Eos, MITF, and PU.1 Recruit Corepressors to Osteoclast-Specific Genes in Committed Myeloid Progenitors^{▿ †}

Antibodies.

The peptide SSGDSSLEKDSL (corresponding to amino acids [aa] 8 to 19 of mouse Eos) was used to make specific antibody against Eos in rabbits (QCB/Biosource, Hopkinton, MA). Antibodies against MITF and PU.1 were previously described (24). Anti-BRG1 antibody was described elsewhere (42). Anti-CBP antibody was a kind gift from Marc R Montminy, San Diego, CA. Other commercially available antibodies include: Flag M2 and hemagglutinin (HA) (Sigma-Aldrich, St. Louis, MO), His₆, glutathione S-transferase (GST), mSin3A, histone deacetylase 1 (HDAC1), CtBP, and Mi-2 (Santa Cruz Biotechnology Inc., Santa Cruz, CA), and histone H3 (Upstate Cell Signaling, Charlottesville, VA).

Cell culture and transfections.

NIH 3T3 cells were transfected using the calcium phosphate procedure. Expression vector pCDNA3-Flag-Eos was a kind gift from Merlin Crossley, Sydney, Australia, and was described earlier (34). The luciferase reporter constructs and expression vector for MITF and PU.1 have been described previously (22). COS-7 cells were transfected using Lipofectamine reagent (Invitrogen Life Technologies, Carlsbad, CA). DNA constructs for Flag-tagged MITF and truncation mutants were recently described (2). HA-tagged PU.1 and deletions were constructed in pCGN vector. To generate pEBG-GST-Eos, full-length Eos cDNA or various truncations of Eos cDNA were PCR amplified from pCDNA3-Flag-Eos plasmid and subcloned into the mammalian GST expression vector (pEBG). All of the amplified sequences were verified by DNA sequencing.

Culture and analysis of osteoclasts.

C57B/6J mice were used to make bone marrow precursors for the in vitro osteoclast differentiation experiments. The Mitf-vga mice were obtained from Lynn Lamoreux (Texas A&M University). The conditional “floxed” Pu.1 allele and conditions for inducible Pu.1 deletion using poly(dI-C)induction of Mx1-Cre were previously described (12). Mice containing the Pu.1 “floxed” allele were a gift from Dan Tenen (Harvard Institutes of Medicine, Harvard Medical School). Use and care of mice in this study were approved by the Ohio State University Institutional Animal Care and Use Committee. Detailed procedures for osteoclast differentiation from bone marrow-derived macrophages (BMMs) prepared from mice have been previously described (25, 26). Briefly, bone marrow was flushed from femurs and cultured for 3 days in the presence of 50 ng/ml CSF-1 on non-tissue-culture dishes. At this point, the nonadherent cell population, containing the committed osteoclast progenitors, were either harvested (zero time point controls) or cultured with 50 ng/ml of CSF-1 (a gift from David Hume, University of Queensland) and 100 ng/ml RANKL (Roche Diagnostics, Indianapolis, IN) for the indicated time periods. Tartrate-resistant acid phosphatase (TRAP)/Acp5 staining was done using a leukocyte acid phosphatase kit (Sigma).

Co-IP, GST pull-down assays, and Western blotting.

Procedures for coimmunoprecipitation (Co-IP), GST pull-down assays, and Western blotting have all been recently described (2). Production of recombinant proteins in Escherichia coli and in vitro GST pull-down assays were performed as previously described (22).

EMSA.

Electrophoretic mobility shift assays (EMSAs) were performed as described previously (21, 22). The sense strand oligonucleotides, representing the mouse Acp5 proximal sequences, was 5′-TTCTGGGGAAGTCCAGTGCTCACATGACCCA-3′. The consensus Eos binding site was mutated at two regions, M1 (GGAA to TTTT) and M2 (GTCC to CAAA).

Retrovirus production and transduction.

MSCV-FlagEos-IRES-GFP (where MSCV is murine stem cell virus, IRES is internal ribosome entry site, and GFP is green fluorescent protein) was constructed by inserting Flag-tagged Eos cDNA into XhoI-digested MSCV-IRES-GFP vector. Retrovirus packaging was performed using the Phoenix cell line. Bone marrow-derived osteoclast progenitors cultured for 2 days with 50 ng/ml CSF-1 were transferred into 12-well plates and transduced using retroviral supernatant as previously described (11). Twenty-four hours after transduction, cells were treated with the combination of CSF-1 and RANKL as described above.

siRNA knockdown of Eos.

Two separate small interfering RNA (siRNA) oligoribonucleotides (Eos siRNA1 [5′-CGGCCAACUUUCAUUGAUCtt3-′] and Eos siRNA2 [5′-CGGCCAACUUUCAUUGAUCtt3′]; lowercase letters distinguish the overhang in the siRNA design that does not have homology with the Eos sequence) directed against exons 6 and 7, respectively, were purchased from Ambion (Austin, TX), along with a control siRNA, which encodes a scrambled sequence with no particular homology to any known sequence. The combination of these two siRNAs at a concentration of 500 nM each in solution T (Amaxa Biosystems) was introduced into 5 × 10⁶ myeloid precursors using program T-020 in a nucleofector (Amaxa Biosystems).

Cells were harvested 72 h posttransfection and analyzed for the Eos knockdown by real-time PCR and Western blot analysis. Effects of Eos knockdown were analyzed by relative expression analysis of Eos targets, such as Ctsk and Acp5.

Analysis of RNA expression.

RNA was extracted using TRIzol kit (Invitrogen) and reverse transcribed by Superscript III reverse transcriptase (Invitrogen). Real-time PCR was conducted, and the relative mRNA expression level was calculated as previously described (43). The primers used for real-time PCR were picked by Oligo v4.0 software, and the sequences used are available upon request.

ChIP and ReChIP.

Chromatin immunoprecipitation (ChIP) assays were performed by the method of Luo et al. (23). Briefly, bone marrow cells were plated at a density of 3 × 10⁶ cells per 10-cm dishes and treated as indicated in the figure legends. Cells were cross-linked, and soluble chromatin with an average size of 200 to 1,000 bp was prepared by sonication. Precleared soluble chromatin was incubated with 5 μg of specific antibodies as indicated in figure legends. Immunoprecipitated DNA-protein complex was extensively washed, eluted, and decross-linked. Precipitated DNA sample was further purified by using a QIAGEN PCR purification kit according to the manufacturer's instructions (QIAGEN). For sequential ChIP (ReChIP) assays, the beads were washed after the first immunoprecipitation with Eos antibody, the immunoprecipitated chromatin was eluted in 10 mM dithiothreitol at 37°C for 30 min and diluted 50-fold in ChIP dilution buffer. This eluted immune complex was divided and immunoprecipitated with the second specific antibody indicated in the figures. Samples were analyzed by real-time PCR either by SYBR green super mix (Bio-Rad) for the Ctsk promoter or by the Roche universal probe library (Roche) probe using the Faststart TaqMan master kit (Roche) for the Acp5, Bcl-2, or c-Fms promoter. The threshold for the promoter being studied was adjusted by that of input values and represented as relative enrichment. All quantitative PCRs (qPCRs) were analyzed by melting curve analysis and agarose gels to confirm the presence of a single specific band.

Eos is downregulated during osteoclast differentiation and can repress Ctsk and Acp5 promoter activity.

Gene expression profiles of osteoclast-like cells (OCL) derived in vitro were determined using DNA microarrays (data not shown). The Ikaros family member Eos was one gene downregulated during osteoclast differentiation initiated by CSF-1 and RANKL that was selected for further analysis. The kinetics of Eos mRNA expression at various stages of OCL differentiation was validated using quantitative reverse transcription-PCR (qRT-PCR). Bone marrow-derived precursors from wild-type mice were cultured for 3 days with CSF-1 only (BMMs) and subsequently stimulated with recombinant CSF-1 and RANKL to induce osteoclast differentiation in vitro (Fig. 1A). Consistent with the microarray data, Eos mRNA expression was highest in OCL precursors treated only with CSF-1 (day 0) or during the early stages of differentiation in cells (day 0.5). Eos mRNA expression was reduced more than fivefold after 3 or 5 days of cytokine stimulation at a time when osteoclast differentiation occurs and expression of MITF/PU.1 target genes, like cathepsin K (Ctsk) and acid phosphatase 5 (Acp5)/tartrate-resistant acid phosphatase (TRAP) are robust (Fig. 1A). Eos protein expression was evaluated using an antibody generated and characterized in our lab (see Materials and Methods), an analysis that demonstrated that Eos protein expression was also significantly reduced during CSF-1/RANKL-initiated osteoclast differentiation (Fig. 1B).

Eos was previously reported to recognize the consensus Ikaros binding site GGGAATRCC and related sequences (29, 34). A potential consensus Eos recognition site overlaps with the functional PU.1 sites present in the Acp5 and Ctsk promoter sequences (Fig. 2A) (22, 27). To examine whether Eos could regulate Acp5 and Ctsk genes through these recognition motifs, transient-transfection assays with luciferase reporters were performed (Fig. 2B). These assays demonstrated that Eos expression repressed both Acp5 and Ctsk reporter activation by MITF and PU.1, either singly or in combination. The c-Fms promoter, which is regulated by PU.1 but not MITF, is not repressed by Eos overexpression (data not shown).

EMSAs were performed with oligonucleotides representing the Acp5 promoter region and recombinant Eos protein containing the N-terminal DNA binding domain (34). The experiments demonstrated that Eos could directly bind to the Acp5 target sequence, and the binding was competed by the addition of unlabeled probe to the EMSA reaction mixture (Fig. 2C). Weaker binding of Eos to the Acp5 probe compared to recombinant PU.1 was observed (Fig. 2D, lane 3 versus lane 2). Unexpectedly, incubation of both PU.1 and Eos resulted in formation of a larger complex that contained both factors (Fig. 2D, lane 4). Mutation of either half of the Eos consensus sequence (M1 [GGAA mutated to TTTT] and M2 [GTCC mutated to CAAA]; Fig. 2A) resulted in loss of Eos binding (Fig. 2D, lanes 6 and 9). PU.1 binding to Acp5 mutation M1 was also significantly reduced (Fig. 2D, lane 5). However, the larger PU.1/Eos complex formed with either Acp5 mutation M1 or M2 (Fig. 2D, lanes 7 and 10). This result indicated that formation of the larger complex might not require Eos to bind directly to DNA.

Eos directly interacts with both PU.1 and MITF through two distinct domains.

The EMSA results suggested that Eos and PU.1 might directly interact, a possibility that was tested by GST pull-down and Co-IP assays (Fig. 3). The various PU.1 and Eos constructs used for these experiments are graphically depicted in Fig. 3A. Eos shares two domains in common with other Ikaros family members: an N-terminal domain containing four zinc fingers crucial for sequence-specific DNA binding and a C-terminal domain with two zinc fingers that is involved in homo- or heterodimerization (29, 34). GST-Eos (101-230) (GST-Eos with aa 101 to 230) protein, containing zinc fingers 1 to 4, was used to pull-down various PU.1 peptides (Fig. 3B). These experiments mapped the Eos interaction region to the C-terminal DNA binding domain (DBD)/ETS domain of PU.1 (Fig. 3B). In a parallel analysis, the C-terminal domain (aa 231 to 532) of Eos fused to GST was unable to pull down either full-length HA-PU.1 or HA-DBD-PU.1 (see Fig. S1A in the supplemental material). The in vitro results were confirmed by Co-IP assays using Flag-tagged Eos (50-230) (Eos with aa 50 to 230) and HA-tagged PU.1 expression constructs expressed in COS-7 cells (Fig. 3C). Identical results were obtained when Flag-tagged full-length Eos protein was used in the Co-IP assays (data not shown).

Whether Eos also directly interacted with MITF was studied using the MITF constructs depicted in Fig. 4A. GST pull-down assays using COS-7 cell extracts cotransfected with GST-Eos and Flag-MITF expression constructs demonstrated that MITF was pulled down with full-length GST-Eos (Fig. 4B, lane 3). Further, the N-terminal region of MITF (1-218) (MITF with aa 1 to 218), but not the C-terminal basic helix-loop-helix-leucine zipper region (aa 219 to 419), formed a complex with GST-Eos (Fig. 4B, lane 4 and 5, respectively). A reciprocal analysis demonstrated that Flag-MITF (1-218) formed a complex with C-terminal Eos (aa 231 to 532, containing zinc fingers 5 and 6) but not N-terminal Eos fragments (Fig. 4C). In vitro GST pull-down assays using recombinant GST-tagged full-length MITF and His-tagged C-terminal Eos (aa 231 to 532) demonstrated that the interaction between these two proteins was direct (see Fig. S1B in the supplemental material). GST-MITF and His-Eos (231-532) could also form a complex in the presence of the Acp5 target DNA sequence as determined by EMSA (see Fig. S1C, lane 3, in the supplemental material), while His-Eos (231-532) did not form a complex with the Acp5 sequence (see Fig. S1C, lane 1, in the supplemental material).

The association of endogenous MITF, PU.1, and Eos was studied in primary bone marrow-derived progenitors by coimmunoprecipitation analysis (Fig. 4D). The analysis demonstrated that Eos, MITF, and PU.1 could be found in a complex in the primary cells (Fig. 4D). Taken together, the results demonstrated that Eos can associate with PU.1 and MITF via distinct N-terminal (aa 101 to 230) and C-terminal (aa 231 to 532) domains, respectively.

Eos/MITF/PU.1 complexes are selectively enriched at target genes in committed osteoclast progenitors.

ChIP was used to determine whether endogenous Eos is present in complexes with MITF and PU.1 at osteoclast target promoters. The regions of Ctsk and Acp5 genes that were analyzed by qPCR following the ChIP procedures are depicted in Fig. 5A. PCR primers were designed to amplify Ctsk (−1548 to −1396) and Acp5 (−210 to −1) promoter regions that have been previously shown to contain functional MITF and PU.1 binding sites (21, 32). In addition to the 5′ regions of the Ctsk and Acp5 genes, a 3′ region corresponding to internal exon/intron regions was used as a negative control (Fig. 5A, left panel). An additional negative control was a reaction in which no primary antibody was added to ChIP reaction mixtures. A representative control experiment is shown for the Ctsk gene with Eos antibody over the time course studied (Fig. 5A, right panel). Following 40 cycles of PCR, no product could be detected in the negative controls. Histone H3 was routinely used as a positive control. The same set of controls was used for all subsequent ChIP experiments presented for both Ctsk and Acp5 genes (data not shown).

The ChIP experiments demonstrated that Eos was enriched at both Ctsk and Acp5 target promoters in osteoclast progenitors grown in the presence of CSF-1 alone for 3 days (Fig. 5B). In contrast, levels of Eos rapidly decreased by 7- to 10-fold following treatment of this same population with both CSF-1 and RANKL for an additional 3 to 5 days (Fig. 5B). The amounts of MITF and PU.1 at Ctsk and Acp5 promoters remained largely unchanged after addition of CSF-1/RANKL, increasing less than twofold after 3 to 5 days of treatment with both cytokines.

Whether Eos was bound to promoters that contained only binding sites for PU.1 or MITF, but not both factors, was next addressed. The proximal c-Fms promoter contains PU.1 binding sites but is devoid of MITF binding sites (10), whereas Bcl2 is an established MITF target that lacks PU.1 sites (28). As predicted, ChIP demonstrated that in cells treated with CSF-1, only PU.1 was enriched at the proximal c-Fms promoter, but neither MITF or Eos could be detected (Fig. 5C). Similarly, MITF was enriched at the Bcl2 promoter, but neither PU.1 nor Eos was detected (Fig. 5C).

To more directly address whether the presence of both MITF and PU.1 were required for enrichment of Eos at target promoters, primary cells from genetic models with reduced levels of MITF and PU.1 proteins were examined by ChIP. Bone marrow cells derived from mice homozygous for the hypomorphic Mitf-vga9 produce 10 to 20% of wild-type MITF protein (see Fig. S2A in the supplemental material). ChIP assays performed on progenitors cultured for 3 days in the presence of CSF-1 demonstrated only three- to five-fold reduction in the levels of MITF, PU.1, and Eos at the Acp5 promoter in theses cells compared to those in the control cells (Fig. 5D). Bone marrow precursors were also obtained for mice homozygous for a previously described PU.1 conditional allele that also contained the interferon-inducible Mx1-cre transgene (12) (see Materials and Methods). PU.1 protein levels were reduced to 10 to 20% of the levels in the control in the bone marrow progenitors cultured for 3 days in the presence of CSF-1 from the PU.1^f/f;Mx1-cre⁺ mice (f for floxed) treated with poly(dI-dC) (see Fig. S2B in the supplemental material). In this genetic model, Eos levels at the Acp5 promoter were reduced sevenfold (Fig. 5D). Similar results were obtained when the Ctsk promoter was studied (data not shown). These results are consistent with the hypothesis that Eos requires the MITF/PU.1 complex for its selective association with target genes.

Eos/MITF/PU.1 complexes recruit corepressors to target promoters in the absence of RANKL signaling.

Eos and other Ikaros family members have been reported to interact with several corepressors, including Sin3, Mi-2, and CtBP (15-18, 33). ChIP assays were used to determine whether these corepressors were enriched at Ctsk and Acp5 promoters in bone marrow progenitors (Fig. 6A, top panels). For this analysis, cells grown in CSF-1 alone for 3 days were compared to cells grown for an additional 3 days with CSF-1 and RANKL, a time when visible differentiation of OCLs is first apparent and robust expression of target genes occurs (25, 26). These experiments demonstrated that corepressors CtBP, HDAC1, and Sin3A were all enriched at Ctsk and Acp5 promoters in cells grown with CSF-1 only, but the levels of their association with these target promoters were significantly reduced following 3 days of CSF-1/RANKL treatment. The Mi-2 protein was also detected at Ctsk and Acp5 promoters, but its association with these promoters upon CSF-1/RANKL stimulation was not significantly different.

The recruitment of the coactivators CBP/p300 and BRG1 was also studied, since these coactivators were previously reported to interact with both MITF and PU.1 (3, 35, 37, 45). This analysis indicated that CBP and BRG1 were not detected at Ctsk and Acp5 promoters in cells grown with CSF-1 alone. However, both of the cofactors were enriched at these target promoters in cells grown for an additional 3 days with CSF-1 and RANKL (Fig. 6A, bottom panels), concurrent with robust expression of these target genes at day 3 (Fig. 1A).

ReChIP assays were used to determine whether MITF, PU.1, Eos, and corepressors Sin3A, CtBP, HDAC1, and Mi-2 were present at the target promoters within the same cells. The first round of immunoprecipitation was carried out with an Eos antibody, and the immunoprecipitated cross-linked DNA-protein complexes were then isolated and disassociated from the beads. Aliquots of this sample were subjected to reimmunoprecipitation using antibodies against Eos, MITF, PU.1, CtBP, HDAC1, Sin3A, and Mi-2. The amount of DNA present in the ReChIP samples using these antibodies was determined by real-time qPCR and compared to the amount of DNA present in the Eos ReChIP sample (Fig. 6B). This analysis indicated that more than 80% of the recovered Eos complex also contained MITF and PU.1 at both Ctsk and Acp5 promoters. In addition, more than 60% of the Eos complex contained CtBP, HDAC1, and Sin3A at the Ctsk promoter. Similar levels of CtBP were seen at the Acp5 promoter, but the levels of both HDAC1 and Sin3A measured at the Acp5 promoter were slightly lower, approximately 40% of complexes (Fig. 6B). The level of Mi-2 complex was much lower than the other corepressor complexes, indicating that Mi-2 may not be strongly associated with the Eos/MITF/PU.1 complex.

Eos overexpression or siRNA knockdown disrupts MITF/PU.1 target gene regulation.

To further study the role of Eos as a negative regulator of MITF/PU.1 target genes, Eos was overexpressed in osteoclast progenitors via retroviral transduction. The bicistronic retroviral construct (MSCV-FlagEos-IRES-GFP) used to express Flag-tagged Eos also expressed the GFP under control of an IRES, allowing individual infected cells to be identified. Viral stocks produced from packaging cells mock transfected, transfected with MSCV-GFP alone (vector control), or transfected with Flag-tagged Eos/GFP vector were used to transduce precursors derived from wild-type mice following 2 days of growth with CSF-1. The efficiency of transduction 24 h postinfection as determined by expression of GFP was 70 to 80% for both vector control and MSCV-Flag-Eos (data not shown). Western blotting was performed to verify expression of exogenous Flag-Eos (Fig. 7A). Retroviral transduction resulted in a modest twofold increase in total Eos protein production as determined by Western blotting and densitometry (Fig. 7A).

Twenty-four hours after viral transduction, cells were treated with both CSF-1 and RANKL for an additional 3 days. Control virus-infected bone marrow precursors were able to differentiate into multinucleated (containing three or more nuclei), TRAP/Acp5-positive osteoclasts. In contrast, multinuclear, TRAP/Acp5-positive cells were rarely detected in cells infected with retrovirus encoding Eos (Fig. 7B). Osteoclast precursors infected with the control retrovirus formed approximately the same number of OCLs as mock-infected cells, while cells infected with Eos virus produced about threefold-fewer OCLs, a reduction that was statistically significant (Fig. 7B, left bar graph, P < 0.01). When only the GFP-positive OCLs were considered, a sixfold reduction was observed in control versus Eos-infected cells (Fig. 7B, right bar graph, P < 0.01).

Supporting these results, qRT-PCR demonstrated that the levels of Ctsk and Acp5 mRNA were reduced six- to eightfold in cells expressing Flag-Eos versus control cells (Fig. 7C). In contrast, Eos overexpression did not affect the expression of two known PU.1 targets, c-Fms and RANK (see Fig. S3A, top panel, in the supplemental material), the receptors for CSF-1 and RANKL, respectively (10, 19). In addition, Bcl2 expression, a target of MITF, was not affected by Eos-overexpressed cells (see Fig. S3A, bottom panel, right, in the supplemental material). Expression of the calcitonin receptor (CTR), another marker gene for osteoclast differentiation, was not affected in cells overexpressing Eos compared with controls (see Fig. S3A, bottom panel, left, in the supplemental material).

In a complementary approach, siRNA was used to knock down Eos expression in primary osteoclast progenitors. When wild-type, bone marrow-derived precursors grown in the presence of CSF-1 alone were transfected with the combination of two Eos siRNAs (see Materials and Methods), both Eos mRNA levels (see Fig. S3B in the supplemental material) and protein levels were significantly reduced 72 h posttransfection (Fig. 7D). This resulted in a 2.5- to 3-fold increase in Acp5 and Ctsk mRNA expression of both Acp5 and Ctsk genes compared to scrambled siRNA-transfected cells (Fig. 7D). Expression of Bcl2 was not affected by Eos siRNA treatment (see Fig. S3B in the supplemental material). In the RAW264.7 cell culture model, cotransfection of the individual Eos siRNAs along with an MITF expression vector resulted in a fourfold increase in expression of basal levels of target genes, even though Eos protein levels were decreased only about 50% by the single siRNAs (see Fig. S3C in the supplemental material).