Arrest-defective-1 protein, an acetyltransferase, does not alter stability of hypoxia-inducible factor (HIF)-1alpha and is not induced by hypoxia or HIF.

The hypoxia-inducible factor (HIF) is a key player in a transcriptional pathway that controls the hypoxic response of mammalian cells. Post-translational modification of the alpha subunit of HIF determines its half-life and activity. Among the multiple reported modifications, acetylation, by an acetyltransferase termed arrest-defective-1 protein (ARD1), has been reported to decrease HIF-1alpha stability and therefore impact on hypoxic gene expression. In contrast, we report that both overexpression and silencing of ARD1 had no impact on the stability of HIF-1alpha or -2alpha and that cells silenced for ARD1 maintained hypoxic nuclear localization of HIF-1alpha. In addition, we show that the ARD1 mRNA and protein levels are not regulated by hypoxia in several human tumor cell lines, including cervical adenocarcinoma HeLa cells, fibrosarcoma HT1080 cells, adenovirus-transformed human kidney HEK293 cells, and human breast cancer MCF-7 cells. Using two model systems ((a) wild-type and HIF-1alpha-null mouse embryo fibroblasts and (b) HeLa cells silenced for HIF-1alpha or -2alpha by RNA interference), we demonstrate that the level of expression of the ARD1 protein is independent of HIF-1alpha and -2alpha. We also demonstrate that ARD1 is a stable, predominantly cytoplasmic protein expressed in a broad range of tissues, tumor cell lines, and endothelial cells. Taken together, our findings demonstrate that ARD1 has limited, if any, impact on the HIF signaling pathway.

A strong relationship between hypoxia, angiogenesis, and tumor progression is emerging (1). The transcription factor, hypoxia-inducible factor (HIF), 1 is the key player in the hypoxic response of cells by regulating the expression of a myriad of genes including those that control angiogenesis and neovascularization, such as the vascular endothelial growth factor (2,3). The stability and, thus, activity of HIF-1␣ is regulated by multiple post-translational modifications, in particular hydroxylation and phosphorylation (4). Hydroxylation by prolyl hydroxylase domain proteins (PHD) has been shown to mediate degradation by the ubiquitin-proteosomal system, and our group showed previously that the PHD2 isoform is a key oxygen sensor in determining HIF-1␣ stability (5). The expression of this isoform was induced under hypoxia, thus suggesting the existence of feedback regulation. It has also been reported that the instability of HIF-1␣ is enhanced by its acetylation, by an acetyltransferase termed arrest-defective-1 protein (ARD1), and that ARD1 expression is repressed in hypoxia (6). We were interested in investigating further the contribution on the function of HIF that this post-translational modification had under different conditions but were surprised to find that ARD1, in contrast to PHD2, had no impact on HIF-1␣ stability and was not hypoxia-or HIF-␣-dependent.
ARD1 was first described in Saccharomyces cerevisiae (7), and yeast mutants for ARD1 were shown to be defective in the mitotic cell cycle. The human and mouse homologs have also been designated TE2 (Swiss-Prot; P41227 and Q9QY36, respectively). Based on amino acid alignment, ARD1 has been assigned to the GNAT (GCN5-related N-terminal acetyltransferase) family, of which there are over 50 members, and it possesses a conserved (Q/R)XXGX(G/A) acetyl-CoA binding motif (8), which is also conserved from yeasts to humans. ARD1 was found to interact with HIF-1␣ in a two-hybrid assay using a Gal4-HIF-1␣-oxygen-dependent degradation domain vector as bait (6). The expressed HIF protein contains a domain, called the oxygen-dependent degradation domain, which is implicated in the regulation of the half-life of the three isoforms of HIF-␣. Matrix-assisted laser desorption ionization time-offlight mass spectrometry analysis showed in vitro acetylation of lysine 532 on a recombinant oxygen-dependent degradation domain fragment of HIF-1␣ when incubated in the presence of recombinant ARD1 (6). Acetylated HIF-1␣ was detected in cell lysates, and mutation of the lysine 532 to arginine resulted in increased stability of overexpressed HIF-1␣ as also observed previously (9). Artificial maintenance of the acetylated state of HIF-1␣ with butyric acid, a global inhibitor of deacetylases, resulted in increased degradation of HIF-1␣ and correlated with an antiangiogenic effect, consistent with diminished levels of HIF-1␣ protein (10). In addition, it has been suggested that the antiangiogenic properties of berberine, an alkaloid component of an herbal Chinese medicine, result from an increase in the acetylation of HIF-1␣ (11). Acetylation of HIF-1␣ was reported to destabilize HIF-1␣ protein by enhancing its interac-tion with a von Hippel-Lindau protein-containing E3 ubiquitin ligase complex (6). Unlike the PHDs, the activity of acetyltransferases is not known to be dependent on the oxygen concentration. It would therefore be expected that ARD1 could acetylate HIF-1␣ in both normoxia and hypoxia. Less acetylated HIF-1␣ was detected in hypoxia in comparison with that observed when HIF-1␣ was stabilized by incubation of HT1080 cells in the presence of the proteasomal inhibitor MG132, in normoxia (6). The explanation for this was that hypoxia down-regulates ARD1 mRNA and protein. However, a recent publication reports comparable normoxic and hypoxic levels of ARD1 mRNA in HepG2 cells (12). The latter authors also report that the activation of HIF-regulated downstream genes such as vascular endothelial growth factor and erythropoietin is not influenced by ARD1, thereby suggesting that ARD1 has little effect on HIF-1.
In the present study, we show that ARD1 overexpression or silencing has no impact on HIF-1␣ and -2␣ protein stability. We go on to clarify the normoxic versus hypoxic regulation of the expression of ARD1 by showing that neither the mRNA nor the protein levels of ARD1 are regulated by either short term or long term hypoxia in several cell types. We also show for the first time that ARD1 expression is independent of HIF-1␣ and -2␣. In addition, we show convincingly that ARD1 is predominantly cytoplasmic and that its silencing does not modify nuclear transfer of hypoxia-stabilized HIF-1␣. We also show that the ARD1 protein is expressed in a broad range of tissues, tumor cell lines, and endothelial cells of different species and that it has a relatively long half-life.
The pSF424-(TE2) mammalian expression vector, kindly provided by Dr. Li-Jung Juan, contains the full-length cDNA of human ARD1 with a FLAG tag on its N-terminal end and is referred to here as pFLAG-ARD1. The cDNA of human ARD1 (NM_003491) was obtained from pSF424-(TE2) using the TaqPCR Master Mix kit (Qiagen) with the following primers: ARD-forward, 5Ј-CGG GGT ACC GCC GCC ACC ATG AAC ATC CGC AAT GCG-3Ј and ARD-reverse, 5Ј-CTA GTC GAC GGA TCT AGA GGA GGC TGA G-3Ј (Eurogentec). The ARD-forward primer introduces a KpnI site for cloning purposes and the Kozak consensus sequence upstream of the initiation codon of ARD1. The ARD-reverse primer abolishes the ARD1 stop codon by the introduction of an XbaI site, which was used for subsequent cloning. The cDNA was inserted into the KpnI and XbaI sites of pcDNA4 TM /TO/myc-His A (Invitrogen) that contains the inserted myc-His tag (MycARD1). All plasmid DNAs were prepared using a Plasmid Maxi Kit (Qiagen). Transfections of pFLAG-ARD1, pcDNA3, pcDNA4, or MycARD1 were performed using the calcium phosphate technique, and cell lysis, SDS-PAGE, and immunoblotting of samples was performed as described below.
Semiquantitative RT-PCR-The mRNA levels of ARD1, HIF-1␣, HIF-2␣, and 36B4 were examined by semiquantitative RT-PCR. Cells were incubated either in normoxia or hypoxia for the indicated times, and the experiment was terminated by simultaneous lysis of all samples. Cells were lysed in TRIzol reagent (Invitrogen), and RNA was isolated using the standard phenol/choloroform/isopropyl alcohol method (18). Reverse transcription of 1 g of total RNA was performed using the Omniscript RT kit (Qiagen). Two independent sets of primers were used to amplify ARD1: set 1 forward primer 5Ј-AAG CTT GCC ATG GAC TAC AAG-3Ј and reverse primer 5Ј-TGT TGG TAT GGC TGA TTA TGA TCA-3Ј. The ARD1 set 2 primers have been published previously (19). The primers for HIF-1␣ were as follows: forward, 5Ј-ACA AGT CAC CAC AGG ACA G-3Ј; reverse, 5Ј-AGG GAG AAA ATCA AGT CG-3Ј. Primers for HIF-2␣ were as follows: forward, 5Ј-GTC ACC AGA ACT TGT GC-3Ј; reverse, 5Ј-CAA AGA TGC TGT TCA TGG-3Ј. Primers for 36B4, used as a control, were forward primer 5Ј-GGC GAC CTG GAA GTC CAA CT-3Ј and reverse primer 5Ј-CCA TCA GCA CCA CAG CCT TC-3Ј. Semiquantitative PCR on cDNA products was performed using the following parameters for the ARD1 primer sets in a Biometra T3 Thermocycler: predenaturation at 94°C for 5 min, 94°C for 30 s, annealing at 55°C for 1 min, elongation at 72°C for 40 s, for 22 cycles, and final elongation at 72°C for 5 min. PCR of an equal volume of cDNA was also performed in a similar manner for the reference mRNA, 36B4, except that only 18 cycles were performed. The ARD1 PCR products were loaded on 3% agarose gels, after volume calibration using the 36B4 reference mRNA PCR product. Gels were stained with ethidium bromide and examined with an ImageMaster VDS (Amersham Biosciences) Gel images were quantified using the GeneGenome chemiluminescent imaging system and GeneTools 3.04 software from SYNGENE (Cambridge, UK).
RNA Interference-HIF-1␣, SIMA (also known as Drosophila HIF-␣), and HIF-2␣ mRNA silencing using siRNA was performed as previously described (5,20). All siRNAs were annealed as per the manufacturer's instructions (Eurogentec). The ARD1 sequence used was as follows: sense, 5Ј-CCA UGG ACA UAU CAC CUC AdTdT-3Ј; antisense, 5Ј-UGA GGU GAU AUG UCC AUG GdTdT-3Ј. Cells at 60% confluence were transfected for 6 h on two consecutive days at a concentration of 20 nM siRNA using the calcium phosphate transfection method. Cells were lysed in SDS sample buffer 2 days after the second transfection.
SDS-PAGE and Immunoblotting-Crude cell extracts for SDS-PAGE were prepared in SDS sample buffer and sonicated, and in general 25-35 g of protein, determined using a Lowry assay (Bio-Rad), was loaded on gels. Proteins were transferred to polyvinylidene difluoride membranes and stained with Amido Black. Membranes were blocked in 5% nonfat milk in TN buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl) and incubated in the presence of the primary and then secondary antibodies in 1% nonfat milk in TN buffer. After washing in TN buffer containing 1% Triton X-100 and then in TN buffer, the immunoreactivity was detected with an ECL kit from Amersham Biosciences. The presented immunoblots are representative of at least two independent experiments often run in duplicate. Immunoblots were quantified as for RT-PCR.
Conventional and Confocal Immunofluorescence Microscopy-Cells were grown to confluence on glass coverslips and then fixed in 3.3% paraformaldehyde for 30 min at room temperature, permeabilized with 0.2% Triton X-100 for 5 min at room temperature, blocked with phosphate-buffered saline containing 0.2% gelatin, 2.0% bovine serum albumin for 30 min at room temperature, and incubated with the polyclonal anti-rabbit ARD1 or HIF-1␣ antibodies (1:1000) in phosphate-buffered saline containing 0.2% gelatin, 2.0% bovine serum albumin for 3 h at room temperature. After washing, cells were incubated in the presence of a biotinylated anti-rabbit secondary antibody conjugated to Alexa 594 (1:1000) and the nuclear fluorescent stain 4Ј,6-diamidino-2-phenylindole (1:1000) for 1 h at room temperature. After washing, coverslips were mounted in Cytifluor (Amersham Biosciences), detection of the fluorescence was performed with a Leica DMR fluorescence microscope, and images were recorded using RSImage software. Laser scan-ning confocal microscopy was performed using a LSM 510 metasystem (Zeiss). The diode laser (405 nm) and the HeNe were switched on and set at 21% output intensity. The multitrack band pass filter (420 -480 nm) and (593-636 nm) configuration was selected. The pinhole diameter was maintained at 1 airy unit. We used a Plan-Apochromat ϫ63/1.4 oil DIC objective. The nontransfected control culture was processed with a scan zoom of 1, and the MycARD1-transfected culture was processed with a scan zoom of 0.7. Images were generated for presentation using ImageJ software.
Cell Fractionation-Confluent HeLa cells were washed with phosphate-buffered saline, scraped into hypotonic buffer (10 mM Tris-HCl, pH 7.9, 1.5 mM NaCl, 10 mM KCl, 0.1% Triton X-100 containing a protease inhibitor mixture; Roche Applied Science), incubated on ice for 10 min, Dounce-homogenized (50 strokes), and centrifuged for 5 min at 2500 rpm. The nuclear pellet was resuspended in hypotonic buffer, vortexed, and centrifuged for 5 min at 2500 rpm. After protein concentration determination using a Lowry assay, SDS sample buffer was added to the supernatant (cytoplasmic fraction) and pellet (nuclear fraction), and 30 g of protein for each sample was analyzed by SDS-PAGE and immunoblotting using either anti-ARD1, anti-Sp1 (nuclear marker), or anti-␣-tubulin (cytoplasmic marker) antibodies.

Characterization of the Specific Immunoreactivity to ARD1
for Use in Immunoblotting and Immunofluorescence-To study the expression and cellular localization of the ARD1 protein, we produced a polyclonal antibody against a peptide of the last 20 C-terminal amino acids of the human ARD1 sequence. This sequence is conserved in the mouse, rat, Xenopus laevis, and partially in zebrafish but not in D. melanogaster, Caenorhabditis elegans, or S. cerevisiae. A single highly immunoreactive band was detected at around 30 kDa by immunoblotting of lysates of HeLa cells (Fig. 1A). This corresponded well to the expected size of human ARD1, a protein of 235 amino acids giving a predicted molecular mass of 26,458 Da. To further validate the immunoreactivity, HeLa cells were transfected with a FLAG-tagged ARD1 expression vector. An additional, single band migrating just above the endogenous ARD1 band was detected (Fig. 1A). The slower migration can be attributed to a modification in the charge and size of the protein due to the presence of the FLAG tag, which was also detected with an anti-FLAG antibody (data not shown). The immunoreactivity to endogenous ARD1 was further confirmed by attenuation of the corresponding band following transient transfection with siRNA targeting ARD1 mRNA (Fig. 1B). Transfection with a control siRNA to SIMA (Similar, Drosophila HIF) is also shown, and ERK2 (42 kDa) detection was used as a loading control. The polyclonal antibody to ARD1 was validated for immunofluorescence in HeLa cells by transfection of cells with an siRNA targeting ARD1 (Fig. 1C). Comparison with the level of the immunoreactive signal from cells transfected with a control siRNA (SIMA) (Fig. 1C, left) showed an almost total loss in the immunoreactivity in the presence of siRNA directed against ARD1 (Fig. 1C, right panel). Thus, detection of exogenous FLAG-ARD1 and attenuation of the endogenous ARD1 validate the specificity of the ARD1 antibody for immunoblotting and immunofluorescence.
Overexpression and Silencing of ARD1 Have No Effect on HIF-1␣ and -2␣ Stability, and Silencing of ARD1 Does Not Modify HIF-1␣ Cellular Localization-To examine the effect of ARD1 on HIF-1␣ and -2␣ stability, we overexpressed ARD1 and incubated cells under normoxic and short and long term hypoxic conditions. HeLa cells were nontransfected, transfected with a control plasmid (pcDNA4), or transfected with a FLAG-tagged ARD1-expressing vector (Fig. 2, A and B). No substantial difference in the protein level for HIF-1␣ and -2␣ at 4 and 8 h was observed in either normoxia or hypoxia between control pcDNA4-transfected cells and cells overexpressing ARD1 when compared with the ERK2 loading control. The level of expression of both the endogenous and exogenous ARD1 remained constant upon exposure of cells to normoxia or hypoxia for 4 and 8 h when the intensity of the bands was compared with ERK2. Similar results were obtained after 24 h of hypoxic treatment (data not shown). There was also no difference in the stability of HIF-1␣ when ARD1-overexpressing HT1080 cells were incubated for 5 h in hypoxia (data not shown). We conclude that the presence of at least a 2-fold increase in the amount of ARD1 has little effect on HIF-1␣ stability in normoxia or hypoxia. In addition, silencing of ARD1 did not have a direct impact on HIF-1␣ or -2␣ protein stability in either normoxia or hypoxia, as ob-FIG. 1. Validation of the anti-ARD1 antibody, for immunoblotting and immunofluorescence, by exogenous expression and silencing of ARD1. A, detection of ARD1 immunoreactivity by immunoblotting of lysates from HeLa cells nontransfected (Ϫ) or transfected with pcDNA3 or with a FLAG-tagged ARD1 plasmid (pFLAG-ARD1) using two different amounts of DNA (5 and 10 g/100-mm diameter Petri dish), performed in duplicate. Cells were lysed 48 h after transfection. B, HeLa cells were nontransfected (Ϫ) or transfected with a control siRNA (SIMA) or siRNA directed against human ARD1 mRNA (20 or 50 nM). Cell lysates were separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes, and immunoblotted with an anti-ARD1 and anti-ERK2 antibody mixture. Detection was performed using ECL. C, to validate the use of the anti-ARD1 antibody for immunofluorescence, HeLa cells were either transfected with a control siRNA (SIMA) (a) or with an siRNA to ARD1 (b). The anti-ARD1 antibody was detected using a secondary antibody conjugated to Alexa 594. Magnification is at ϫ63, and the bar represents 60 m.

ARD1 mRNA and Protein Expression Is Not Regulated by Short and Long Term Hypoxia or by Hypoxic Mimetics-Given
that the level of the mRNA of ARD1 were reported to be down-regulated in HT1080 cells after a 2-or 4-h hypoxic (1.0% O 2 ) treatment (6), we were interested to know if this was a general phenomenon in tumor cells and if it was maintained under long periods of hypoxia. To this aim, we investigated regulation of the ARD1 mRNA and protein in several cell lines exposed to short and long term hypoxia. The mRNA levels were examined by semiquantitative RT-PCR in breast cancer MCF-7 and fibrosarcoma HT1080 cells exposed to 4, 24, and 48 h of hypoxia (1.2% O 2 ). We did not observe substantial changes in the mRNA expression levels when comparing cells in normoxia to cells in short or long term hypoxia for MCF-7 cells (Fig. 4A) and HT1080 cells (Fig. 4B). Expression of 36B4 was used as a loading control. To confirm hypoxic regulation in these cells at the mRNA level, we examined the same samples for the hypoxic induction of two known hypoxia response element-containing genes: carbonic anhydrase 9 (CA9) (Fig. 4A) and the proapoptotic member of the Bcl-2 family (BNIP3) (Fig. 4B). As expected, these genes were up-regulated in hypoxia. The lack of hypoxic regulation of ARD1 was further confirmed by quanti- tative real time PCR. 2 To test the possibility that the previously reported decrease in expression in hypoxia was dependent on the cell density, we seeded cells at different densities and examined ARD1 expression in hypoxia. However, no difference in the mRNA level was detected (data not shown). A second set of primers for PCR amplification of the full-length cDNA for ARD1, which was reported to show decreased mRNA expression in HEK293 cells incubated in hypoxia (1% O 2 ) for 16 h (19), was also tested on the same RNA panel presented in Fig. 4A. Again no difference in expression levels was detected (data not shown). Thus, based on these experiments, which were reproduced repeatedly, ARD1 mRNA does not appear to be influenced either by short or long term hypoxia or by confluence.
We then examined whether the protein levels of ARD1 were regulated by hypoxia, using samples generated in parallel to those used for the RT-PCR experiments described in Fig. 4B. Exposure of HT1080 cells to 4, 24, and 48 h of hypoxia did not notably modify the expression levels of the ARD1 protein (Fig.  4C). The timing of this experiment was staggered so that all time points were lysed simultaneously to ensure equal final cell densities. The efficacy of the hypoxic treatment was confirmed by the detection of HIF-1␣ under these conditions. Similar results were obtained for MCF-7 cells exposed to identical hypoxic conditions (data not shown), and no difference in the expression level of the ARD1 protein was observed for HeLa cells as shown in Fig. 2A. In addition, incubation of LS 174T cells under hypoxic conditions with 0.2 or 3.0% O 2 instead of 1.2% for 24 h did not lead to changes in the expression level of the ARD1 mRNA when compared with normoxia (data not shown). Thus, variable degrees of hypoxia have no effect on ARD1 mRNA expression.
To further investigate possible hypoxia-dependent regulation of ARD1, two hypoxic mimetics known to induce HIF-1␣ stability were examined. The hypoxic mimetic cobalt chloride stabilizes HIF-1␣ and -2␣ by inhibiting the PHDs, and thereby the HIF-␣ subunit escapes degradation (21). Exposure to cobaltous ions did not modify the ARD1 protein level in HeLa or RCC4 cells (Fig. 4D). Incubation of MCF-7 cells with another hypoxia mimetic, 2,2Ј-dipyridyl (100 M), also did not modify the protein expression levels (data not shown). Since RCC4 cells are deficient in VHL, HIF-␣ is constitutively stable and hydroxylated under normoxic conditions by PHDs and FIH-1 (factor inhibiting HIF-1) (22). Thus, these results also suggest that the presence or absence of hydroxylated HIF-␣ does not influence the level of ARD1 protein. We again questioned whether confluence could explain the discrepancies between our findings in several cell types and those published for the HT1080 cells, since confluence has also been reported to result

ARD1 Is Not Hypoxia-inducible Factor-dependent
in stabilization of HIF-1␣ (23); however, we did not note a difference in the ARD1 protein levels, analyzed by immunoblotting or immunofluorescence, in normoxia compared with hypoxia for the various cell confluences tested (data not shown).
ARD1 Expression Is Not HIF-1␣-or HIF-2␣-dependent-To directly examine the possible influence that HIF-1␣ may have on ARD1 protein expression, we examined wild-type (HIF-1␣ ϩ/ϩ ) and HIF-1␣-null (HIF-1␣ Ϫ/Ϫ ) mouse embryo fibroblasts. The detection of HIF-1␣ in hypoxic lysates of HIF-1␣ ϩ/ϩ and control HeLa cells and its absence in HIF-1␣ Ϫ/Ϫ cells confirmed the status of these cells with respect to HIF-1␣ (Fig. 5A). It is of note that HIF-2␣ has been reported to be transcriptionally inactive in mouse embryo fibroblasts (24,25). Similar levels of ARD1 protein were observed in both HIF-1␣ ϩ/ϩ and HIF-1␣ Ϫ/Ϫ cells. To confirm the lack of involvement of HIF-␣ in ARD1 mRNA and protein regulation, we silenced HIF-1␣ and HIF-2␣ in HeLa cells using siRNA and examined the level of ARD1 mRNA (Fig. 5B) and protein (Fig. 5C). This siRNA approach has been validated previously, for HIF-1␣, in our laboratory (5) and in a previous publication for HIF-2␣ (20). Transfection   FIG. 5. The expression of the ARD1 mRNA and protein is independent of HIF-1␣ and -2␣. A, lysates of HIF-1␣ ϩ/ϩ and HIF-1␣ Ϫ/Ϫ mouse embryo fibroblasts and HeLa cells, incubated in normoxia (N) or hypoxia (H) (4 h), were examined by immunoblotting with anti-HIF-1␣ and anti-ARD1 antibodies. ERK2 immunoreactivity was used as a loading control. B, semiquantitative RT-PCR of RNA obtained from HeLa cells transfected with either control siRNA (SIMA), siRNA to HIF-1␣, or siRNA to HIF-2␣ for detection of HIF-1␣, HIF-2␣, and ARD1. Cells were lysed 48 h after the second transfection. 36B4 was used as a loading control for all panels. C, lysates of HeLa cells transfected with a control siRNA to SIMA or to HIF-1␣ or -2␣. The immunoreactivity to HIF-1␣, HIF-2␣, and ARD1 was examined together with ERK2; the latter was used as a loading control. D, the cellular localization of ARD1 was examined by immunofluorescence using an anti-ARD1 antibody and a secondary antibody conjugated to Alexa 594 in cells incubated either in normoxia (a, c, and e) or hypoxia (b, d, and f) and transfected with either a control siRNA (SIMA) (a and b) or an siRNA to HIF-1␣ (c and d) or HIF-2␣ (e and f). Magnification was ϫ63, and the bar represents 60 m.
with their respective siRNAs resulted in a moderate decrease in the HIF-1␣ mRNA and a substantial decrease in HIF-2␣ mRNA when compared with the endogenous RT-PCR signal for the sample transfected with the control siRNA (SIMA) (Fig.  5B). A parallel RT-PCR analysis with the same RNA showed no change in the level of expression of ARD1 when either HIF-1␣ or -2␣ levels were diminished (Fig. 5B). The high level of silencing obtained for both HIF␣ proteins did not influence the level of expression of the ARD1 protein (Fig. 5C). In addition, HeLa cells silenced for HIF-1␣ or HIF-2␣ retain the same ARD1 cytoplasmic localization as compared with those cells silenced with control siRNA, SIMA (Fig. 5D). These results strongly suggest that the expression and the subcellular localization of ARD1 are independent of HIF-1␣ and -2␣.
ARD1 Is a Predominantly Cytoplasmic Protein-Examination of the ARD1 protein sequence on the protein data base PSORTII (prediction of protein sorting signals and localization sites in amino acid sequences) gave a very low nuclear localization signal score, suggesting predominant cytoplasmic localization, as observed above for cells transfected with the control siRNA (Fig. 1C, left panel). However, since the proposed substrate for ARD1, HIF-1␣, is predominantly localized to the nucleus in cells exposed to hypoxia, we examined the immunofluorescence to ARD1 after incubation of cells in normoxia and hypoxia. Hypoxic conditions did not modify either the intensity, compared with normoxia, or the localization of the ARD1 immunofluorescence (Fig. 6A). Given the faint immunoreactivity detected across the nucleus of cells during immunofluorescence analysis using conventional microscopy, these samples were additionally analyzed by confocal microscopy. Confocal analysis confirmed that ARD1 is predominantly localized to the cytoplasm; however, a small percentage of staining was identified within the nucleus (Fig. 6B, left). A predominantly cytoplasmic localization was favored even when exogenous Myc-ARD1 was overexpressed with high efficiency (Fig. 6B, right). (Note that because of the high intensity of staining in transfected cells, the exposure time was substantially reduced compared with the nontransfected image; thus, the endogenous signal in nontransfected cells was only slightly detected). To further confirm the cytoplasmic localization of ARD1, we performed subcellular fractionation, SDS-PAGE, and immunoblotting. ARD1 was only detected in the cytoplasmic fraction (Fig.  6C). The nuclear protein Sp1 and the cytoplasmic marker ␣-tubulin were used to confirm the identity of the respective nuclear and cytoplasmic fractions.
ARD1 Is a Relatively Stable Ubiquitously Expressed Protein-Examination of the stability of the ARD1 protein showed it to be relatively stable, with a half-life of at least 72 h, which is comparable with that for ERK2 (Fig. 7). The stability of the protein Paip2, a protein involved in the turnover of vascular endothelial growth factor mRNA, with a half-life of 12 h was examined in parallel (16). The same stability profile was observed for ARD1 when cells were incubated in hypoxia (data not shown). Since HIF-1␣ is known to play a major role in tumor progression and angiogenesis, we examined the expression of ARD1 in tumor and endothelial cells. We found that the ARD1 protein is expressed in a broad range of human tumor cell lines and in human, bovine, and porcine endothelial cells  panels (b and d) show staining in the presence of the anti-ARD1 antibody. Magnification is at ϫ63, and the bar represents 60 m. B, ARD1 localization in nontransfected and MycARD1-transfected cells was analyzed by confocal microscopy using a secondary antibody conjugated to Alexa 594 (red). Cell nuclei were also stained with 4Ј,6-diamidino-2-phenylindole (blue). A ϫ63 objective was used, and the bar represents 60 m. The nontransfected control cells were processed with a scan zoom of 1, and the MycARD1-transfected cells were processed with a scan zoom of 0.7. C, cytoplasmic (Cyt) and nuclear (Nuc) fractions were prepared and analyzed by SDS-PAGE and immunoblotting with the anti-ARD1 antibody. To verify the fractionation procedure, fractions were analyzed for the nuclear protein Sp1 and the cytoplasmic protein ␣-tubulin using the corresponding antibodies. FIG. 7. ARD1 is a relatively stable protein. HeLa cells were incubated in the absence (Ϫ) or (ϩ) presence of the protein synthesis inhibitor cycloheximide (CHX; 10 g/ml) for 15, 24, 48, and 72 h. Cell lysates were examined by immunoblotting using the antibody to ARD1. Immunoreactivity to ERK2 was used as a loading control, and the immunoreactivity to the relatively short-lived protein Paip2 was examined to determine the efficacy of the cycloheximide treatment. ERK2 immunoreactivity was detected in parallel for each blot. (Fig. 8A). Immunoreactivity to the ubiquitously expressed ERK2 is shown for comparison. Immunoblotting with the ARD1 antibody of lysates of murine tissues showed that ARD1 protein was expressed in all of the tissues studied with the exception of the skin (Fig. 8B). Again the immunoreactivity to ERK2 is shown by way of comparison. (The upper band seen in some tissues is cross-reactivity to ERK1). These data open the way to the study of the relevance of ARD1 function in endothelial cell biology and in tumor progression.

DISCUSSION
In contrast to a report by Jeong et al. (6), we demonstrate that neither overexpression nor silencing of ARD1 by RNA interference affects HIF-1␣ stability. The report by Jeong et al. (6) showed a decrease in the half-life of HIF-1␣ in ARD1overexpressing cells and a substantial stabilization of HIF-1␣ to hypoxic levels in HEK293 cells treated with an ARD1 antisense-expressing vector. However, we did not observe a substantial decrease in the level of hypoxia-stabilized HIF-1␣ or -2␣ when ARD1 was overexpressed. In addition, we demonstrate that silencing of ARD1 does not modify the level of HIF-1␣ or -2␣ in normoxia or hypoxia when detected by immunoblotting, unlike silencing for PHD2 that results in stabilization of HIF-1␣ in normoxia. Our findings are supported by those of Fisher et al. (12), who showed that the mRNA levels of the downstream HIF-regulated genes, erythropoietin and vascular endothelial growth factor, were unchanged in HepG2 and HEK293 cells silenced for ARD1 by RNA interference.
Several of the proteins that modify HIF-1␣ are regulated by hypoxia, either negatively or positively, which suggests the existence of feedback loops. For example, two of the three human PHD isoforms that hydroxylate HIF-1␣ are hypoxiaregulated (5,26,27). Two previous publications have indicated that ARD1 mRNA is regulated by hypoxia, whereas two other more recent reports show no hypoxic regulation. The first that indicated hypoxic regulation, by Jeong et al. (6), showed downregulation of the mRNA in HT1080 cells incubated in hypoxia (2 and 4 h) or in the presence of the hypoxia mimetic cobalt chloride. However, only a minimal effect was observed with another mimetic, 2,2Ј-dipyridyl. A decrease in the level of the protein was also reported by these authors but concerned only overexpressed protein. The second report, by Chun et al. (19), showed down-regulation of ARD1 mRNA and protein in HEK293 cells incubated in hypoxia for 16 h. In contrast, a more recent report showed no hypoxic (time not specified) regulation of ARD1 at the mRNA level in HepG2 cells and no effect, or a minimal effect, on the mRNA level in Hep3B, HEK293, and K562 cells treated with cobalt chloride (12). However, not all of their data were consistent, since another hypoxia mimetic, FG2229, which has been shown to inhibit the PHDs but also collagen prolyl hydroxylase (28), did result in a decrease in ARD1 mRNA in HepG2, Hep3B, K562, and HEK293 cells. Our more extensive data concern not only the mRNA but also endogenous and exogenous ARD1 protein expressed in HT1080, HeLa, MCF-7, and HEK293 cells incubated in short or long term hypoxia or in the presence of the hypoxia mimetics cobalt chloride and 2,2Ј-dipyridyl. These data, obtained from several repeated experiments, indicate that ARD1 mRNA and protein are not regulated in these cells by either short or long term hypoxia or by hypoxia mimetics in the studied cells. In addition, we discarded confluence and the oxygen level as reasons to explain possible differences with reported results. Our results are also supported by another recent report, a microarray study by Manalo et al. (2), which examined a total of 22,283 gene probes, including those for ARD1, in human pulmonary artery endothelial cells incubated in normoxia or hypoxia and overexpressing HIF-1␣. This array did not identify ARD1 as being repressed or induced.
We further indicate that ARD1 expression is not dependent on HIF-1␣ or -2␣ by demonstrating that the ARD1 levels are the same in wild-type and HIF-1␣-deficient cells or in cells in which HIF-1␣ or -2␣ were silenced by RNA interference.
Little information is available concerning the status of ARD1 in angiogenesis and cancer, let alone on its in vivo impact in overall cell biology. ARD1 was initially suggested to be involved in cell cycle control in yeast (7), to be involved in neuronal development in mice (29), and to be implicated in human hepatocyte cell proliferation (12). To form a functional N-terminal acetyltransferase, ARD1 must interact with another N-terminal acetyltransferase, NAT1 (29 -32). From a cancer standpoint, it is of interest to note that the mouse N-terminal acetyltransferase is implicated in vascular development (33)(34)(35) and therefore may play an active role in angiogenesis. The human orthologue, tubedown-100, also termed NATH or Ga19, has been shown to be implicated in transcription of the osteocalcin gene (36) and is overexpressed in papillary thyroid carcinoma (37) and gastric cancer (38). The role, if any, that ARD1 plays in these processes as a NAT-interacting partner remains unstudied. Thus, we were interested in examining the expression of ARD1 in mouse tissues, tumor cell lines, and, in particular, endothelial cells, which are central to angiogenesis. We observed that ARD1 is expressed at highly detectable levels in a broad range of tissues, tumor cell lines, and endothelial cells of different species origin. We also demonstrate that endogenous ARD1 is a stable protein that, when analyzed by both conventional and confocal microscopy, was localized predominantly to the cytoplasm. However, since a small amount of ARD1 was seen in the nucleus in nuclear sections obtained with the confocal microscope, we cannot exclude the possibility that it can shuttle in and out of the nucleus, possibly by simple diffusion, since it is of small size and does not appear to possess a nuclear localization signal. We also show that the ARD1 cytoplasmic localization remains unchanged in hypoxia and when HIF-1␣ or -2␣ are silenced by RNA interference or when ARD1 is highly overexpressed. Exogenously expressed mouse and human ARD1 has also been shown by cell fractionation of HT1080 cells (6) and by immunofluorescence of rat kidney fibroblasts (29) to be localized to the cytoplasm, whereas exogenous and endogenous human ARD1 were reported to be cytoplasmic and nuclear in HeLa cells visualized by conventional microscopy (30). Given the silencing of the cytoplasmic immunofluorescence and the lack of nonspecific immunoreactivity on immunoblots for cells transfected with siRNA directed against ARD1 mRNA in this study, we are confident in reporting that the majority of ARD1 is cytoplasmic in HeLa cells. These data suggest that ARD1 would act primarily in the cytoplasm on either the low levels of HIF-1␣ present under normoxic conditions or on hypoxia-stabilized HIF-1␣ before its rapid transfer to the nucleus. Indeed, normoxic acetylation of HIF-1␣ was reported by Jeong et al. (6).
In conclusion, our study (a) shows for the first time that efficient ARD1 silencing by RNA interference and overexpression of ARD1 do not influence HIF-1␣ or -2␣ stability, (b) demonstrates that ARD1 silencing does not modify HIF-1␣ subcellular localization, (c) clarifies the situation with respect to hypoxic regulation of ARD1 and its cytoplasmic localization, (d) provides new data indicating that ARD1 is not an HIF-1␣ or -2␣-dependent gene, and (e) shows ubiquitous expression of stable ARD1. These data suggest that the role of ARD1 in angiogenesis does not implicate directly HIF-1␣ regulation. However, since highly detectable levels of ARD1 are present in numerous tumor cell lines and endothelial cells, it is possible that ARD1 plays a significant, yet to be determined, role in angiogenesis and tumor progression.