Gene Expression Profiling Identifies HOXB4 as a Direct Downstream Target of GATA-2 in Human CD34+ Hematopoietic Cells

Aplastic anemia is characterized by a reduced hematopoietic stem cell number. Although GATA-2 expression was reported to be decreased in CD34-positive cells in aplastic anemia, many questions remain regarding the intrinsic characteristics of hematopoietic stem cells in this disease. In this study, we identified HOXB4 as a downstream target of GATA-2 based on expression profiling with human cord blood-derived CD34-positive cells infected with control or GATA-2 lentiviral shRNA. To confirm the functional link between GATA-2 and HOXB4, we conducted GATA-2 gain-of-function and loss-of-function experiments, and HOXB4 promoter analysis, including luciferase assay, in vitro DNA binding analysis and quantitative ChIP analysis, using K562 and CD34-positive cells. The analyses suggested that GATA-2 directly regulates HOXB4 expression through the GATA sequence in the promoter region. Furthermore, we assessed GATA-2 and HOXB4 expression in CD34-positive cells from patients with aplastic anemia (n = 10) and idiopathic thrombocytopenic purpura (n = 13), and demonstrated that the expression levels of HOXB4 and GATA-2 were correlated in these populations (r = 0.6573, p<0.01). Our results suggested that GATA-2 directly regulates HOXB4 expression in hematopoietic stem cells, which may play an important role in the development and/or progression of aplastic anemia.


Introduction
The pathogenesis of aplastic anemia (AA) is complex and many questions remain unanswered. It has been proposed that immunological injury in hematopoietic stem cells (HSCs) leads to reduced numbers of stem cells in the bone marrow (BM) [1,2,3]. Evidence supporting this suggestion was obtained from in vitro analysis of samples obtained from AA patients, as well as from clinical evidence that immunosuppressive therapy with drugs such as anti-thymocyte globulin (ATG) and cyclosporin A is effective in 75% of AA cases [2,3]. On the other hand, the observation that some patients do not respond to immunosuppressive therapy, and may even develop myelodysplastic syndrome or acute leukemia as a consequence of clonal evolution [4], suggests that the immunological aberration is not the sole mechanism of the disease. However, only limited information has emerged regarding potential intrinsic abnormalities of HSCs in AA, as most studies have focused on the extrinsic immunological abnormalities.
Hematopoiesis is intricately controlled by numerous transcription factors that control and coordinate the expression of lineagespecific genes [5]. Foundational studies defined involvement of members of a family of developmental regulators, the GATA transcription factors [6,7,8,9]. GATA-1, GATA-2, and GATA-3 are termed as hematopoietic GATA factors, based on their important activities in controlling distinct and overlapping aspects of hematopoiesis [8]. GATA-1 functions to promote the development of erythrocytes, megakaryocytes, eosinophils, and mast cells, and GATA-3 plays a role in T-lymphocyte development [5,8,9,10,11]. On the other hand, GATA-2 functions early in hematopoiesis and is required for maintenance and expansion of HSCs and/or multipotent progenitors [9,12,13,14,15]. Therefore, it is possible that changes in the GATA-2 expression level could lead to aberrant proliferation and differentiation of HSCs, and may be responsible for the development of AA. Previously, we demonstrated decreased expression of GATA-2 in CD34-positive cells in AA [16]. This observation was confirmed in another study, which involved microarray-based transcriptional profiling in CD34-positive cells with AA [17]. Therefore, we speculated that some stem cell-specific genes are aberrantly expressed as a consequence of GATA-2 downregulation in HSCs in AA, which could contribute to the development and/or progression of the disease.
To examine this possibility, we conducted microarray analysis with human cord blood-derived CD34-positive cells infected with lentiviral shRNA specific to GATA-2. Among GATA-2-regulated genes, we focused on the HOXB4 gene as one of the causative candidate genes for AA. In the present study, we investigated the functional link between GATA-2 and HOXB4, and also examined correlation between GATA-2 and HOXB4 expression based on clinical specimens.

Clinical samples
Patients involved in the analysis were diagnosed with AA (n = 10) or idiopathic thrombocytopenic purpura (ITP) (n = 13) in our institute or related hospitals [16]. The severity of AA at the time of diagnosis was moderate (n = 7) or severe (n = 3). Diagnosis and assessment of disease severity were performed according to published criteria [18]. Clinical samples were collected after   Cell lines and isolation of CD34-positive cells K562 cells were maintained in RPMI medium containing 10% fetal bovine serum. Methods for establishment of K562 cell lines stably expressing exogenous GATA-2 were described previously [19]. GATA-2 expression vector was constructed using the coding region of human GATA-2 cloned into pcDNA(2)/Myc-His vector (Invitrogen) [19].

Real-time quantitative PCR and expression profiling
Total RNA was purified with TRIzol (Invitrogen), and cDNA was synthesized from 1.5 ug of purified total RNA using Superscript II (Invitrogen). Reaction mixtures (20 uL) for realtime quantitative RT-PCR consisted of 2 uL of cDNA, 10 uL of Power SYBR green master mix (Applied Biosystems) and appropriate primers. Primer sequences are available upon request. Product accumulation was monitored by measuring SYBR Green fluorescence and normalized relative to GAPDH or 28S mRNA.
For expression profiling, Human Oligo chip 25k (Toray) was used according to the manufacturer's protocol. Aliquots of 1 ug of total RNA were amplified with an Ambion Amino Allyl aRNA kit (Ambion), labeled with Amersham Cy5 Mono-Reactive Dye (GE Healthcare), and hybridized to the Human Oligo chip 25k array. Data were collected and normalized using a 3D-Gene Scanner 3000 system (Toray). For global normalization, background value was subtracted, and subsequently adjusted to the average signal value of 25. Each probe on the microarray was linked with specific Gene Ontology (GO) terms based on Oligo MicroArray DataBase system (OMAD) (Operon).

Promoter assay
The DNA fragment of the HOXB4 gene promoter region, including 2 transcriptional start sites [20], was obtained from human genomic DNA of a healthy volunteer by PCR using the following primers: forward: 59-CGCAGGTACCCTATG-TAAATCCTGGTGT-39 (spanning positions 2262 to 2279 from the translational start site); reverse: 59-AC-CCCTCGAGCGTTTTCCTGTTTCCGAA-39 (255 to 272 from the translational start site). These primers contain artificial restriction enzyme sites (KpnI and XhoI, respectively), and the amplified DNA fragments were inserted into the KpnI/XhoI sites of pGL3-Basic (Promega). GATA deletion construct was generated using a QuickChange TM Site-Directed Mutagenesis Kit (Stratagene) using the mutagenesis primer 59-GTGAGCA-GATTTCCTCGGGAATCGCAGGCG-39.
To assay HOXB4 transcriptional activity, aliquots of 5610 5 K562 cells (ATCC CCL-243) [19] were transfected with 1 ug of HOXB4 promoter construct and 50 ng of a Renilla luciferase reporter plasmid with FuGene 6 (Promega). After 24 h of incubation in culture media, the cells were harvested, and both firefly and Renilla luciferase activities in the cell extracts were determined using the Dual Luciferase Reporter Assay System (Promega).

Quantitative ChIP analysis
Real-time-PCR-based quantitative chromatin immunoprecipitation (ChIP) analysis was conducted as described [23,24]. Primer sequences are available upon request.

Antibodies
Antibodies for GATA-2 (H-116), Myc (9E10), and His probe (G-18) were obtained from Santa Cruz Biotechnology. Alpha-Tubulin (CP06) was obtained from Calbiochem. GATA-2 (ab22849) and rabbit control IgG were obtained from Abcam. exogenous Myc/His-tagged GATA-2 in K562 cells. The results of 3 representative clones (A1, A4, and A6) and 2 control clones are shown. Alpha-Tubulin was used as a loading control. (D) Correlation between GATA-2 and HOXB4 mRNA in each clone. mRNA expression was normalized relative to that of GAPDH, and the correlation of these genes was assessed by Spearman's rank correlation method. (E-F) GATA-2 knockdown in K562 cells. (E) Anti-GATA-2 Western blotting analysis of whole-cell extracts and (F) Quantitative RT-PCR analysis of GATA-2 and HOXB4 (mean 6 SE, n = 3), from K562 cells transfected with siRNA against human GATA-2 or control siRNA. Alpha-Tubulin was used as a loading control. mRNA levels were normalized relative to GAPDH. *p,0.05. (G) MSCV retroviral vector-mediated GATA-2 overexpression in cord blood-derived CD34-positive cells. Quantitative RT-PCR analysis was performed to detect GATA-2 and HOXB4 expression (mean 6 SE, n = 3). mRNA levels were normalized relative to 28S.

Western blotting analysis
Whole-cell extracts were prepared by boiling cells for 5 min in SDS sample buffer, and the samples were resolved by SDS-PAGE and transferred onto Immobilon-P membranes (Millipore), as described previously [24]. Detection was performed with ECL plus (GE Healthcare) and BioMAX XAR Film (Kodak).

Statistics
Statistical significance was assessed by two-sided Student's t test. Potential correlations between HOXB4 and GATA-2 expression levels were examined by Spearman's rank correlation test. Z-score was calculated to rank the GO terms by the relative changes in gene expression.

Expression profiling identified HOXB4 as a GATA-2 downstream target gene
To identify GATA-2-regulated genes that may contribute to the pathophysiology of AA, we conducted expression profiling using human cord blood-derived CD34-positive cells infected with control and GATA-2 lentiviral shRNA. Western blotting and quantitative RT-PCR revealed knockdown of GATA-2 protein and mRNA expression (Figures 1A, 1B). We first selected genes that showed 1) changes of more than 1.4-fold, and 2) more than 20 of normalized expression value, which yielded 1126 downregulated and 572 upregulated genes, respectively (Table S1). Quantitative RT-PCR confirmed significant downregulation of representative GATA-2 target genes, such as KIT and GFI1B ( Figure 1B) [25]. Gene Ontology analysis revealed significant enrichment for ''signal transduction,'' ''regulation of transcription, DNA-dependent,'' and ''Metabolic process'' ( Figure 1C). In the present study, we focused on the HOXB4 gene for further analysis ( Figure 1B), because HOXB4 has been reported to play an important role in inducing HSC expansion [26,27]. Furthermore, it has also been reported that restoring HOXB4 protein in HSCs of AA patients rescues growth potential [28].
To further confirm whether GATA-2 regulates HOXB4 expression in human cord blood-derived CD34-positive cells, we conducted both overexpression and knockdown of GATA-2, using retroviral vector (MSCV) and siRNA, respectively, and confirmed that GATA-2 regulates HOXB4 expression ( Figure 2G and Figure S1).

GATA-2 binds directly to the HOXB4 promoter and regulates its activity
Our recent ChIP-seq (chromatin immunoprecipitation and sequencing) analysis revealed a significant GATA-2, but not GATA-1, peak at the HOXB4 promoter in K562 cells ( Figure 3A) [24], suggesting that GATA-2 directly regulates HOXB4 transcription through its promoter. Thus, we evaluated the sequence of the HOXB4 promoter and focused on the GATA sequence at 2160/2157 ( Figure 3B), which is conserved between human and mouse [20], and also across species based on the UCSC genome browser (http://genome.ucsc.edu/). However, the surrounding sequence (GGATAA) does not completely coincide with a ''canonical'' GATA-1 binding motif, such as (A/T)GATA(A/G) or (C/G)(A/T)GATAA(G/A/C)(G/A/C) [6,9,24]. As it has been demonstrated that the incomplete GATA motif, including GATA(A/G), could be bound by GATA-1 [24], we hypothesized that GATA-2 binds to the GATA sequence at 2160/2157 of the HOXB4 promoter and confers a significant promoter activity. To test this possibility, we first conducted luciferase reporter assay in K562 cells using wild-type and GATA-deleted HOXB4 promoter constructs. As shown in Figure 3C, deletion of the GATA sequence significantly decreased reporter activity as compared to the wild-type promoter activity. Also, we analyzed if GATA-2 depletion could affect promoter activity of HOXB4. K562 cells were infected with control and GATA-2 lentiviral shRNA, and subsequently puromucin was added in culture media for selection. Western blotting and quantitative RT-PCR revealed knockdown of GATA-2 protein and mRNA expression ( Figures 3D). As shown in Figure 3E, GATA-2 knockdown significantly reduced the promoter activity of HOXB4.
Next, to assess whether GATA-2 binds to the GATA sequence of the HOXB4 promoter, EMSA analysis was conducted with nuclear extracts from K562 cells transiently overexpressing GATA-2. As shown in Figure 4A, a single major band was detected after incubation of nuclear extracts of K562 cells with a FITC-labeled oligonucleotide containing the 2160/2157 GATA sequence (lane 2). The intensity of this band increased in proportion to the extent of GATA-2 overexpression (lanes 3-4) and the band intensity decreased considerably when a 100-fold excess of unlabeled oligonucleotide was included in the binding reaction (lane 5). In contrast, a 100-fold excess of a GATAmutated oligonucleotide (TATC to AAAC) did not affect DNA binding (lane 6). Quantitative ChIP analysis in both K562 and cord blood-derived CD34-positive cells revealed GATA-2 occupancy at the HOXB4 promoter region ( Figure 4B). As a reference, we also examined GATA-2 occupancy at GATA-2 21.8 kb and NECDIN promoter regions, which are considered positive and negative control sites, respectively ( Figure 4B) [24]. These results suggested that GATA-2 directly regulates HOXB4 expression via binding to the 2160/2157 GATA sequence.

Significant correlation between GATA-2 and HOXB4 expression in clinical samples
To examine correlation between GATA-2 and HOXB4 expression, CD34-positive cells were isolated from BM mononuclear cells from 10 patients with AA and 13 patients with ITP after obtaining informed consent. We analyzed HOXB4 and GATA-2 mRNA levels by quantitative RT-PCR. As shown in Figure 5, we demonstrated that the expression of HOXB4 was significantly correlated with that of GATA-2 (r = 0.6573, p,0.01).

Discussion
In addition to immunological destruction of HSCs, in principle, altered gene expression in HSCs in AA may be an important determinant of AA pathology. Previously, we reported the decreased expression of GATA-2 in CD34-positive BM cells [16], and similar results were observed in an independent study [17]. In the present study, we identified the HOXB4 gene as a direct target of GATA-2 in HSCs. HOXB4 is predominantly expressed in HSCs and is particularly important for controlling HSC proliferation [26,27,28,29]. Therefore, we consider that downregulation of HOXB4 in CD34-positive cells appears to be related to one of the pathophysiological aspects of AA, i.e., a reduced HSC number.
We demonstrated that GATA-2 regulates HOXB4 mRNA level via direct binding to the 2160/2157 GATA site of the HOXB4 promoter (Figures 2,3,4). However, as it appears that GATA-2 has relatively modest effect on HOXB4 transcription in our analysis, additional factors may also affect HOXB4 expression. It has been reported that the HOXB4 promoter region from 2263 to 2116 relative to the translational start site is important for the activity in K562 cells [20], and that thrombopoietin [30], NF-Y [31], and USF1/2 [20,32] regulate HOXB4 transcription through the promoter region. It is still unknown whether these factors, including GATA-2, regulate HOXB4 transcription redundantly or cooperatively in HSCs. NF-Y is known to be a constitutive and Bone marrow cells were collected from 10 patients with AA and 13 patients with ITP after obtaining informed consent [16]. Mononuclear cells were isolated by Ficoll-Hypaque density gradient centrifugation. CD34-positive cells were then collected using the MACS system. GATA-2 and HOXB4 mRNA levels were examined by quantitative RT-PCR analysis, and were normalized relative to GAPDH mRNA expression. ubiquitous transcription factor [33], but it may participate in the regulation of some promoters in a lineage-specific manner by interacting with GATA-1 or GATA-2, such as those of globins, Gfi1b, KLF13, and FcgammaRIIA [34,35,36,37,38]. Thus, GATA-2 may function in a cooperative manner with NF-Y to induce HOXB4 in HSCs. Next, we demonstrated significant correlation between the expression of HOXB4 and GATA-2 in CD34-positive cells from patients of AA and ITP (Figure 5), suggesting a functional link between GATA-2 and HOXB4 in clinical samples. Furthermore, in the same series of samples, we also found that both HOXB4 and GATA-2 expression levels were significantly lower in CD34positive cells from AA patients as compared to those from ITP patients ( Figure S2). However, ITP patients might have a more active stem and progenitor compartment than normal controls, which would cause the differences in GATA-2 and HOXB4 expression to be exaggerated. Thus, further analyses are required to compare the expression of GATA-2 and HOXB4 in AA with ITP as opposed to CD34+ cells from healthy donors.
The mechanism underlying the decrease in GATA-2 expression in CD34-positive cells with AA is unclear. To date, several transcription factors and signaling molecules have been reported to induce and/or maintain GATA-2 expression in HSCs, such as Scl/TAL1 [24], Evi1 [39], Notch1 [40], BMP, and Wnt signaling [41]. Furthermore, it has been reported that GATA-2 could be a target of epigenetic inactivation by DNA methylation [42]. However, the significance of these factors in the pathogenesis of AA has not yet been evaluated. As shown in Figure 5, the significant correlation between GATA-2 and HOXB4 expression also suggested that HOXB4 might regulate GATA-2 expression.
Among several studies enumerating candidate target genes of HOXB4 [43,44,45,46,47], Oshima et al [46] and Fan et al [47] recently demonstrated that HoxB4 functions as a master regulator of hematopoiesis by directly regulating multiple hematopoietic transcription factors, including Runx1, Scl/TAL1 and Gata2, through integrated analysis of HoxB4 binding and gene expression profiling in murine embryonic stem (ES) cells. Therefore, the pathogenesis of AA might be caused by a defect in regulatory network among hematopoietic transcription factors, which would result in the decrease of GATA-2 expression. Alternatively, it is also possible that some cytokines, which are thought to play an role in immune-mediated suppression of HSCs, participate in the regulation of GATA-2. In support of this, Xu et al. recently demonstrated that GATA-2 expression is reduced by interferongamma in an in vitro analysis based on mesenchymal stem cells [48]. Further studies are required to elucidate the basis of GATA-2 downregulation in AA.
In summary, our results suggested that GATA-2 directly regulates HOXB4 expression in hematopoietic stem cells, which may play an important role in the development and/or progression of AA. Figure S1 siRNA-mediated knockdown of GATA-2 in CD34+ cells derived from cord blood. siRNA-mediated GATA-2 knockdown in cord blood-derived CD34-positive cells. Quantitative RT-PCR analysis was performed to detect GATA-2 and HOXB4 expression (mean 6 SE, n = 3). mRNA levels were normalized relative to 28S. *p,0.05. (EPS) Figure S2 HOXB4 expression is significantly decreased in CD34-positive cells in patients with AA and ITP. GATA-2 and HOXB4 mRNA levels were examined by quantitative RT-PCR analysis using CD34-positive cells from patients with AA (n = 10) and ITP (n = 13) ( Figure 5). The levels of expression of these genes were normalized relative to GAPDH mRNA expression. Student's unpaired t test was used for statistical analyses.

(EPS)
Table S1 Gene expression profiling by GATA-2 knockdown in cord blood-derived CD34-positive cells. A total of 1,126 upregulated and 572 downregulated genes after GATA-2 knockdown in cord blood-derived CD34-positive cells were shown. Each value was shown as global normalization value as described in Materials and Methods. (XLSX)