag and undergo receptor editing. If this fails to produce a functional, innocuous receptor, the cells die by apoptosis. These processes are inhibited by PI3K (2, 6, 15, 35) and promoted by Foxo family members (11, 12). Upregulation of Rag in BCR-stimulated immature B cells is mediated primarily by Foxo1. Although overexpression of constitutively active Foxo3 promotes BCR-induced Rag expression (12), shRNA targeting of Foxo1 inhibits this process almost completely (11). Knockdown of Foxo3 has little effect on Rag expression (11), and Foxo3^−/− pre–B cells express normal Rag levels (14). Increased Igλ usage and RS recombination in Igλ⁺ B cells from Foxo3^−/− mice suggest that receptor editing is not only unimpaired in the absence of Foxo3 but is actually enhanced. Apoptosis is reduced in anti-IgM–stimulated Foxo3^−/− immature B cells, however, indicating a unique role for Foxo3 in this process. This likely explains our observation that immature B cell numbers are normal in Foxo3^−/− mice despite a reduced number of pre–B cells (14).

Together, these studies support the following model for Foxo regulation and function in immature B cells. Pre-existing Foxo protein is largely phosphorylated and sequestered in the cytoplasm, where it may be degraded (11, 15). Disruption of tonic BCR signaling or engagement of the BCR with self-antigen leads to the loss of Akt activity (2). Foxo proteins, no longer phosphorylated by Akt, enter the nucleus (11, 15). Foxo1 then activates Rag transcription, initiating L chain receptor editing (11). If receptor editing is successful, tonic signaling by the newly generated BCR restores Akt activity, drawing the Foxo proteins out of the nucleus and averting cell death. If editing is unsuccessful, however, and receptors remain nonfunctional or autoreactive, then Foxo3 promotes the death of these cells via apoptosis.

Although Foxo3 is known to control the expression of Bim in other cell types (7), and Bim is thought to contribute to the deletion of autoreactive immature B cells (28, 29), we did not observe a consistent change in Bim expression in Foxo3^−/− immature B cells. This suggests that Foxo3 promotes immature B cell apoptosis via Bim-independent mechanisms, an idea supported by two recent studies. Foxo3 binding sites in the Bim locus were shown not to be required for Bim to mediate apoptosis in response to cytokine withdrawal (36), and mice lacking Bim specifically in the B lineage did not have an increase in the bone marrow immature B cell pool (37). Foxo3 can also promote the expression of another proapoptotic BH3 family member PUMA (38), which could possibly contribute to immature B cell apoptosis. In addition, Foxo3 has been reported to inhibit NF-κB (39, 40). NF-κB activity is enhanced in cells undergoing receptor editing (41), and its loss results in a cell intrinsic reduction of immature B cell numbers (42). It is thus possible that the balance of Foxo3 and NF-κB activity plays a role in determining whether receptor editing or apoptosis occurs in immature B cells, with NF-κB favoring editing and Foxo3 favoring apoptosis.

Despite impaired BCR-induced apoptosis in Foxo3^−/− immature B cells, Foxo3^−/− mice do not develop autoantibodies. Delayed apoptosis in the absence of Foxo3 may extend the window for Foxo1 to successfully edit autoreactive receptors via its promotion of Rag expression. The increase in the frequency of Igλ-expressing B cells and the increase in RS recombination in these cells in Foxo3^−/− mice is consistent with this idea. This scenario suggests a system of checks and balances to maintain B cell tolerance; when Foxo3-mediated deletion is reduced, Foxo1-mediated editing is increased. Because the block in immature B cell apoptosis is not complete in Foxo3^−/− mice, some autoreactive B cells may still be eliminated by deletion in vivo. We show that this is indeed the case for high-affinity B cells in the anti-HEL Ig × mHEL model. In addition, the increased frequency of Igλ-expressing B cells drops off between the transitional and mature B cell stage in the spleens of Foxo3^−/− mice, suggesting that some Igλ⁺ cells that remain autoreactive and escape the bone marrow may be eliminated in the periphery. Finally, we did not observe enhanced secretion of autoantibodies by Foxo3^−/− B cells in vitro, nor did we observe an increase in the T3 population, known to be enriched in autoreactive, anergic B cells, in Foxo3^−/− mice. Taken together, these data suggest that loss of Foxo3 does not breach B cell tolerance despite impairing immature B cell apoptosis.

These studies were performed in whole body Foxo3^−/− mice. It is likely that the effects we observe on immature B cell survival and receptor editing are B cell intrinsic, as defects in BCR-induced apoptosis are observed in sort-purified immature B cells, and we believe that the increase in receptor editing results from increased survival. However, we cannot rule out a role for Foxo3 in other cell types in our results. Foxo3 is known to limit inflammatory cytokine production in myeloid cells (43–45). In T cells, it can promote pathogenic Th1 differentiation (46), limit the maintenance of CD4 and CD8 T cell memory (47–50), and contribute to the development of T regulatory cells (51). Although these parameters are unlikely to affect central B cell tolerance checkpoints, alterations in T or myeloid cell function in the absence of Foxo3 may alter the ability of autoreactive B cells to survive or become activated in the periphery. Future studies using mice with B cell–specific deletion of Foxo3 will shed light on this issue. It is also important to consider that the majority of these studies were performed with Foxo3^−/− mice on the FVB background. It is possible that other genetic backgrounds may reveal a more dramatic effect of Foxo3 deficiency on B cell tolerance. We have begun to address this issue by backcrossing the FVB.Foxo3^−/− mice to the C57BL/6 background. Preliminary analysis of these mice after four to five generations of backcrossing suggests that the increased frequency of Igλ-expressing B cells, particularly among immature and transitional cells, is retained (Supplemental Figs. 1C, 2D).

Foxo3 levels are reduced in B cells from murine lupus models (34). Similarly, analysis of publicly available gene expression profiling data (26, 27) indicates that Foxo3 levels are reduced in B cells from SLE patients with high levels of anti-dsDNA Abs. There are several models, not mutually exclusive, that reconcile our data with these observations. A B cell intrinsic defect upstream of Foxo3 that also affects Foxo1 levels could result in a more severe loss of central B cell tolerance in SLE patients than in Foxo3^−/− mice because of an impairment of both editing and apoptosis. Indeed, central tolerance defects have been observed in SLE patients (52, 53). Furthermore, SLE is a complex disease which results from a combination of increased autoantigen availability, loss of adaptive immune tolerance, and innate immune system hyperactivity (54). It is thus possible that preexisting low levels of Foxo3 in a subset of patients preferentially promote the survival of low-affinity autoreactive B cells that remain after extensive editing. These cells could then become activated to secrete autoantibodies in the context of additional SLE-associated defects, which would likely not have been present in the Foxo3^−/− mice. Finally, low levels of Foxo3 in B cells may be a consequence of increased peripheral B cell activation (55) or a more inflammatory environment in patients who have elevated anti-DNA Abs or are undergoing disease flares, rather than a cause of autoantibody production. Consistent with this idea, reduced expression of PTEN, an inhibitor of the PI3K pathway, has been observed in some SLE patients’ B cells in the periphery and correlates with disease activity (56). Ongoing studies to address the effect of Foxo3 deficiency in murine lupus models will address this issue.

We thank Arturo Menchaca, Lyndsay Joson, Hansaa Abbasi, Elizabeth Curry, Julia McLouth, Ian Matthews, and Angela Mobley for excellent technical assistance.

The authors have no financial conflicts of interest.