Tyrosine phosphorylation of HoxA10 decreases DNA binding and transcriptional repression during interferon gamma -induced differentiation of myeloid leukemia cell lines.

The DNA binding affinity of HoxA10 is increased by partnering with Pbx proteins. A consensus sequence for Pbx1-HoxA10 DNA binding has been derived, but genuine target genes have not been identified. We noted that the derived Pbx-HoxA10 DNA-binding consensus is similar to a repressor element in the CYBB promoter. The CYBB gene, which encodes the respiratory burst oxidase component gp91(phox), is expressed only in myeloid cells that have differentiated beyond the promyelocyte stage. In these studies, we demonstrate that interferon gamma (IFN-gamma)-induced differentiation of myeloid cell lines abolishes in vitro Pbx-HoxA10 binding to either the derived consensus or the similar CYBB sequence. We also demonstrate that HoxA10, overexpressed in myeloid cell lines, represses reporter gene expression from artificial promoter constructs with Pbx-HoxA10 binding sites. We determine that HoxA10 has endogenous repression domains that are not functionally altered by IFN-gamma treatment. However, IFN-gamma-induced differentiation of myeloid cell lines leads to HoxA10 tyrosine phosphorylation, which decreases in vitro DNA binding to Pbx-HoxA10 binding sites. Therefore, these investigations identify the CYBB gene as a potential target for HoxA10 and define repression of genes expressed in mature myeloid cells as a novel role for HoxA10 during myeloid differentiation.

The 39 human and murine HOX genes encode homeodomain transcription factors necessary for embryogenesis (1,2) and definitive hematopoiesis (3). Genes of the HOX A and B paralog groups are preferentially expressed in CD34ϩ bone marrow progenitor cells and are activated 3Ј to 5Ј during hematopoiesis (3). Expression of 3Ј HOX A and B genes increases in early CD34ϩ cells and decreases in CD34ϩ committed progenitors. In contrast, transcription of the 5Ј genes (i.e. HOX9-13) is invariant in CD34ϩ cells, and decreases in mature phagocytes (3).
In comparison with normal, mature myeloid cells, expression of HOXA10 is increased in acute myeloid leukemia, chronic myeloid leukemia, or myelodysplasia (4). Consistent with this, overexpression of HoxA10 in murine bone marrow induces a myeloproliferative disorder, which evolves to acute leukemia (5). These results suggest that HoxA10 is involved in progres-sion of myeloid differentiation. Although the most abundant HoxA10 transcript in human myeloid cells encodes a 406-amino acid protein (predicted molecular mass of 50 kDa) (6), alternatively spliced transcripts have been described at various stages of murine embryogenesis (7) and in transformed cell lines (4). In myeloid leukemia cell lines, a HoxA10 transcript is present encoding a protein that initiates 20 amino acids Nterminal to the homeodomain (6). It is hypothesized that the 80-amino acid (15-kDa) "short A10" may contribute to immortalization of myeloid cell lines, although the role of short A10 in normal myelopoiesis is unknown.
Similar to the other Abd-like Hox proteins (Hox9 -13), DNA binding affinity of HoxA10 is increased by partnering with Pbx proteins (8). Although consensus sequences for Pbx-HoxA10 binding have been derived (8 -10), genuine Pbx-HoxA10 target genes have not been identified. It has been hypothesized that HoxA10 regulates myeloid differentiation by activating transcription of genes that are necessary for progression of myelopoiesis. Conversely, HoxA10 might repress transcription of genes characteristic of differentiated myeloid cells, or HoxA10 might activate transcription at one stage of myelopoiesis and repress transcription at another, as has been described for homologous Drosophila proteins, during embryogenesis (11).
We have been studying regulation of genes encoding the respiratory burst oxidase proteins, gp91 phox (the CYBB gene) (12) and p67 phox (the NCF2 gene) (13). These genes are transcribed in cells differentiated beyond the promyelocyte stage and therefore provide a model for gene regulation during late myelopoiesis. Both the CYBB and NCF2 genes contain sequences similar to the Pbx-HoxA10 binding consensus sequence. One of these CYBB sequences is within a previously described repressor element (14), which binds the CCAAT displacement protein (CDP) 1 in electrophoretic mobility shift assays (EMSA) with HeLa or K562 nuclear proteins (14,15). In NIH 3T3 cells, overexpression of CDP represses an artificial promoter construct containing the CYBB element (16).
In contrast, our previous investigations demonstrated that CDP is not a component of the complex binding to the CYBB repressor element in EMSA with nuclear proteins from the myeloid lines PLB985 and U937 (17). Since HoxA10 mRNA is present in these myeloid cell lines, but not in HeLa or K562 cells (4), our current studies investigate the hypothesis that, in committed myeloid progenitors, HoxA10 interacts with repressor elements and suppresses transcription of some myeloid-specific genes until later stages of myelopoiesis. Previously, we demonstrated that IFN-␥-induced myeloid differentiation decreases in vitro protein binding to the CYBB repressor element, coincident with increased CYBB transcription (17). Therefore, we also investigate the effect of IFN-␥-induced differentiation on HoxA10 DNA binding and functional activity in myeloid cells.

Plasmids and Site-directed Mutagenesis: Reporter Constructs and
Plasmids for Protein Expression-Artificial promoter/reporter constructs were generated as described previously (18), in the minimal promoter/reporter vector, p-TATACAT (19) (obtained from Dr. A. Kraft, University of Colorado, Denver). Constructs were generated with three copies (in the forward orientation) of the consensus sequence for HoxA10-Pbx binding (p-a10TATACAT) (8) or four copies (in the forward direction) the Ϫ94 to Ϫ134 bp sequence from the CYBB promoter (p-cybba10TATACAT). This CYBB promoter sequence has previously been demonstrated to function as a repressor element in myeloid cell lines (16). An artificial promoter construct with five copies of the Gal4 DNA binding site and a minimal promoter from the thymidine kinase gene linked to a chloramphenicol acetyltransferase (CAT) reporter (p-gal4TKCAT) was obtained from T. Gabig (Indiana University, Indianapolis).
The cDNAs for human HoxA10 and "short A10" were obtained from C. Largman (University of California, San Francisco) and subcloned in to the mammalian expression vector pSR␣ (20). The human Pbx1 cDNA and a FLAG epitope-tagged HoxA10 cDNA were obtained from M. Cleary (Stanford University, Stanford, CA) (8) and also subcloned into the pSR␣ vector. The cDNAs for HoxA10 and "short A10" were also subcloned into the vector pM2(GAL4), for expression as a fusion protein with the DNA binding domain of GAL4 (21).
Cell Culture-All cell lines were of human origin. The promyelocytic leukemia cell line PLB985 was obtained from Thomas Rado (University of Alabama, Birmingham). The myelomonocytic cell line U937 (23) was obtained from Andrew Kraft (University of Colorado, Denver). Cell lines were maintained and differentiated as described (17,18). U937 cells were treated with 200 units/ml human recombinant IFN-␥ (Roche Molecular Biochemicals).
EMSAs-Nuclear extract proteins were prepared by the method of Dignam et al. (24), with protease and phosphatase inhibitors, as described (18). Oligonucleotides probes were prepared, and EMSA and antibody supershift assays were performed as described (18). Antiserum to HoxA10 (not cross-reactive with other Hox proteins) was obtained from Covance Research Products (Richmond, CA). Pbx antibodies (C-20 and P-20), blocking peptides, and an antibody to CDP were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Rabbit polyclonal antiserum raised to whole CDP protein (purified from HeLa cells) was a generous gift of Ellis Neufeld (Boston Children's Hospital, Boston, MA).
In Vitro Translated Proteins and Tyrosine Dephosphorylation-In vitro transcribed HoxA10, short A10, and Pbx1 mRNA were generated from linearized template DNA, using the Riboprobe System, according to the manufacturer's instructions (Promega, Madison, WI). In vitro translated proteins were generated in rabbit reticulocyte lysate, according to the manufacturer's instructions (Promega). Control (unprogrammed) lysates were generated in similar reactions in the absence of input RNA. In vitro translated proteins and nuclear proteins were tyrosine-dephosphorylated with Yop protein-tyrosine phosphatase (New England Biolabs, Beverly, MA). Proteins (either 10 l of in vitro translated protein or 2 g of nuclear proteins) were incubated 30 min at 30°C, in a 20-l reaction volume with 50 units of Yop and 1ϫ reaction buffer, according to the manufacturer's instructions. Control proteins were incubated, similarly, in 1ϫ reaction buffer without Yop.
EMSA with the in vitro translated proteins was performed as described (18). Binding assays with in vitro translated proteins and the dscybbA10 oligonucleotide were performed in the presence of a 200-fold molar excess of the urccaat oligonucleotide. The urccaat oligonucleotide competes for binding of CP1, found in reticulocyte lysate, to the dscybbA10 probe.
In other experiments, U937 cells were transfected with 30 g of pSR␣, HoxA10/pSR␣, or HoxA10(FLAG)/pSR␣. The cells were incubated for 48 h at 37°C, 5% CO 2 and harvested for extraction of nuclear proteins or total cellular RNA.
Immunoprecipitation and Western Blotting-Western blots were performed with 30 g of nuclear proteins extracted from U937 cells, with or without 48 h IFN-␥ incubation. Proteins were separated on 12% SDS-PAGE, transferred to nitrocellulose, and probed with HoxA10 antiserum (Covance Research Products) or control rabbit pre-immune serum, and proteins were detected by chemiluminescence, according to the manufacturer's instructions (Amersham Pharmacia Biotech). In other experiments, Western blots of nuclear proteins from U937 cells transiently transfected with pSR␣ or FLAG epitope-tagged HoxA10/ pSR␣ were performed with anti-FLAG antibody.
To determine HoxA10 phosphorylation state, U937 cells, with or without 48 h of IFN-␥ differentiation, were incubated for 4 h at 37°C with 32 P-orthophosphate, as described (27). Cells were lysed, under denaturing conditions, and proteins were immunoprecipitated for 4 h at 4°C, with 2 l of HoxA10 antiserum or 2 l of control rabbit preimmune serum, followed by a 1-h incubation with 30 l of 50% staph protein A-Sepharose bead slurry, as described (27). Immunoprecipitated proteins were washed with radioimmune precipitation buffer, eluted in SDS sample buffer, and identified by autoradiography of 12% SDS-PAGE.
Immunoprecipitation experiments were also performed with 100 g of nuclear proteins extracted from U937 cells, with or without 48-h IFN-␥ incubation. Nuclear proteins were diluted into radioimmune precipitation buffer, with protease and phosphatase inhibitors, as described (28), and incubated with either 1 l of anti-phosphotyrosine antibody (4G10; Upstate Biotechnology, Inc., Lake Placid, NY) or irrelevant antibody (mouse anti-rabbit IgG), for 4 h at 4°C, followed by a 1-h incubation with 15 l of 50% staph protein A-Sepharose bead slurry. Beads were washed with radioimmune precipitation buffer, and proteins were eluted in SDS sample buffer, separated on 12% SDS-PAGE, and transferred to nitrocellulose. Blots were probed with HoxA10 antibody (Covance Research Products) or control rabbit preimmune serum.
Similar experiments were performed to determine tyrosine phosphorylation of in vitro translated HoxA10. In vitro translated proteins (10 l), with or without Yop treatment, and control lysate were diluted into radioimmune precipitation buffer and immunoprecipitated as described above. The proteins were detected by autoradiography of SDS-PAGE.
RNA Extraction and Northern Blotting-Total cellular RNA was extracted from U937 cells, as described (17), 48 h after transfection with HoxA10/pSR␣, FLAG epitope-tagged HoxA10/pSR␣, or control pSR␣. Northern blots were performed with 20 g of total cellular RNA, as described (17). Probes for Northern blots were generated by random primer labeling of cDNAs encoding human gp91 phox , p67 phox and chicken ␥-actin, as described (17).

HoxA10 DNA Binding Decreases during IFN-␥-induced
Myeloid Differentiation-The consensus sequence for HoxA10 binding to DNA as a heterodimer with Pbx1 (8) consists of a Pbx1 site adjacent to a HoxA10 site: 5Ј-ATGATTTATGA-3Ј (HoxA10 site in boldface type, the Pbx site in italics) (9). Inspection of the promoter regions of the genes encoding gp91 phox (the CYBB gene) and p67 phox (the NCF2 gene) identified sequences similar to the Pbx-HoxA10 consensus (Fig. 1A). The CYBB sequence includes the core sequence preferred in HoxA10 binding site selection experiments (TTAT) and the Pbx consensus (7). However, unlike sequences identified by binding site selection, there was overlap of the Pbx and Hox binding cores. The NCF2 sequence includes an alternative HoxA10 core, identified by binding site selection (TAAT) (7), and a Pbx consensus, altered at position 3.
The Pbx-HoxA10-like sequence in the CYBB promoter is within a 30-bp region previously demonstrated to function as a repressor element in undifferentiated cells (16). In EMSA with nuclear proteins from the promyelocytic leukemia cell line PLB985, an unidentified, specific protein complex interacts with this CYBB repressor element (referred to as complex A) (14,15). We previously demonstrated that IFN-␥-induced dif-ferentiation of PLB985 cells abolishes in vitro binding of complex A to the repressor element, coincident with an increase in CYBB transcription (17).
We hypothesized that HoxA10 is a component of complex A binding the CYBB repressor element and that myeloid differentiation decreases HoxA10 DNA binding. To pursue this hypothesis, we investigated whether in vitro protein binding to the derived Pbx-HoxA10 consensus sequence decreases during IFN-␥-induced myeloid differentiation. In these experiments, we used U937 cells, a monocyte-committed cell line that expresses HoxA10 mRNA (4). IFN-␥ treatment of U937 cells results in monocyte differentiation and increased CYBB and NCF2 transcription (17). EMSAs were performed, using nuclear proteins from U937 cells and a radiolabeled probe with the derived consensus sequence for Pbx-HoxA10 binding (referred to as dsA10 oligonucleotide) (Fig. 1B). A complex binds to dsA10 that is of similar mobility to complex A generated by U937 nuclear proteins and the homologous CYBB sequence FIG. 1. Sequences in the CYBB and NCF2 genes, similar to the derived Pbx-HoxA10 consensus, have cross-competitive binding specificities. A, sequence analysis identifies sequences from the CYBB and NCF2 promoters similar to the derived consensus for Pbx1-HoxA10 binding. The HoxA10 binding core is indicated in boldface type, and the Pbx core is shown in italics. Note that, in the CYBB promoter sequence, there is a 1-bp overlap between the HoxA10 and Pbx core sequences. B, a specific protein complex, binding to the derived Pbx-HoxA10 binding site consensus, decreases during IFN-␥-induced U937 differentiation and is competed for by the sequences from the CYBB and NCF2 genes. EMSA was performed with the dsA10 probe and nuclear proteins from U937 cells (2 g) without (lanes 1-5) and with (lane 6) 48-h IFN-␥ differentiation in the presence of unlabeled, synthetic oligonucleotide competitor (200-fold molar excess). Lane 1, no competitor; lane 2, homologous dsA10 oligonucleotide; lane 3, dscybbA10 oligonucleotide; lane 4, dsncf2A10 oligonucleotide; lane 5, urccaat unrelated oligonucleotide; lane 6, no competitor. The arrowhead indicates the A 1 complex. C, binding of a specific protein complex (complex A) to the dscybbA10 probe is decreased during IFN-␥-induced U937 differentiation. EMSA was performed with the dscybbA10 probe and nuclear proteins from U937 cells (2 g) without (lane 1) and with (lane 2) 48-h IFN-␥ differentiation. The arrowhead represents binding of specific complex A (17), and the asterisk shows binding of protein complex, previously demonstrated to represent the classical CCAAT binding complex, CP1 (14). D, binding of complex A to the dscybbA10 probe is competed for by the Pbx-HoxA10 binding consensus and other similar sequences. EMSA was performed with the dscybbA10 probe and nuclear proteins from U937 cells (2 g) in the presence of unlabeled, synthetic oligonucleotide competitor (200-fold molar excess (the dscybbA10 oligonucleotide). Binding of this complex to the dsA10 probe (referred to as complex A 1 ) is competed for by excess, unlabeled, homologous oligonucleotide and by oligonucleotides with the similar sequences from the CYBB and NCF2 genes (the dsncf2A10 oligonucleotide) but not by several dissimilar oligonucleotides (Fig. 1B).
Binding of both complex A 1 to the dsA10 probe and of A to the dscybbA10 probe is decreased in EMSA with nuclear proteins from IFN-␥-treated U937 cells, in comparison with nuclear proteins from undifferentiated cells (Fig. 1, B and C). These results are consistent with our previous data with nuclear proteins from PLB985 cells and the dscybbA10 probe (17). In EMSA with U937 nuclear proteins, binding of the A complex to the dscybbA10 probe was competed for by homologous oligonucleotide and the dsncf2A10 and dsA10 oligonucleotides, but not by unrelated oligonucleotides or by dscybbA10 or dsncf2A10 oligonucleotides with mutations in the TTAT or TAAT sequences (Fig. 1D).
However, there are differences in the complexes shifted by the dsA10 and dscybbA10 probes. Binding of complex A 1 to the dsA10 probe is of lower affinity than complex A binding to the dscybbA10 probe (in terms of pmol of shifted probe/g of nuclear proteins; Fig. 1, compare B and C). Also, the dsA10 probe binds several higher mobility protein complexes, in addition to A 1 , which are competed for by homologous oligonucleotide or dscybbA10 but not dsncf2A10 (Fig. 1B). These higher mobility complexes may represent proteins encoded by alternatively spliced HoxA10 transcripts (6, 7). Alternatively, these complexes may represent proteolytic fragments of HoxA10, although U937 nuclear proteins used in these experiments were tested for integrity in other EMSAs with characterized binding sites.
To determine if HoxA10 is a component of either complex A 1 binding to the dsA10 probe or complex A binding to the dscybbA10 probe, we used an antibody raised to the HoxA10 C terminus that does not recognize other Hox proteins. In preliminary experiments, we determined that this antibody does not recognize recombinant CDP (data not shown). In EMSA with the dsA10 probe and nuclear proteins from U937 cells, binding of complex A 1 is disrupted by HoxA10 antibody but not preimmune serum ( Fig. 2A). Antibody to HoxA10 also disrupts binding of complex A to the dscybbA10 probe (Fig. 2B). Identical results were obtained with nuclear proteins from PLB985 cells (not shown). In addition, HoxA10 antibody disrupts two high mobility complexes binding the dsA10 probe, suggesting that they contain the HoxA10 C terminus, consistent with previously described, alternatively spliced messages (6,7). Interestingly, the dscybbA10 probe does not bind high mobility species, cross-immunoreactive with HoxA10.
EMSAs were also performed to determine if Pbx1 is a component of either complex A 1 (binding to dsA10) or complex A (binding to dscybbA10). EMSAs were performed with U937 nuclear proteins and Pbx antibodies, with or without blocking peptides. In EMSA with an antibody to a peptide in the N terminus of Pbx1, inconsistent disruption of both complexes A and A 1 is demonstrated (not shown). However, in EMSA with an antibody to a peptide in the Pbx C terminus, binding of complex A 1 to the dsA10 probe and of complex A to the dscybbA10 probe is disrupted (Fig. 1, C and D). However, these results do not exclude the possibility that additional, uniden-FIG. 2. HoxA10 from U937 nuclear proteins interacts in vitro with DNA probes containing the derived Pbx-HoxA10 DNA-binding consensus or the similar CYBB promoter sequence. A, a U937 nuclear protein, cross-immunoreactive with HoxA10, binds in vitro to the dsA10 probe. EMSA was performed with the dsA10 probe and nuclear proteins from U937 cells (2 g), preincubated with the following. Lane 1, rabbit preimmune serum (2 l); lane 2, HoxA10 specific rabbit serum (2 l). The arrowhead indicates complex A 1 , and the double arrows indicate higher mobility complexes, also cross-immunoreactive with HoxA10. B, a U937 nuclear protein, cross-immunoreactive with HoxA10, binds in vitro to the dscybbA10 probe. EMSA was performed with the dscybbA10 probe and nuclear proteins from U937 cells (2 g), preincubated with the following. Lane 1, rabbit preimmune serum (2 l); lane 2, HoxA10-specific rabbit serum (2 l). The arrowhead indicates complex A; the asterisk indicates CP1 binding to a CCAAT box in the probe. C, a U937 nuclear protein, cross-immunoreactive with Pbx1, binds in vitro to the dsA10 probe. EMSA was performed with the dsA10 probe and nuclear proteins from U937 cells (2 g), preincubated with the following. Lane 1, Pbx antibody (2 g) and blocking peptide (1 l); lane 2, Pbx antibody alone (2 g). The arrowhead indicates complex A 1 . D, a U937 nuclear protein, cross-immunoreactive with Pbx1, binds in vitro to the dscybbA10 probe. EMSA was performed with the dscybbA10 probe and nuclear proteins from U937 cells (2 g), preincubated with the following. Lane 1, Pbx antibody (2 g) and blocking peptide (1 l); lane 2, Pbx antibody alone (2 g). The arrowhead indicates complex A; the asterisk indicates CP1 binding to a CCAAT box in the probe.
tified proteins bind to dsA10 or dscybbA10 and participate in these complexes.
Consistent with our previously reported EMSA with PLB985 nuclear proteins and the dscybbA10 probe (17), neither complex A binding to the dscybbA10 probe nor complex A 1 binding to the dsA10 probe is disrupted by either of two antibodies to CDP (not shown).
HoxA10 Represses Transcription of Artificial Promoter Constructs with Pbx-HoxA10 Binding Sites-To determine if HoxA10 represses transcription in myeloid cells, U937 cells were transfected with artificial promoter constructs containing multiple copies of the derived Pbx-HoxA10 binding site (p-a10TATACAT) or the repressor element from the CYBB gene (p-cybba10TATACAT). Transfectants with these constructs demonstrated increased promoter activity that is statistically significant in comparison with control, empty vector (p-TATA-CAT) transfectants (p Ͻ 0.05 for both) (Fig. 3A). Reporter gene expression from U937 transfectants with the p-a10TATACAT construct is significantly repressed by overexpression of HoxA10 (p ϭ 0.012, n ϭ 6). Overexpression of HoxA10 also represses the p-cybba10TATACAT construct in U937 transfectants (p ϭ 0.001, n ϭ 5). However, HoxA10 has no significant effect on p-TATACAT control vector expression (p Ͼ 0.5, n ϭ 6) (Fig. 3A).
FIG. 3. In undifferentiated U937 cells, HoxA10 represses transcription from artificial promoter constructs containing a Pbx-HoxA10 binding site. A, overexpression of HoxA10 in U937 cells represses reporter gene expression from artificial promoter constructs with either the derived Pbx-HoxA10 binding consensus or the similar CYBB promoter sequence. U937 cells were transfected with an artificial promoter construct containing either three copies of the derived consensus for Pbx-HoxA10 binding, a minimal promoter, and a reporter (p-a10TATACAT), four copies of the similar CYBB promoter sequence (p-cybba10TATACAT), or control vector (p-TATACAT) (70 g); a vector to overexpress either HoxA10 (HoxA10/pSR␣), short A10 (SA10/pSR␣), Pbx (Pbx/pSR␣), control vector (pSR␣) (30 g), or HoxA10 and Pbx1 (15 g each); and a vector to control for transfection efficiency (CMV/␤gal) (15 g). Results are reported as absolute CAT activity (in cpm), and each experiment was repeated at least four times. B, overexpression of HoxA10 in IFN-␥-treated U937 cells does not repress transcription from artificial promoter constructs with either the derived Pbx-HoxA10 binding consensus or the similar CYBB promoter sequence. U937 transfections were performed as in A, except that the transfectants were treated with IFN-␥ (200 units/ml) for 48 h. Note the difference of the x axis scale in comparison with Fig. 3A. C, HoxA10 either has endogenous repression domains or recruits repressor proteins. U937 cells were transfected with an artificial promoter construct with five copies of the DNA-binding site for the GAL4 transcription factor (p-gal4TKCAT) (30 g); a vector to over express the DNA-binding domain of GAL4 as a fusion protein with HoxA10 (A10Gal4) or short A10 (sA10Gal4) or vector control (20 g); and a vector to control for transfection efficiency, p-CMV/␤gal (15 g). Results were reported as absolute CAT activity (in cpm), and each transfection was performed at least four times. D, increased amounts of HoxA10 fusion protein, relative to reporter construct, resulted in increased reporter repression in U937 transfectants. U937 cells were transfected with an artificial promoter construct with five copies of the DNA-binding site for the GAL4 transcription factor (p-gal4TKCAT) (3 g); a vector to overexpress the DNA-binding domain of GAL4 as a fusion protein with HoxA10 (A10Gal4) or short A10 (sA10Gal4) or vector control (20 g); and a vector to control for transfection efficiency p-CMV/␤gal (15 g). Results were reported as absolute CAT activity (in cpm), and each transfection was performed three times. Transfectants were assayed with and without 48-h IFN-␥ treatment. Note the difference in the x axis scale in comparison with Fig. 3C.
Since the "short A10" form of HoxA10, present in myeloid cell lines, contains the DNA-binding homeodomain, we investigated whether overexpressed short A10 represses transcription. U937 cells were co-transfected with either p-a10TATACAT, p-cybbTATACAT, or control p-TATACAT and a vector to overexpress short A10. Short A10 also significantly represses reporter gene expression from the p-a10TATACAT and p-cybba10TATACAT constructs (p Ͻ 0.05 for both) but less than full-length HoxA10 (Fig. 3A).
In contrast, overexpression of Pbx1, in U937 cells co-transfected with either p-a10TATACAT, p-cybba10TATACAT, or control p-TATACAT, did not significantly alter reporter gene expression (Fig. 3A). However, in U937 cells co-transfected with either p-a10TATACAT or p-cybba10TATACAT and vectors to overexpress both HoxA10 and Pbx1, repression of reporter gene expression is significantly greater than with HoxA10 alone (Fig. 3A, difference in reporter gene activity with HoxA10 versus HoxA10 plus Pbx, p ϭ 0.015 for p-a10TATACAT and p ϭ 0.026 for p-cybba10TATACAT). Short A10 does not include the HoxA10 Pbx1 interaction domain (6). Consistent with this, repression of the two HoxA10-Pbx-containing constructs in U937 transfectants overexpressing short A10 and Pbx1 is not significantly different from the repression with short A10 alone (data not shown).
Since IFN-␥ treatment of U937 cells decreases Pbx-HoxA10 binding to both the derived Pbx-HoxA10 consensus sequence and the CYBB repressor element, we investigated whether overexpressed HoxA10 represses transcription in IFN-␥treated U937 cells. U937 cells were co-transfected with p-a10TATACAT, p-cybba10TATACAT, or empty vector control and a vector to overexpress HoxA10, short A10, Pbx1, or empty vector control. U937 transfectants incubated for 48 h with IFN-␥ (200 units/ml) were compared with transfectants incubated for the same time without IFN-␥. Treatment with IFN-␥ significantly increases expression from the p-a10TATACAT and p-cybba10TATACAT constructs (p Ͻ 0.05) but not p-TATA-CAT control transfectants (Fig. 3B).
In contrast to undifferentiated U937 transfectants, overexpression of HoxA10 did not significantly repress reporter gene expression from these constructs in IFN-␥-treated cells, with or without Pbx1 (p Ͼ 0.40 for all combinations in comparison with empty expression vector control). However, overexpressed short A10 represses reporter gene expression from p-a10TATACAT and p-cybba10TATACAT constructs in IFN-␥treated U937 transfectants (Fig. 3B). Reporter gene activity in short A10 overexpressing U937 cells co-transfected with p-a10TATACAT is decreased 78.3% without versus 64.1% with IFN-␥ treatment. In U937 transfectants with p-cybbTATACAT, overexpression of short A10 decreases reporter activity 49.3% without versus 88.9% with IFN-␥ treatment. Overexpression of short A10 does not repress p-TATACAT reporter expression in U937 cells with IFN-␥ treatment.

HoxA10 Contains Transcriptional Repression Domains That Are Not Functionally Altered by IFN-␥-To determine if
HoxA10 protein possesses endogenous repression domains, the full-length protein and short A10 were expressed as fusion proteins with the DNA binding domain of GAL4 (A10gal4DB and SA10gal4DB, respectively). U937 cells were co-transfected with these fusion protein constructs (or empty gal4DB vector control) and an artificial promoter construct with multiple copies of a GAL4 DNA-binding site, linked to a minimal promoter and a CAT reporter (p-gal4TKCAT) (20 g). Reporter gene expression is significantly repressed by overexpression of A10gal4DB (p ϭ 0.00029, n ϭ 11, Fig. 3C). This repression is not significantly altered by 48 h of IFN-␥ treatment (p ϭ 0.000017, n ϭ 8; p value for reporter activity with or without IFN-␥ was 0.949). However, overexpression of SA10gal4DB does not significantly repress reporter gene activity in U937 transfectants, with or without 48 h of IFN-␥ (without IFN-␥, p ϭ 0.50, n ϭ 7; with IFN-␥, p ϭ 0.77, n ϭ 11; p value for reporter activity with or without IFN-␥ was 0.65, Fig. 3C). U937 transfections were also performed with one-tenth the amount of reporter plasmid and the same amounts of A10gal4DB, SA10gal4DB, and control gal4DB plasmids. In these experiments, A10gal4DB repressed reporter expression to a greater extent than in the previous transfections with 10-fold more input reporter plasmid DNA (Fig. 3D).
Overexpressed HoxA10 Repressed Endogenous gp91 phox and p67 phox Expression in U937 Cells-RNA isolated from U937 transfectants was analyzed by Northern blot to determine if overexpression of HoxA10 was associated with decreased abundance of gp91 phox or p67 phox mRNA. U937 cells were transfected with either control pSR␣ expression vector, HoxA10/ pSR␣, or FLAG epitope-tagged HoxA10/pSR␣. Total cellular RNA from the U937 transfectants was analyzed for abundance of gp91 phox , p67 phox , or ␥-actin control mRNA by serially probing Northern blots. Message abundance of both gp91 phox and p67 phox was decreased in U937 transfectants with either HoxA10/pSR␣ (Fig. 4A) or FLAG epitope-tagged HoxA10/pSR␣ (not shown). Both tagged and full-length, untagged proteins were overexpressed because of the possibility that epitope tagging alters the function of HoxA10.
To demonstrate that HoxA10 protein was expressed in the U937 transfectants, nuclear proteins from epitope-tagged HoxA10/pSR␣ and control pSR␣ transfectants were analyzed by Western blot. Nuclear proteins from transfectants with HoxA10/pSR␣, but not control pSR␣, demonstrated a 50-kDa FIG. 4. Overexpressed HoxA10 represses gp91 phox and p67 phox expression in U937 cells. A, overexpression of HoxA10 in U937 cells decreases abundance of gp91 phox and p67 phox mRNA. U937 cells were transfected with HoxA10/pSR␣ or control pSR␣ (30 g) and harvested 48 h later for extraction of total cellular RNA. RNA (20 g) was analyzed by Northern blot, probed for gp91 phox , p67 phox , and ␥-actin mRNA as indicated. U937 cells transfected with HoxA10/pSR␣ demonstrated decreased gp91 phox and p67 phox mRNA abundance in comparison with control transfectants. Blots were probed for ␥-actin mRNA to control for loading. B, overexpressed HoxA10 is detected by Western blot in U937 transfectants. U937 cells were transfected with FLAG epitope-tagged HoxA10/pSR␣ or control pSR␣ (30 g) and harvested 48 h later for extraction of nuclear proteins. Nuclear proteins (30 g) were separated by 12% SDS-PAGE and Western blots performed with anti-FLAG antibody. epitope-tagged protein by Western blot (Fig. 4B). No immunoreactive protein species were detected when the blot was probed with irrelevant antibody (mouse anti-rabbit IgG) (not shown).
HoxA10 Is Tyrosine-phosphorylated during IFN-␥-induced U937 Differentiation-Several mechanisms might decrease HoxA10 DNA binding, and therefore transcriptional repression, during myeloid differentiation. Decreased HoxA10 abundance might result in successful competition for an adjacent or overlapping element by transcriptional activators. Conversely, increased activator abundance, during differentiation, might result in successful competition for the DNA binding site, or post-translational modification of HoxA10, such as phosphorylation, might decrease HoxA10 affinity for the DNA-binding site.
To investigate the effect of myeloid differentiation on HoxA10 abundance, U937 cells were treated with for 48 h with IFN-␥. By Western blot, nuclear proteins from treated and untreated U937 cells demonstrate three HoxA10 cross-immunoreactive species: a 50-kDa species, the predicted size of the protein encoded by the major transcript in myeloid cells (4); a 42-kDa species, the predicted size of a protein encoded by an alternatively spliced HoxA10 mRNA described in murine tissues (7); and a 15-kDa species, the predicted size of short A10 (6). IFN-␥ treatment of U937 cells does not alter the abundance of any of these HoxA10 species (Fig. 5A), although binding to Pbx-HoxA10 consensus sequences is decreased by IFN-␥ treatment in EMSA with the same nuclear proteins (Fig. 1, B and  C). No protein species were detected when the blot was probed with rabbit pre-immune serum (not shown).
To determine if IFN-␥-induced differentiation results in HoxA10 phosphorylation, both IFN-␥-treated and undifferentiated U937 cells were 32 P-labeled, and lysate proteins were analyzed. IFN-␥ treatment of U937 cells increases the phosphorylation of anti-HoxA10 immunoprecipitable 50-and 42-kDa HoxA10 species, but not of short A10 (Fig. 5B). Since HoxA10 includes tyrosine, threonine, and serine residues (4), we next investigated whether HoxA10 was tyrosine-phosphorylated during myeloid differentiation. U937 cells were lysed, after 48 h of incubation with or without IFN-␥, and lysate proteins were immunoprecipitated with an anti-phosphotyrosine antibody or irrelevant control antibody. In Western blots of immunoprecipitated proteins, IFN-␥ treatment of U937 cells increased immunoprecipitable, tyrosine-phosphorylated 50and 42-kDa HoxA10 (Fig. 5C). In contrast, no immunoreactive species were detected when the blot was probed with rabbit preimmune serum (not shown).
HoxA10 Tyrosine Phosphorylation Decreases DNA Binding Affinity-To determine if HoxA10 tyrosine phosphorylation influences DNA binding affinity, in vitro translated proteins were dephosphorylated and used in EMSA with Pbx-HoxA10 binding probes. In preliminary experiments, in vitro translated HoxA10 was immunoprecipitable with anti-phosphotyrosine antibody (Fig. 6A). To generate tyrosine-dephosphorylated HoxA10, in vitro translated protein was treated with the specific tyrosine phosphatase, Yop. Yop-treated in vitro translated HoxA10, was not immunoprecipitated by anti-phosphotyrosine antibody, although Yop treated protein was intact (Fig. 6A).
Tyrosine-dephosphorylated HoxA10 and control proteins were used in EMSA with the dsA10 and dscybbA10 probes. As controls, in vitro translated HoxA10 was incubated in Yop reaction buffer without enzyme, and control reticulocyte lysate was incubated with and without Yop. Tyrosine-dephosphorylated HoxA10 demonstrates increased binding to the Pbx-HoxA10 consensus probe (Fig. 6B). Similar results were obtained with the dscybbA10 probe (Fig. 6C). Binding of in vitro translated HoxA10 to the dscybbA10 probe was competed for by unlabeled homologous oligonucleotide, by the dsA10 and dsncf2A10 oligonucleotides, and by an oligonucleotide representing another CYBB promoter sequence (dscybb5ЈA10), similar to the derived Pbx-HoxA10 consensus (Fig. 6D).
To determine if HoxA10 tyrosine dephosphorylation influences interaction with Pbx, in vitro translated HoxA10, with and without Yop treatment, was incubated in binding reactions with in vitro translated Pbx1. Both tyrosine dephosphorylated HoxA10 and control HoxA10 interact with Pbx1 to form a low mobility complex with the Pbx-HoxA10 consensus probe. However, binding affinity of the Pbx-HoxA10 complex is reproducibly increased with Pbx1-tyrosine-dephosphorylated HoxA10, in comparison with Pbx1-control HoxA10 (Fig. 6B). Identical , were analyzed by Western blot with a specific antibody to HoxA10. B, HoxA10 is phosphorylated during IFN-␥-induced U937 differentiation. U937 cells, untreated or after 48 h of IFN-␥ treatment (as indicated), were labeled by incubation with [ 32 P]orthophosphate. Cell lysate proteins were immunoprecipitated with HoxA10 antibody or control rabbit preimmune serum, and phosphorylated proteins were detected by autoradiography of SDS-PAGE. C, HoxA10 is tyrosine-phosphorylated during IFN-␥-induced U937 differentiation. Nuclear proteins isolated from U937 cells (100 g) that were either untreated or treated for 48 h with IFN-␥ (as indicated), were immunoprecipitated with an antiphosphotyrosine antibody or irrelevant control antibody (mouse antirabbit Ig). Western blot of the immunoprecipitated proteins was probed with a HoxA10-specific antibody.

FIG. 6. Tyrosine phosphorylation decreases DNA binding affinity of HoxA10 for Pbx-HoxA10 binding sites. A, in vitro translated
HoxA10 is tyrosine-phosphorylated. In vitro translated HoxA10 (10 l), either with or without Yop tyrosine phosphatase treatment, was immunoprecipitated with anti-phosphotyrosine antibody or irrelevant control antibody. Unprogrammed reticulocyte lysate was included as a control. Immunoprecipitated proteins, separated by 12% SDS-PAGE, were detected by autoradiography as indicated. B, tyrosine phosphatase treatment of in vitro translated HoxA10 increases DNA binding to the derived Pbx-HoxA10 consensus sequence, with and without Pbx1. EMSA was performed with the Pbx-HoxA10 consensus sequence probe (dsA10) and in vitro translated HoxA10 or control rabbit reticulocyte lysate, with and without in vitro translated Pbx1. Binding reactions were incubated in the presence of HoxA10 antibody (␣HoxA10), rabbit preimmune serum, or no antibody, as indicated. HoxA10 antibody disrupted both complexes generated by HoxA10 with or without Pbx1, binding to the dscybbA10 probe. The arrowheads indicate HoxA10 cross-immunoreactive complexes. F, tyrosine phosphatase treatment of "short A10" does not increase DNA binding affinity. EMSAs were performed with the CYBB promoter sequence probe (dscybbA10) and in vitro translated results were obtained with the dscybbA10 probe (Fig. 6C). In addition, the complexes generated by binding of in vitro translated HoxA10 and HoxA10 plus Pbx1 to the dscybbA10 probe were disrupted by antibody to HoxA10, but not by pre immune serum (Fig. 6E). Identical results were also obtained with the dsA10 probe (data not shown).
We also generated in vitro translated short A10 protein, in reticulocyte lysate, and performed similar experiments. In contrast to our results with HoxA10, binding of short A10 to the dscybba10 probe is not increased by Yop treatment (Fig. 6F). Identical results were obtained with the dsA10 probe (not shown).
Since HoxA10 is phosphorylated during IFN-␥-induced U937 differentiation, we hypothesized that tyrosine dephosphorylation of endogenous HoxA10, from IFN-␥-treated U937 cells, would restore binding to the dsA10 and dscybbA10 probes. U937 nuclear proteins from undifferentiated U937 cells and from U937 cells treated for 48 h with IFN-␥ were incubated with Yop. Control extracts were incubated under the same conditions in the absence of the enzyme. Consistent with our hypothesis, Yop treatment of nuclear proteins from IFN-␥treated U937 cells increases binding of the HoxA10-containing protein complex to the dsA10 and dscybbA10 probes (Fig. 7, A  and B). This complex was verified to contain immunoreactive HoxA10 in EMSA with the HoxA10-specific antibody (not shown). Yop treatment of nuclear proteins from undifferentiated U937 cells also increases abundance of the HoxA10-containing protein complex binding these probes. DISCUSSION Previous investigations suggest that HoxA10 increases proliferation and blocks differentiation during early myelopoiesis. HoxA10 function may be difficult to determine if there is variation in protein-protein interactions, protein-DNA interactions, or activity of functional domains during myelopoiesis. Our investigations determined that tyrosine phosphorylation of HoxA10 occurs during IFN-␥ induced myeloid differentiation and that tyrosine-phosphorylated HoxA10 has decreased DNA binding affinity. Additionally, we demonstrate, in U937 myeloid cells, that HoxA10 represses expression from artificial promoter constructs with Pbx-HoxA10 binding sites. We determine that HoxA10 has endogenous repression domains, not affected by IFN-␥-induced myeloid differentiation. Perhaps most interestingly, we identify the CYBB gene as a potential target for Pbx-HoxA10 repression, in undifferentiated myeloid cells.
We found that IFN-␥-induced myeloid differentiation decreases in vitro HoxA10 DNA binding to the derived Pbx-HoxA10 consensus, and to a similar CYBB promoter sequence. The CYBB sequence includes Pbx (5Ј-atgat-3Ј) and HoxA10 (5Ј-ttat-3Ј) cores, previously identified by binding site selection (7,8). However, unlike the derived consensus, there was a 1-bp overlap between the two sites. Despite this difference, the Pbx-HoxA10 consensus and CYBB sequence have cross-competitive binding specificities, although the derived consensus binds the complex with lower affinity. Also, the complex shifted by the CYBB probe migrates as a broader band than the complex shifted by the derived Pbx-HoxA10 consensus, suggesting that the CYBB sequence recruits additional proteins to the binding site. Other investigators found that HoxA9 interacts simultaneously with Pbx1 and Meis1 at DNA-binding sites (29,30). It is similarly possible that a Meis protein participates in Pbx-HoxA10 binding.
We identified a similar sequence in the NCF2 gene (31), with cross competitive binding specificity to the derived Pbx-HoxA10 consensus. The NCF2 sequence has a bp change in position 3 in the Pbx core (5Ј-ataat-3Ј) and an alternative HoxA10 core (5Ј-taat-3Ј) (7). The NCF2 gene encodes the respiratory burst oxidase protein p67 phox and is transcriptionally activated at the same point in myelopoiesis as the CYBB gene, suggesting that HoxA10 interacts with multiple genes activated during late myeloid differentiation. Previous investigations identified two other sequences in the CYBB promoter with cross-competitive binding with the CYBB sequence investigated in the current studies (17,32). Therefore, HoxA10 may exert an effect on transcription by interacting with multiple promoter sites. We are currently investigating the significance of these other CYBB promoter sequences, and several similar NCF2 sequences (31).
In our investigations, overexpression of HoxA10 represses reporter gene expression from constructs with the derived Pbxshort A10, or rabbit reticulocyte control. Comparison of DNA binding affinity, with and without Yop treatment was made.  A, tyrosine phosphatase treatment of nuclear proteins from U937 cells increases HoxA10 DNA binding to the derived Pbx-HoxA10 consensus. EMSAs were performed with the dsA10 probe and nuclear proteins (3 g) isolated from U937 cells that were either treated for 48 h with IFN-␥ (lanes 1 and 2), or untreated (lanes 3 and 4). Nuclear proteins were either Yop-treated (lanes 1 and 3) or sham-incubated in Yop buffer (lanes 2 and 4). The arrowhead indicates binding of the A 1 complex. B, tyrosine phosphatase treatment of nuclear proteins from U937 cells increases HoxA10 DNA binding to the CYBB promoter sequence, similar to the derived Pbx-HoxA10 consensus. EMSAs were performed with the dscybbA10 probe and nuclear proteins (3 g) isolated from U937 cells that were either treated for 48 h with IFN-␥ (lanes 1 and 2), or untreated (lanes 3 and 4). Nuclear proteins were either Yop-treated (lanes 1 and 3) or sham-incubated in Yop buffer (lanes 2 and 4). The arrowhead indicates binding of the A complex, and the asterisk indicates CP1 binding to the CCAAT box in the probe.
HoxA10 consensus or the similar CYBB promoter sequence in U937 cells. Although Pbx1 overexpression augments HoxA10 repression, Pbx1 alone did not repress reporter gene expression. Since U937 cells have endogenous HoxA10, this suggests that Pbx1 is not rate-limiting. Our results differ from U937 transfection experiments by other investigators with HoxA9 and an artificial promoter construct with the Pbx-Meis-HoxA9 consensus (30). In those studies, HoxA9 overexpression did not alter reporter expression. Differences in experimental design may explain the discrepancy, including shorter post-transfection incubation (6 versus 48 h), less transfected DNA (0.5 g/ 10 6 cells versus 3.0 g/10 6 cells), and differences in the minimal promoter. Or there may be differences in HoxA9 and HoxA10 function, despite similarity in protein-protein interactions and DNA binding specificity. In these investigations, we determined that overexpression of HoxA10 decreases endogenous gp91 phox and p67 phox mRNA abundance. Although this result is reassuringly consistent with our reporter gene assays, we cannot exclude the possibility that HoxA10 influences abundance of these transcripts indirectly by altering transcription of other genes or disrupting differentiation.
We determined that IFN-␥-induced U937 differentiation is accompanied by HoxA10 tyrosine phosphorylation, which decreases DNA binding but does not alter endogenous HoxA10 repression domains. Therefore, HoxA10 tyrosine phosphorylation, reversible by tyrosine phosphatases, is a mechanism for reversible gene regulation in response to cytokines. We found that Yop tyrosine phosphatase treatment of U937 nuclear proteins increases binding of the Pbx-HoxA10-containing complex to both the derived consensus and similar CYBB sequence. However, Yop-treated proteins from IFN-␥ differentiated U937 cells do not generate the Pbx-HoxA10 complexes as efficiently as proteins from Yop-treated, undifferentiated U937 cells. This is consistent with the hypothesis that IFN-␥ treatment increases kinase activity in U937 cells and antagonizes endogenous (and also exogenous) phosphatase activity.
We also investigated the significance of the "short A10" transcript (6). We identified a HoxA10 cross-immunoreactive protein the predicted size of short A10 but were unable to demonstrate IFN-␥-induced phosphorylation in U937 cells. Also, DNA-binding of in vitro translated short A10 is not increased by tyrosine phosphatase treatment. Consistent with this, short A10-induced repression of reporter expression from constructs with Pbx-HoxA10 binding sites is not decreased by IFN-␥ treatment. Since short A10 has insignificant endogenous repression activity, these results suggest that repression is due to successful competition with transcriptional activators for the same (or adjacent) DNA sequences.
Therefore, the mechanism of HoxA10 transcriptional repression is 2-fold: repression due to endogenous HoxA10 domains and binding site competition with transcriptional activators. Our studies also imply a function for short A10 in immortalized cell lines. Since short A10 represses transcription but is not modulated by differentiation-induced phosphorylation, it might contribute to transformation by repressing transcription of genes that are necessary for differentiation progression, or characteristic of the differentiated phenotype.
Previous studies of the CYBB promoter indicated that CDP represses transcription by binding within a 30-bp sequence that includes the Pbx-HoxA10-like site (15,16). CDP binds to this sequence in EMSA with nuclear proteins from the epithelial cell line HeLa and the erythroleukemia cell line K562. Additional studies determined that CDP DNA binding is modulated by another homeodomain protein, SATB1 (33). These factors are nuclear matrix-associated, and relative abundance regulates DNA interactions (33). These studies, in combination with our current observations, suggest a mechanism for transcriptional repression by homeodomain proteins in cells of various lineages. In nonmyeloid cells, transcription of CYBB may be repressed by CDP. In committed myeloid progenitors, SABT1 increases, decreasing CDP association with the repressor element, coincident with nuclear matrix disassociation. Transcriptional repression in myeloid progenitors would be maintained by Pbx-HoxA10 binding until later in differentiation. This interaction could be rapidly (and reversibly) modulated by HoxA10 tyrosine phosphorylation, during differentiation or the inflammatory response (both increasing CYBB and NCF2 transcription).
Our investigations suggest that one role of HoxA10 during myeloid differentiation is repression of transcription of genes characteristic of mature myeloid cells, such as components of the phagocyte respiratory burst oxidase. It will be of interest to investigate other genes also transcriptionally activated during late myeloid differentiation (or actively transcribed during the immune response) to determine the significance of this HoxA10 function.