Dual Mechanism of Rag Gene Repression by c-Myb during Pre-B Cell Proliferation

ABSTRACT

Developing B lymphocytes undergo clonal expansion following successful immunoglobulin heavy chain gene rearrangement. During this proliferative burst, expression of the Rag genes is transiently repressed to prevent the generation of double-stranded DNA (dsDNA) breaks in cycling large pre-B cells. The Rag genes are then reexpressed in small, resting pre-B cells for immunoglobulin light chain gene rearrangement. We previously identified c-Myb as a repressor of Rag transcription during clonal expansion using Abelson murine leukemia virus-transformed B cells. Nevertheless, the molecular mechanisms by which c-Myb achieved precise spatiotemporal repression of Rag expression remained obscure. Here, we identify two mechanisms by which c-Myb represses Rag transcription. First, c-Myb negatively regulates the expression of the Rag activator Foxo1, an activity dependent on M303 in c-Myb's transactivation domain, and likely the recruitment of corepressors to the Foxo1 locus by c-Myb. Second, c-Myb represses Rag transcription directly by occupying the Erag enhancer and antagonizing Foxo1 binding to a consensus forkhead site in this cis-regulatory element that we show is crucial for Rag expression in Abelson pre-B cell lines. This work provides important mechanistic insight into how spatiotemporal expression of the Rag genes is tightly controlled during B lymphocyte development to prevent mistimed dsDNA breaks and their deleterious consequences.

INTRODUCTION

Expression of the recombination-activating genes Rag1 and Rag2 (referred to here as Rag) during lymphocyte development is essential for generation of the diverse B and T cell receptor repertoires required for effective adaptive immune responses (1). This process involves the generation of double-stranded DNA breaks in the antigen receptor loci (2). If generated during S phase, these breaks have the potential to be repaired by homologous recombination, resulting in chromosomal translocations that sometimes lead to oncogenic transformation (3). Thus, both RAG protein and Rag mRNA expression in developing lymphocytes, which undergo periods of proliferation and clonal expansion, are restricted to cells in the G₀-G₁ phases of the cell cycle (4, 5). Identifying the molecular mechanisms responsible for this spatiotemporal control of Rag expression is critical for our understanding of tumor suppression and proper immune system development.

Developing pro-B lymphocytes in the bone marrow undergo a period of clonal expansion following successful immunoglobulin heavy chain [Ig(H)] locus recombination (5). While the mechanism of RAG protein downregulation in these large, cycling pre-B cells is well characterized (6), the mechanisms by which Rag mRNA expression is repressed are less defined (discussed below). Moreover, little is known about how the activities of these repressive factors are controlled as cells cease to proliferate, differentiate into small pre-B cells, and reexpress the Rag genes for recombination of the immunoglobulin light chain [Ig(L)] loci.

Interleukin 7 receptor (IL-7R) signaling in large pre-B cells has been shown to activate the phosphatidylinositol 3-kinase (PI3K)–Akt pathway, resulting in phosphorylation and nuclear exclusion of Foxo1 (7), a crucial activator of Rag transcription (8, 9). Additionally, our laboratory has described Gfi1b and Ebf1 as negative regulators of Foxo1 expression (10, 11). Together the negative regulation of Foxo1 at the protein and mRNA levels may explain in large part how Rag gene repression occurs in trans. However, the possibility that transcription factors work in cis to diminish Rag mRNA levels during this proliferative burst to act as an additional safeguard against aberrant Rag expression has been minimally explored. Our group described Gfi1b binding to a region 5′ of the B cell-specific Erag enhancer (12), where it deposits repressive chromatin marks (10). Experiments using stably integrated reporter constructs showed that this region antagonizes Erag function in cis (10). Another study suggested that IL-7R signaling drives Stat5 binding to an uncharacterized cis element located 6 kb upstream of Rag1 to repress Rag expression (13), though no definitive evidence for direct repression was provided. Thus, other than Gfi1b, direct negative regulators of Rag transcription that act by binding to defined cis elements in the Rag locus in the context of cellular chromatin have yet to be described.

We previously used Abelson murine leukemia virus (AMuLV)-transformed B cell lines to screen for novel repressors of Rag transcription. The v-Abl oncogene selectively transforms developing B cells in a large cycling pre-B cell-like state in which Rag transcription is repressed. This developmental block can be reversed by inhibiting v-Abl kinase activity with STI-571 (STI), which induces cell cycle exit, differentiation to a small pre-B cell-like state, and robust Rag transcription (14). We identified Ebf1 and c-Myb, two well-studied transcription factors in the context of B cell development, as repressors of Rag transcription in these highly proliferative cells (11). Short hairpin RNA (shRNA) knockdown of either factor alone was sufficient to induce Rag transcription in AMuLV B cells independent of v-Abl inhibition with STI. Additional experiments suggested that Ebf1 does not repress Rag transcription directly, but rather through controlling expression of Foxo1 and Gfi1b, two factors previously identified as positive and negative regulators of Rag transcription, respectively (8–10). However, the mechanism by which c-Myb repressed Rag transcription (directly or via other factors) was less clear.

Here, we set out to understand the mechanism of Rag repression by c-Myb using AMuLV-transformed B cells. To do so, we compared the repressive activity of wild-type (WT) c-Myb to that of a c-Myb transactivation domain mutant (M303V) that has been shown to have impaired repressive activity in other contexts (15, 16). Experiments comparing the activities of the WT and mutant proteins revealed that c-Myb represses Rag transcription by a dual mechanism: indirectly through repression of Foxo1 and directly by occupying the Erag enhancer (12) in the Rag locus and antagonizing Foxo1 binding to a consensus forkhead site in this cis-regulatory element crucial for Rag expression in pre-B cells. Thus, this work identifies c-Myb as a direct negative regulator of Rag transcription that acts in cis in a stage-specific manner during B cell development to prevent aberrant Rag expression. It also shows how distinct but complementary mechanisms of gene repression by c-Myb cooperate to enforce the precise spatiotemporal Rag expression required for proper B cell development and generation of a diverse B cell receptor repertoire.

RESULTS

M303V mutation impairs the ability of c-Myb to repress Rag transcription.

Just as we found previously for its upstream regulator, Ebf1, overexpression of c-Myb in primary B-lineage progenitors was sufficient to significantly reduce Rag1 and Rag2 transcript levels (Fig. 1A). To investigate the mechanism of Rag repression by c-Myb, we used a previously described AMuLV-transformed Rag1-green fluorescent protein (GFP) reporter B cell line (9, 17). Rag1-GFP levels are low in these cells, and treatment with STI induces differentiation and robust Rag expression (Fig. 1B). We previously showed that protein levels of the Rag repressor Ebf1 decrease drastically upon STI-induced differentiation (11), allowing the upregulation of Rag expression. In contrast, c-Myb protein expression levels are similar in untreated and STI-treated cells (Fig. 1C), suggesting that mechanisms exist to control c-Myb's repressive activity in proliferating versus resting pre-B cells.

FIG 1.

M303V mutation impairs the ability of c-Myb to repress Rag transcription. (A) qPCR measuring Rag1 and Rag2 expression in primary B-lineage progenitors transduced with VEC or WT c-Myb retrovirus. The data were pooled from two independent experiments. *, P < 0.05, and ***, P < 0.005 versus the VEC control sample (Student's t test). (B) GFP expression in an AMuLV-transformed B cell line generated from heterozygous Rag1-GFP knock-in mice, where GFP reports transcription from the endogenous Rag locus. Analysis was performed on untreated (UT) or cells treated with STI-571 (+STI). The numbers above the gate indicate the percentages of GFP⁺ cells in untreated (top) or STI-treated (bottom) samples. (C) Immunoblot measuring c-Myb protein levels in the UT and +STI cells from panel B. The lamin immunoblot served as a loading control. The data from two independent experiments were quantified and pooled. The Student t test P value is shown. (D) GFP expression in Rag1-GFP reporter cell cultures transduced with WT c-Myb or M303V c-Myb retrovirus. Uninfected cells were distinguished from transduced cells in each culture by staining with anti-hCD4 (retroviral marker). Analysis was performed on UT and +STI cells. The numbers above the gates indicate the percentages of GFP⁺ cells among uninfected (top) or transduced (bottom) cells. The data are representative of 3 independent experiments. (E) GFP expression in sorted Rag1-GFP reporter cells transduced with 3×FLAG WT c-Myb or 3×FLAG M303V c-Myb retrovirus. Analysis was performed on UT and +STI cells. The numbers above the gates indicate the percentages of GFP⁺ cells in WT-transduced (top) or M303V-transduced (bottom) samples. The data are representative of 3 independent experiments. (F) FLAG immunoblot measuring 3×FLAG WT c-Myb and 3×FLAG M303V c-Myb protein levels in cells from panel E. The data from three independent experiments were quantified and pooled. The Student t test P value is shown. The error bars represent the standard deviations of two independent immunoblots.

We hypothesized that interactions with corepressors might control c-Myb activity. A previous study showed that B cell development in mice homozygous for a c-Myb transactivation domain (TAD) point mutation (M303V) is blocked at the exact stage during which cells undergo a proliferative burst and Rag transcription is repressed (large cycling pre-B cell, or Hardy fraction C′) (15). A subsequent study of a similar c-Myb TAD mutant (L302A) showed that this mutation abrogates the ability of c-Myb to both activate and repress target genes (16), suggesting this region of c-Myb mediates interactions with coactivators and corepressors. Thus, we wanted to test if the M303V mutation impaired the ability of c-Myb to repress Rag transcription in AMuLV-transformed B cells.

Figure 1D shows AMuLV-transformed Rag1-GFP reporter B cell cultures transduced with WT or M303V c-Myb retroviruses. Looking at GFP expression in untreated cells (top row), while WT c-Myb overexpression strongly decreased GFP levels compared to control untransduced cells in the same culture (top left), the effect was not as strong as for M303V c-Myb (top right), suggesting the M303V mutant has impaired repressive activity. This is more evident upon STI-induced differentiation (bottom row). While Rag induction was repressed in cells transduced with WT c-Myb (bottom left), more robust Rag induction was observed in cells transduced with M303V c-Myb (bottom right).

To rule out the possibility that differences in expression of the WT and mutant proteins accounted for this phenotype, we transduced cells with 3×FLAG-tagged WT and M303V c-Myb retroviruses and sorted the infected cells to obtain pure cultures. The sorted 3×FLAG M303V c-Myb cells had higher Rag expression levels than the sorted 3×FLAG WT c-Myb cells under untreated conditions (Fig. 1E, left), and Rag induction upon STI treatment was much more robust in the 3×FLAG M303V c-Myb cells than in the 3×FLAG WT c-Myb cells (Fig. 1E, right). FLAG immunoblotting using lysates prepared from these sorted cell lines showed the expression levels of the 3×FLAG-tagged c-Myb proteins were similar (Fig. 1F). Together, these experiments show that the M303V mutation impairs the ability of c-Myb to repress Rag transcription.

M303V c-Myb is defective in repression of Foxo1.

Foxo1 is a potent activator of Rag transcription during pro- and pre-B cell development (8, 9). Our previous work suggested that c-Myb represses Rag transcription, at least in part, through repression of Foxo1, as WT c-Myb overexpression reduces Foxo1 protein levels (11). To test whether M303V c-Myb also represses Foxo1, we transduced Rag1-GFP reporter cells with empty vector (VEC) or with 3×FLAG WT c-Myb (WT) or 3×FLAG M303V c-Myb (M303V) retrovirus and sorted the infected cells to purity; Rag1-GFP levels in these cell lines under untreated conditions are shown in Fig. 2A. Foxo1 immunoblotting showed that WT-c-Myb-transduced cells had significantly reduced Foxo1 protein levels compared to control empty-vector-transduced cells, confirming that WT c-Myb can repress Foxo1 (Fig. 2B). However, Foxo1 protein levels in the M303V c-Myb cell line were not reduced and were comparable to levels in the control cell line (Fig. 2B). Thus, the M303V mutation impairs the ability of c-Myb to repress Foxo1 expression.

FIG 2.

M303V c-Myb is defective in repression of Foxo1. (A) GFP expression in sorted Rag1-GFP reporter cells transduced with VEC or with 3×FLAG WT c-Myb or 3×FLAG M303V c-Myb retrovirus. The numbers above the gate indicate the percentages of GFP⁺ cells for each cell line under untreated conditions. The data are representative of 2 independent experiments. (B) Representative immunoblot measuring Foxo1 protein levels in cell lines from panel A with a FLAG immunoblot to confirm expression of 3×FLAG-tagged c-Myb proteins. The data from two independent experiments were quantified and pooled, with lamin-normalized Foxo1 levels in the VEC control cell line set to 1. *, P < 0.05 versus the VEC control sample (Student's t test). (C) Pro-B cell H3K4me1 and H3K4me3 ChIP-seq tracks (18) showing the locations of Foxo1 promoter (Pfoxo1) and enhancer (Ifoxo1B and Ifoxo1A) sites assessed for c-Myb binding. (D) ChIP using WT and M303V cell lines from panel A with anti-FLAG or IgG control antibody. qPCR was performed with the recovered chromatin to assess WT and M303V c-Myb occupancy in regions in the Foxo1 and Gfi1b loci described previously (11) compared to a control region in the Rag1 promoter (Pbrag1). The data are representative of 3 independent experiments. The error bars represent the standard deviations of triplicate reactions.

Using published chromatin immunoprecipitation sequencing (ChIP-seq) data (18, 19), our previous work identified a region in the Foxo1 promoter (Pfoxo1) and regions in a putative Foxo1 enhancer (Ifoxo1A and Ifoxo1B) bound strongly by Ebf1 in AMuLV cells, suggesting that Ebf1 directly represses Foxo1 expression (11) (Fig. 2C). Given that Ebf1 and c-Myb overexpression results in a decrease in Foxo1 protein levels, we hypothesized that c-Myb might also bind to these regions to negatively regulate Foxo1 in a direct manner. To test this, we performed FLAG ChIP using the 3×FLAG WT c-Myb and 3×FLAG M303V c-Myb cell lines. Both WT and M303V c-Myb bound the regions in the Foxo1 locus, but not a negative-control region in the Rag1 promoter (Pbrag1) (Fig. 2D). This result suggests that c-Myb directly represses Foxo1 expression. It also suggests that the defect in Foxo1 repression observed for M303V c-Myb is not due to a reduction in DNA binding ability, as occupancy levels at regions in the Foxo1 locus were comparable between the WT and M303V c-Myb proteins. This is consistent with previous gel shift experiments showing the two proteins have similar DNA binding abilities in vitro (15).

Foxo1 expression is also negatively regulated by the transcriptional repressor Gfi1b (10). We previously demonstrated that short hairpin RNA (shRNA) knockdown of either Ebf1 or c-Myb resulted in a decrease in Gfi1b mRNA levels, suggesting these factors positively regulate Gfi1b expression (11). FLAG ChIP clearly showed that both WT and M303V c-Myb proteins are strongly bound to the region in the Gfi1b promoter (Pgfi1b) previously shown to be bound by Ebf1 in AMuLV cells (Fig. 2D) (11), suggesting that positive regulation of Gfi1b by c-Myb is direct. As observed for the DNA elements in the Foxo1 locus, Gfi1b promoter binding levels are comparable between the WT and M303V c-Myb proteins.

In summary, these experiments showed that, unlike WT c-Myb, M303V c-Myb is unable to repress Foxo1 expression, providing an explanation for the defect in Rag repression observed for the mutant protein. And because comparable levels of DNA binding to these regions are observed for both WT and M303V c-Myb, this suggests M303V c-Myb's defect in Foxo1 repression is not due to a decrease in DNA binding ability, but rather, is due to the inability of M303V c-Myb to recruit corepressors to the Foxo1 locus.

Both WT and M303V c-Myb antagonize Foxo1 binding to the Erag enhancer.

Despite its mutation, M303V c-Myb possesses some residual repressive activity, as Rag1-GFP expression in the sorted M303V c-Myb cells was reduced compared to control empty-vector-transduced cells under untreated conditions (22% versus 34%) (Fig. 2A). This residual repressive activity was also evident upon STI treatment of these cell lines. While Rag1-GFP induction was much greater in M303V c-Myb cells (63%) than in WT c-Myb cells (22%), it was still reduced compared to control empty-vector-transduced cells (80%) (Fig. 3A). Quantifying the Rag1-GFP mean fluorescence intensity (MFI) in these STI-treated cell lines showed that the repressive activity of M303V c-Myb, while not as strong as that of WT c-Myb, is significant (Fig. 3B). Moreover, we found that, like WT c-Myb, overexpression of M303V c-Myb in primary B-lineage progenitors was sufficient to significantly repress Rag1 and Rag2 expression (Fig. 3C). Thus, we sought to understand this activity.

FIG 3.

M303V c-Myb antagonizes Foxo1 binding to the Erag enhancer to repress Rag expression. (A) GFP expression in sorted Rag1-GFP reporter cells transduced with VEC or with 3×FLAG WT c-Myb or 3×FLAG M303V c-Myb retrovirus and treated with STI. The numbers above the gate indicate the percentages of GFP⁺ cells for each cell line following STI treatment. The data are representative of 2 independent experiments. (B) Rag1-GFP MFI in sorted cell lines following STI treatment. The data from two experiments with independently generated cell lines were pooled. **, P < 0.01, and ***, P < 0.005 versus the VEC control cell MFI (Student's t test). (C) qPCR measuring Rag1 and Rag2 expression in primary B-lineage progenitors transduced with VEC or M303V c-Myb retroviruses. The data were pooled from two independent experiments. *, P < 0.05, and ***, P < 0.005 versus the VEC control sample (Student's t test). (D) Pro-B cell H3K4me1 and H3K4me3 ChIP-seq tracks (18) showing the locations of the Erag enhancer region and control regions for ChIP experiments. (E) ChIP using untreated WT and M303V cells from panel A with anti-FLAG or IgG control antibody. qPCR was performed with the recovered chromatin to assess WT and M303V c-Myb occupancy in the Erag enhancer region indicated in panel D compared to a control region in the Rag1 promoter. The data are representative of 3 independent experiments. (F) ChIP using STI-treated cells from panel A with anti-Foxo1 or IgG control antibody. qPCR was performed with the recovered chromatin to assess Foxo1 occupancy in the Erag enhancer region indicated in panel D compared to a control region between Erag and the Rag2 promoter. *, P < 0.05 versus the VEC control cells (Student's t test). The data are representative of 2 independent experiments with independently generated cell lines. The error bars represent the standard deviations of triplicate reactions.

We previously showed that Foxo1 inducibly binds to the well-characterized Erag enhancer following STI treatment and that WT c-Myb overexpression reduced Foxo1 binding to this regulatory element (11, 12) (Fig. 3D). We hypothesized that the ability of M303V c-Myb (which cannot repress Foxo1 expression) to repress Rag could be a direct effect, as M303V c-Myb and Foxo1 may compete for binding to the Erag enhancer. To test if WT and M303V c-Myb can bind to Erag, we performed FLAG ChIP using the 3×FLAG WT and 3×FLAG M303V c-Myb cell lines. Both WT and M303V c-Myb clearly bound to Erag but not to a control region in the Rag1 promoter (Pbrag1) (Fig. 3E). Thus, as with the previous ChIP experiments for regions in the Foxo1 and Gfi1b loci, the Erag ChIP signal showed that WT and M303V c-Myb possess similar DNA binding abilities.

To test if M303V c-Myb was capable of antagonizing Foxo1 binding to Erag, we performed Foxo1 ChIP in the STI-treated cell lines. Robust Foxo1 binding to the Erag enhancer but not to a control region between the enhancer and the Rag2 promoter was observed in the STI-treated empty-vector control cell line (Fig. 3F). As shown previously, Foxo1 binding to Erag was significantly reduced in the WT c-Myb cell line. Interestingly, Foxo1 binding was also significantly reduced in the M303V c-Myb cell line. This suggests that while the mutant protein is unable to repress Foxo1 expression, M303V c-Myb retains the ability to antagonize STI-induced Foxo1 binding to the Erag enhancer through its own ability to occupy this regulatory element, thus explaining the mutant protein's residual repressive activity on Rag expression. It also suggests that WT c-Myb antagonizes Rag expression by two independent mechanisms: indirectly through repression of Foxo1 and directly by occupying Erag and antagonizing Foxo1 binding.

Targeting an artificial transcriptional repressor to predicted c-Myb binding sites in Erag represses STI-induced Rag expression.

Targeted deletion of Erag, a 1.7-kb enhancer element lying 23 kb upstream of the Rag2 promoter, results in impaired B cell development, owing to reduced Rag expression and subsequent defects in V(D)J recombination (12). Similar defects in Rag expression and V(D)J recombination are observed when Foxo1 is conditionally deleted at different stages of B cell development in vivo (8). Accordingly, experiments in our laboratory have shown Rag expression in AMuLV B cells is entirely Foxo1 dependent, as genetic deletion of Foxo1 abolishes STI-induced Rag expression (20). A consensus forkhead binding site exists within the Erag ChIP amplicon analyzed in this study, and both gel shift and ChIP-seq experiments suggested that the site is bound by Foxo1 in pre-B cells in the Irf4^−/− Irf8^−/− IL-7 withdrawal system of pre-B cell differentiation (7) (Fig. 4A). However, the same ChIP-seq analysis showed that Foxo1 also binds to other regions in the Rag locus that may be important for activating Rag gene expression in pre-B cells (7, 11).

FIG 4.

Targeting a transcriptional repressor to c-Myb binding sites in Erag represses STI-induced Rag expression. (A) Rag locus schematic showing CRISPR/Cas9 transcriptional interference strategy. A catalytically inactive Cas9 protein fused to a KRAB repressor domain (dCas9-KRAB) was targeted to the Erag enhancer through coexpression of custom sgRNAs. The sequence of the Erag ChIP amplicon from Fig. 3 is shown with a previously described consensus Foxo1 binding site (7) and four adjacent c-Myb binding sites (5′-ACCNGNC-3′) in boldface and underlined. The lines above the sequence indicate sgRNA target sequences. (B) GFP expression in untreated and STI-treated Rag1-GFP reporter cells expressing dCas9-KRAB only or dCas9-KRAB and sgRNAs targeting regions indicated in panel A. The numbers above the gates indicate the percentages of GFP⁺ cells for each cell line following STI treatment. The V5 immunoblot with a lamin loading control shows V5-tagged dCas9-KRAB expression between the cell lines with relative V5/lamin levels quantified. (Note that the sgRNA targeting the c-Myb 1 site contained a seed sequence mutation.) (C) GFP expression in Rag1-GFP reporter cells expressing dCas9-KRAB only or dCas9-KRAB and sgRNAs targeting the Foxo1 Erag site, two other Rag locus Foxo1 binding sites (Irag1 and Irag2) described previously (7), or a control site between Erag and the Rag2 promoter. The approximate locations of these sites are indicated in panel A. The numbers above the gates indicate the percentages of GFP⁺ cells for each cell line following STI treatment. All the data are representative of two independent experiments.

We sought to test the importance of Foxo1 binding to this particular site within Erag, where c-Myb acts as a binding antagonist, for Rag expression. To do so, we utilized CRISPR/Cas9 transcriptional-interference technology (21). We created cell lines in which a catalytically inactive Cas9 protein (dCas9) fused to a Kruppel-associated box (KRAB) repressor domain is targeted to Erag by coexpressed small guide RNAs (sgRNAs) (Fig. 4A, top). More specifically, we designed the sgRNAs so that the dCas9-KRAB repressor was targeted to the Foxo1 Erag site or to four adjacent predicted c-Myb binding sites (5′-ACCNGNC-3′) (Fig. 4A, bottom). As a control, we created cells expressing the fusion protein but no sgRNA so that no dCas9-KRAB would be targeted to Erag.

dCas9-KRAB was expressed at similar levels in each of the cell lines, and STI treatment induced robust Rag transcription in the cells expressing dCas9-KRAB but no sgRNA (Fig. 4B). In the cells where dCas9-KRAB was targeted to the Foxo1 binding site in Erag, the induction of Rag expression upon STI treatment was strongly repressed. The same repression of Rag induction was observed when dCas9-KRAB was targeted to the c-Myb binding sites adjacent to the Foxo1 site, with targeting to c-Myb binding sites 2 and 3 nearest to the forkhead site having the strongest effect.

To support the notion that the effects of the targeted dCas9-KRAB protein on Rag expression were specifically due to perturbation of Foxo1 binding to this Erag site, we created cell lines in which the dCas9-KRAB repressor was targeted to two other previously described Foxo1 binding regions in the Rag locus also marked with H3K4me1 (Irag1 and Irag2) (Fig. 3D) (7), or to a control region between Erag and the Rag2 promoter, by coexpressed sgRNAs (Fig. 4A). Targeting dCas9-KRAB to these sites had no effect on STI-induced Rag expression (Fig. 4C), suggesting Foxo1 binding to Erag and not other putative enhancer regions bound by Foxo1 in the locus is the crucial step for activation of Rag transcription in differentiating pre-B cells.

Targeting a catalytically inactive dCas9 protein to predicted c-Myb binding sites in Erag represses STI-induced Rag expression.

dCas9-KRAB targeting to Erag could trigger epigenetic silencing by recruiting the corepressor KAP1 and other repressive chromatin-modifying enzymes to the enhancer (22). Based on our experiments shown in Fig. 3, we hypothesized that M303V c-Myb might repress Rag expression simply though steric hindrance of Foxo1. Thus, to test whether steric inhibition of transcription factor binding would be sufficient to repress Rag, we targeted a dCas9 protein lacking the KRAB repressor domain (dCas9-ONLY) to the Foxo1 Erag site or to four adjacent c-Myb binding sites (Fig. 5A). While targeting dCas9-ONLY to a control region between Erag and the Rag2 promoter had no effect on STI-induced Rag expression, targeting to the Foxo1 Erag site strongly repressed Rag induction (Fig. 5B). Likewise, targeting dCas9-ONLY to the c-Myb sites in Erag repressed STI-induced Rag expression to varying degrees, with targeting to c-Myb binding sites 2 and 3 once again having the greatest effects. These results show that targeting a protein incapable of corepressor recruitment to c-Myb binding sites adjacent to the Erag Foxo1 binding site is sufficient to repress STI-induced Rag expression. This supports the hypothesis that M303V c-Myb's ability to antagonize Foxo1 binding to Erag is largely a steric effect.

FIG 5.

Targeting a catalytically inactive dCas9-ONLY protein to c-Myb binding sites in Erag is sufficient to repress STI-induced Rag expression. (A) Rag locus schematic showing dCas9-ONLY targeting to Foxo1 and c-Myb binding sites in the Erag enhancer or to a control region between Erag and the Rag2 promoter, using custom sgRNAs. The lines above the sequence indicate sgRNA target sequences. (B) GFP expression in untreated and STI-treated Rag1-GFP reporter cells coexpressing dCas9-ONLY and sgRNAs targeting the sites indicated in panel A. The numbers above the gates indicate the percentages of GFP⁺ cells for each cell line following STI treatment. A V5 immunoblot with a lamin loading control shows V5-tagged dCas9-ONLY expression between the cell lines with relative V5/lamin levels quantified. All the data are representative of two independent experiments.

Preventing Foxo1 binding to Erag is sufficient to repress Foxo1-dependent Rag transcription.

While STI treatment drives robust Foxo1 binding to Erag and Rag induction in AMuLV B cells, it likely has other effects on transcriptional regulators of Rag expression and pre-B cell differentiation. To focus on and study only the Foxo1-dependent events involved in the induction of Rag expression, we utilized the Foxo1-estrogen receptor (ER) fusion protein system (9). In this system, treatment of AMuLV B cells with 4-hydroxytamoxifen (4-OHT) induces robust Rag expression by driving the nuclear translocation of a Foxo1-ER ligand binding domain fusion protein (Fig. 6A). Following 4-OHT treatment, ChIP using an antibody targeting the ER ligand binding domain can detect robust Foxo1-ER binding to Erag (Fig. 6B). Moreover, Foxo1-ER activation and Erag binding are sufficient to drive hallmarks of enhancer activation at Erag, including increased p300 and RNA polymerase (Pol) II occupancy and increased H3K27ac (Fig. 6B). This suggests Foxo1 binding to Erag is the crucial step in activation of this enhancer.

FIG 6.

Preventing Foxo1 binding to Erag is sufficient to repress Foxo1-dependent Rag transcription. (A) GFP expression in sorted Rag1-GFP reporter cells transduced with Foxo1-ER retrovirus. The numbers above the gate indicate the percentages of GFP⁺ cells in UT and 4-OHT-treated cultures. (B) ChIP using UT and 4-OHT-treated cell lines from panel A. qPCR was performed with the recovered chromatin to assess Foxo1-ER, p300, and Pol II occupancy, along with levels of H3K27ac, at Erag. *, P < 0.05 versus the corresponding ChIP in the UT cells (Student's t test). (C) GFP expression in untreated and 4-OHT-treated Rag1-GFP Foxo1-ER cells coexpressing dCas9-ONLY or dCas9-KRAB and either a control sgRNA or sgRNAs targeting c-Myb binding sites in Erag, as in Fig. 4A and 5A. The numbers above the gates indicate the percentages of GFP⁺ cells for each cell line following 4-OHT treatment. (D) ChIP using the 4-OHT-treated cell lines from panel C with anti-V5 antibody or IgG control. qPCR was performed with the recovered chromatin to assess dCas9-KRAB occupancy at Erag. *, P < 0.05 versus the corresponding ChIP in the control sgRNA cells (Student's t test). (E) ChIP using the 4-OHT-treated cell lines from panel C with anti-ER ligand binding domain antibody or IgG control antibody. qPCR was performed with the recovered chromatin to assess Foxo1-ER occupancy at Erag. *, P < 0.05 versus the corresponding ChIP in the control sgRNA cell line. All the data are representative of two independent experiments. The error bars represent the standard deviations of triplicate reactions.

Interestingly, in Foxo1-ER AMuLV B cells expressing either dCas9-ONLY or dCas9-KRAB, along with sgRNAs targeting the c-Myb 2 and c-Myb 3 binding sites in Erag, Foxo1-driven Rag induction was strongly repressed compared to cells expressing a control sgRNA (Fig. 6C). ChIP in the dCas9-KRAB Foxo1-ER cells confirmed the targeting of the dCas9-KRAB protein to Erag by the c-Myb 2 and 3 sgRNAs (Fig. 6D) and that the binding of dCas9-KRAB is inversely correlated with the amount of Foxo1-ER Erag binding in these 4-OHT-treated cells (Fig. 6E). This experiment clearly showed that the prevention of Foxo1 binding to this particular forkhead site in the Erag enhancer is sufficient to repress Rag expression during pre-B cell differentiation. Thus, Foxo1 binding to a crucial activator site in Erag is necessary for Foxo1-dependent Rag transcription, and c-Myb's antagonism of Foxo1 binding to Erag is a bona fide mechanism by which c-Myb represses Rag transcription.

Based on our experiments, we proposed that c-Myb represses Rag transcription in proliferating pre-B cells by a dual mechanism (Fig. 7). First, c-Myb indirectly represses Rag by negatively regulating expression of the Rag activator Foxo1, an activity dependent on M303 and likely involving the recruitment of corepressors. Second, c-Myb directly represses Rag by binding to the Erag enhancer and antagonizing Foxo1 binding to this important cis-regulatory element. This activity may involve recruitment of corepressors by c-Myb that modify chromatin and restrict DNA accessibility. However, steric hindrance by c-Myb may account for some of this repressive activity, as we showed that proteins unable to recruit corepressors (M303V c-Myb and dCas9-ONLY) are capable of antagonizing Foxo1 Erag binding and repressing Rag expression. Regardless of the exact mechanism, our studies provide the first description of the direct negative regulation of Rag expression in B cells by c-Myb working in cis at the Erag enhancer. Together, the distinct but complementary actions of c-Myb cooperate to ensure Rag transcription is strongly repressed in proliferating pre-B cells.

FIG 7.

Model for dual repression of Rag by c-Myb during pre-B cell differentiation. Proproliferative signals in large pre-B cells drive c-Myb binding to regulatory sites in the Foxo1 locus to limit Foxo1 protein levels and to the Erag enhancer to prevent Foxo1 binding and Rag transcription. Upon differentiation, these signals are lost, resulting in reduced c-Myb binding to sites in the Foxo1 locus and Erag. This allows more robust Foxo1 expression and frees the Erag enhancer to allow Foxo1 binding, which drives high levels of Rag transcription in small pre-B cells necessary for Rag1/2 production and recombination at the Ig kappa locus.

DISCUSSION

c-Myb plays an important role in B lymphoid priming through activation of lineage-specific genes in uncommitted progenitor cells (23–25). Using the AMuLV-transformed B cell system, we uncovered a new role for c-Myb at the large-to-small pre-B cell transition, where we identified c-Myb as a bona fide repressor of Rag transcription in cells undergoing clonal expansion (11). Our new findings reveal that repression of Rag transcription by c-Myb proceeds via a dual mechanism that ensures stringent spatiotemporal control of Rag expression during this critical developmental transition.

Regarding indirect Rag repression, our data suggest that c-Myb is both a direct negative regulator of Foxo1 and a direct positive regulator of Gfi1b. Together, these activities would result in a reduction of Foxo1 protein levels by c-Myb. Because the M303V mutation clearly impairs the ability of c-Myb to repress Foxo1 (Fig. 2B) but does not affect c-Myb binding to elements in the Foxo1 and Gfi1b loci (Fig. 2C), we postulate that the M303V mutation prevents c-Myb from recruiting coactivators and corepressors to the Foxo1 and Gfi1b loci to achieve repression and activation, respectively. This is consistent with the observation that a similar c-Myb mutant (L302A) is unable to either activate or repress direct target genes in the context of myeloid differentiation (16). Although M303 and surrounding TAD residues are part of an amphipathic helix crucial for c-Myb's interaction with the KIX domain of the coactivators p300 and CBP (15, 26), the mechanism by which this region mediates Myb's interactions with corepressors such as histone deacetylases (27, 28) is unknown and warrants further investigation.

With regard to direct repression, we found that, like WT c-Myb, the M303V c-Myb mutant can antagonize Foxo1 binding to the Erag enhancer upon STI-induced differentiation (Fig. 3F). Since exogenously supplied WT and M303V c-Myb bind equally well to Erag (Fig. 3E), we surmise that the repressive activity of the mutant protein lies in its ability to sterically interfere with Foxo1 binding to Erag. Thus, these experiments comparing WT and M303V c-Myb were crucial in that they unexpectedly revealed a distinct mode of direct Rag repression by c-Myb. Without this comparison, we would have attributed the reduced Foxo1 binding to Erag in the WT c-Myb-overexpressing cell line simply to reduced Foxo1 levels in these cells (Fig. 2B).

Several important questions remain unanswered. Unfortunately, were unable to extend our c-Myb ChIP experiments to endogenous c-Myb in either Abelson pre-B cells or primary B-lineage progenitors. Difficulty with endogenous c-Myb ChIP in murine cells has been noted in the literature (29) but was performed by others (30). Thus, optimizing c-Myb ChIP conditions in these contexts will be an important step toward support of a model where c-Myb antagonizes Foxo1 Erag binding during pre-B cell development in vivo. Moreover, if c-Myb does indeed bind Erag in these contexts, whether repression by c-Myb requires the recruitment of corepressors or is achieved solely through steric inhibition of Foxo1 is unclear. Finally, the signals that regulate c-Myb binding to Erag during pre-B cell differentiation are unknown (Fig. 7). Given that pre-B cell differentiation involves proliferation followed by cell cycle exit, we speculate that c-Myb binding and its repressive activity at Erag might be controlled by regulators of the cell cycle. Indeed c-Myb activity is regulated by cyclin D1 and CDK4/6 in T cells (31), and c-Myb can be repositioned to distinct target gene promoters during different cell cycle phases in both primary hematopoietic stem cells and transformed cells (32). One tempting hypothesis is that the same proproliferative IL-7R signaling axis that inhibits Foxo1 nuclear localization during clonal expansion might also drive c-Myb's repressive activity (Fig. 7). Such a mechanism could work through posttranslational modifications that alter c-Myb activity or localization, since c-Myb protein levels are the same in large and small pre-B cells (Fig. 1C).

Whatever the exact mechanism of direct repression by c-Myb, it is interesting that inhibiting Foxo1 binding to a single site in the Rag locus is sufficient to repress Rag expression upon pre-B cell differentiation. Early studies seeking to identify cis-regulatory elements controlling Rag expression in B lymphocytes revealed multiple regions in the Rag locus that display DNase hypersensitivity and enhancer activity in cell lines and transgenic mice (12, 33, 34). More recent ChIP studies have identified several regions in the Rag locus where chromatin marks of cell-type-specific enhancers such as H3K4me1 are present (18) (Fig. 3B) and mapped the binding of Rag activators, including Foxo1 and Pax5, to multiple regions in the locus (7). Our experiments suggest Foxo1 binding to a single site in Erag is the essential first step in enhancer activation and Rag expression in pre-B cells (Fig. 6B). Moreover, our experiments show that CRISPR/Cas9 transcriptional interference provides a powerful interrogative tool to perturb transcription factor binding at enhancers in the context of native chromatin to identify activator sites crucial for enhancer activity and target gene expression. Given its importance, it will be interesting to see if noncoding single nucleotide polymorphisms (SNPs) associated with diseases of defective Rag expression (such as severe combined immunodeficiency syndrome [35]) are discovered for this Erag Foxo1 site.

p300 is best known as a coactivator that interacts with sequence-specific transcription factors and enhances gene expression. However, experiments with one of the originally described c-Myb mutants (L302A) defective for p300 interactions revealed that, in the context of myeloid differentiation, the mutation affects the ability of c-Myb to both activate and repress target genes (16). Further support of the notion that c-Myb and p300 cooperate for gene repression came in a study showing p300 and CBP KIX domain mutations affect the ability of c-Myb to both activate and repress genes in mouse embryonic fibroblasts (MEFs) and T cells (36). We were unable to establish a role for p300 in Rag gene repression in our system, as overexpression and knockdown of p300 in AMuLV B cells showed no effect (unpublished observations). Moreover, ChIP experiments showed a positive relationship between p300 and Rag, as p300 occupancy increases significantly at Erag upon both STI treatment (unpublished data) and Foxo1-ER activation (Fig. 6B). Thus, if c-Myb and p300 are cooperating for Rag repression, it is not achieved through c-Myb-dependent recruitment of p300 to the Erag enhancer. As mentioned previously, no studies have investigated if the M303V mutation also disrupts c-Myb's interactions with corepressors. Coimmunoprecipitation and mass spectrometry experiments could be used to identify differential interacting proteins between the WT and the M303V c-Myb mutant to generate a list of candidates that could be tested for roles in Rag gene repression by c-Myb in proliferating pre-B cells.

Finally, M303V c-Myb failed to rescue cell survival upon deletion of endogenous c-Myb in an AMuLV B cell line with floxed c-Myb alleles (unpublished observations). In the context of B cell development, the failure to rescue is consistent with the observation that c-Myb^M303V/M303V mice completely lack cells in the large cycling pre-B cell/Hardy fraction C′ compartment (15). Thus, in addition to its role in B-lineage specification in uncommitted progenitor cells, c-Myb and its M303-dependent activities are crucial for the survival and clonal expansion of committed pre-B cells. In the context of leukemic transformation, this rescue failure suggests that the M303-dependent activities of c-Myb are required for maintaining the AMuLV-transformed state. This is in line with studies describing the requirement for c-Myb in transformation by ABL oncogenes and in cancers, including pre-B cell acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML) (37–40). Thus, disrupting M303-dependent interactions in c-Myb-driven cancers is a tempting therapeutic strategy. Indeed, two recent reports describe small-molecule inhibitors of c-Myb–p300 interactions that induce differentiation and apoptosis of cancer cell lines (41, 42). A more complete understanding of the M303-dependent protein interactions, and the identity of M303-dependent target genes, will provide more insight into the role of c-Myb in oncogenic transformation and likely identify additional therapeutic targets.

MATERIALS AND METHODS

Cell culture and chemicals.

AMuLV-transformed B cells were cultured in RPMI 1640 (Gibco) supplemented with 5% (vol/vol) fetal calf serum (FCS) (Gemini), 100 mg/ml penicillin and streptomycin (Gibco), and 55 nM 2-mercaptoethanol (Gibco). GP2 retroviral packaging cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Gibco) supplemented with 5% (vol/vol) FCS, 100 mg/ml penicillin and streptomycin, and 1 mM sodium pyruvate (Gibco). All the cells were grown at 37°C in 5% CO₂. STI (Cayman) was used at a final concentration of 2.5 μM, and 4-OHT (Sigma) was used at a final concentration of 10 μM. Treatment durations were 16 to 20 h for all STI and 4-OHT experiments.

Retroviral expression plasmids.

Murine stem cell virus (MSCV)-based cDNA retroviral expression constructs were previously described (9). The murine c-Myb transcript variant 2 (RefSeq mRNA accession no. NM_010848.3) open reading frame (ORF) was PCR amplified from a primary B cell cDNA library with Platinum Pfx DNA polymerase (Invitrogen) and cloned into a multiple-cloning site (MCS) upstream of an internal ribosome entry site (IRES). This IRES is followed by coding sequence for human CD4 (hCD4) cell surface protein to mark infected cells.

The m303V c-Myb mutant was created by cloning WT c-Myb into a pBSK plasmid. Mutagenesis was then performed using a QuikChange multisite-directed mutagenesis kit (Agilent) according to the manufacturer's instructions with the following mutagenesis primer: 5′-GCTGGAGTTGCTCCTGGTGTCAACAGAGAACGA-3′.

C-terminal 3×FLAG-tagged WT c-Myb and M303V c-Myb constructs were created by PCR amplifying the ORFs lacking a stop codon and cloning them upstream of a 3×FLAG sequence inserted into the MCS in the MSCV IRES hCD4 retroviral vector.

Retrovirus production and infection.

GP2 packaging cells were transfected with retroviral and VSVG plasmids at a 5:3 ratio using Lipofectamine 2000 (Invitrogen) and placed at 37°C. The cells were moved to 32°C at 24 h. At 48 h, the retroviral supernatant was collected and concentrated (1.5 h; 16,800 × g; 4°C) using an SW41 Ti rotor and an L8-M ultracentrifuge (Beckman). The retroviral pellets were resuspended in medium supplemented with 4 μg/ml Polybrene (Sigma-Aldrich). Target cells were cultured in concentrated retrovirus at 32°C for 24 h and then moved to 37°C and expanded. Analysis and sorting were performed 2 to 4 days postinfection.

c-Myb overexpression in primary B-lineage progenitors.

All experiments were approved by the Animal Care and Use Committee at the University of California, Berkeley. The handling of animals was in accordance with protocol R253-0405.

Bone marrow was isolated from 6- to 12-week-old male C57/B6 mice (Jackson), and B cell progenitors were expanded via culture in RPMI 1640 with 10% (vol/vol) FCS and supplemented with 5 ng/ml recombinant mouse IL-7 (Shenandoah Biotechnology). The cells were transduced with retrovirus as described above on day 3 and sorted on day 7 as described below directly into TRIzol LS (Invitrogen). RNA was isolated using a Direct-zol mRNA miniprep kit (Zymo), and cDNA was prepared with SuperScript III (Invitrogen) for quantitative real-time PCR (qPCR).

Cell sorting and flow cytometry.

Single-cell suspensions were prepared, and when necessary, cells were labeled with anti-hCD4–phycoerythrin (PE), anti-CD19–peridinin chlorophyll protein 5.5 (PerCP5.5), and anti-IgM–allophycocyanin (APC) antibodies (eBioscience) using standard techniques. An FC500 (Beckman Coulter) or LSRII (BD Biosciences) flow cytometer was used for analysis, while a MoFlo or Influx cell sorter (Dako-Cytomation) was used for sorting. The data were analyzed with FlowJo software (Tree Star).

CRISPR/Cas9 transcriptional interference.

Small guide RNAs were designed using the MIT CRISPR Design Tool (http://crispr.mit.edu/) and custom oligonucleotides ordered from Integrated DNA Technologies. The oligonucleotides were annealed, phosphorylated, and directionally cloned into the BsmBI site of a modified PiggyBac dual expression vector downstream of a U6 promoter and upstream of an EF1α promoter driving expression of either a catalytically inactive Streptococcus pyogenes Cas9 protein alone (dCas9-ONLY) or fused to a Kruppel-associated box repressor domain (dCas9-KRAB). Individual sgRNA-dCas9 constructs were cotransfected with Super PiggyBac transposase expression vector using Lipofectamine 2000 (Invitrogen), and cells were G418 selected (800 μg/ml) for 2 weeks to generate stable cell lines. The sgRNA oligonucleotides are listed in Table 1.

TABLE 1.

sgRNA oligonucleotides used in the study

Oligonucleotide	Sequence (5′–3′)
	Top	Bottom
Erag c-Myb site 1	CACCGCGTTTCCAACTTCCTCCAGC	AAACGCTGGAGGAAGTTGGAAACGC
Erag c-Myb site 2	CACCGATCGCCTGCTGGAGGAAGT	AAACACTTCCTCCAGCAGGCGATC
Erag c-Myb site 3	CACCGCAACTGGCAGATCGCCTGC	AAACGCAGGCGATCTGCCAGTTGC
Erag Foxo1 site	CACCGTCAAAACAATGCTAAGCCCT	AAACAGGGCTTAGCATTGTTTTGAC
Erag c-Myb site 4	CACCGTCGACAGAAAATAACTGCGC	AAACGCGCAGTTATTTTCTGTCGAC
Irag1	CACCGAATGGGGCTATTAATAGATC	AAACGATCTAATAATAGCCCCATTC
Irag2	CACCGTCTCTCTAAGTGCTTTACAC	AAACGTGTAAAGCACTTAGAGAGAC
Controla	CACCGGGCAGATTTCTCTGAGTTTG	AAACCAAACTCAGAGAAATCTGCCC

^aSite between Erag and the Rag2 promoter.

Immunoblotting.

Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer supplemented with Complete protease inhibitor cocktail (Roche). Cell debris was cleared by centrifugation, and soluble protein was quantified with Bradford reagent (Bio-Rad). Laemmli SDS loading buffer was added to 20 μg of protein per sample prior to separation on SDS-8% PAGE gels. Following transfer to Immobilon-FL polyvinylidene difluoride (PVDF) (Millipore) or Protran BA85 nitrocellulose (GE Whatman) membranes and blocking with 5% (vol/vol) milk-PBS solution, the blots were probed with primary and secondary antibodies and analyzed with the Odyssey infrared imaging system (Li-Cor Biosciences). The primary antibodies used were as follows: anti-Myb (Millipore; 05-175), anti-Foxo1 (Cell Signaling; L27), anti-lamin B1 (Abcam; ab16048), anti-FLAG M2 (Sigma-Aldrich), and anti-V5 (GeneTex; GTX117997). Infrared dye-conjugated secondary antibodies were from Molecular Probes-Invitrogen. Quantification of immunoblots was performed using ImageJ64 software.

Chromatin immunoprecipitation.

ChIP was performed as previously described (43). Briefly, approximately 100 million cells per experimental condition were harvested, fixed, lysed, and sonicated using a Covaris S220 ultrasonicator. Following centrifugation to remove insoluble material, the chromatin was quantified, and equal amounts were used in experimental and control immunoprecipitations. Five milligrams of anti-FLAG M2 (Sigma-Aldrich), anti-Foxo1 (Abcam; ab70382), anti-p300 (Santa Cruz; SC-585 X), anti-ER (Santa Cruz; MC-20), anti-V5 (Thermo Fisher Scientific; R960-25), anti-H3K27ac (Abcam; ab4729), anti-Pol II (Abcam; 8WG16), or IgG control (GeneScript) antibodies was conjugated to Protein G Dynabeads (Invitrogen) and added to samples. Following overnight immunoprecipitation, the beads were collected and washed three times with low-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris, pH 8.1, 150 mM NaCl), once with high-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris, pH 8.1, 500 mM NaCl), once with LiCl buffer (250 mM LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris, pH 8.1), and twice with Tris-EDTA (TE) buffer. DNA-protein complexes were eluted from the beads with TE elution buffer (TE with 1% SDS), and cross-links were reversed overnight at 65°C. Following RNase A (Fermentas) and proteinase K (Invitrogen) treatment, chromatin was isolated using QIAquick columns (Qiagen) and subjected to qPCR analysis.

qPCR.

qPCR was carried out on an Applied Biosystems 7300 thermocycler using SYBR Select master mix (Applied Biosystems). The PCR conditions were as follows: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 15 s and 60°C for 1 min. For ChIP qPCR, data are presented as a percentage of the ChIP input. Primer sequences are listed in Table 2.

TABLE 2.

Primers used in the study

Primer	Sequence (5′–3′)
	Forward	Reverse
ChIP
Pfoxo1	GAAAAATACCCCACCGCCCC	AATGGACGCGCGAAGTCTC
Ifoxo1A	ACTCCCTGTGCCTTTTTAGCC	CTTCGACATTGTTCGTGGCG
Ifoxo1B	TCCACCTTGGAGGAAATAAGGC	GGAAGCTAGACACGCCAAGT
Pgfi1b	GGGAGCTGTCCCTCTTCTGA	CTCATAACGTTGACCGAGCC
Erag	CGTTTCCAACTTCCTCCAGC	GCCCTGCGCAGTTATTTTCT
Pbrag1	GGAAGTTTAGCTGGGGGACC	CCACCGTAGGCATTCTCAGG
Rag locus
Control	CTCTGTATAGCCCTGGCTGTCCAG	TGAAGCCGGGCAGTGGTG
Gene expression
Rag1	CATTCTAGCACTCTGGCCGG	TCATCGGGTGCAGAACTGAA
Rag2	TTAATTCCTGGCTTGGCCG	TTCCTGCTTGTGGATGTGAAAT
Hprt	CTGGTGAAAAGGACCTCTCG	TGAAGTACTCATTATAGTCAAGGCA