Identification of Transcriptional Targets of HOXA5

The homeobox gene HOXA5 encodes a transcriptional factor which has been shown to play important roles in embryogenesis, hematopoiesis and tumorigenesis. In order to decipher downstream signaling pathways of HOXA5, we utilized oligonucleotide microarray analysis to identify genes which are differentially expressed in HOXA5 induced cells compared to uninduced cells. Comparative analysis of gene expression changes after 9 hours of HOXA5 induction in Hs578T breast cancer cells identified 306 genes whose expression was modulated at least two-fold. Ten out of these 306 genes were also upregulated by at least two-fold at 6 hours post-induction. The expression of all of these ten genes was confirmed by semi-quantitative RT-PCR. Among these ten genes, which are most likely to be direct targets of HOXA5, we initiated an investigation into the pleiotrophin gene by first cloning its promoter. Transient transfection assays indicated that HOXA5 can specifically activate the pleiotrophin promoter. Promoter deletion, chromatin immunoprecipitation assay and gel shift assays were performed to show that HOXA5 can directly bind to one binding site on the pleiotrophin promoter. These data strongly suggest that microarray analysis can successfully identify many potential direct downstream genes of HOXA5. Further functional analysis of these targets will allow us to better understand the diverse functions of HOXA5 in embryonic development and tumorigenesis.


INTRODUCTION
HOX genes, a subset of the homeobox genes which were originally identified in Drosophila, encode a family of transcriptional factors essential for axial and appendicular patterning and organogenesis (1). Recently, many HOX genes were found to be aberrantly expressed in a variety of cancers, including those of the breast, kidney and skin, suggesting that they may also contribute to the progression of tumors (2)(3)(4)(5). All 39 HOX transcriptional factors identified in vertebrates have in common a 60-amino acid DNAbinding homeodomain that binds DNA and that has been conserved with respect to sequence, structure, and mechanism of DNA binding (6). In vitro assays show that all HOX proteins can bind to similar DNA motifs with a core sequence of TAAT (7,8). In sharp contrast to relatively simple binding motifs revealed in in vitro binding assays, each HOX gene has a variety of functions which cannot be completely compensated by other HOX genes. In order to understand the complexity of HOX gene functions extending from embryo development to hematopoietic maturation and tumorigenesis, it is important to identify the specific targets for each HOX gene.
The evidence for an important role of Hoxa5 in development comes mainly from the study of the Hoxa5 knockout mice (9,10). The resultant heterozygous mutant mice were viable and indistinguishable from their wild type littermates. However, more than 60% of the homozygous pups died within 4 days of birth. In homozygous newborn mutants, improper tracheal and lung morphogenesis lead to tracheal occlusion, and to respiratory distress associated with a marked decrease in the production of surfactant protein (10). Loss of Hoxa5 also perturbed intestinal maturation and transiently affects thyroid development (11,12). Analysis of the skeletal structures in the homozygote revealed consistent abnormalities affecting the region between sixth cervical vertebra (C6) and the first lumbar vertebra. In human, HOXA5 has been shown to play a role in blood cell differentiation (13,14). Constitutive expression of HOXA5 expression in human hematopoietic progenitor cells causes a significant shift toward myeloid differentiation and away from erythroid differentiation (13). In addition, HOXA5 expression was lost in more than 60% breast cancer cell lines and primary breast carcinoma cells (15). Overexpression of HOXA5 in breast cancer cell line, MCF7, induced apoptosis by upregulating the expression of p53 (15). However, in the p53mutant breast cancer line Hs578T, we have presented evidence that HOXA5-mediated apoptosis by activating caspase-2 and caspase-8 (16). HOXA5 and TNFα acted synergistically to induce apoptosis (16). These studies indicated that HOXA5 may behave as a tumor suppressor gene in breast cells.
Despite knowledge of a variety of functions of the HOXA5 gene that have been found to date in development and tumorigenesis, very few specific targets genes have been identified. The identity of the target genes could better define the multiple pathways through which HOXA5 may positively or negatively modulate growth. In this study, we utilized microarray analysis of an inducible HOXA5 breast cancer cell line, HS578T, to identify 306 genes whose expressions are significantly modified after HOXA5 induction.
Further, we demonstrated that HOXA5 can directly bind to and activate the promoter of one these genes, pleiotrophin (PTN). Further sorting and functionally analyzing the candidate targets of HOXA5 will shed light on the underlying mechanisms of HOXA5 function. In the present study, we performed 4 pairwise comparisons for each time point (induced, n=2 and uninduced, n=2). Only those altered genes that appeared in 4 out of 4 comparisons were selected. This conservative analytical approach was used to limit the number of false-positives. The ESTs obtained in the data were searched for their recent annotation using the 'Analysis Center' at the Affymetrix site (www.netaffx.com). The microarray data has been deposited to the GEO database (Series entry: GSE2241; http://www.ncbi.nlm.nih.gov/geo).
Quantitative Real Time RT-PCR Analysis: Ten up-regulated genes as identified by microarray were selected for real time RT-PCR analysis to verify the array results.
Reverse transcription reaction was performed as follows: 1ug of DNase treated total RNA, 0.5 ug of anchored oligo-dT15 primer and 500 M dNTPs (NEB) were heated for 5 min at 65°C; 1× first strand buffer (Invitrogen, La Jolla, CA), 0.01 M DTT, and 200 units Superscript II (Invitrogen) were added and reverse transcription was carried out, in a 20ul reaction, for 50 min at 42°C and terminated by heating for 15 min at 70°C. To assess for potential contamination of solutions, a control containing all reagents, but devoid of RNA was included. In addition, a control containing all reagents, except the Superscript II, was included for each sample in order to monitor for possible residual genomic DNA in the RNA preparations.
The Q-RTPCR was performed using the fluorescent dye SYBR Green Master Mix following standard protocols on an ABI PRISM 7900 sequence detection system (Applied Biosystems, CA). The primers used for the PCR reaction were shown in Table 1.
Native gel electrophoresis was used to characterize the final products. The data were first analyzed using the Sequence Detector Software SDS 2.0(Applied Biosystems, CA).
Results were calculated and normalized relative to the GAPDH control by using the Microsoft Excel program. All of the PCR were performed in triplicate and mean values were shown in figures.
Western Blot Analysis: Twenty µg of protein was fractionated in a 4-12% NuPAGE gel and transferred to PVDF membranes. The membrane was blocked with 100 ml of TBS buffer (10 mM Tris-base pH7.5, 0.9% NaCl) containing 5% dry milk and 0.1% Tween-20 for 1 hour on the shaker at room temperature or overnight in cold room. The membrane was rinsed once with TBS buffer before incubating with an appropriate dilution of the 1˚ antibody in TBS buffer containing 5% milk and 0.02% Tween-20 on shaker for 1 h. The primary antibody bound membrane was washed with TBS buffer containing 0.1% Tween-20 four times and then incubated with the 2˚ antibody (from Amersham ECL kitanti-rabbit or anti-mouse) at about 1:1000 dilution for 1 to 1.5 hours on a shaker. The filter was developed by using the ECL-Plus reagent (Amersham). Rabbit anti-peptide antibodies to HOXA5 were provided by Zymed Inc. (San Francisco, CA) Transient Transfection Assay: 2×10 5 SKBR3 cells were seeded onto each well of 6well plate 24 h prior to transfection. 1 to 2 ug of plasmids was transfected into cells using Gene Jammer (Stratagene Corp., La Jolla, CA ) according to the manufacturer's instruction. 24 h post-transfection, cells were harvested for luciferase and β-Gal assay using luciferase activity measuring kit (Promega, Madison WI) and β-Gal assay kit (ICN Biomedicals Inc., Aurora, OH) according to the manufacturer's instruction. The luciferase activities were normalized to the β-Gal activities for each sample. The foldactivation was calculated as the ratio of normalized luciferase activities in the cells transfected with the reporter plasmid in the presence of HOXA5-expressing plasmid to that in the absence of HOXA5-expressing plasmid. Each transfection was repeated at least three times and the averaged data is shown in the figures.
Chromatin Immunoprecipitation (ChIP) assay: Chip assays were performed using Chromatin Immunoprecipitation (ChIP) Assay Kit (Upstate Inc., Lake Placid, NY) according to the manufacturer's instruction. In brief, 1 x 10 6 vector and HOXA5-induced cells were cross-linked by adding formaldehyde directly to culture medium. Cells were harvested and sonicated to shear DNA to lengths between 200 and 1000 base pairs. After centrifuging samples for 10 minutes at 13,000 rpm at 4°C, the supernatant was precleared with 75µl of salmon sperm DNA/Protein A agarose-50% slurry for 30 minutes at 4°C with agitation. 2 µg of HOXA5 antibody was then added to the supernatant fraction for incubation overnight at 4°C with rotation. Then, 60µl of salmon sperm DNA/Protein A agarose was added to collect the antibody/histone complex. The protein A agarose/antibody/histone complex was extensively washed for five minutes as suggested and heated at 65°C for 4 hours to reverse histone-DNA crosslinks. The DNA was recovered by phenol/chloroform extraction and ethanol precipitation. The PCR were performed using two pairs of primers (control: nt-1084 5'-CTA CTT GCC ACA AGA   CAA TG and nt-725 5'ACG CTA AGG CAA TGC ATA GG-3' ; HBS: nt-271 5'-GAG  ATC TGG CTT TGC ACT CAT CTG AA-3' and nt-8 5'-GCA TAT GGA GAA TGG   GAG GGA ATG A-3').
Gel-shift assay: Nuclear extracts from HOXA5-inducible cells were prepared and gelshift assays were performed as described previously (15,17). The reaction was carried out in a final volume of 20-ul. One microliter of nuclear extract (~2 ug total protein) was added to binding buffer containing 2 ug of ploy [d(I-C)] (Amersham Biosciences Corp., Piscataway, NJ ), 20 mM HEPES-HCl, pH 7.9, 50 mM KCl, 1 mM EDTA, 10 mM MgCl 2 , 6% glycerol and 2×10 4 cpm of 32 P-end-labeled oligonucleotide. After 15 min on ice, loading buffer was added, and the Protein-DNA complexes were resolved in nondenaturing 5% polyacrylamide gels by electrophoresis at 100V for 3-4 h in 0.25×TBE (1×TBE: 0.089M Tris base, 0.089M boric acid, 2mM EDTA). The gels were dried and exposed to Kodak-RP film (Kodak, New Haven, CT). The 20-mer oligonucleotides containing the canonical HOXA5-binding site within the PTN promoter and the mutated binding-sites, were synthesized and purified by high performance liquid chromatography.
Double-stranded oligonucleotides were end-labeled with [γ-32 P] ATP. To ascertain specificity of binding, unlabeled competitor wild-type, or mutated oligonucleotides were incubated with the protein extract prior to the addition of the labeled oligonecleotide. A supershift assay was also performed by incubating the DNA-protein complex with 2 ug of rabbit polyclonal HOXA5 antiserum (Zymed Inc., San Francisco, CA) for 10 min on ice.

Identification of genes which are differentially expressed after induction of HOXA5 expression
The tet-off HOXA5 inducible Hs578T cell line was established as described previously (16). In this system, HOXA5 expression is tightly controlled by the tetracycline-responsive promoter and can be rapidly induced by removal of Doxycycline (a tetracycline analog) from the culture medium (Fig. 1). Similar to the parental cells (P), under uninduced condition (0h time point), the expression of HOXA5 was undetectable by western blot analysis. At 3 h post-induction, the expression of HOXA5 became detectable and continued to increase over the time course of induction until 6-9 h postinduction. At 12 hours post-induction, cells began to undergo apoptosis (16).
These data suggested that HOXA5-mediated gene expression had occurred during the first few hours of induction. The earliest responding genes are more likely to be the direct targets of HOXA5. To this end, we harvested cells at 0, 6 and 9 h after induction.
Total RNA was purified from these cells and subjected to oligonucleotide microarray analysis using Affymetrix chips.
The microarray experiments were repeated once with separately purified RNA samples for each of the time points. The gene expression profiles of cells induced for 6 and 9 h were compared to that under uninduced conditions (0 h). The differentially expressed genes in 4 out of 4 comparisons (6 h_#1 versus 0 h_#1 and _#2; 6 h_#2 versus 0 h_#1 and _#2; 9 h_#1 versus 0 h_#1 and _#2; 9 h_#2 versus 0 h_#1 and _#2) are listed in the Tables (Supplementary Data). At 9 h post-induction, when we observed the maximal number of genes that were differentially expressed, we identified 262 genes whose expression was up-regulated by at least 2-fold and 44 genes whose expression was down-regulated by at least 2-fold. Since HOXA5 is well-established as a positive regulator of genes, we have focused only on the up-regulated genes in this study.
Among the 262 up-regulated genes, 10 genes or ESTs were also up-regulated at 6 h post-induction (Table 2). In contrast, only one gene which is upregulated at 6 h postinduction did not appear in the list of upregulated gene at 9 h post-induction.

Validation of microarray data by real time PCR analysis
The ten genes whose expression was upregulated at both 6 h and 9 h postinduction are more likely to be the direct targets of HOXA5. Therefore, we examined the expression pattern of these genes over the time course of HOXA5 induction. Similar to the continued increase in the expression of HOXA5 during the time course of induction, we observed that compared to parent cells and uninduced cells, expression of most of these ten genes increased in parallel to the expression of HOXA5 (Fig. 2). Hence, the data generated by the microarray was confirmed for each of these genes by real time PCR, thereby validating the robustness of the analysis.

HOXA5 specifically activates the promoter of the pleiotrophin gene
Next, a search was performed of the promoter region of these ten HOXA5induced candidate genes. Among these, we observed the presence of multiple HOXA5 core binding motifs (TAAT) in the promoter sequences of the pleiotrophin (PTN) gene.
While the core binding sequences of HOX genes are TAAT, the specificity of binding for each HOX genes has been shown, in one study, to be attributable to the flanking sequences (8). But no consensus HOXA5-specific binding sites beside the TAAT core sequence were identified. PTN encodes a 136-amino acid cytokine which is a important contributor to growth, differentiation, and maintenance of viability within the mammalian nervous system during development (21,22). Hoxa5 is also highly expressed in the nervous system during mouse embryogenesis. Moreover, in primary breast cancers, we analyzed multiple array databases and determined that PTN expression is lost in breast cancer (www.oncomine.com). PTN expression level was significantly lower in breast tumor samples in three out of three studies involving a total of 32 normal breast samples and 149 breast carcinoma samples. The adjusted P-values for these three studies were 0.002, 0.1 and 5.6×10 -6 respectively. Based upon the data that there is loss of HOXA5 Since many HOX genes are functionally redundant, we next tested the specificity of HOXA5 on activation of PTN promoter. PTN promoter was cotransfected with expression constructs of either, HOXB1, HOXB3, HOXB5 or HOXD9. We found that among the HOX genes tested, only HOXD9 weakly activated the promoter of PTN (Fig.   3E). These results suggested that, among the HOX genes tested, it is likely that HOXA5 acts as a transcriptional regulator upstream of PTN.

Homeodomain of HOXA5 is required for its ability to transactivate the PTN promoter
The homeodomain of the HOX protein, which binds to DNA, is generally required for its transactivation abilities. However, HOX proteins can also regulate gene expression by interacting with other transcriptional factors independent of their binding domain (25). To investigate whether the DNA binding domain of HOXA5 is required for activation of PTN expression, we constructed a mutant HOXA5 plasmid which encoded a truncated HOXA5 protein, lacking most of its homeodomain (Fig. 3A). Transient transfection assays indicated that this truncated HOXA5 protein completely lost its ability to activate the promoter of PTN (Fig. 3B). The inability of truncated HOXA5 to activate gene expression was not due to failure of expression since Western analysis revealed that both wild type and truncated protein were expressed at similar levels ( Fig. 3C). These results suggested that HOXA5 DNA binding domain is required for its ability to regulate the PTN promoter.
In vitro assays have shown that HOX gene family transcription factors bind to very similar binding sites with core sequences such as TAAT (8). The specificity for transcriptional regulation by HOX genes is often achieved either by interaction with cofactors, or by the DNA context around the binding site (26-29). The N-terminal part of the HOX protein is known to play an important role in protein-protein interaction. For example, the pentapeptide (YPWMR), which is found N-terminal to the homeodomain of many HOX proteins, is the determinant for interaction with the cofactor PBX1A (30).
Interaction with PBX1A has been shown to greatly increase the DNA binding affinity of HOX proteins, including HOXA5, and also altered DNA binding specificity (30-32). We hypothesized that the truncated HOXA5 protein (that had lost its binding capacity but retained its ability to interact with cofactors) may competitively inhibit the transactivation of the PTN promoter mediated by wild type HOXA5. As predicted, by cotransfecting cells with wild type HOXA5 and different concentrations of mutant HOXA5, we found that the truncated HOXA5 protein inhibited the transactivation ability of wild-type HOXA5 in a dose-dependent manner. In the presence of ten-fold excess of mutant HOXA5-expressing plasmids, the transactivation of PTN promoter by wild-type HOXA5 was inhibited by more than 50% (Fig. 3D). These results provide further evidence that HOXA5 can, with some specificity, increase the promoter activity of PTN.

Mapping the binding sites of HOXA5 on the promoter of PTN
The DNA binding domain of HOXA5 is required for HOXA5-mediated activation of PTN promoter. We next attempted to define the binding sites of HOXA5 in the promoter by serially deleting the PTN promoter. Within about 1kb of PTN promoter sequences immediately upstream of the transcriptional start site, we found 7 potential HOXA5 binding sites. We constructed a series of deletions aimed at removing these sites in a stepwise manner. As shown in Fig Although it still carries two upstream TAAT sites, the mutated plasmid completely lost its ability to be activated by HOXA5 (Fig. 4A). These findings provided additional lines of evidence that HOXA5 bind to the PTN promoter, and that the putative HBS is most likely a bona fide HOXA5 binding site in the PTN promoter.
To test whether HOXA5 can bind directly to the putative HBS site, we performed gel-shift assays with synthetic probes designed according to the sequence of HBS (Fig.   4B). In most previous studies on HOX protein-DNA binding assay, purified HOX proteins were used for the gel-shift assay (8,26). The strong binding sites found in these in vitro assays usually confer very weak transactivation ability to the consensus binding site-containing promoters in in vivo transfection assay (8,26). Besides co-factors affecting the binding affinity in vivo, post-transcriptional modifications of HOX protein also alter their DNA binding abilities (34). We used nuclear extracts from HOXA5-  4D). Adding HOXA5-specific antibody to the reaction resulted in a shift of the HOXA5specific bands (lanes 2, 3), but had no effect on the migration of nonspecific bands (lanes 4,5).
To study if HOXA5 binds to HBS in intact cells, we performed a chromatin immunoprecipitation (ChIP) assay using a HOXA5-specific antibody. Consistent with the results of the gel-shift assay, the DNA fragment encompassing the HBS was specifically In this study, we utilized oligonucleotide microarray analysis to identify the downstream targets of HOXA5. We found that induction of HOXA5 expression can rapidly modulate the expression of a large group of genes representing a wide variety of functional categories. Semi-quantitative RT-PCR was able to confirm the gene expression changes identified by the microarray analysis in all 10/10 genes tested. In breast cancer cells, it was initially reported that PTN mRNA was detected in 25% of breast cancer cell lines and about 60% of breast carcinomas (23). However, the expression of PTN in normal breast tissue was not examined. The authors concluded that PTN was overexpressed in a large proportion of breast carcinomas. Since tumors that did not express PTN also contained similar proportions of normal tissue, they argue that the expression of PTN in normal breast tissue is too low to be detected by the RNase protection assay. In recent years, gene expression profiles of breast cancer using large numbers of clinical samples have been performed using microarray analysis (49-51). We searched through these gene expression databases at www.oncomine.com and found that, in contrast to the published data (23), PTN expression level was significantly lower in breast tumor samples in three out of three studies involving a total of 32 normal breast samples and 149 breast carcinoma samples. Our data suggests that PTN expression was lost in breast tumors. Our previous study showed that HOXA5 expression was also lost in more than 60% primary breast carcinomas. Identification of PTN as a direct target of HOXA5 implies that loss of HOXA5 expression may contribute to the loss of PTN expression in breast tumors. Both PTN and HOXA5 play diverse roles in development and tumorigenesis. PTN was reported to stimulate the proliferation of fibroblast, endothelial and epithelial cells and act more like an oncogene (23). On the other hand, HOXA5 was found to induce apoptosis in breast cancer cells when it was overexpressed (15,16). However, both of them have also been shown to promote cellular differentiation (13,14,42). Currently, we can not integrate these diverse functions into a simple model.
Further studies are needed to address the functional relationship between these two genes in breast tumorigenesis.
As mentioned above, overexpression of HOXA5 in breast cancer cells triggers apoptotic pathways (15,16). Interestingly, when we searched through genes which were upregulated at 9 h post-induction, we identified many genes related to receptor-mediated apoptosis. Among them are two TNF receptor family members, TNFR9 and TNFR10b.
TNF10b (also named as DR5) is a receptor for TRAIL, which is a well-documented apoptotic inducer (52). Among the 12 gene probes which displayed unregulated genes at 6 h post-induction, three of them represent one single gene GADD45β. GADD45β is a putative target of NF-κB (53). IL-8, which is another putative target of NF-κB (54), is also upregulated at 6 h post-induction. These results strongly suggest that NF-κB pathway may be activated downstream of HOXA5. Our recent data showed that p65 protein was translocated into nucleus 30min after induction of HOXA5 expression (data not shown), which has been shown to be one of the most important criteria for NF-κB activation (55,56). Activation of NF-κB antagonizes apoptosis by numerous triggers, including the ligand engagement of 'death receptor' such as tumor-necrosis factor (TNF) receptor (57). Consistent with these findings, we have recently shown that HOXA5 and TNFα can synergistically induce apoptosis, strongly suggesting that the receptormediated apoptotic pathway was activated (16). In conclusion, HOXA5 may trigger receptor-mediated apoptotic pathways which involve activation of NF-κB signaling pathway.
In summary, we have successfully identified many direct or indirect targets of HOXA5. The HS578T-tet-inducible system allowed us easily to identify genes whose expression was upregulated early, which are more likely to be the direct targets and may initiate HOXA5 downstream signaling pathways. Further comprehensive sorting and characterization of these downstream targets will help us better understand how the

Table 1. Primers used in the quantitative real time RT-PCR analysis
Gene Name Primer Sequence     Fig. 4 B. D.

C.
No

Control
Vector HOXA5

Supplementary Data
The following four tables are gene lists which included differentially expressed genes with at least two-fold changes (Sig log-average >1) in 4 out of 4 comparisons as described in the text.