BCR editing in the bone marrow contributes to B cell tolerance by orchestrating secondary Ig rearrangements in self-reactive B cells. We have recently shown that loss of the BCR or a pharmacologic blockade of BCR proximal signaling pathways results in a global “back-differentiation” response in which immature B cells down-regulate genes important for the mature B cell program and up-regulate genes characteristic of earlier stages of B cell development. These observations led us to test the hypothesis that self-Ag-induced down-regulation of the BCR, and not self-Ag-induced positive signals, lead to Rag induction and hence receptor editing. Supporting this hypothesis, we found that immature B cells from xid (x-linked immunodeficiency) mice induce re-expression of a Rag2-GFP bacterial artificial chromosome reporter as well as wild-type immature B cells following Ag incubation. Incubation of immature B cells with self-Ag leads to a striking reversal in differentiation to the pro-/pre-B stage of development, consistent with the idea that back-differentiation results in the reinduction of genes required for L chain rearrangement and receptor editing. Importantly, Rag induction, the back-differentiation response to Ag, and editing in immature and pre-B cells are inhibited by a combination of phorbol ester and calcium ionophore, agents that bypass proximal signaling pathways and mimic BCR signaling. Thus, mimicking positive BCR signals actually inhibits receptor editing. These findings support a model whereby Ag-induced receptor editing is inhibited by BCR basal signaling on developing B cells; BCR down-regulation removes this basal signal, thereby initiating receptor editing.
Receptor editing was first described by Nemazee and Weigert (1, 2) in conventional transgenic systems where B cells altered the specificity of Ig receptors away from autoreactivity in response to self-Ag. The design of the κ L chain locus, with V and J gene segments but no diversity D regions (such as those found at the H chain locus), allows for the editing of L chains through a series of nested deletions or inversions. Editing of H chain alleles by VH gene replacement has also been described (3, 4) but is probably relatively uncommon compared with L chain editing. A series of anti-DNA and anti-MHC class I Ig knock-in models (4, 5, 6, 7, 8) provided further evidence that secondary recombination at L chain loci was important for maintaining tolerance to self-Ags and suggested that the editing was highly efficient (7, 9). For example, as many as half or more of normal B cells in bone marrow (BM)5 appear to undergo some degree of editing (10, 11, 12). Editing has also been suggested to occur in the periphery (13, 14), although there has been debate about whether this represents reinduction of Rag1/2 in mature B cells or the recruitment of immature cells to the spleen (13, 14, 15, 16, 17).
Immature cells exposed to self-Ag are induced to re-express Rag1 and Rag2 (or continue Rag expression from the pre-B stage) (18, 19). Studies with mice carrying a Rag2-GFP bacterial artificial chromosome reporter transgene indicated that Rag expression is still apparent in immature B cells in the BM (16, 17), and there is evidence that Rag2-GFPnegIgMhi (where neg is negative and hi is high) immature B cells remain competent to reinitiate Rag1/2 gene transcription following receptor cross-linking by Ag (20). The kinetics of L chain replacement was examined in an experiment where the mouse Cκ constant region was replaced with human Cκ (hCκ) (12). Heterozygote hCκ mice with a knock-in anti-HEL L chain and lacking a prerearranged Ig H chain demonstrated the appearance of hCκ+ cells with an ∼2-h delay compared with mCκ+ cells. How much of this L chain replacement was due to self-reactivity vs poor pairing of the anti-HEL L chain with endogenous H chains is not clear. Similar kinetic data were observed in a knockin λ L chain model (21).
Several of the current models for receptor editing emphasize the important role for signals transduced through the cross-linked BCR in turning on or keeping on Rag proteins and the recombination machinery (22, 23, 24). However, there are several studies consistent with the idea that positive signaling through the BCR may be required to turn off Rag proteins and maintain allelic exclusion. For instance, mice carrying knockin receptors for the 3-83 H chain and L chain show very poor allelic exclusion with a single dose of H chain and L chain that is only corrected by breeding to homozygosity (25, 26). One interpretation of these data is that a single dose of the receptor is unable to provide a sufficiently strong basal signal. B cells deficient in CD19 show strong spontaneous up-regulation of Rag, loss of allelic exclusion at L chain loci, and impaired positive selection (27), suggesting that the lack of positive signaling through the BCR is associated with the induction of Rag and failure to progress in development. Similarly, immature B cells lacking the protein tyrosine kinase syk show loss of L chain allelic exclusion (28). Together, these data support the hypothesis that developing B cells require threshold signals from Ig receptors expressed on the cell surface to block further Ig gene rearrangements.
We recently reported that deletion of the BCR from immature B cells generated in IL-7 BM cultures following Cre-mediated excision of a floxed H chain led to the induction of Rag expression and a new L chain rearrangements (29). Incubation of immature B cells with a tyrosine kinase inhibitor or a PI3K inhibitor led to a similar phenotype. Surprisingly, both loss of the BCR and blockade of proximal signaling pathways resulted in a global “back-differentiation” response where cells turned off many genes important for the mature B cell program (e.g., CD22 and class II MHC) and turned on genes characteristic of earlier stages in B cell development (e.g., IL-7R and TdT).
These observations led us to test the hypothesis that a threshold for basal signaling may also contribute to the induction of Rag genes and receptor editing in immature B cells in the HEL model system following self-Ag exposure. Our data support a model whereby basal signaling from the BCR expressed on the cell surface is required to suppress further recombination activity. In this model, self-Ag induces editing by modulating BCR surface expression and hence down-regulating basal BCR signaling. In the absence of a sufficient basal BCR signal, Rag proteins are induced and allow for further rearrangements at Ig loci until a nonautoreactive receptor is produced.
Materials and Methods
Rag2-GFP bacterial artificial chromosome transgenic (Tg) (16) and human Cκ knockin mice were provided by Dr. M. Nussenzweig (Rockefeller University, New York, NY). Rag2-GFP animals were bred with MD4 anti-HEL Ig (HEL-Ig) Tg mice (30) to generate double Tg animals as described (29). MD2/L chainKI/hCk/KLK4 membrane-bound HEL (mHEL) Tg mice were generated by breeding mice carrying the hCκ L chain knockin allele to the animals previously described (31). Floxed B1-8 H chain knockin (B1-8f) (32) and 3-83κ L chain knockin (33) mice were bred to Rag2-GFP Tg animals to generate B1-8f/3-83κ/Rag2-GFP animals. xid/CbaN mice (34) were obtained from The Jackson Laboratory. Xid females were bred with HEL-Ig/Rag2-GFP males to generate HEL-Ig/Rag2-GFP/xid−/y male mice, and xid males were bred with HEL-Ig/Rag2-GFP females to generate HEL-Ig/Rag2-GFP (wild-type (WT)-CbaN) male mice. For some experiments, control mice for xid experiments were HEL-Ig/Rag2-GFP mice of non-CBA background. Lyn-deficient mice on the B6 background were obtained from The Jackson Laboratory and bred with HEL-Ig/Rag2-GFP mice for two generations to obtain homozygous lynnull/Rag2-GFP animals. The HEL-Ig/Rag2-GFP and B1-8f/3-83 mice were crossed onto C57BL/6J mice for at least six generations. All mice were maintained in specific pathogen-free conditions and were generally between 6 and 12 wk of age at the time of the experiments. All experiments were approved by the University of Minnesota Institutional Animal Care and Use Committee (Minneapolis, MN).
Single cell suspensions of BM cells from the various mice were prepared (35), and placed into IL-7 BM culture (19, 36, 37) as previously described (29). For secondary Ag cultures, cells were recultured in complete medium with HEL (Sigma-Aldrich), herbimycin A (Calbiochem), Ly290042 (Calbiochem), 4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrazolo[3,4-d]pyrimidine (PP2; Calbiochem), wortmannin (Calbiochem), PMA (Sigma- Aldrich), and/or ionomycin (Calbiochem) for the times indicated. The HELΔRD and HELΔRDGN mutants and duck egg lysozyme (DEL) have been described previously (35, 38). TAT-Cre mediated deletion of the floxed B1-8 H chain was as described (29).
Flow cytometry and cell sorting
BM cells harvested at the end of IL-7 or secondary cultures were stained in FACS buffer (PBS, 2.5% FBS, and 0.2% sodium azide) with either FITC-, PE-, CyChrome-, allophycocyanin-, or biotin-conjugated mAbs to B220, IgM, IgMa, IgMb, IgDa, CD19, CD22, CD23, CD24, CD43, CD69, CD86, integrin α4, integrin β7, PIRA/B, and IA/IE (BD Pharmingen). Staining with biotinylated Abs was revealed by SAv-PE or SAv-allophycocyanin (BD Pharmingen). In some staining conditions, annexin-V-PE and/or 7-aminoactinomycin D (BD Pharmingen and Calbiochem) was used to exclude dead cells. For Ag culture cell sorting, cells were stained with mAbs to IgMa and B220 in staining buffer (1× PBS with 10% FBS) and sorted by a FACSVantage apparatus (BD Biosciences). For purification of splenic B cells, cells were enriched using MACS separation with the B cell isolation kit and LS columns as recommended by the manufacturer (Miltenyi Biotec). Flow cytometry analyses were performed using CellQuest (BD Biosciences) and FlowJo (Tree Star) software.
Genomic DNA was isolated from fresh spleen or IL-7 culture-derived cells as described (39). Endogenous V-Jκ1 rearrangements were detected using a real-time PCR assay modified from the method previously described (39, 40). The V-Jκ1 PCR was performed with an upstream Vκ degenerate primer (5′-GGCTGCAG(G/C)TTCAGTGGCAGTGG(A/G)TC(A/T)-3′) and a primer that annealed just downstream of Jκ1 (5′-GCCACAGACATAGACAACGGAAGAA-3′) (18), with a TaqMan probe covering the Jκ constant region (5′-TTGCCTTGGAGAGTGGCCAGAATC). All reactions were run using the following cycling conditions: 2 min at 50°C, 10 min at 95°C, then 15 s at 95°C, and 60°C for 55 cycles. Duplicate samples were analyzed and normalized using an 18S ribosomal genomic TaqMan probe (Applied Biosystems).
Microarray gene chip analysis
Total RNA was extracted from the in vitro cultured cells, and biotinylated cRNA probes were synthesized and hybridized to U74Av2 probe arrays following standard Affymetrix protocols (Affymetrix Expression Analysis Technical Manual P/N 700218 revision 2) as described (29). Three or four independent sorts were performed for each sample. All statistical analyses were performed using MS Excel (Microsoft Office X) as previously described (29). A given transcript was considered to be undetected if two of three of the arrays within the group had detection p values of >0.05, and transcripts were included for further analysis only if they were present in both control and HEL-treated groups. To identify genes that were differentially expressed between control (Ctrl) (i.e., BM 48h Ctrl, GFP Ctrl, and Cre Ctrl) and experimental populations (labeled with asterisks i.e., HEL 48h, GFP HA, and CreMlo, where HA is herbimycin A and lo is low), we used the Student’s t test analysis with the assumptions that the sample groups followed a two-tailed distribution and had unequal variance. For visualization of the arrays, each expression value was divided by the mean of the expression values for the relevant IgM+ samples. These ratios were transformed into log2 space and subjected to centered average linkage clustering using Cluster and visualized by TreeView software (41). The microarray data have been deposited in the Gene Expression Omnibus (GEO) repository (www.ncbi.nlm.nih.gov/geo/) with accession number GSE2227. The populations included are B6 IgMneg (B220+ IgMneg pro- and pre-B cells from polyclonal C57BL/6 IL-7 cultures; GEO name B6 Mneg), fraction D (Fr. D) (B220+ CD43neg IgMneg pre-B cells from BALB/c BM; GEO name FxD), fraction E (Fr. E) (B220+ IgM+ IgDneg newly formed B cells from BALB/c BM; GEO name FxE), CreMlo (B220+ IgMa lo, B1-8f H chain, 3-83 L chain Mx-Cre Tg BM cultured 48 h with IFN to delete the H chain; GEO name Cre Mlo), Cre Ctrl (B220+ IgMhi, B1-8f H chain, 3-83 L chain KI cultured 48 h with IFN; GEO name Ctrl Mhi), HEL 48h (B220+ IgMa loGFP+, HEL-Ig Tg IL-7 cultured BM incubated 48 h in 1 μg/ml HEL; GEO name HEL 48h), BM 48h control (B220+ IgMa hiGFPneg, HEL-Ig Tg IL-7 cultured BM incubated 48 h in medium; GEO name BM 48h Ctrl), HEL IgMlo (B220+ IgMa loIgDa neg immature B cells from HEL-Ig Tg IL-7 BM cultures; GEO name HEL Mlo), HEL IgMhi (B220+ IgMa hiIgDa neg immature B cells from HEL-Ig Tg IL-7 BM cultures; GEO name HEL Mhi), GFP HA (IgM+ GFP+, Rag2-GFP HEL-Ig IL-7 cultured BM incubated 24 h in 400 ng/ml herbimycin A; GEO name GFPpos), and GFP control (IgM+ GFPneg, Rag2-GFP HEL-Ig IL-7 cultured BM incubated 24 h in medium; GEO name GFPneg).
Delayed kinetics of Rag2-GFP expression following Ag stimulation
Mice carrying a BCR transgene specific for HEL-Ig were crossed with Rag2-GFP Tg mice to generate HEL-Ig/Rag2-GFP double Tg animals. The Rag2-GFP transgene allows the expression of Rag2 to be quantified at the single-cell level by flow cytometry, and Rag2-GFP expression strongly correlates with the induction of new endogenous Ig L chain rearrangements in HEL-Ig B cells (20, 29). To test the hypothesis that Ag-induced receptor editing could be a consequence of the down-regulation of surface Ig M (sIgM) and the subsequent loss of basal signaling rather than an “activation” response due to acute cross-linking of the BCR, we generated xid/HEL-Ig/Rag2-GFP mice. xid, x-linked immunodeficiency, is a mutation in the pleckstrin homology domain of Bruton’s tyrosine kinase (btk) that impairs the recruitment of Btk to the plasma membrane and severely reduces signaling through the BCR. However, receptor internalization is not impaired in xid mice (42). If an “activation” response through positive signaling downstream of the BCR is necessary for receptor editing, then that process should be impaired in xid/HEL-Ig/Rag2-GFP mice. In contrast, if removal of a basal signal via down-regulation of sIgM is essential for receptor editing, then this process should be unperturbed or possibly enhanced in these mice.
BM from WT and xid HEL-Ig/Rag2-GFP Tg mice was cultured in IL-7 for 5 days to generate large numbers (∼60–80 × 106) of highly purified populations of IgM+ immature B cells (≥98% of WT B220+ cells are IgMa+IgDa neg compared with ≥90% for xid), as previously described (35). Cells were washed, stimulated with 1 μg/ml HEL (a concentration that saturates the HEL-Ig BCR (43), and harvested at 3, 6, or 12 h for flow cytometry.
Immature B cells treated with Ag undergo an initial “activation” response as measured by global gene expression (L.E.T., unpublished data) and a rapid up-regulation of the cell surface activation proteins CD69 and CD86, together with a down-regulation of sIgM levels over the first 12 h of culture (Fig. 1, A and B). Although immature B cells do not induce CD69 and CD86 to the extent seen in mature B cells (44), they still induce robust responses that can be used to indicate the activation status of immature B cells. The decrease in sIgM levels in response to Ag was rapid in both WT and xid B cells (Fig. 1,E). The expression of CD69 and CD86 in WT cells was ∼70% of the 12-h maximum after a 1.5-h pulse of Ag (not shown) and 82% of the maximum after a 3-h Ag pulse. Compared with WT cells, xid B cells showed impaired induction of the activation Ags CD69 and CD86 at all of the time points tested (Fig. 1, A, B, and D). Despite the strong “activation” signal provided by Ag, there were only background numbers of Rag2-GFP+ cells through 6 h of culture in both WT and xid cells, with small responses observed in both cultures at 12 h (Fig. 1 C). Rag expression can be rapidly induced (within 3 h) by acute pharmacologic inhibitors of BCR-dependent PI3K activation (e.g., wortmannin and Ly290042; data not shown). Therefore, the failure to see Rag2-GFP expression is not due to the fact that Rag2 induction is an inherently later response than CD69 or CD86 expression. Rather, we favor the hypothesis that prolonged down-regulation of BCR basal signaling is required to induce Rag gene expression.
To determine whether a prolonged absence of the BCR signal was necessary for Rag induction, a series of “pulse-chase” experiments was performed. WT and xid B cells were incubated with Ag for periods of time before washing and returning to culture in the absence of Ag, followed by analysis at 24 and 48 h. Despite impaired up-regulation of the activation Ags CD69 and CD86 due to defective BCR signaling, xid B cells showed Rag2-GFP responses that were comparable to those observed in control B cells (Fig. 2, A and B). We observed that the induction of Rag2-GFP lagged significantly behind the “activation” response, as measured by CD69 and CD86 (Figs. 1 and 2). Cells that received a short pulse of HEL (e.g., 3 h) showed high-level induction of the activation Ag CD69 but had only modest Rag2-GFP up-regulation. A prolonged incubation of immature B cells with Ag is necessary for optimal Rag2-GFP up-regulation in this system, suggesting that basal BCR signals must drop below a specific threshold to induce Rag expression, but not the expression of CD69 or CD86. In WT immature B cells, treatment with the PI3K inhibitor Ly290042 for 3 h induces Rag expression (data not shown), suggesting that when the BCR signal is completely ablated, Rag induction occurs quite rapidly. We conclude that xid B cells have the ability to undergo vigorous receptor editing responses following incubation with Ag despite impaired signaling downstream of the BCR. The discordance between Ag-induced activation signals and the expression of Rag2-GFP in xid B cells prompted us to further analyze the signal downstream of the BCR and its role in receptor editing.
Elevated Rag2-GFP expression in lynnull immature B cells
We next examined immature HEL-Ig/Rag2-GFP B cells deficient for the Src family kinase lyn, a key mediator of proximal BCR signaling. lyn appears to be involved in positive regulation of BCR signaling in immature B cells by phosphorylating ITAMs of Igα and Igβ, and lyn-deficient immature B cells show impaired allelic exclusion and developmental progression (28, 45). We predicted that if positive signaling is necessary for Rag expression, then it should be impaired in lynnull/HEL-Ig/Rag2-GFP mice. In contrast, if removal of a basal signal is essential, then Rag expression and receptor editing should be unperturbed or possibly enhanced in immature B cells from these mice.
We used IL-7 BM cultures to generate immature B cells from lynnull mice. At the end of 5 days in culture there were elevated numbers of Rag2-GFP+ cells (on average, 6-fold) in lynnull/HEL-Ig/Rag2-GFP immature B cells compared with WT B cells (Fig. 2,C). Lyn+/neg heterozygote B cells also had increased numbers of Rag2-GFP positive cells compared with controls. A potential caveat to these studies is that lyn has also been shown to have a negative regulatory role in BCR signaling in mature B cells; whether this is true for immature B cells in the BM remains unclear (46, 47). Given the potential dual role of Lyn in regulating BCR signaling and the redundant role of Fyn and Blk in mediating positive BCR-dependent signals, we used the Src family kinase inhibitor PP2 to abrogate Src family kinase signaling downstream of the BCR. Consistent with the findings in lynnull cells, treatment of HEL-Ig/Rag2-GFP immature B cells with PP2 resulted in elevated GFP expression as compared with medium alone controls (Fig. 2 D). PP2 treatment resulted in higher levels of Rag2-GFP induction than were observed in lynnull B cells, suggesting that other Src family members may partially compensate in the absence of lyn. Together, these findings suggest that Src family kinases contribute to the basal signal in immature B cells that suppresses Rag2-GFP.
Rag2-GFP expression correlates with IgM receptor down-regulation
Tolerance has been suggested to result from the endocytosis of surface Ig after incubation with self-Ag (48, 49, 50). To assess the effects of Ag affinity and concentration on the kinetics of Rag2-GFP induction, we used two site-directed mutants of HEL with reduced affinity for the HEL receptor (HEL, ∼10neg9 M affinity; HELΔRD,∼10neg8 M affinity; and HELΔRDGN,∼10neg7 M) as well as DEL, which binds with even lower affinity (∼10neg6 M) (35, 38). After a 5-day IL-7 BM culture, HEL-Ig/Rag2-GFP immature B cells were incubated with HEL or the various analogues at 2.5, 0.5, or 0.1 μg/ml and then analyzed by flow cytometry for CD69 expression at 2 h, IgM expression at 24 h, and Rag2-GFP responses at 48 h (Fig. 3,A). All doses of HEL elicited CD69 activation, prolonged down-regulation of IgMa, and maximal Rag2-GFP induction. The HELΔRD mutant induced nearly maximal up-regulation of CD69 at all concentrations, whereas only the highest dose led to a prolonged down-regulation of IgMa and a near maximal induction of Rag2-GFP. The highest dose of HELΔRDGN induced CD69 to relatively high levels but resulted in only a minimal Rag2-GFP response. There was no significant Rag2-GFP induction at any of the doses of DEL. These data further demonstrate the discordance between the activation of immature B cells by Ag, as measured by CD69, and the induction of Rag2-GFP in this system. Importantly, prolonged down-regulation of sIgM was associated with the induction of Rag2-GFP regardless of the degree of cell activation. As illustrated in Fig. 3 B, it appears that sIgM levels must drop below a certain threshold for efficient Rag2-GFP expression. Modest reduction in sIgM levels resulted in minimal Rag2-GFP induction. In contrast, once sIgM levels dropped below a threshold (in these experiments an IgM MFI of ∼1000) Rag2-GFP was efficiently induced. Thus, we conclude that prolonged sIgM down-regulation below a specific threshold is required to induce receptor editing.
Immature B cells undergo a back-differentiation response following Ag exposure
Immature B cells that lose surface BCR expression due to Cre-mediated deletion of Ig H chain show a loss of L chain allelic exclusion characterized by the induction of Rag genes and new L chain gene rearrangements (29). Furthermore, these cells undergo a striking reversal in development with new onset expression of pro-/pre-B cell genes and extinction of many genes of the mature B cell program. A similar picture emerged when immature B cells were treated with pharmacologic agents that block proximal signaling pathways downstream of the BCR (29). These data suggested that basal BCR signaling provides tonic signals that suppress Rag expression and hold the immature B cell genetic program in place.
We were interested in determining whether Ag-stimulated immature B cells, which undergo a prolonged down-regulation of cell surface BCR expression, might show a similar back-differentiation response with the expression of genes indicative of an earlier stage in B cell development. Microarray analyses were used to compare gene expression profiles of HEL-Ig/Rag2-GFP double Tg immature B cells incubated with either medium alone or with 1 μg/ml HEL for 2 days. Cells were stained with Abs against B220 and IgMa, and untreated (B220+, IgMa+, Rag2-GFPneg; Ctrl 48h) and HEL treated (B220+, IgMa lo, Rag2-GFP+; HEL 48h) cells were sorted by flow cytometry. Total RNA was isolated and converted to biotinylated cRNA probes that were then hybridized to Affymetrix murine U74Av2 chips. Data were analyzed using Affymetrix Microarray Suite 5.0 software (see Materials and Methods).
The analysis identified 212 transcripts that were differentially expressed between medium-alone control (Ctrl 48h) and HEL-treated cells (HEL 48h) at 48h (see Materials and Methodsfor details regarding filtering criteria). We compared the patterns of gene expression in Ag-treated cells with the data from additional arrays generated from several sorted B cell populations: pro- and pre-B cells generated from polyclonal C57BL/6 IL-7 BM cultures (B6 IgMneg); immature B cells generated from HEL-Ig Tg IL-7 BM cultures (HEL IgMhi); pre-B cells, Fr. D as described by Hardy et al. (51) sorted from BM of BALB/c mice (Fr. D); and newly formed B cells, Fr. E as described by Hardy et al. (51) also sorted from the BM of BALB/c mice (29). Expression values for each gene were divided by the means of IgMhi cell populations (HEL IgMhi, Fr. E, Ctrl 48h). These ratios were then log2 transformed, and unsupervised clustering was performed using Cluster, with the data visualized using TreeView (41) (Fig. 4). Further detailed description of the sorting schemes and array analysis is provided in Materials and Methods.
Previous reports have shown that Rag mRNA and protein are induced in immature B cells after BCR cross-linking (23). Similarly, cells treated with HEL for 48 h showed strong induction of transcripts for Rag1 and Rag2 (Table I). In addition, a number of other interesting transcripts were up-regulated including the nuclear proteins/transcription factors Jun, Myb, Ku70, Lef-1, Ezh2, and Ets2, the signaling molecules Fyn and IκBα, and the cell surface proteins IL-7R and CXCR4. Down-regulated transcripts included the nuclear protein/transcription factor XBP-1, the signaling molecules SHP-1, Hck, Slap, PKCγ, Fes, and Igα, and many cell surface molecules such as CD20, CD22, CD32, PIR-A, PIR-B, Ia invariant chain, CXCR5, and integrin α4. Strikingly, the array profile of immature B cells treated with HEL (HEL 48h) clustered together with data from pro- (B6 IgMneg) and pre-B cells (Fr. D) rather than with immature B cell populations (HEL IgMhi, Fr. E) (Fig. 4). Many of the genes up-regulated in cultures containing self-Ag are characteristically expressed in pro-B and pre-B cells, whereas many of the down-regulated transcripts are expressed as part of the mature B cell gene program. We conclude that Ag treatment of immature B cells leads to a “back-differentiation” of the cells, similar to that observed when immature B cells lose basal signaling due to loss of BCR.
|GenBank Accession No. .||Transcript .||Mean Expression Valuea .||.||Fold Differenceb .||Nominal p Valuec .|
|.||.||Ctrl 48h .||HEL 48h .||HEL 48h/Ctrl 48h .||.|
|Cell surface molecules|
|Intracellular signaling effectors|
|Nuclear molecules/transcription regulators|
|GenBank Accession No. .||Transcript .||Mean Expression Valuea .||.||Fold Differenceb .||Nominal p Valuec .|
|.||.||Ctrl 48h .||HEL 48h .||HEL 48h/Ctrl 48h .||.|
|Cell surface molecules|
|Intracellular signaling effectors|
|Nuclear molecules/transcription regulators|
Mean expression values of selected transcripts from the 212 total transcripts identified as significantly differentially expressed between immature B cells untreated (Ctrl-48 h) or treated with HEL (HEL-48 h) for 48 h. Data represent the mean of three individual sorts for each cell population.
Fold change of mean expression values from HEL-treated and untreated immature B cells. Positive values denote upregulation and negative values denote downregulation.
Student’s t test calculated assuming two-tailed distribution with unequal variance.
Surface protein expression of immature B cells incubated with self-Ag
We next sought to validate the array data by investigating whether the changes in gene expression were reflected at the protein level. At time 0 and after 2 or 3 days of culture with medium alone or with HEL, cells were stained with various mAbs to cell surface markers and analyzed by flow cytometry. Freshly isolated non-Tg spleen and BM cells from C57BL/6 mice were analyzed in parallel. The cell surface phenotype of B220+ IgM+ cells immediately after expansion in IL-7 confirmed that these cells expressed many of the surface markers characteristic of in vivo immature B cells (Fig. 5). After incubation with HEL, immature B cells showed decreased surface levels of B220, CD22, PIR-A/B, and integrin α4 and increased levels of CD43 and class II IA-IE (Fig. 4 B). CD23 and integrin β7 showed no changes in surface protein levels. In general, these protein data mirrored, and thus validated, the results obtained in the microarray analysis. Note that the populations of cells were moving together as a “unit,” with all cells demonstrating the changes in surface protein levels. We have previously shown that there is no significant proliferation of immature B cells in this system after Ag treatment and just an average 25–30% total loss of cells over the course of a 2-day Ag culture (20, 35). Together, these data rule out the possibility that the gene expression and protein changes observed are due to selective survival and/or expansion of the small numbers (<3%) of non-IgM-bearing B cells typically present at the initiation of the cultures.
Ag-stimulated immature B cells show global gene expression changes consistent with a reversal in development
To define the developmental stage of Ag-incubated immature cells, we performed unsupervised hierarchical clustering of the gene expression data with data from previously described B cell populations (29). These populations included cells derived from mice carrying a floxed B1-8 knockin H chain (B1-8f), a 3-83 knockin L chain, either with (CreMlo) or without (Cre Ctrl) the IFN inducible Mx-Cre after 48 h of IFN treatment (20), as well as immature HEL-Ig/Rag2-GFP B cells cultured for 24 h in the absence (GFP Ctrl) or presence of 400 ng/ml herbimycin A (GFP HA) and sorted based on B220, IgMa, and Rag2-GFP expression. We also included in this analysis the following additional sorted cell populations: early IgMlo immature B cells generated from HEL-Ig Tg IL-7 BM cultures (HEL IgMlo) and IgM+ immature B cells from BM of BALB/c mice (Fr. E). Expression values for each transcript were divided by the mean expression level for all IgMhi cell populations (HEL IgMhi, BM 48h Ctrl, GFP Ctrl, Fr. E, and Cre Ctrl) to provide a baseline from which to compare gene expression levels. These fold difference ratios were log2 transformed, with clustering and visualization as described above.
As shown in Fig. 6 A, Ag-incubated cells (HEL 48h) clustered with IL-7 culture generated pro- and pre-B cells (B6 IgMneg) and normal pre-B cells (Fr. D), as well as with IgMneg H chain-deleted cells (Cre Mlo), and herbimycin-treated HEL-Ig B cells (GFP HA). The clustering also showed that the HEL-treated cells were more similar to pre-B (Fr. D) and pro-/pre-B (B6 IgMneg) cells than to early immature (HEL IgMlo) cells, indicating that the cells were clearly back-differentiating to an earlier B cell stage.
To further address the correlation between HEL-treated immature B cells, pre-/pro-B cells and early immature B cells, a separate clustering analysis was performed using 155 transcripts that best discriminated normal pro-/pre-B cells (B6 IgMneg) from early immature B cells (HEL IgMlo). Analysis of this gene list also showed that HEL-treated immature B cells BM (HEL 48h) clustered together with pre-B (Fr. D) and pro-/pre-B (B6 IgMneg) cells rather than with early immature (HEL IgMlo) cells (Fig. 6 B). We conclude that immature B cells incubated with self-Ag for 48 h show a striking global back-differentiation to an early stage of B cell development characterized by the down-regulation of many mature B cell genes and the up-regulation of genes normally expressed in pro-B and pre-B cells.
Suppression of Rag responses and L chain rearrangements by phorbol ester and ionophore
If the loss of basal signaling from the BCR in immature B cells drives receptor editing, we hypothesized that the addition of agents that mimic BCR signals should suppress Rag expression and block the ability of Ag to induce Rag and new L chain endogenous rearrangements. We tested this idea using PMA, a phorbol ester that activates protein kinase C molecules by mimicking endogenous diacylglycerol (reviewed in Refs.52 and 53), and ionomycin, a calcium ionophore that activates downstream calcium-dependent signaling pathways (54). Combinations of PMA and ionomycin have long been used to initiate activation signals in lymphocytes (55, 56).
To determine the influence of PMA and ionomycin on baseline Rag2-GFP expression in immature B cells, we performed a series of dose response experiments. As single agents and in combination there was no evidence for strong induction of Rag2-GFP in these cells by PMA and ionomycin (Fig. 7 A) across broad concentration ranges. Cell counts showed that the combination of PMA/ionomycin was not toxic to the cells (data not shown). Quantitation of the absolute number of Rag2-GFP+ cells showed no increases with any combination of PMA and ionomycin (data not shown). In contrast, treatment of immature B cells with HEL consistently resulted in a >6-fold induction in the absolute number of GFP+ cells compared with medium alone. We conclude that treatment of immature B cells with PMA and ionomycin does not induce Rag2-GFP expression. This observation is consistent with our hypothesis that it is the absence of a basal signal that initiates Rag expression and receptor editing, not a positive signal through the BCR that PMA and ionomycin mimics.
The addition of PMA and ionomycin to HEL-treated immature B cells at the initiation of the cultures blocked the up-regulation of Rag2-GFP levels at 24 and 48 h (Fig. 7,B). Similarly, the induction of Rag2-GFP following incubation of cells with the tyrosine kinase inhibitor herbimycin A was also suppressed (Fig. 7,B). As measured by flow cytometry, PMA and ionomycin blocked the back-differentiation of cells induced by either Ag treatment or incubation with the tyrosine kinase inhibitor herbimycin A (as well as with PI3K inhibitors; data not shown). IL-7R up-regulation was blocked (not shown) and the levels of PIR-A were, in fact, higher on cells treated with Ag and PMA/ionomycin, consistent with continued progression through development (Fig. 7,B). CD22 levels were similarly up-regulated by PMA/ionomycin (not shown). Surface expression of the activation marker CD86 was up-regulated at 12 h by HEL incubation (Fig. 7 B) and was strongly induced by the combination of PMA and ionomycin, confirming that the cells were receiving strong “activation” signals.
We next assayed cells cultured with HEL or HEL plus PMA/ionomycin for new endogenous L chain rearrangements by using a quantitative PCR assay and found that PMA/ionomycin completely prevented the induction of rearrangements observed after HEL incubation for 48 h (Fig. 7,C). To confirm these results using an additional model system, immature B cells carrying a floxed H chain allele and Rag2-GFP (B1-8f/3-83κ/Rag2-GFP) were cultured with TAT-Cre (57) for 2 days with or without PMA and ionomycin (Fig. 7 D). The up-regulation of Rag2-GFP in cells undergoing Cre-mediated deletion of the BCR was also blocked by the combination of PMA and ionomycin. We conclude that PMA and ionomycin, presumably by activating protein kinase C and calcium-dependent pathways, block Rag induction and the onset of new endogenous L chain rearrangements in these systems.
PMA and ionomycin suppress L chain receptor editing in pre-B cells
Finally, we tested the influence of PMA and ionomycin on pre-BCR editing to self-Ag in a newly developed in vivo model system (31). In this model, mice that carry an anti-HEL H chain transgene, an anti-HEL L chain knockin allele, and a mHEL transgene have an expanded pre-B cell compartment and show vigorous L chain editing responses that lead to the accumulation of high numbers of non-HEL-binding (edited) B cells in the periphery (31). For the experiments described here, a human Cκ knockin allele (12) was introduced into these mice (generating H chainTg/+/L chainKI/hCk/mHEL animals) so that we could monitor the frequency of editing to the alternate L chain allele by staining cells for hCκ.
We sorted editing pre-B cells from the BM of these mice (B220+ CD43negIgMneg) to high purity (>98%) (Fig. 8, A and B), and then cultured the cells in vitro with the survival factor BAFF alone or with BAFF plus PMA/ionomycin for 18 h. Over this short culture period there were no significant differences in overall cell numbers as determined by live cell gating (Fig. 8,C) and trypan blue staining. Importantly, pre-B cells incubated with PMA/ionomycin generated increased numbers of immature B cells that bound HEL in a sandwich assay and ∼5-fold reduced numbers of hCκ-edited cells (Fig. 8 C). Thus, PMA/ionomycin suppresses ongoing L chain receptor editing responses in pre-B cells in this system.
Despite many elegant studies demonstrating that receptor editing is an efficient mechanism to modify the Ag specificity of BCRs expressed on developing B cells, the mechanisms that guide and control the process remain poorly understood. The data reported here suggest that an important signal for receptor editing in response to self-Ag is the absence or reduction of basal signaling from surface BCRs on developing B cells. In light of these findings, current models that describe the regulation of receptor editing may need to be revised.
The induction of Rag1 and Rag2 mRNA and protein in response to BCR cross-linking in immature B cells is well established (22), and previous reports have emphasized the role of positive signaling through the BCR following exposure to self-Ag as a key stimulus for editing. One aspect of the “positive signaling” model that remains unexplained, however, is the slow kinetics of Rag induction observed when IL-7 BM culture-derived immature B cells are incubated with specific Ags (18). As shown in Fig. 1, despite a rapid “activation” response following BCR stimulation as measured by CD69 and CD86 induction, Rag2-GFP positive cells were present at low levels before 24 h and then peaked at 48 h. Cells stimulated with Ag for short pulses (∼e.g., 3 h) showed significant activation responses; however, they did not undergo strong induction of Rag2-GFP. Similarly, in time course microarray studies we have observed that Rag induction does not occur in parallel with the generalized activation response that follows Ag incubation (peaking at ∼6 h) but instead parallels the back-differentiation response (that begins around 24 h), culminating in an induction of the pro-/pre-B genetic program and a relative extinction of the mature B cell program. Together with our observations that B cells lacking the proximal signaling molecule btk show normal Rag2-GFP induction following Ag incubation despite impaired activation responses and that lynnull immature B cells show elevated basal Rag2-GFP expression, these data support the hypothesis that a major driving factor for Rag induction by Ag in this model system is a reduction in basal or tonic BCR signaling due to down-regulated levels of surface BCR. Moreover, we propose that basal signaling must drop below a certain threshold to turn off Rag expression and that the lynnull immature B cells are below this threshold while the xid immature B cells are not.
This idea is further supported by our finding that the gene expression profile of cells stimulated with Ag is highly similar to that of cells that have lost BCR expression following Cre-mediated deletion of H chain or following treatment with the protein tyrosine kinase inhibitor herbimycin. In each case, these cells turn on scores of genes characteristic of pro- and pre-B cells and turn down or off many other genes that comprise the mature B cell program. The parallels observed between the three experimental models suggest that similar mechanisms may be at work.
One prediction from our “reduced signaling” model for receptor editing is that editing should be inhibited if cells are provided with surrogates for positive signaling. This prediction was borne out by the PMA/ionomycin data, where it was shown that PMA/ionomycin blocked the induction of Rag2-GFP+ cells following treatment with Ag or a tyrosine kinase signaling inhibitor or in response to the loss of the BCR through Cre-mediated deletion. Furthermore, in a novel transgenic model system where all pre-B cells are undergoing editing to membrane self-Ag, PMA/ionomycin-treated cells showed reduced levels of ongoing L chain editing (Fig. 7). This experiment was particularly informative, as it more closely mimics true in vivo receptor editing and suggests that positive signaling suppresses editing in pre-B cells. PMA/ionomycin also effectively suppressed Rag mRNA in Jurkat T cell signaling mutants with elevated basal Rag2 levels (58). Interestingly, neither PMA alone nor ionomycin alone suppressed Rag as efficiently as the combination (B.R.S., unpublished observations), suggesting that a complex downstream signal may be required for efficient Rag suppression and then positive selection of immature B cells into the mature pool. Our observation that cells treated with PMA/ionomycin expressed higher levels of surface proteins such as PIR-A and CD22 than cells not stimulated is consistent with the idea that signaling induces positive selection of immature B cells to the mature cell stage. That the same signal required for Rag suppression might also be used to stimulate positive selection is not surprising and would mirror the situation observed in T cell-positive selection (59, 60, 61). A potential caveat we cannot exclude is that the signals generated by PMA and ionomycin may interfere with a positive tolerogenic signal set by differential BCR activation of downstream signaling pathways such as those involving ERK, JNK, NFAT, and NF-κB. We feel that this is unlikely because PMA and ionomycin treatment also blocked Rag2-GFP induction following TAT-Cre-mediated receptor deletion or herbimycin A treatment. Neither of these latter situations is subject to the above caveat, as these situations involve inhibition of all downstream BCR-dependent signals. Thus, the simplest explanation of the PMA/ionomycin results is that they simply mimic basal BCR signaling and thereby prevent Rag2-GFP induction.
Based on these data, we propose the following model for the regulation of L chain receptor editing, where basal signaling through the BCR in early immature B cells is required to suppress Rag gene expression and then signal for the positive selection of cells. Developing B cells with a successful H chain rearrangement at the pro-B stage express H chain in concert with the invariant surrogate L chains V-pre-B and λ5, undergo several rounds of clonal expansion, and then enter the resting, pre-B stage (62). Prolonged signaling through the surrogate BCR (perhaps involving the NF-κB pathway (63) induces κ germline transcription and activation of Rag transcription and translation. L chain gene rearrangements then initiate, generally at κ, and continue until a functional L chain is produced that can pair with the H chain, move through the secretory pathway, and be expressed as a functional BCR on the cell membrane.
In this model, a critical step in the pre-B to immature-B transition is the accumulation of sufficient Ig on the cell surface that is competent to produce tonic BCR signals. Of note, the levels of IgM expressed by immature B cells is very high, perhaps higher than at any other stage of B cell development, and we suggest that this high level expression is required to produce an adequate basal signal. Furthermore, immature B cells have little or no expression of many inhibitory receptors, such as CD22 and PIR-A, which are present on mature B cells. This may facilitate positive signaling through the BCR at the immature B cell stage.
Cells bearing Ig receptors with high levels of self-reactivity likely never display significant levels of surface Ig (i.e., they are endocytosed immediately after cell surface expression) and L chain alleles continue to rearrange, generally on the activated allele initially (64). For receptors with low-level self-reactivity, i.e., insufficient to signal for receptor down-regulation, the increased signaling through the receptor may “assist” the basal signal and result in a termination of Rag and progression in development. L chains that assemble poorly or that fail to pair with the H chain in a way that promotes adequate basal signaling would be edited, again because a sufficient basal BCR signal is not delivered. Thus, in this model the signal for receptor editing in response to self-Ag is not primarily driven by the cross-linking of surface BCRs but rather by the absence of a sufficient basal signal due to prolonged Ag-induced BCR down-regulation.
It seems likely that the level of signaling a B cell receives at this critical step in development, either basal or “assisted” by self-Ag, may be important for further differentiation steps toward the various alternate fates of B cells, i.e., anergy or the follicular, marginal zone, or B-1 pathways. We suggest that occasionally cells will express high levels of BCR and only later, during the immature stage, encounter strongly self-reactive Ags (20). These cells might receive a strong initial signal from self-Ag, but the primary consequence of that signal would be a down-modulation of sIg. The subsequent prolonged loss of basal signal would result in the reinitiation of Rag1/2 and the back-differentiation program, thereby providing an opportunity to revise the receptor by editing. Importantly, this model does not rule out the possibility that signaling through the receptor, perhaps under certain costimulation circumstances, may drive the induction of Rag. We believe that our proposed model is consistent with the data available and provides a number of testable hypotheses for future work.
What is the nature of the basal signal driving the cessation of Rag expression at the pre-B/immature B cell transition? It is apparent from the work of Monroe and colleagues that the minimal cellular machinery required to drive B cell development is the Ig αβ heterodimer (65). We hypothesize that the basal signal is transmitted through the following series of interacting proteins: the BCR, Igαβ, Syk, Lyn, Btk, and PI3K. Knockouts of syk (28), lyn (45), PI3K p85α (66), PI3K p110δ (67), and btk (Fig. 1) and the inhibition of PI3K with inhibitors (29) result in a phenotype compatible with a role for basal signaling in suppressing Rag induction. Of interest, a recent report showed that a high percentage of B cells that “squeak” through development in human X-linked agammaglobulinemia, a disorder characterized by mutations in btk, have a peripheral B cell repertoire enriched for autoimmune BCRs (68). The authors propose that inefficient signaling through the BCR may favor the development of autoreactive B cells, which is consistent with the model proposed here. Further downstream molecules conveying the basal signal may include phospholipase C-γ and protein kinase C-β; inhibitor studies and further genetic dissection will be required to explore the role of these molecules. Recently, Verkoczy et al. showed that phospholipase C-γ2 played a significant role in down-regulating RAG expression and B cell-positive selection (66).
These findings have significant implications for our understanding of allelic exclusion, whereby each B cell expresses a single BCR (62). In general, the mechanisms that initiate and maintain allelic exclusion are not well understood. H chain allelic exclusion requires the expression of a functional membrane-bound H chain protein, because mice lacking the Cμ transmembrane domain show a complete block in B cell development at the pro-B stage, and B cells fail to establish H chain allelic exclusion (69). H chain allelic exclusion also requires the Ig receptor-associated signaling proteins Igα and Igβ (70, 71, 72). Less is known about the signaling requirements for L chain allelic exclusion, where the situation is complex due to the presence of two κ and two λ alleles and the potential for multiple rearrangements at each locus.
There is evidence that the accessibility of chromatin to the V(D)J recombination machinery is important for controlling Ig (and TCR) rearrangements (reviewed in Ref. 73). Factors contributing to accessibility include the initiation of germline transcripts, the remodeling of histones, and the methylation status of individual alleles. Two basic models have been proposed to account for allelic exclusion: the regulated model and the stochastic model. In the regulated model, a functional rearrangement on one allele provides a signal that prevents, through feedback inhibition, further rearrangements on the other allele (74, 75). In the stochastic model, the probability and efficiency of V(D)J rearrangement is sufficiently low that allelic exclusion is obtained by “default.”
Although experimental data examining rearrangement status in single cells seems to favor the “regulated” model (76), a study from Schlissel and colleagues using a knockin GFP allele into Jκ1 provides support for a stochastic model to explain the initiation of L chain allelic exclusion (64). A key finding was that only ∼5% of pre-B cells showed sterile κ transcripts, and expression was mono-allelic. Furthermore, culturing of these GFP+ sterile κ+ pre-B cells led to preferential rearrangement of the GFP allele. The authors propose that the low frequency of cells bearing sterile transcripts for κ in the pre-B compartment may reflect competition for limiting numbers of transcription factors that can drive sterile κ transcription. As noted (76), these data are also consistent with a regulated model where pre-B cells activate sterile κ transcription only for a brief time and in a developmentally regulated fashion. In this model, cells with germline transcripts either undergo functional rearrangement and developmental progression or are arrested in the compartment, perhaps to undergo a second round of locus activation at a later time. The issues of initiation and then the maintenance of allelic exclusion are central to the concept of receptor editing, where allelic exclusion is maintained in the setting of the potential for multiple rearrangements at a locus. The data presented here are consistent with the regulated model.
In summary, we have now shown that three distinct perturbations of immature B cells, i.e., deletion of the BCR, treatment with proximal signaling inhibitors, and Ag treatment, result in a highly similar back-differentiation of cells to an earlier stage in B cell development. The plasticity exhibited by these cells is remarkable and suggests that the genetic program of B cells may be particularly flexible. This idea is further supported by the work of Busslinger and colleagues, who have shown that Pax5 is important for the maintenance of the B cell program (77). Early B cells that lose Pax5 show the ability to back-differentiate into stem cells and reprogram for the T lineage. Mature B cells with induced deletions of Pax5 “lose their place” in development and back-differentiate to the pro-B stage. Among the target genes for Pax5 are the Igα and Igβ signaling molecules, and we speculate that impaired transcription of these and perhaps other crucial signaling proteins may lead to a loss of basal BCR signaling and the developmental regression observed. Because mature B cells are poised to differentiate following appropriate signaling into highly efficient Ig-secreting plasma cells, the developmental plasticity observed in B cells may reflect a “ready-state” at the level of chromatin remodeling or other epigenetic alterations that allow for rapid genetic reprogramming.
We thank M. Nussenzweig for providing Rag2-GFP and human Cκ knock-in mice, K. Rajewsky for B1-8f/3-83κ mice and TAT-Cre, F. Batliwalla, P. Gregersen, W. Ortmann, E. Gillespie, A. Becker, and J. Plumb-Smith for assistance with the microarray experiments, D. Corcoran, H. Jecklin, A. Illes, and D. Fods for assistance in genotyping and animal husbandry, J. Peller for assistance in cell sorting, and L. Heltemes Harris, K. Murphy, and M. Schlissel for discussion and suggestions.
Following completion of this manuscript, Dr. Tim Behrens moved to Genentech where he is currently Director of Exploratory Clinical Development.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by grants from the National Institutes of Health.
Abbreviations used in this paper: BM, bone marrow; Ctrl, control; DEL, duck egg lysozyme; Fr. D, fraction D; Fr. E, fraction E; hCk, human Cκ (constant region); HA, herbimycin A; HEL, hen egg lysozyme; hi (superscript), high; lo (superscript), low; mHEL, membrane-bound HEL; neg (superscript), negative; PP2, 4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrazolo[3,4-d]pyrimidine; sIg, surface Ig; Tg, transgenic; WT, wild type; XID, x-linked immunodeficiency.