A Chromatin Immunoprecipitation Screen Reveals Protein Kinase C (cid:1) as a Direct RUNX1 Target Gene*

RUNX1 (also known as AML1) is a DNA-binding transcription factor that functions as a tumor suppressor and developmental determinant in hematopoietic cells. Target promoters have been identified primarily through the use of differential expression strategies and candidate gene approaches but not biochemical screens. Using a chromatin immunoprecipitation screen, we identified protein kinase C (cid:1) as a direct RUNX1 target gene and demonstrate that endogenous RUNX1 binds the chromatinized protein kinase C (cid:1) promoter of U937 cells. A phy-logenetically conserved RUNX1-binding site within the PKC (cid:1) promoter binds RUNX1 in electrophoretic mobility shift analyses and confers RUNX1 responsiveness on a heterologous promoter. Changes in RUNX1 activity affect endogenous protein kinase C (cid:1) expression, and a domi-nant-negative form of RUNX1 protects U937 cells from apoptotic stimuli previously shown to be dependent on protein kinase C (cid:1) . This protection can be reversed by the ectopic expression of protein kinase C (cid:1) . Together these findings demonstrate that protein kinase C (cid:1) is a direct, downstream target of RUNX1 and links RUNX1 to a myeloid apoptotic pathway. Three make TGTGGT GGAGTTAGAGACATA), were annealed and labeled with digoxigenin dUTP (Roche). The was incubated with U937 cell extracts for 15 min at room temperature and loaded onto a 6% non-reducing mobility shift gel (Invitrogen). was transferred to Nylon membrane, positively charged (Roche), cross-linked using a Stratal-inker (Stratagene, La Jolla, CA), and developed using a digoxigenin wash and buffer chemiluminescence (Roche). Competitor analyses were performed using 50 unlabeled double-stranded oligomers. Wild-type using

Three RUNX transcription factors make up the mammalian family named for homology with the Drosophila pair-rule gene, RUNT (1). The RUNX proteins are DNA-binding transcription factors with the context-specific capability of activating or repressing gene expression (1,2). The transcription factors are important developmental determinants with oncogenic potential (1). Loss of function results in malignancy, in the cases of RUNX1 and RUNX3, and dysplasia in the case of RUNX2 (3)(4)(5)(6)(7). Gain of function also appears to have oncogenic effects, because aberrant activation of any of the RUNX genes contributes to dysregulated proliferation (8 -12). Thus, the transcription factor family has complex roles in growth and development, and either loss or gain of activity can predispose a cell to maturation defects or transformation.
Identification of RUNX target genes is essential for understanding the cellular effects of the transcription factors. RUNX1 has been the most extensively analyzed family mem-ber, and numerous target pathways have been identified (13)(14)(15). Most RUNX1 target genes have been revealed by identifying consensus motifs within candidate target promoters or by performing differential expression screens (13,15). These approaches have been invaluable for identifying RUNX1 target pathways. However, candidate gene approaches may overlook targets with one or more base deviations from the perfect TG(T/C)GGT consensus or uncharacterized regulatory elements distant from promoters regions (16,17). Differential expression strategies likely omit target genes that undergo small changes in expression or genes that change only under select culture conditions. It is, therefore, likely that many RUNX1 target genes important for cellular function remain undefined.
Chromatin immunoprecipitation (ChIP) 1 is a revolutionary biochemical technique for demonstrating direct interactions between transcription factors and their target promoters (18). The assay allows for the biochemical enrichment of chromatin fragments associated with endogenous factors and is commonly used for candidate target gene validation (18). Recently, ChIP was shown to be a potent method for identifying novel target promoters (19 -22). Immunoprecipitated chromatin fragments were cloned and sequenced to reveal transcription factor targets (19). The unbiased ChIP screen is distinct from other target identification strategies because it does not rely on changes in gene expression or the disproportionate presence of transcription factor consensus sites within known regulatory regions. The assay, thus, promises to identify a unique subset of target genes that would be missed by conventional strategies. Additionally, because ChIP exclusively identifies direct biochemical targets, the screen is insulated from the indirect and often confusing cellular effects that result from alterations of transcriptional pathways.
Using a RUNX1 ChIP screen, we identified protein kinase (PKC␤) as a direct RUNX1 target. Changes in RUNX1 activity affect PKC␤ expression and a well characterized, PKC␤-dependent apoptotic pathway of myeloid cells. Thus, RUNX1 regulation of PKC␤ is an important aspect of normal and pathological biology of myeloid cells. mg/ml geneticin (Invitrogen). Cells transduced with MigR1-based retroviral vectors were sorted to purity using a FACS Vantage cell sorter (BD Biosciences) 24 h after transduction.
Constructs-A sequence-encoding FLAG sequence was added to the 5Ј end of RUNX genes by PCR. RUNX1, RUNX1 DN (i.e. AML1a (25)), and PKC␤ (ATCC) were cloned into LXIN (Clontech) or MigR1 (24) bicistronic proviral vectors. The FLAG-RUNX DNA-binding domain construct encodes the first 177 amino acids of RUNX1 and was cloned into LXIN. The PKC␤ promoter construct encompassing bases Ϫ17 to Ϫ1003, with respect to the translational start site, was cloned into PGL2-basic vector (Promega, Madison, WI) using PCR. Complementary oligomers containing tandem repeats of the bases Ϫ562 to Ϫ532 were annealed and cloned into the PGL2-promoter vector (Promega). Mutant oligomers contain the sequence GCTAGC in place of the RUNX1 consensus site, TGTGGT.
Chromatin Immunoprecipitation Screen-The Chromatin immunoprecipitation screen was modified from the previously described strategy (19,20). Chromatin was prepared from 10 8 control U937 cells transduced with LXIN or cells transduced with LXIN expressing the FLAG-tagged RUNX1 DNA-binding domain. The fixation was performed using 1% formaldehyde for 15 min at room temperature on a rocker platform. Fixation was terminated by adding glycine to 0.125 M and washing twice in phosphate-buffered saline. Lysis and sonication to 500-bp fragments were performed as described previously (19,20). Chromatin samples were diluted 20-fold in immunoprecipitation buffer (0.1 M KCl, 20 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 0.1% Tween 20) and incubated with anti-FLAG M2 agarose (Sigma) at 4 o C, on an end-over-end platform, overnight. After extensive washing with immunoprecipitation buffer, complexes were eluted from beads using 100 g/ml of FLAG peptide (Sigma), according to the manufacturer's instructions. Samples were de-cross-linked at 65 o C, treated with proteinase K, and end-repaired as described previously (19,20). DNA fragments were purified with a QIAquick spin kit (Qiagen, Valencia, CA) and cloned into TOPO blunt vectors (Invitrogen) (19,20). Equal volumes of TOPO ligation samples were transformed into DH5␣ and plated on LB agar containing kanamycin (50 g/ml). Sequence from clones was obtained using a T7 primer.
Chromatin Immunoprecipitation-Aliquots of chromatin samples prepared as described above were immunoprecipitated using FLAG M2 agarose beads (Sigma) or anti-AML1 (Calbiochem) as appropriate. Immunoprecipitated samples were prepared for PCR as described and analyzed using primers for the PKC promoter (forward primer, ATC-CCATTGGTCATTCTGCA; reverse primer, TATTGATCTACT-GAAATCCTTCCTC) or actin promoter (forward primer, GAGCACA-GAGCCTCGCCTTT; reverse primer, AGACAAAGACCCCGCCGGTT) (18). Amplification was performed using titanium Taq polymerase (Clontech) at thermocycler settings of: one cycle at 95 o C for 10 min, followed by 35 cycles of 95 o C for 1 min, 60 o C for 1 min, and 68 o C for 1 min. Aliquots were analyzed by agarose gel electrophoresis.
Electrophoretic Mobility Shift Analyses-Complementary oligomers, including the putative RUNX1-binding motif (CCAAGAATCTTCTC TGTGGT GGAGTTAGAGACATA), were annealed and labeled with digoxigenin dUTP (Roche). The probe was incubated with U937 cell extracts for 15 min at room temperature and loaded onto a 6% nonreducing mobility shift gel (Invitrogen). Probe was transferred to Nylon membrane, positively charged (Roche), cross-linked using a Stratalinker (Stratagene, La Jolla, CA), and developed using a digoxigenin wash and buffer chemiluminescence (Roche). Competitor analyses were performed using 50ϫ unlabeled double-stranded oligomers. Wild-type oligomers were identical to the probe. GCTAGC replaced the RUNX1 consensus TGTGGT in mutant oligomers.
Apoptosis Assays-Pure cell line populations expressing the appropriate genes were obtained by selection on 1 mg/ml G418 (for LXIN or PKC LXIN) and cell sorting (for MigR1 or RUNX1 DN MigR1) was performed using a FACS Vantage cell sorter (Becton Dickinson). U937 cells (10 6 ) were plated on six-well plates. Samples were treated with 10 nM PMA for 12 h and analyzed for annexin binding using Apop-TACS flow cytometry kit (R&D Systems, Minneapolis, MN) and streptavidinallophycocyanin antibodies (Caltag Laboratories, Burlingame, CA). Samples were analyzed on a FACSCalibur (BD Biosciences). Histograms of annexin-positive cells represent the propidium iodide-negative population. TNF␣ was measured by enzyme-linked immunosorbent assay (R&D Systems).

RESULTS
Chromatin Immunoprecipitation Screen-To identify novel, direct RUNX1 gene targets, we modified a recently described, unbiased ChIP screen (Fig. 1A) (19,20). We stably expressed a FLAG-tagged version of the RUNX1 DNA-binding domain in U937 myeloid cells using a bicistronic LXIN retroviral vector. The system of anti-FLAG immunoprecipitation and FLAG peptide elution allows for the gentle isolation of protein complexes in a single precipitation (26). Control LXIN cells and FLAG- RUNX1 LXIN cells were fixed with formaldehyde, and chromatin was prepared. Samples were sonicated to equivalent sizes of ϳ500 bp (Fig. 1B). Chromatin immunoprecipitation was performed using anti-FLAG agarose beads, and RUNX1-chromatin complexes were eluted using FLAG peptide. This allowed for effective pull-down and elution of FLAG-RUNX1 from formaldehyde-fixed and sonicated samples (Fig. 1C). RUNX1-chromatin complexes were then processed to allow DNA fragments to be cloned. Transformation of DH5␣ with cloned fragments demonstrated a 10-fold enrichment in clones from cells expressing FLAG-RUNX1 compared with controls (Fig. 1D). This is consistent with enrichment for RUNX1 target-DNA fragments by the FLAG immunopurification.
PKC␤ Is a Direct RUNX1 Target-We sequenced 10 clones derived from the FLAG-RUNX1 cell line and five clones from the control cell line. Three clones from the FLAG-RUNX1 line appeared to be from promoter regions, a frequency comparable with that of previously described cloning strategies (19). Furthermore, eight of the 10 clones from the FLAG-RUNX1 cell line represented unique genomic loci, whereas only one of five control clones was unique (data not shown). We pursued in more detail the most intriguing clone, which corresponded to bases Ϫ265 to Ϫ1003 of the promoter for the PKC␤ gene (27).
PKC␤ is expressed in myeloid cells and is associated with a well characterized apoptotic pathway, making it a potentially important leukemogenic target. To validate the FLAG-RUNX association with the PKC␤ promoter, we performed conventional ChIP-PCR analyses. First, we repeated the ChIP using anti-FLAG immunoprecipitation and FLAG peptide elution, using primers designed from the cloned PKC␤ promoter fragment. PKC␤ promoter sequences were amplified specifically from the FLAG-RUNX1 samples, confirming the enrichment for these sequences by the FLAG immunopurification strategy ( Fig. 2A). No enrichment of the control ␤-actin promoter was observed. To rule out differences in chromatin preparations as a cause of the positive ChIP results, we performed immunoprecipitations on a single chromatin sample from the cell line expressing FLAG-RUNX1. Anti-FLAG antibodies enriched for PKC␤ promoter sequences significantly more than control rabbit IgG (data not shown). These PCR results indicate that FLAG-RUNX1 associates with the chromatinized PKC␤ promoter in U937 cells.
We next performed immunoprecipitations on chromatin samples from wild-type U937 cells using anti-RUNX1 antibodies to identify an association between endogenous RUNX1 and the endogenous PKC␤ gene. PCR analysis confirmed that immunoprecipitation of endogenous RUNX1 enriches chromatin fragments from the PKC␤ promoter, in comparison to a control immunoprecipitation (Fig. 2B). This demonstrates that endogenous RUNX1 associates with the chromatinized PKC␤ promoter of wild-type U937 cells.
Localization of the RUNX1 Binding Site in the PKC␤ Promoter-The observed association of RUNX1 with sequences upstream of PKC␤ suggested that the promoter might be RUNX1-responsive. We, therefore, cloned the PKC␤ promoter upstream of a firefly luciferase reporter gene and transfected control U937 cells and cells overexpressing RUNX1 or dominant-negative RUNX1 (Fig. 3C). RUNX1 increased luciferase expression by greater than 3-fold (Fig. 3A). Conversely, transfection of cells expressing dominant-negative RUNX1 resulted in reduced luciferase expression (Fig. 3B). These changes demonstrate that the PKC␤ promoter confers RUNX1 responsiveness on a heterologous reporter gene and are consistent with

FIG. 2. Chromatin immunoprecipitation analysis showing interaction of RUNX1 with the PKC␤ promoter.
Samples were analyzed by PCR using primers amplifying PKC␤ or actin promoter sequence. A, ChIP analysis of U937 cell lines used in the screen described in Fig. 1 (CONT versus FLAG-RUNX1). Immunoprecipitation was performed using anti-FLAG antibodies, and elution was performed using FLAG peptide. B, ChIP analysis of endogenous RUNX1 in wild-type U937 cells. Immunoprecipitation was performed using control IgG or anti-RUNX1 antibodies, and elution was performed using 1% SDS, 0.1 M HCO 3 . the observed association between RUNX1 and the PKC␤ promoter in the ChIP assay.
The cloned PKC␤ promoter fragment has one motif (positioned at Ϫ549 to Ϫ544 with respect to the transcriptional start site) that perfectly matches the RUNX1 consensus-binding site (TG(T/C)GGT) (27). Additionally, five sites deviating by 1 bp are scattered throughout the cloned region. To identify sequences potentially bound by RUNX1, we aligned the human and mouse sequences (Fig. 4A). Of the six potential regulatory regions, only the TGTGGT consensus is conserved between species, suggesting that it is potentially a transcription factorbinding site. To test this hypothesis, a 35-bp probe including the conserved TGTGGT motif was incubated with U937 cell nuclear extract and subjected to electrophoretic mobility shift analysis (EMSA). The extract resulted in a shift of the probe, demonstrating an interaction of a factor in U937 cells with this region (Fig. 4B). The shifted probe could be competed by an excess of unlabeled oligonucleotide but not by an oligonucleotide carrying a mutation of the TGTGGT motif to GCTAGC. Incubation of the probe with extracts plus anti-RUNX1 antibody supershifted the probe, confirming that RUNX1 is, indeed, able to interact with the TGTGGT motif (Fig. 4C). Furthermore, a tandem repeat of the wild-type oligomer placed upstream of a heterologous promoter-reporter construct is sufficient to confer RUNX1 responsiveness in transfected cells (Fig. 4D). The mutant oligomer that failed to bind RUNX1 in the EMSA lacks this activity in the transfection assay. These findings confirm the ChIP results by demonstrating that RUNX1 interacts with and activates a binding site from the PKC␤ promoter.
PKC␤ Expression Studies-The validation of RUNX1 interactions with the PKC␤ promoter suggested that the expression of endogenous PKC␤ should be affected by changes in RUNX1 activity. We used the MigR1 retrovirus to overexpress wildtype RUNX1 or a DN protein (RUNX1 DN) consisting of the DNA-binding domain without the co-regulator interaction domains (25, 28 -30). The retroviruses express the RUNX genes and green fluorescent protein from a bicistronic message allowing for the production of unselected, multiclonal populations of transduced cells by sorting. Quantitative RT-PCR analysis indicated that DN RUNX1 reduced PKC␤ expression by 50% with respect to controls expressing green fluorescent protein alone (Fig. 5A). These findings are consistent with the protein levels observed by immunoblot analysis (Fig. 5B, left). Conversely, overexpression of wild-type RUNX1 increased the endogenous PKC␤ message by greater than 3-fold (Fig. 5A). PKC␤ protein levels, accordingly, increased in these cells (Fig. 5B, right). These findings indicate that changes in RUNX1 activity affect PKC␤ expression and are consistent with RUNX1 regulating the PKC␤ promoter.
We expressed DN RUNX1 in U937 cells by retroviral transduction with the MigR1 vector and obtained pure populations by cell sorting. We hypothesized that these cells would have defective PKC␤ pathways and would be resistant to PMAinduced apoptosis. In agreement with previous studies, we found that control cells treated with PMA readily underwent apoptosis (Fig. 6A) (31)(32)(33)(34)(35)(36). By contrast, we found that cells expressing DN RUNX1 were protected from apoptosis (Fig. 6, A  and B). Similar results were obtained with stable U937 cell lines produced using LXIN and RUNX DN LXIN retroviral vectors and selected on G418-containing media (data not shown). Resistance to apoptosis was correlated with a significant reduction in TNF␣ (from 4.04 Ϯ 1.35 to 1.55 Ϯ 0.35 ng/ml, p Ͻ 0.02), consistent with TNF␣ functioning downstream of PKC␤ (data not shown). Additionally, treatment of either control cells or cells expressing RUNX DN with TNF␣ induced apoptosis (data not shown). This demonstrates that cells expressing the RUNX DN protein retain an intact apoptotic pathway. Furthermore, this finding suggests that the blunted apoptotic phenotype results from altered PMA signaling. These findings are consistent with a defective PKC␤ pathway in cells expressing DN RUNX1.
Rescue of Apoptosis-The correlation between the expression of dominant-negative RUNX1 and a defect in the PKC␤ apoptotic pathway suggests that reduced PKC␤ expression could be the direct cause of this defect. To determine whether the reductions in PKC␤ produced the resistance to PMA, we stably expressed PKC␤ in U937 cells and transduced these cells with MigR1 control retrovirus or the dominant-negative RUNX1 retrovirus. We hypothesized that the apoptotic phenotype of cells expressing dominant-negative RUNX1 would be restored by PKC␤ expression. Indeed, following PMA treatment, we observed wild-type levels of apoptosis in the DN RUNX1 cells ectopically expressing PKC␤ (Fig. 6, A and B). Immunoblot analysis confirmed that PKC␤ protein levels reflect the sensitivity to PMA (Fig. 6C). This indicates that the reduction in apoptosis resulting from expression of dominant-negative RUNX1 is a direct result of reduced PKC␤ expression. DISCUSSION Our ChIP screen revealed that RUNX1 binds the PKC␤ promoter in myeloid cells. Candidate gene approaches and differential expression screens have not previously identified PKC␤ as a RUNX1 target gene in myeloid cells. ChIP is especially well suited to target identification because it is an unbiased biochemical technique. Targets are revealed without a prior need to identify affected pathways. Furthermore, the indirect effects of transcription factors are filtered because ChIP is based on DNA-binding properties instead of an ability to disrupt cellular biology. Additionally, prior knowledge of regulatory elements is not required, and elements are not assumed to fall within convenient proximal promoter regions. Regulatory regions are potentially revealed whether they are positioned at promoters, within distant enhancer elements, or even within introns. In support of the identification of PKC␤ as a novel RUNX target gene, we have shown that the PKC␤ gene promoter is activated by RUNX1 and contains a canonical binding site to which RUNX1 specifically binds. This binding site is sufficient for RUNX1 activation of a heterologous promoter. Mutation of the RUNX1-binding site in the context of the full-length PKC␤ promoter does not fully eliminate RUNX1 responsiveness of the promoter (data not shown), indicating that additional RUNX1-binding sites likely contribute to the regulation of this gene. Nevertheless, it is clear that endogenous RUNX1 binds to the endogenous PKC␤ gene and that PKC␤ gene activity is regulated by RUNX1. Interestingly, microarray analysis of lymphoid leukemia specimens recently revealed PKC␤ as a marker in samples from patients with the TEL-AML1 (i.e. TEL-RUNX1) translocation but not other forms of acute lymphoid leukemia (37).
The PKC␤ pathway has been studied extensively in myeloid cells. Several myeloid lines, including a U937-derived line, express reduced levels of PKC␤ (38 -40). This deficiency results in blunted PMA-induced apoptosis (31)(32)(33)(34)(35). PKC␤ is the only deficient member of the PKC family identified in these cells, and the apoptotic phenotype can be restored by ectopic expression of PKC␤ (31)(32)(33)(34)(35). Our demonstration that RUNX1 regulates PKC␤ provides a plausible and novel mechanism to explain previous observations linking RUNX1 to apoptosis. For example, the RUNX1-opposing fusion core-binding factor ␤-smooth-muscle myosin heavy chain inhibits apoptosis in myeloid cells (41). Furthermore, reduction in the RUNX1-interfering AML1-ETO oncoprotein in Kasumi cells results in apoptosis, suggesting that the oncoprotein protects cells from death-inducing background genes present in these lines (42). The present finding that cells expressing DN RUNX1 essentially recapitulate the phenotype of PKC␤-deficient U937 cells and that this can be rescued by PKC␤ supports the hypothesis that PKC␤ regulation underlies at least some of the effects of RUNX1 on apoptosis The link between RUNX factors and apoptosis may not be limited to RUNX1 in myeloid cells, because loss of RUNX3 function produces gastric carcinoma, possibly by inhibiting the apoptotic response to transforming growth factor ␤ (3). The link is complicated, however, because in some settings interference with RUNX activity has been associated with increased apoptosis (43), and for unknown reasons dominant-negative RUNX1 in the form of the RUNX DNA-binding domain sometimes behaves differently than AML1-ETO (29). In these settings, the role of PKC␤ regulation may be overshadowed by competing phenotypic effects of other RUNX target genes.