Wild-type (WT) C57BL/6 mice were purchased from Jackson Laboratory. Ets1^−/− mice (RRID:MGI: 3833458) and littermate WT controls used in this study were bred in our facility and maintained on a mixed genetic background of C57BL/6 × 129Sv because, on a pure C57BL/6 genetic background, the loss of Ets1 is lethal perinatally. The mutation in the Ets1 locus of these mice has previously been described (23, 38). Mice carrying the Prdm1-green fluorescent protein (GFP) allele that inactivates Blimp1 were obtained from Dr. Stephen Nutt (Walter and Eliza Hall Institute of Medical Research, Parkville, VIC, Australia) (39).

Wild-type mouse B cells were purified from spleens of 3-month-old C57BL/6 male mice using negative selection with the Easysep mouse B cell isolation kit (Stem Cell Technologies). Purified B cells were rested for 1 h in complete media [RPMI 1640 + 10% fetal bovine serum, 1% pen/strep, 1% Glutamax (Gibco), and 50 μM β-mercaptoethanol] in a tissue culture incubator. After resting, cells were cross-linked by adding formaldehyde to a final concentration of 0.25% for 8 min. Fixed B cells were lysed and chromatin prepared according to the manufacturer’s protocol for the ChIP-IT High Sensitivity Kit (Active Motif). Chromatin was sonicated to yield fragments of an average size ~200–700 bp and immunoprecipitated with a rabbit polyclonal anti-Ets1 antibody (sc-350X, Santa Cruz) that has previously been used in chromatin immunoprecipitation assays (40–42). Two biological replicates were separately prepared and analyzed. The number of uniquely mapped reads for ChIP-seq was between 10 and 27 million for the Ets1 ChIP-seq and between 7 and 10 million for sequencing of the input.

Libraries were generated from the purified chromatin, and ChIP-sequencing was performed on input chromatin and Ets1-precipitated chromatin using an Illumina Hiseq2500 Sequencing System. The ChIP-seq data were found to be of good quality using the normalized strand coefficient and the relative strand correlation parameters as described previously (43). The reads were aligned to mouse mm9 genome assembly using Bowtie (44).

We identified Ets1-bound regions using the MACS2 program (45). Peaks identified from each biological replicate were compared to input controls using the irreproducible discovery rate to identify reproducible Ets1-binding sites of which 10,391 were detected. Ets1-bound regions were annotated for enrichment at intergenic regions, promoters, exons, or introns using ChIPseeker (46). Overrepresented motifs in the peaks were analyzed using the findmotifgenome function in Homer (47). Locations of binding sites with respect to potential target genes were visualized in GenomeBrowser (https://genome.ucsc.edu/cgi-bin/hgGateway).

Ets1 consensus binding sites and other transcription factor-binding sites in the Ets1-bound regions were identified using PscanChIP (48). The regions were then aligned centered on the Ets1 consensus motif. ChIP-seq data for acetylated and methylated lysine variants of histone H3 in primary B cells were obtained from the GEO database and are reported in Ref. (49–51). The program ArchTex (52) was then used to map average histone modifications surrounding the Ets1 consensus-binding sites. As a control, ArchTex was also used to align histone modifications surrounding 60,000 random consensus Ets1-binding sites from regions not bound by Ets1 in ChIP-seq assays.

To test if Ets1 might co-regulate some of the same genes as Pax5, Tcf3 (E2A), or Irf4, we obtained ChIP-seq data for these factors in primary mouse splenic B cells, as reported in Ref. (51, 53, 54). We used Bedtools to identify the co-occurrence between peaks and common target sites for these transcription factors.

Spleens from female WT (C57BL/6) and Ets1 knockout mice (C57 × 129Sv) were used to prepare single cell suspensions, and quiescent follicular B cells were sorted from each population based on the following markers: B220⁺ CD23^hi CD21^low CD80^neg IgA^neg IgE^neg IgG1^neg IgG2a^neg IgG2b^neg IgG3^neg using FACSAria II cell sorter. Dead cells were gated out from the sorted population using Live/Dead Fixable Aqua dead cell stain (Molecular Probes). Total RNA was isolated from two biological replicates of sorted B cells and subjected to RNA-sequencing on an Illumina Hiseq2500 Sequencing System. Sequence reads (24–49 million per sample) were aligned to the mm9 genome assembly using the Tophat program, and differences in gene expression between samples were analyzed using Cuffdiff (55). For further analysis, we chose genes that showed a fold change between WT and Ets1^−/− B cells of at least twofold [log₂(FC) = 1.0 or more] and a q value of 0.05 or less. We excluded genes with a reported Fragments Per Kilobase of transcript per Million mapped reads (FPKM) of less than 1.0 in either genotype. CummeRbund was used to visualize the Cuffdiff analysis, including generating heat maps of gene expression levels.

Genes that showed an alteration in expression as described above were analyzed for their functions by DAVID software (56) to identify the most-relevant gene ontology (GO) terms that were significantly enriched. We used an EASE score equal to 0.05 and a count threshold of 5 to identify enriched pathways.

To validate RNA-seq data, we purified B cells from the spleens of 3-month-old male and female WT and Ets1^−/− mice (C57BL/6 × 129Sv genetic background) by first using magnet beads (Stem Cell Technologies mouse B cell purification kit) followed by sorting mature naïve follicular B cell subset B220⁺ CD23^hi CD21^low CD11b^neg CD80^neg IgA^neg IgE^neg IgG1^neg IgG2a^neg IgG2b^neg IgG3^neg. RNA was extracted from sorted B cells (n = 4 of each genotype) using Direct-zol RNA MiniPrep kits (Zymo Research). cDNA was synthesized from equal amounts of RNA using QuantiTect Reverse Transcription kit (Qiagen). Quantitative RT-PCR was performed using iQ SYBR Green Supermix (Bio-Rad). The sequences of primers used for RT-PCR were: Slamf6 (Forward-CAGCTAATGAATGGCGTTCTAGG, Reverse-CTTAGGTTGATAACGAGGGCAG), Egr1 (Forward-AACCGGCCCAGCAAGACACC, Reverse-TGGCAAACTTCCTCCCACAAAT), Ptpn22 (Forward-AGCAAGCCTACAGAACGTG, Reverse-TCCAGAGGTGCGTTACATATTC), Stat4 (Forward-TGGCAACAATTCTGCTTCAAAAC, Reverse-GAGGTCCCTGGATAGGCATGT), Prdm1 (Forward-TGTGGTAATGTCGGGACTTTG, Reverse-TTCCTTTTGGAGGGATTGGAG), Hprt (Forward-CCTCATGGACTGATTATGGACAG, Reverse-TCAGCAAAGAACTTATAGCCCC), Gapdh (Forward-AATGGTGAAGGTCGGTGTG, Reverse-GTGGAGTCATACTGGAACATGTAG), and Actb (β-actin) (Forward-GCAGCTCCTTCGTTGCCGGTC, Reverse-TTTGCACATGCCGGAGCCGTTG). Gene expression was normalized to all three housekeeping genes (Gapdh, Hprt, and Actb) using Bio-Rad CFX Manager Software.

Plasmids encoding mouse Stat4 (57) or human myc-tagged Ptpn22 (58) were obtained from Dr. John O’Shea (National Institutes of Health, Bethesda, MD, USA) or Dr. Dimitar Efremov (International Centre for Genetic Engineering and Biotechnology, Rome, Italy), respectively. The cDNAs were cut from the original plasmids and sub-cloned into the MIGR1 retroviral plasmid, which contains an internal ribosomal entry site (IRES) followed by GFP to allow easy identification of virally infected cells. The resulting plasmids were confirmed by sequencing. The MIGR1 plasmid encoding mouse Ets1 has been described previously (22). The plasmid used for Egr1 knockdown was generated by cloning a shRNA against Egr1 into the microRNA-30-adapted shRNAmir retroviral vector (MSCV-lmp) from Open Biosystems. Oligonucleotides with sense and antisense strands of an shRNA targeting nucleotides 2,314–2,336 of mouse Egr1 mRNA (NM_007913.5) were cloned into MSCV-lmp and confirmed by sequencing. MSCV-lmp also contains an IRES-GFP module that can be used for detecting transduction. A retroviral plasmid containing a shRNA against Prdm1 in MSCV-lmp, where the GFP gene was substituted with a GFP variant Ametrine (59), was obtained from Dr. Matthew E. Pipkin (The Scripps Research Institute, Jupiter, FL, USA).

Retroviral plasmids were used to transfect the Plat-E packaging cell line along with the plasmid pCL-Eco (which contains additional copies of the viral structural genes) using Fugene-6. The retroviral supernatant was harvested after 48 h of transfection and used to infect WT or Ets1^−/− B cells that were purified using anti-B220 microbeads (Miltenyi Biotec) and stimulated with 10 μg/ml of lipopolysaccharide (LPS) for 24 h. The infected B cells were subsequently returned to fresh medium containing LPS. Forty-eight hours later, B cells were stained for flow cytometry with antibodies to B220 and CD138. Just prior to flow cytometry, 7AAD was added to allow exclusion of dead cells from analysis. Flow cytometry data were collected on a BD LSR II flow cytometer and analyzed using FlowJo software.

To test viral gene expression, infected cells were harvested to make lysates. For the Stat4-virus infected cells, we detected expression in the packaging cell line Plat-E as well as in sorted GFP⁺ virally-infected B cells. For the other constructs, we used unsorted, infected B cells to make lysates. Protein expression was analyzed by Western blotting using the following monoclonal antibodies: anti-Ptpn22 (clone D6D1H; Cell Signaling Technology), anti-Egr1 (clone T.126.1; Thermo Fisher Scientific), anti-Stat4 (clone C46B10; Cell Signaling Technology), anti-Blimp1 (clone 6D3; EMD Millipore), and anti-GAPDH (clone 6C5; EMD Millipore).

Statistical analysis for qPCR and ELISPOT assays were performed using GraphPad Prism software. p-Value was calculated using unpaired Student’s t-test with two-tailed p-value or with the Mann–Whitney U-test.

To gain insight into how Ets1 mechanistically regulates B cell differentiation, we assessed its genome-wide occupancy by chromatin immunoprecipitation followed by deep sequencing (ChIP-seq) using chromatin derived from mouse mature B cells. Purified splenic B cells showed enhanced phosphorylation of Ets1 based on SDS-PAGE mobility when compared to whole spleen (Figure 1A). This shift in mobility arises from calcium-induced CAM kinase-dependent serine phosphorylation (34, 60) and results in inhibition of Ets1 DNA binding (61). In order to restore Ets1-binding activity, we rested the B cells for 1–2 h, which resulted in normalization of Ets1 phosphorylation status (Figure 1A). After the resting period, purified B cells were fixed and sonicated.

B cell chromatin was immunoprecipitated with an anti-Ets1 antibody that has been used previously in chromatin immunoprecipitation (40–42). We identified 10,391 reproducible peaks of Ets1 binding, which target 8,975 genes in B cells (GEO dataset: GSE83758). Among Ets1-bound regions, approximately half are in the promoters of the genes (within 3 kb of the transcriptional start site). An additional 20% are in the introns of genes and about 27% are located in distal intergenic regions that may represent enhancers or silencers (Figure 1B). Less than 10% of Ets1-binding sites were detected in coding exons, the 5’ or 3’ UTRs of genes or within 3 kb downstream of target genes. We identified the consensus Ets1-binding motif in 68% of target sites (Figure 1C). Additional transcription factor motifs that were enriched in the Ets1-bound regions include combined ETS-RUNX sites, combined PU.1-IRF sites, IRF2 sites, STAT3 sites, and PAX5 sites (Figure 1C).

Pax5 is a key transcription factor that regulates B cell identity and limits differentiation to ASCs (62, 63). We have previously implicated Ets1 in controlling Pax5 levels in B cells (22, 29). Indeed, examination of the ChIP-seq peaks of Ets1 showed several peaks in the Pax5 gene including one in the proximal promoter and several in introns of the gene including two strong peaks in intron 5 where a B cell-specific enhancer of Pax5 has been described (64) (Figure S1 in Supplementary Material). These observations support the idea that Ets1 may directly regulate expression of the Pax5 gene. Several strong Ets1-binding peaks were also detected near the Prdm1 gene, which encodes Blimp1 (Figure S2 in Supplementary Material). Although these regions are not associated with any known regulatory elements, they are enriched in H3K4me1 and H3K27Ac marks indicative that they may represent functional regulatory elements. When examining other genes that have been previously described as Ets1 targets [such as those encoding Cd79a (Igα), H-2Aa and H-2Eb1 (isoforms of MHC II), and Nfkb1 (p50)], we found that each gene contained nearby Ets1-binding sites (Figure S3 in Supplementary Material) and, therefore, is potentially regulated by Ets1 in vivo.

To understand how Ets1 might regulate B cell responses, we analyzed pathways associated with Ets1 binding. Ets1-binding peaks were enriched in genes associated with immune response, antigen processing and presentation, immune signaling pathways, and mature B cell differentiation (Figure 1D). Interestingly, Ets1-binding peaks are located near a large number of genes involved in the BCR signaling cascade (Figure S4 in Supplementary Material). Additional non-immune pathways enriched in Ets1-binding sites included genes involved in processing of non-coding RNA, protein folding, and apoptosis (data not shown).

The ENCODE project has mapped histone modifications and transcription factor-binding sites for various human and mouse cell lines (65, 66). In mature primary mouse B cells, data are available for the H3K27me3 modification (49). In addition, two research groups have mapped histone modifications (H3K4me1, H3K4me2, H3K4me3, H3K9Ac, H3K27Ac, and H3K27me3) in primary mouse B cells (50, 51). We tested whether any of these histone marks showed enrichment near the Ets1-binding peaks as compared to random chromatin. We found that peaks of H3K4me3 (associated with promoters) were strongly enriched on either side of the peak Ets1-binding locus (Figure 1E). Similarly, peaks of H3K4me2, H3K9Ac, and H3K27Ac (associated with both promoters and enhancers) were also strongly enriched flanking the Ets1-binding sites. Weaker enrichment of H3K4me1 (associated with enhancers) was also observed flanking Ets1-binding sites, while there was no enrichment of H3K27me3 (associated with repressed genes) near Ets1-binding peaks. None of the histone marks showed specific enrichment when random chromatin regions were used (not shown). These data indicate that binding of Ets1 is strongly associated with active promoters and enhancers of genes, but is associated weakly if at all with repressed sites.

ChIP-seq datasets are also available for Pax5 in mouse mature splenic B cells (51). We compared the binding sites of Pax5 with those bound by Ets1 to identify genes that might be co-regulated. There were 4,590 overlapping peaks common between Ets1 and Pax5 (Figure 2A), representing ~44% of all Ets1-bound regions and ~45% of all Pax5-bound regions. Among the genes having peaks for both Ets1 and Pax5 are some that encode genes that are crucial for B cell biology including CD79a, ICOS ligand, Foxp1, PLCγ2, and Vav1. Ontology enrichment analysis of the identified common targets of Ets1 and Pax5 indicate that they are associated with immune system activation and effector processes (not shown). This suggests that there is a strong overlap in function of Ets1 and Pax5 in B cells.

Binding sites for Tcf3 (E2A) (53) and Irf4 (54) have also been mapped in primary B cells. We found that ~47% of E2A-binding sites overlap with those of Ets1 (Figure 2B). These overlapping binding sites include genes such as Cd19, Blnk, and Rasgrp1. With Irf4, relatively few (865) binding sites were detected in primary naïve B cells (Figure 2C). However, of these 865 sites, ~58% overlap with Ets1-binding sites, including sites in important B cell genes such as Cxcr4, Cxcr5, Lyn, Pax5, Ebf1, and Foxo1. Overall, these studies show that Ets1 is targeted to promoters and/or enhancers of genes important for B cell activation and differentiation and can potentially co-regulate such genes along with other B cell transcription factors like Pax5, E2A, and Irf4.

The presence of a binding site for a transcription factor does not always correlate with its ability to activate or repress transcription from the gene. To better identify genes in B cells whose expression directly depends on Ets1, we performed RNA-sequencing using RNA isolated from WT and Ets1^–/− splenic B cells (GEO Dataset: GSE83797). One potential complicating factor is that Ets1^−/− mice have different B cell composition in the spleen, because they lack marginal zone B cells and their follicular B cells (1) have a more activated phenotype, (2) undergo altered isotype-switching, and (3) differentiate at higher rates to ASCs than their WT counterparts (23, 25, 26, 36, 67). Therefore, to compare similar populations of B cells from WT and Ets1^−/− mice, we used flow cytometry to isolate follicular B cells (B220⁺ CD23^hi CD21^low) that were not activated (CD80^neg) and had not undergone class-switching (IgA^neg IgE^neg IgG1^neg IgG2a^neg IgG2b^neg IgG3^neg) (Figure 3A). We avoided sorting B cells based on surface levels of IgM or IgD, because we did not want to activate the cells by cross-linking the antigen receptor. Total RNA was prepared from sorted B cell subsets from each genotype and subjected to deep sequencing.

We identified a list of genes that showed at least a twofold change in expression between WT and Ets1^−/− B cells with a q-value of 0.05 or less and with a FPKM value of at least 1.0 in either genotype. Using these criteria, 484 genes showed altered expression in the absence of Ets1 (Figure 3B). Of these, the expression of 208 genes (43%) was downregulated in the absence of Ets1, suggesting that Ets1 activates the expression of those genes. The expression of 276 genes (57%) was upregulated in the absence of Ets1, suggesting that Ets1 might repress these genes (Figure 3B). Overall, the changes in expression of potential Ets1 target genes were mostly moderate (typically in the range of twofold to fourfold differences between WT and Ets1^−/−). Unexpectedly, we found that the Pax5 gene was not one of those genes whose expression was altered in the absence of Ets1 (Figure S5 in Supplementary Material). Functional annotation of the molecular pathways that are overexpressed in Ets1^−/− B cells (i.e., pathways repressed by Ets1) showed that they were predominantly associated with inflammatory responses and chemotaxis (Figure 3C; Figure S6A in Supplementary Material). In contrast, the molecular pathways that are under-expressed in Ets1^−/− B cells (pathways activated by Ets1) were associated with immune cell signal transduction and activation (Figure 3D; Figure S6B in Supplementary Material). Thus, Ets1 is an important regulator of B cell functions by both positively and negatively regulating various aspects of the immune response.

We compared the ChIP-seq dataset that contained binding sites for Ets1 to the RNA-seq dataset that showed which genes changed expression in Ets1^−/− B cells to find genes that might be directly regulated by Ets1. There were 263 genes that contained a nearby Ets1-binding peak and showed an alteration in expression in the absence of Ets1 (Figure 4A). This represents ~3% of genes containing Ets1-binding peaks. Among these genes, 137 (~52%) are activated by Ets1 and 126 (~48%) are repressed by Ets1. Examining the list of 263 genes, we identified numerous putative Ets1 target genes where SNPs have been associated with autoimmune disease susceptibility (Table 1). Additional Ets1 target genes include ones important for immune function such as Il5ra, Tlr1, Traf4, and Ltk, but that not yet been implicated as susceptibility loci for autoimmune disease. In further studies, we focused on those genes listed in Table 1, because of their known associations with immune regulation and autoimmune disease susceptibility.

We selected a subset of potential Ets1 target genes and designed primers to test their expression in qRT-PCR. The genes selected for further analysis were Egr1, Stat4, Prdm1, Slamf6, and Ptpn22 as shown in Figure 4B and Figure S7 in Supplementary Material. qRT-PCR analysis showed that Egr1, Stat4, Prdm1, and Ptpn22 all displayed a pattern of expression consistent with the RNA-seq results (Figure 4C). On the other hand, expression of Slamf6 did not appear to change in the absence of Ets1 in this analysis.

Stat4 and Ptpn22 showed decreased expression in the absence of Ets1, while Egr1 and Prdm1 are elevated. Therefore, Ets1 may promote the expression of Stat4 or Ptpn22, while suppressing expression of Egr1 or Prdm1. Alternatively, these changes in Stat4, Ptpn22, Egr1, and Prdm1 may instead represent secondary changes in gene expression that reflect the fact that Ets1^–/− B cells are primed to become activated and differentiate to ASCs, rather than being important drivers of the Ets1^–/− B cell phenotype. If Ptpn22 or Stat4 were target genes that mediated the effects of Ets1 on B cell differentiation, then we would expect that expression of Stat4 and Ptpn22 in B cells might in part mimic the effects of expression of Ets1, such as its ability to suppress ASC generation. To test this, we cloned cDNAs encoding Stat4 and Ptpn22 into a retroviral vector so that we could restore expression of these genes to Ets1^−/− B cells and overexpress them in WT B cells. WT or Ets1^−/− splenic B cells were stimulated with LPS and infected with control virus (MIGR1), virus-expressing Ets1 (MIGR1-Ets1), or viruses-expressing Stat4 or Ptpn22 (MIGR1-Stat4 or MIGR1-Ptpn22). We monitored cellular differentiation to ASCs (B220^low CD138⁺) using flow cytometry in the GFP⁺ virally infected cells.

Western blot using extracts of virally infected B cells showed that B cells infected with Ptpn22 virus overexpress Ptpn22 (Figure 5A). Retrovirally expressed Stat4 was also easily detectable in the packaging cell line (Figure 5A), but its overexpression in B cells was less obvious due to the high levels of endogenous Stat4 in this cell type (Figure 5A). Viruses encoding Stat4 or Ptpn22 reproducibly resulted in low percentages of GFP⁺ virally infected B cells (5–10%) and low intensity of GFP staining, while empty virus or virus encoding Ets1 resulted in high percentages (40–80%) of GFP⁺ cells and high intensity of GFP staining (Figure 5B). Expression of Ets1 suppresses development of B220^loCD138⁺ plasmablasts compared to B cells infected with empty vector (MIGR1) (Figure 5C). On the other hand, we found that there was a dramatic increase in the B220^loCD138⁺ plasmablast cell numbers in B cells expressing Stat4 or Ptpn22 (Figure 5D), indicating that these genes stimulate the development of ASCs.

Unlike Stat4 and Ptpn22, Egr1 and Prdm1 were upregulated in Ets1^−/− B cells by about twofold to threefold, suggesting that Ets1 represses expression of these genes. We used a retroviral vector encoding shRNAs to knockdown expression of these genes in B cells (Figure 6A) (59). Expression of shRNA against Egr1 was effective in reducing levels of Egr1 protein in stimulated B cells (Figure 6B), but did not impair formation of B220^lowCD138⁺ plasmablasts (Figures 6C–E). The Prdm1 shRNA was also able to knockdown expression, although it was less efficient in Ets1^−/− B cells (Figure 6B and data not shown). The Prdm1 shRNA did not alter ASC differentiation in WT or Ets1^−/− B cells (Figures 6C–E), despite the fact that Blimp1 is known to be essential for plasma cell generation. This is likely because sufficient Blimp1 is still expressed to allow ASC differentiation. To further assess the role of Prdm1 in the phenotype of Ets1^−/− B cells, we crossed Ets1^−/− mice to mice carrying a GFP knock in allele in the Prdm1 locus that disrupts expression of the Prdm1 gene and leads to reduced levels of full-length functional Blimp1 protein (104). Homozygous Prdm1 knockout mice carrying this allele die embryonically due to a combination of developmental defects, but heterozygous mice are viable (104). We generated Ets1^−/−Prdm1^gfp/+ mice that carry a single copy of Prdm1 and express reduced levels of Blimp1 (Figure 7A). The numbers of ASCs in these mice and control mice was quantitated using ELISPOT, which showed that reduced levels of Prdm1 was not sufficient by itself to restrain excess ASC formation in the absence of Ets1 (Figure 7B). In summary, the changes in expression of the genes we tested (Stat4, Ptpn22, Egr1 and Prdm1) cannot by themselves explain the effects of Ets1 on ASC formation.