Novel Roles of GATA1 in Regulation of Angiogenic Factor AGGF1 and Endothelial Cell Function*

AGGF1 is an angiogenic factor, and its deregulation is associated with a vascular malformation consistent with Klippel-Trenaunay syndrome (KTS). This study defines the molecular mechanism for transcriptional regulation of AGGF1 expression. Transcription of AGGF1 starts at two nearby sites, −367 and −364 bp upstream of the translation start site. Analyses of 5′- and 3′-serial promoter deletions defined the core promoter/regulatory elements, including two repressor sites (from −1971 to −3990 and from −7521 to −8391, respectively) and two activator sites (a GATA1 consensus binding site from −295 to −300 and a second activator site from −129 to −159). Both the GATA1 site and the second activator site are essential for AGGF1 expression. A similar expression profile was found for GATA1 and AGGF1 in cells (including various endothelial cells) and tissues. Electrophoretic mobility shift assay and chromatin immunoprecipitation assays demonstrated that GATA1 was able to bind to the AGGF1 DNA in vitro and in vivo. Overexpression of GATA1 increased expression of AGGF1. We identified one rare polymorphism −294C>T in a sporadic KTS patient, which is located in the GATA1 site, disrupts binding of GATA1 to DNA, and abolishes the GATA1 stimulatory effect on transcription of AGGF1. Knockdown of GATA1 expression by siRNA reduced expression of AGGF1, and resulted in endothelial cell apoptosis and inhibition of endothelial capillary vessel formation and cell migration, which was rescued by purified recombinant human AGGF1 protein. These results demonstrate that GATA1 regulates expression of AGGF1 and reveal a novel role for GATA1 in endothelial cell biology and angiogenesis.

AGGF1 is an angiogenic factor, and its deregulation is associated with a vascular malformation consistent with Klippel-Trenaunay syndrome (KTS). This study defines the molecular mechanism for transcriptional regulation of AGGF1 expression. Transcription of AGGF1 starts at two nearby sites, ؊367 and ؊364 bp upstream of the translation start site. Analyses of 5and 3-serial promoter deletions defined the core promoter/regulatory elements, including two repressor sites (from ؊1971 to ؊3990 and from ؊7521 to ؊8391, respectively) and two activator sites (a GATA1 consensus binding site from ؊295 to ؊300 and a second activator site from ؊129 to ؊159). Both the GATA1 site and the second activator site are essential for AGGF1 expression. A similar expression profile was found for GATA1 and AGGF1 in cells (including various endothelial cells) and tissues. Electrophoretic mobility shift assay and chromatin immunoprecipitation assays demonstrated that GATA1 was able to bind to the AGGF1 DNA in vitro and in vivo. Overexpression of GATA1 increased expression of AGGF1. We identified one rare polymorphism ؊294C>T in a sporadic KTS patient, which is located in the GATA1 site, disrupts binding of GATA1 to DNA, and abolishes the GATA1 stimulatory effect on transcription of AGGF1. Knockdown of GATA1 expression by siRNA reduced expression of AGGF1, and resulted in endothelial cell apoptosis and inhibition of endothelial capillary vessel formation and cell migration, which was rescued by purified recombinant human AGGF1 protein. These results demonstrate that GATA1 regulates expression of AGGF1 and reveal a novel role for GATA1 in endothelial cell biology and angiogenesis.
The AGGF1 gene, previously known as VG5Q, encodes an angiogenic factor with 714 amino acid residues (1). AGGF1 was identified through genetic analysis of Klippel-Trenaunay syn-drome (KTS, MIM #149000), 2 which is a congenital vascular disorder composed of capillary malformations, venous malformations or varicose veins, and hypertrophy of the affected tissues (2)(3)(4)(5). KTS is a congenital disorder, but most cases are sporadic. The genetic basis of KTS is complex and may involve multiple genes, environmental factors, and their interactions (6). To date, identification of susceptibility genes associated with KTS has relied upon gross cytogenetic defects reported in KTS patients. Three chromosomal abnormalities have been identified in three separate KTS patients: two balanced translocations t(5.11)(q13.3;p15.1) and t(8,14)(q22. 3;q13), and an extra supernumerary ring chromosome 18 (7)(8)(9). Chromosomal breakpoints involved in KTS translocation t(5;11)(q13.3; p15.1) have been fully characterized. No gene has been identified within a 100-kb region flanking the chromosome 11p15.1 translocation breakpoint. In contrast, the chromosome 5p13.3 breakpoint is located in the promoter/regulatory region of the AGGF1 gene and leads to increased transcriptional activation of AGGF1 by 3-fold (1). The results suggest that deregulation of AGGF1 is associated with KTS. However, the molecular mechanism for the deregulation is not known. In this study, we defined the promoter of AGGF1 and important cis-acting DNA elements or trans-acting nuclear factors in the regulation of the AGGF1 gene. We show that translocation t(5:11) increases transcription of AGGF1 by removing cis-acting DNA elements that repress expression of AGGF1. AGGF1 protein contains a N-terminal coiled-coil motif, an OCRE motif, a forkhead-associated domain, and a C-terminal G-patch domain (1,10). Purified human recombinant AGGF1 protein promotes angiogenesis as potently as VEGF (1). AGGF1 protein is released outside endothelial cells when angiogenesis starts (1). It binds strongly to endothelia cell surface, and may act in an autocrine fashion (1). Strong expression of AGGF mRNA was detected in cells relevant to KTS, including endothelial cells, vascular smooth muscle cells, and MG-63 osteoblasts (1). Tissue immunostaining studies with an anti-AGGF1 antibody identified strong AGGF1 protein expression in blood * This study was supported in part by a Scott Hamilton CARES research grant vessels embedded in various tissues including the heart, kidney, tail, and limb and co-localized with an endothelial specific marker CD31 as well as a vascular smooth muscle cell-specific marker, smooth muscle cell ␣-actin (1). In a small case control study, the frequency of a single nucleotide polymorphism (SNP) in AGGF1, E133K, was found to be greater in cases (3.8%) than in controls (1), but later studies found that E133K showed a frequency of 2.2-3.3% in other general control populations (6,11,12). These results argue that SNP E133K is unlikely to confer a risk of KTS. On the other hand, a recent large scale case control study employing a STRUCTURE program demonstrates that two common SNPs in AGGF1, exonic SNP rs7704267 and intronic SNP rs13155212, are significantly associated with susceptibility of KTS even after adjustment of population structural parameters of the cases and controls (6). Therefore, AGGF1 remains a strong candidate gene associated with risk of KTS.
GATA factors are important transcription factors that mediate cell-specific gene expression. There are six members in the GATA family of transcription factors. GATA1 is a key transcription factor that is central to the differentiation, proliferation, and/or apoptosis of erythroid (13), megakaryocytes (14), eosinophilic cells (15), and mast cells (16). However, the potential role of GATA1 in endothelial cells has not been studied. In this study, we uncovered a novel role of GATA1 in endothelial cells through the promoter analysis of AGGF1. We identified an AGGF1 promoter SNP, Ϫ294CϾT, that affects a cis-acting DNA element at this location that preferentially interacts with GATA1. SNP Ϫ294CϾT disrupted GATA1 binding to DNA in endothelial cells and markedly reduced transcription activation of AGGF1. Furthermore, siRNA against GATA1 effectively knocked down expression of GATA1, reduced expression of AGGF1, resulted in cell apoptosis, and subsequently inhibited endothelial cell vessel formation and cell migration. The effects by GATA1 siRNA were rescued by recombinant human AGGF1 protein. Together, these results suggest that GATA1 regulates expression of AGGF1 in endothelial cells and is involved in AGGF1-mediated angiogenesis and other endothelial cell phenotypes.

EXPERIMENTAL PROCEDURES
Study Subjects-185 KTS patients were enrolled in North America for this study. The diagnosis of KTS was based on published reports (2)(3)(4). This study has been approved by the Cleveland Clinic Foundation and Mayo Clinic Institutional Review Boards on Human Subject Research. Informed consent was obtained from all participants according to the standards established by the local Institutional Review Boards.
Identification of SNPs in AGGF1-SNP identification was carried out using direct DNA-sequencing analysis. The 2-kb promoter/regulatory region/5Ј-untranslated region of AGGF1 was PCR-amplified using two pairs of primers (supplemental Table 1) and sequenced.
TaqMan SNP Assays-The frequency of SNP Ϫ294CϾT variant in the AGGF1 promoter/regulatory region was analyzed in 1764 non-KTS control samples using the TaqMan 5Ј-allelic discrimination assay as described previously (17)(18)(19)(20). The assay probes (supplemental Table 1) were ordered using the Assay-By-Design service from the Applied Biosystems.
Further deletions were created based on the 1.1kb-AGGF1pluc reporter gene using a PCR-based method. For the 5Ј-deletion series, the forward primers were designed and contained a unique NheI site. The reverse primer was designed based on the vector sequences after the unique XhoI site. Each PCR fragment was cut with NheI and XhoI and cloned into the NheI/ XhoI-cut 1.1-kb AGGF1p-luc plasmid. For the 3Ј-deletion series, the forward primer was designed based on the sequence spanning the NheI site. The reverse primers were designed to contain a unique XhoI site. Each PCR fragment was cut with NheI and XhoI and cloned into the NheI/XhoI-cut 1.1kb-AGGF1p-luc plasmid.
SNP-294CϾT in the AGGF1 promoter/regulatory region was introduced into the core Ϫ536-bp-AGGF1p-luc reporter gene or Ϫ8.4kb-AGGF1p-luc reporter gene by site-directed mutagenesis using the mega-primer PCR-based method (21). The deletions involving the two activator sites were created by PCR and subcloning. The PCR primers used for creating deletions and SNP Ϫ294CϾT are shown in supplemental Table 1. All mutant constructs were verified by DNA sequencing analysis.
Cells Culture and Transfection-Human umbilical vein endothelial cells (HUVECs) and other cells were cultured as described (1) and transfected by electroporation using a Nucleofector device and HUVEC kits (Amaxa, Inc., Cologne, Germany). Transfection of HEK293 cells was carried out using Lipofectamine 2000 (Invitrogen) as described previously (22,23).
Transcriptional Assays-The transcription activation activity was measured by the luciferase assay as described (24,25).
RT-PCR and Western Blot Analyses, and Immunofluorescence Staining-Total RNA was isolated from HUVECs and other cells using the TRIzol reagent (Invitrogen), treated with DNase I (Roche Applied Science, Indianapolis, IN), and used for RT-PCR analysis as described (1). Western blot analysis and immunostaining studies were performed from various cells and mouse tissues as described previously (1).
Primer-extension Analysis-The exact transcription start sites of AGGF1 were determined by primer extension analysis using the Primer Extension System AMV Reverse Transcriptase Kit (Promega) and as described before (26). The primer extension products were run in parallel with a DNA sequence ladder obtained by cycle sequencing using the same [␥-32 P]ATP-labeled primer extension primer with the AGGF1p-luc plasmid DNA as the template as described before (27).
Preparation of Nuclear Lysates and EMSAs-Nuclear extracts for HUVECs or transfected HEK 293 cells were prepared using NE-PER Nuclear and Cytoplasmic Extract kits (Pierce). The probes for EMSA were designed based on the sequences from the AGGF1 promoter/regulatory region (supplemental Table 1). Positive control probes for GATA1 binding and TFII-I binding sites were described (28,29) and are shown in supplemental Table 1. EMSA was carried out as described (30,31).
For supershift EMSA, a rat anti-GATA1 antibody (sc-266) and goat anti-GATA2, GATA3, GATA4, and GATA6 antibodies (1 g, Santa Cruz Biotechnologies, Santa Cruz, CA) were added to the reaction mixture, and the mixture was incubated on ice for 15 min before addition of the probe. For competition EMSA experiments, excessive unlabeled probes were added to the binding reaction mixture before addition of the labeled probes.
Chromatin Immunoprecipitation Assays-ChIP assays were carried out with solutions prepared following the protocol from Upstate Biotechnology (Lake Placid, NY). Chromatin was sheared by sonication for 15 min to short fragments of ϳ200 to 1000 bp in a water bath with generation of high power ultrasound (15 cycles of 30 s on and 30 s off, 1 cycle/min) at the maximum power. To reduce nonspecific background, protein A-agarose (Pierce) was presaturated with herring-sperm DNA (Sigma). Immunoprecipitation was performed with 1 g of a rat anti-GATA1 antibody (Santa Cruz Biotechnology). The normal anti-rat IgG was used as a negative control. After immunoprecipitation, the mixture was extracted with phenol/chloroform and precipitated with ethanol. Immunoprecipitated DNA was analyzed by PCR. ChIP assays were replicated three times. The PCR primers used for ChIP assays were designed based on the core promoter/regulatory region of AGGF1 and are shown in supplemental Table 1.
Matrigel Endothelial Vessel Formation Assays-The property of HUVECs to spontaneously form capillary vessels in Matrigel basement membrane matrix (BD Biosciences, Oxford, UK) was assessed as described previously (1,32). For the AGGF1 rescue experiments, purified recombinant human AGGF1 protein was mixed with Matrigel at 4°C, which was placed back to an incubator for 30 min, resulting in solid Matrigel ready for Matrigel vessel formation assays.
Analysis of Cell Migration by a Scratch Assay-HUVECs treated with or without siRNA were plated on polylysine-coated two-well chamber slides (BD Bioscience) at 5 ϫ 10 4 cells/well in endothelial basal medium supplemented with EGM-2 SingleQuots (IGF1, epidermal growth factor, and VEGF) (Lonza, Walkersville, MD), FIGURE 1. Mapping of the TSSs of the AGGF1 gene. A, the nucleotide sequence of the human AGGF1 promoter/regulatory region is shown. Only the region from Ϫ1045 bp to ϩ9 bp from the translation start site is shown.
The A residue of the start codon ATG is designated as "ϩ1." B, identification of the TSSs of AGGF1 by RT-PCR analysis. RT-PCR was performed using total RNA isolated from HUVECs. R, reverse primer; P1 to P7, a series of forward primers with locations indicated by the number of bp from the translation start site; top panel, results from RT-PCR; bottom panel, positive control for PCR primers (regular PCR with genomic DNA). C, precise mapping of the TSSs for human AGGF1 gene using primer extension analysis. Primer extension reactions were performed with total RNA samples isolated from HUVECs. The extended products were analyzed with a 6% denatured urea polyacrylamide gel together with a sequencing ladder generated using the same primer and plasmid DNA samples. The nucleotide sequence readout is shown on the right. The TSSs are shown with stars.
which was changed every 24 h. Two days after plating, a scratch was applied using a 20-l pipette tip. Chambers were washed with endothelial growth media and replaced with endothelial basal medium supplemented with EGM-2 SingleQuots (Lonza). 16 h after the scratch, cells were photographed. For the AGGF1 rescue experiments, purified recombinant human AGGF1 protein or bovine serum albumin control was coated on slides, followed by the scratch assay as described above.
Apoptosis Assays-HUVECs were treated with GATA siRNA or control scramble siRNA. After 24 h, cells were harvested, and apoptosis was analyzed using flow cytometry that detects DNA breaks labeled by a fluorescein anti-bromodeoxyuridine antibody and total cellular DNA labeled by propidium iodide (APO-BRDUTM Kit, BD Pharmingen).
Statistical Analysis-Data are shown as mean Ϯ S.E. Statistical analysis was performed using Student's t test for comparing two groups and analysis of variance for comparisons among groups. A p value of 0.05 was considered to be significant.

Identification of the Transcription Start Sites of the AGGF1
Gene in HUVEC-BLAST searches of public databases, including the NCBI data base, identified five cDNA or expressed sequence tags clones that match the AGGF1 genomic sequence, including HSU84971, AI939311, AA311507, BX426365, and BX442568. The 5Ј-start sites of these five clones are 289, 327, 334, 338, 349, or 360 bp from the translation start site, respectively. The data suggest that the transcription start site (TSS) of AGGF1 is at least 360 bp from the translation start site ATG (we designate the position of the A residue of codon ATG as ϩ1 throughout the text) (Fig. 1A).
RT-PCR analysis was used to experimentally map the TSS of AGGF1. We designed a reverse primer (R) located at the position of Ϫ206 and a series of forward primers at positions of Ϫ625 (P7), Ϫ559 (P6), Ϫ507 (P5), Ϫ431 (P4), Ϫ400 (P3), Ϫ324 (P2), and Ϫ256 (P1), respectively (Fig. 1B). RT-PCR analysis of total RNA isolated from HUVECs with primer combinations P1/R and P2/R yielded positive signals, whereas other combinations did not produce any PCR signal. The data suggests that the TSS of AGGF1 is located between positions Ϫ400 and Ϫ324.
To map the TSS of AGGF1 more precisely, primer extension analysis was carried out. As shown in Fig. 1C, transcription of AGGF1 starts at two sites that are 2 bp apart. The first TSS is at the position of Ϫ367 and the second one is at Ϫ364 (Fig. 1, A and C). The second TSS appears to be used more frequently than the more upstream one (Fig. 1C).
Structural Characteristics of the Core Promoter of AGGF1-DNA sequences for a region of 1045 bp upstream from the AGGF1 translation start site are shown in supplemental Fig. S1. Notably, the core promoter/regulatory region of AGGF1 is highly GC-rich (62%). There are more than 50 CpG dinucleotide repeats, including 5 HpaII/MspII restriction sites (CCGG) (supplemental Fig. S1).
Identification of Two Upstream Cis-acting DNA Elements That Repress the Transcription of AGGF1-A series of six 5Ј-deletions were created for the AGGF1 promoter in 8.4kb-AGGF1p-luc (Fig. 2). The deletions were transiently transfected into HUVEC, and transcriptional activity was measured. Compared with the promoter-less reporter, the Ϫ1971 AGGF1p-luc exhibited a 600-fold increase of transcriptional activity.
Removal of sequences from Ϫ8391 to Ϫ7521 increased transcription activity of AGGF1 by 2-fold, suggesting that there is a cis-acting DNA element at the region that represses expression of AGGF1. Removal of sequences from Ϫ3990 to Ϫ1971 increased AGGF1 transcription activity by an additional 3-fold, indicating the presence of the second repressor element within the promoter/regulatory region of AGGF1 (Fig. 2).
Identification of a Cis-acting DNA Element from Ϫ129 to Ϫ159 That Increases Transcription of AGGF1-To map the region responsible for the basal promoter activity, additional nine 5Ј-serial promoter truncations, including Ϫ1036, Ϫ936, Ϫ836, Ϫ736, Ϫ636, Ϫ536, Ϫ286, Ϫ236, and Ϫ36 deletions, were created and analyzed (Fig. 3A). Removal of sequences between Ϫ536 and Ϫ286 drastically reduced transcription of the AGGF1 promoter. These data suggest that the basal promoter of AGGF1 is located between Ϫ536 and Ϫ286 from the translation start site, which is consistent with earlier results that the transcription start sites (Ϫ367 and Ϫ364) of AGGF1 are located in this region. A deletion of the region from Ϫ286 to Ϫ136 reduced the AGGF1 expression by another 2-fold, which may implicate a weak promoter in the region. The luciferase activities for the Ϫ136 and Ϫ36 deletions were low (5.5 Ϯ 0.7 and 5.0 Ϯ 0.3, respectively), but still 5-fold higher than the calibration value of 1 for the empty vector, which may reflect nonspecific activation or implicate another weak promoter in the region.
Analyses of additional ten 3Ј-serial promoter deletions revealed an interesting cis-acting DNA element that is essential for the expression of AGGF1 (Fig. 3B). The cis-acting DNA element is located between Ϫ129 to Ϫ159 from the translation start site (Fig.  3B, 3C).
Identification of an SNP, Ϫ294CϾT, at a GATA1 Binding Site in the Promoter/Regulatory Region of AGGF1 in a Patient with KTS-One novel SNP, which changes C to T at the Ϫ294, was identified in the promoter region of AGGF1 in a sporadic male KTS patient (Fig. 4A). The patient was an adopted child who was affected with a capillary malformation of the abdomen, a large capillary malformation of the right leg, hypertrophy of the right leg, and important venous malformations. The patient underwent several operations and hospitalizations because of symptoms of KTS. SNP Ϫ294CϾT was not present in 1764 control individuals, suggesting that it is a rare SNP. SNP Ϫ294CϾT occurs at a consensus binding site for GATA1 (Ϫ295 to Ϫ300) or TFII-1 (Ϫ293 to Ϫ298), but later studies indicate that it is a GATA1 binding site (Fig. 3C).

GATA1 Interacts Directly with the AGGF1 Promoter in Endothelial
Cells-An EMSA was performed using a double-stranded oligonucleotide encompassing the region from Ϫ301 to Ϫ284 (EMSA1, Fig.  4B) and HUVEC nuclear extracts. As shown in Fig. 4C, incubation of nuclear extracts from HUVECs with the 32 P-labeled EMSA1 resulted in the formation of a DNAprotein complex. The DNA-protein interaction appeared to be specific because it was eliminated by addition of 50-fold excess of unlabeled EMSA1 but not affected by addition of 50-fold excess of three nonspecific control doublestranded oligonucleotides (NS1, NS2, and NS3) (supplemental Fig.  S2). A 50-fold excess of an unlabeled GATA1 oligonucleotide from the human ␤-globin promoter (EMSA2) (28) eliminated the DNA-protein complex, but even a 300-fold excess of a consensus TFII-I oligonucleotide from the c-fos promoter (EMSA3) (29) did not have any effect (supplemental Fig. S2). These data suggest that the DNA-protein complex is a complex with GATA1. This conclusion is further confirmed using supershift EMSA studies. The specific DNA-protein complex was shifted in EMSA by using nuclear extracts preincubated with an anti-GATA1 antibody, but not with antibodies against other GATA factors expressed in endothelial cells, including GATA2, GATA3, GATA4, GATA6, and negative control IgG (Fig. 4D). Together, these data suggest that the interval from Ϫ301 to Ϫ284 of the AGGF1 promoter/regulatory region can interact preferably with GATA1.
To verify that the endogenous, native GATA1 protein binds to the AGGF1 promoter in vivo, a conventional ChIP assay was performed using a specific monoclonal GATA1 antibody and specific primers for the AGGF1 promoter region (for primer sequences, see supplemental Table 1). In ChIP assays using HUVEC extracts, GATA1 specifically binds to the AGGF1 promoter (Fig. 4E). The same results were obtained  2-12) or without (lane 1) nuclear extracts from HEK293 cells transfected with GATA1. Lanes 3-7, EMSA with preincubation of a rat anti-GATA1 or goat anti-GATA2, -GATA3, -GATA4, and -GATA6 antibodies, respectively. Lanes 8 and 9, negative controls. SS complex: supershifted complex. Similar results were obtained with nuclear extracts from HUVECs (data not shown). E, ChIP analysis detected binding of the GATA1 protein from HUVEC or HEK293 cells with transient expression of GATA1 to the AGGF1 promoter in vivo. Primers that can amplify the AGGF1 promoter were used for the PCR analysis. GATA1 Ab, presence of an anti-GATA1 antibody. 1kb, ChIP with PCR primers covering the AGGF1 promoter fragment with the GATA1 binding site (located within a 1-kb region upstream from the translation start site); 4kb and 8kb, ChIPs with PCR primers covering other AGGF1 promoter fragments without the GATA1 binding site (located 4 kb and 8 kb upstream from the translation start site, respectively). using protein extracts from HEK293 cells transiently transfected with a GATA1 expression plasmid (Fig. 4E). These data further indicate that GATA1 can interact with a cis-DNA element in the AGGF1 promoter/regulatory region.
GATA1 mRNA and Protein Are Expressed in Endothelial Cells-Immunostaining studies showed that GATA1 was expressed strongly in the nuclei of some representative endothelial cells, including HUVECs and HBMECs (human brain microvascular endothelial cells) (Fig. 5). Similarly, GATA1 was strongly colocalized with von Willebrand factor in the endothelium of large vessels in the mouse heart and human aortas (Fig. 5).
Semi-quantitative RT-PCR and Western blot analyses showed that GATA1 was expressed in various tissues including the heart, liver, brain, lung, kidney, aorta, and bone marrow, and its expression was higher in the aorta than in other organs tested (Fig. 6, A and B). GATA1 expression was detected in three cell lines (HEL, THP1, and U937) and three types of endothelial cells (HUVECs, HBMECs, and human coronary artery endothelial cells (Fig. 6, A and B)). The expression levels in HEL cells were higher than that in endothelial cells.
Western blot analysis revealed that expression of the GATA1 protein was restricted to the liver in neonatal mice at the age of P2 (Fig. 6C) and P3 (data not shown), consistent with the results by Northern blot analysis (32). Starting at P5, GATA1 expression is detected in other tissues/organs (Fig. 6C). Similar expression patterns were observed for AGGF1 except that a low level of AGGF1 expression was detected in the heart, brain, and kidney at the age of P2 (Fig. 6C).
GATA1 Trans-activates the AGGF1 Promoter-To assess the effect of GATA1 on the AGGF1 promoter activity, co-transfection experiments were performed using the core Ϫ536-AGGF1p-luc construct (containing nucleotides from Ϫ536 to Ϫ129 of the AGGF1 promoter/regulatory region) and an expression construct containing the GATA1 cDNA or an empty expression vector as control. Co-transfection of the GATA1 expression construct into HUVECs resulted in markedly increased transcription activation of the AGGF1 promoter compared with the empty expression vector (Fig. 7A). No luciferase activity was detected with the parental pGL3-basic luciferase vector together with the GATA1 expression vector, indicating that the transactivation of the AGGF1 promoter requires the presence of the DNA binding site for GATA1. Similar results were obtained with a longer version of the AGGF1pluc reporter gene, the Ϫ8.4kb-AGGF1p-luc luciferase reporter (Fig. 7A, right panel).
GATA2 Does Not Regulate Expression of AGGF1-GATA2 is highly expressed in endothelial cells and control expression of VEGFR2 (33,34), thus we determined whether GATA2 could regulate expression of AGGF1. Knockdown of GATA2 by siRNA does not affect expression of AGGF1 in HUVEC by both RT-PCR and Western blot analyses (supplemental Fig. S3A). HUVECs with overexpression of GATA2 did not increase the transcriptional activity of AGGF1 promoter (supplemental Fig.  S3B). These data indicate that GATA2 does not affect expression of AGGF1.
SNP Ϫ294CϾT Weakens the GATA1-DNA Complex-To determine whether the formation of GATA1-DNA complex is affected by the AGGF1 promoter SNP Ϫ294CϾT identified in a KTS patient, we carried out EMSA experiments using a mutant EMSA1 probe containing the SNP (Fig. 7B). As shown in Fig.  7B, more GATA1-DNA complex was formed with the wildtype probe with the C allele than the mutant probe with the T allele.
SNP Ϫ294CϾT Reduces Expression from the AGGF1 Promoter-Transcriptional activity of the AGGF1 promoter was markedly reduced by 75.2% after introduction of the Ϫ294T allele (Fig. 7A). The AGGF1p-luc reporter was not responsive to increased expression of GATA1 (Fig. 7A). These data suggest that AGGF1 promoter SNP Ϫ294CϾT reduces AGGF1 expression. Both the GATA1 Site and the Second Activator Site from Ϫ129 to Ϫ159 Are Essential for Expression of AGGF1-A deletion of the GATA1 site reduced expression of AGGF1 by more than 4-fold, and so does the deletion of the second activator site from Ϫ129 to Ϫ159 (Fig. 7C). A deletion of both sites abolished the transcriptional activity of the AGGF1 promoter to almost the basal level (Fig. 7C). These data suggest that the GATA1 site and the second activator site are essential for expression of AGGF1.
Knockdown of Endogenous Expression of GATA1 by siRNA Impairs Endothelial Vessel Formation and Endothelial Cell Migration and Induces Apoptosis-To determine the cellular roles of GATA1-mediated AGGF1 expression, we knocked down expression of GATA1 in HUVECs and studied its potential roles in endothelial tube formation, migration, and apoptosis. A specific siRNA targeted to GATA1 or a scramble siRNA was transfected into HUVECs. Semi-quantitative RT-PCR analysis showed that, compared with the control scramble siRNA, GATA1 siRNA reduced expression of GATA1 by 80%. The expression level of AGGF1 was also decreased by 70%, but expression of control genes, including VEGF, GATA2, and GAPDH, was not affected by GATA1 siRNA (Fig. 8A). A similar level of inhibition of GATA1 and AGGF1 expression by the GATA1 siRNA was achieved at the protein level (Fig. 8B).
It has been reported previously that an antisense oligonucleotide and siRNAs targeted AGGF1 inhibited endothelial vessel formation in an in vitro Matrigel angiogenesis assay (1). Because GATA1 regulates the expression of AGGF1, we examined the role of GATA1 siRNA in endothelial vessel formation. 4 M GATA1 siRNA was electroporated into HUVECs and plated on Matrigel-coated plates to allow capillary vessel formation. The formation of mature, well connected vessels was counted under an inverted phasecontrast microscope (40ϫ magnifi-cation) in 12 randomly selected fields in 3 wells. GATA1 siRNA inhibited endothelial vessel formation (Fig. 8C).
GATA1 siRNA was transfected into HUVECs, and its effects on endothelial cell migration and apoptosis were also examined. GATA1 siRNA reduced HUVEC migration (Fig. 8D) and increased apoptosis (Fig. 8E). The control scramble siRNA did not affect HUVEC migration or apoptosis (Fig. 8, D and E).
Purified Recombinant Human AGGF1 Rescues Effects by GATA1 siRNA-The AGGF1 protein is released outside of HUVECs when angiogenesis starts, and purified human AGGF1 protein can promote strong angiogenesis in a chicken embryo angiogenesis assay (1). Here we assessed whether purified recombinant human AGGF1 protein can rescue the effects of GATA1 siRNA (Fig. 9A). HUVECs were transfected with GATA1 siRNA and used for endothelial vessel formation, migration, and apoptosis in the presence of 6.4 g of recombinant human AGGF1 protein or control bovine serum albumin. Interestingly, more mature endothelial vessels were formed with AGGF1/Matrigel mixture than with control bovine serum albumin/Matrigel (Fig. 9B), suggesting that inhibition of endothelial vessel formation by GATA1 siRNA can be rescued by recombinant human AGGF1 protein applied externally. In the scratch cell migration assay, recombinant human AGGF1 protein rescued the inhibition of HUVEC migration mediated by knockdown of GATA1 expression by siRNA (Fig. 9C). In the endothelial cell apoptosis assay, purified human AGGF1 protein was able to rescue GATA1 siRNA-medicated HUVEC apoptosis (Fig. 9D).

DISCUSSION
AGGF1 plays a role in angiogenesis and altered expression of AGGF1 is associated with vascular malformations consistent with KTS. Here, we demonstrate that GATA1 is involved in transcriptional activation of the AGGF1 gene. We identified a consensus DNA binding site for GATA1 centered at Ϫ295 bp to Ϫ298 bp from the translation start site. Both EMSA and ChIP studies demonstrated that GATA1 interacted specifically with the GATA1 binding site (Fig. 4). Overexpression of GATA1, but not GATA2, increased transactivation of the AGGF1 promoter ( Fig. 7 and supplemental Fig. S3). Knockdown of GATA1 expression, but not GATA2 expression, by siRNA decreased expression of AGGF1 ( Fig. 8 and supplemental Fig. S3). These data indicate that GATA1 is an important regulator of the AGGF1 gene.
Previous studies showed that purified human AGGF1 promoted angiogenesis as potently as VEGF and knockdown of AGGF1 expression inhibited endothelial vessel formation (1). Similarly, knockdown of GATA1 expression reduced expression of AGGF1 and resulted in inhibition of endothelial vessel formation in an in vitro Matrigel angiogenesis assay (Fig. 8).
The current study shows that knockdown of GATA1 expres-sion inhibited endothelial cell migration and induced endothelial cell apoptosis (Fig. 8). The effects of GATA1 siRNA on endothelial vessel formation, endothelial cell migration, and apoptosis can be rescued by purified recombinant human AGGF1 protein (Fig. 9). These data suggest that the effects of GATA1 on endothelial cell phenotypes may be through regulation of the expression of AGGF1. Significantly, our results suggest that the function of GATA1 is not necessarily restricted to the hematopoietic linkage cells, and on the contrary, this study uncovers a novel role of GATA1 in endothelial cell biology and angiogenesis.
The AGGF1 promoter/regulatory region have several interesting features. First, it lacks the TATA box and has two transcription start sites located Ϫ367 bp and Ϫ364 bp from the translation start site. Thus, the AGGF1 promoter joins a growing list of vascular genes that use a TATA-less promoter and possess multiple, closely spaced transcription initiation sites, including the promoters for GPIIb1, ␣ 2 and ␣ 5 integrins, platelet GPIX, CD11a, and CD11b (35)(36)(37)(38)(39)(40). Second, the sequences flanking the TSS and translation start site of the AGGF1 gene are highly C/G-rich and contain more than 50 CpG islands. The CpG islands can become the target of DNA methylation. Also, they differ from other chromatin in several other respects, including a reduction in histone H1 and a general absence of nucleosomes in the region (41). Methylation may be one mechanism by which the expression of the AGGF1 gene is regulated at the "pre-transcriptional" level. Third, the AGGF1 gene uses a functionally analogous initiator element (first described for the terminal deoxytransferase gene promoter (42) to direct transcription initiation.
Functional results in this study demonstrate that, in addition to GATA1, several other putative trans-acting factors may regulate expression of AGGF1. It is possible that the AGGF1 promoter is regulated both positively and negatively by other transacting factors. First, there are two cis-acting DNA elements that repress expression of AGGF1, one located from Ϫ8391 bp to Ϫ7512 bp (repressor 1) from the translation start site, and the other located from Ϫ3990 bp to Ϫ1971 bp (repressor 2) (Fig.  2B). Future studies are needed to precisely define the minimum sequences for the two repressors and to identify potential proteins that bind to the two sites. Second, the 536-bp DNA fragment from AGGF1 translation start site is capable of driving the highest expression of the AGGF1 gene (Fig. 3). The full expression of AGGF1 requires GATA1 that binds to a consensus GATA1 DNA binding site centered from Ϫ295 bp to Ϫ300 bp, and a cis-acting element located from Ϫ159 to Ϫ129 bp from the translation start site (Fig. 3). Both the GATA1 site and the second activator site are essential for expression of AGGF1 (Fig.  7C). The AGGF1 Ϫ295 to Ϫ300 GATA site consists of GGATAA, a deviated version of the canonical GATA consen-FIGURE 6. Detection of GATA1 expression in a variety of tissues and cells. A, RT-PCR analysis. Control, a mammalian expression plasmid for GATA1; No RT, negative control with RNA but without reverse transcriptase in the reaction. B, Western blot analysis. Control, nuclear extracts from HEK293 cells transfected with a mammalian expression plasmid for GATA1. The tissues samples, including the heart, liver, brain, lung, kidney, aorta, and bone marrow were from mice. Note that the lanes for the heart and liver are reversed in B compared with A and C. HEL, THP1, and U937 are different types of human cells. HUVEC, human umbilical vein endothelial cells; HBMEC, human brain microvascular endothelial cells; COAEC, human coronary artery endothelial cells. Housingkeeping gene GAPDH was used as a loading control in both RT-PCR and Western blot analyses. The bands from RT-PCR and Western blot analyses were scanned, quantified, and plotted after calibration with GAPDH bands. C, expression profile of GATA1 and AGGF1 in neonatal mice at the age of P2 and P5 by Western blot analysis. sus sequence, (A/T)GATA(G/A). Interestingly, such deviation appeared to abrogate the binding of GATA2, -3, -4, and -6 but not GATA1, as demonstrated by the supershift assay with GATA antibodies (Fig. 4D). Thus, unlike other members of the GATA family, GATA1 possesses the unique binding specificity to GGATAA in the AGGF1 promoter. It is such a unique binding affinity of GATA1 to the AGGF1 promoter that defines GATA1 as the specific regulator of AGGF1 expression. However, this could not explain why the KTS promoter polymorphism reduced GATA1 binding, because the KTS polymorphism resulted in the reversal of the deviated sequence back to the canonical GATA consensus sequence AGATAA. It would be interesting for future studies to determine whether the flanking sequence is involved in regulating GATA1 binding to AGGF1 promoter. It is unknown how the second cis-acting activator element regulates the expression of AGGF1, but it is notable that it contains three direct CA(G/T)GG repeats (5-GTGAGTTTCAGGGCGTCATGGCCAGGGG-CCA-3Ј). This cis-acting element may increase the expression of AGGF1 by binding a transcription factor or by its unique structure feature. Third, the core AGGF1 promoter showed potential binding sites for a number of well known transcription factors, including GR, HIF-1, C/EBPb/a, NF-B, Elk-1/c-Ets, MZF1, Th1/E4, STATX, and NKX2. Whether these factors regulate expression of AGGF1 or not is a question that needs to be addressed in the future.
The genetic basis of KTS largely is unknown. Molecular characterization of a translocation t(5;11) associated with KTS has led to the molecular cloning of AGGF1. We previously reported that translocation t(5:11) increased the transcription activity of AGGF1 by 3-fold. The chromosome 5p13.3 translocation breakpoint is located Ϫ1644 from the translation start site ( Fig.  2A). Translocation t(5;11) is expected to remove the two repressor sites (Ϫ8391 to Ϫ7512; Ϫ3990 to Ϫ1971), resulting in increased AGGF1 expression. The present study identifies a Ϫ294CϾT polymorphism in the AGGF1 promoter/regulatory region in a severe KTS case (large capillary malformations, important venous malformations, hypertrophy of the right leg, multiple surgeries, and hospitalizations). The Ϫ294CϾT polymorphism reduces binding of GATA1 to the AGGF1 promoter and dramatically decreases transactivation of AGGF1, indicating that it is a functional variant. The Ϫ294CϾT polymorphism was identified in one of 185 KTS patients, but in none of 1764 non-KTS controls. Fisher's exact test revealed a trend, but not significant association with KTS (two-tailed p ϭ 0.095). The KTS patient was an adopted child, and studies on family members are not possible. One final note is that Ϫ294CϾT acts by a loss-of-function mechanism, whereas translocation t(5:11) acts by a gain-of-function mechanism. One interpretation of the data is that Ϫ294CϾT is merely a rare polymorphism that is not associated with KTS. The alternative interpretation is that both FIGURE 7. Regulation of AGGF1 expression by the GATA1 site and the second activator site. A, overexpression of GATA1 increases expression of AGGF1 and SNP Ϫ294CϾT reduces transcription activation of the AGGF1 promoter. Transcriptional activity for wild-type Ϫ294C and mutant Ϫ294T Ϫ536-AGGF1p-luc reporter genes (left panel) or longer 8.4kb-AGGF1p-luc reporter genes (right panel) are shown. HUVECs were co-transfected with a reporter genes and pcDNA3-GATA1 mammalian expression plasmid. Transcriptional activities are shown as relative luciferase activities on y axis. GATA1 strongly activates transcription of the wild-type core AGGF1 promoter, but not the mutant promoter. Western blot analysis showed an increased expression level of GATA1 in HUVECs transfected with a GATA1 expression construct compared with HUVECs transfected with the expression vector. GAPDH was used for loading control. B, SNP Ϫ294CϾT affects binding of GATA1 to DNA. Sequences of EMSA probes for wild-type (C) or mutant GATA1 binding sites (left) and results of EMSA (right) are shown. Lane 1, 32 P-labeled C probe alone; lane 2, EMSA for 32 P-labeled C probe and HUVEC nuclear extracts; lane 3, EMSA for 32 P-labeled mutant T probe and HUVEC nuclear extracts. C, transcriptional activity assays for mutant AGGF1 promoters with a deletion of the GATA1 site (M1), the second activator site (M2), or both (M3).
loss-of-function and gain-of-function mechanisms of AGGF1 are associated with risk of KTS. There are precedents that both loss-of-function and gain-of-function mutations in the same gene cause one disease. For example, both loss-of-function and gain-of-function mutations in TBX5 cause Holt-Oram syndrome (24,25,43). Expression of a key angiogenic factor like AGGF1 is under delicate regulation in cells, and either up-regulation or down-regulation of AGGF1 can have a deleterious effect and increased risk of KTS. Heterozygous VEGF knockout mice died during embryogenesis (E9.5), and 2-to 3-fold overexpression of VEGF also led to embryonic lethality (E12.5-E14) (44). It should be noted that the in vivo effect of SNP Ϫ294CϾT in human tissue could not be assessed due to lack of human samples from the SNP carrier. Future studies are needed to test whether the association between SNP Ϫ294CϾT and KTS can be further established.
In conclusion, the present study defines molecular mechanisms for the transcriptional regulation of the expression of the AGGF1 gene. We defined the precise TSSs and mapped the regulatory motifs within the AGGF1 promoter/regulatory region. We identified one rare SNP in the AGGF1 promoter, Ϫ294CϾT, associated with one sporadic KTS patient. Functional analysis of the SNP led to the finding that GATA1 is a key regulator of the AGGF1 gene. Further studies revealed a novel role of GATA1 in endothelial cells. Our results indicate that GATA1 plays important roles in endothelial vessel formation, endothelial cell migration, and apoptosis, likely by regulation of expression of AGGF1 as exogenous recombinant human FIGURE 8. Identification of new cellular roles of GATA1 in endothelial cells. A, RT-PCR analysis was used to determine the expression levels of GATA1, AGGF1, VEGF, GATA2, and GAPDH with treatments of HUVEC with phosphate-buffered saline buffer (mock), scramble siRNA, and GATA1 siRNA. Quantification was based on three independent experiments (n ϭ 3). B, Western blot analysis showed that the GATA1 siRNA reduced expression of GATA1 and AGGF1. GAPDH serves as a loading control. Quantification was based on four independent experiments (n ϭ 4). C, in vitro Matrigel endothelial vessel formation assay. HUVECs from a confluent monolayer were induced to form sprouts on the Matrigel (capillary vessel morphogenesis). HUVECs were transfected with electroporation buffer (Mock) (a), scramble siRNA (b), and GATA1 siRNA (c). The number of well connected tubes was counted and is shown in the right graph. D, HUVEC migration by a scratch assay. E, HUVEC apoptosis assay by TUNEL. AGGF1 protein can rescue the defects caused by GATA1 siRNA.