Histone Deacetylase 7 Associates with Hypoxia-inducible Factor 1 (cid:1) and Increases Transcriptional Activity*

Hypoxia-inducible factor (HIF)-1 (cid:1) is a transcription factor that controls expression of genes responsive to low oxygen tension, including vascular endothelial growth factor ( VEGF ), erythropoietin, and glycolytic en-zymes. The activity of HIF-1 (cid:1) is regulated by binding to the transcriptional co-activator cAMP-response ele-ment-binding protein-binding protein (CBP)/p300. Using the yeast two-hybrid screening system, we found that the inhibitory domain of HIF-1 (cid:1) strongly interacted with the C-terminal domain of histone deacetylase (HDAC) 7. The o -nitrophenyl (cid:2) - D -galactopyranoside assay revealed that regions containing amino acids 735– 785 of HIF-1 (cid:1) and amino acids 669–952 of HDAC7 were minimum contact sites of the interaction. The binding of HDAC7 with HIF-1 (cid:1) was reproduced in HEK293 cells grown under normoxic and hypoxic conditions (2% O 2 ). HDAC7 bound solely to HIF-1 (cid:1) among other HIF- (cid:1) family members, including HIF-2 (cid:1) and HIF-3 (cid:1) , whereas HIF-1 (cid:1) only interacted with HDAC7 in the class II HDAC family. Although HDAC7 p300 a Complex— The C-TAD of HIF-1 been shown with transcriptional co-activa-tors via the CH1 domain the N-terminal domain of CBP p300 To determine whether HDAC7, HIF-1 and p300/ complex or whether there is competition for binding between these proteins, yeast three-hybrid performed. three-hybrid

Activation of HIF-responsive genes requires recruitment of transcriptional co-activators such as p300, CBP, SRC-1 and TIF-2. These co-activators bind to the N-TAD and C-TAD domains of HIF-1␣ (13,14). CBP and p300 are paralogous, multidomain proteins that bind to the transactivation domains of a vast array of transcription factors and are components of the general transcriptional apparatus (15). Current protein structural analysis of HIF-1␣ and CBP/p300 reveals that the cysteine/histidine-rich 1 (CH1) domain of CBP and p300 binds the C-TAD of HIF-1␣ (16,17). Oxygen-dependent hydroxylation of asparagine residue 803 in HIF-1␣ C-TAD by factor inhibiting HIF-1/asparaginyl hydroxylase inhibits the binding of CBP/ p300 to HIF-1␣ and thereby causes suppression of HIF-1␣-dependent transcription (18 -21). The unique inhibitory domain (ID) of HIF-1␣, located between the N-TAD and the C-TAD, is reported to inhibit the transcriptional activities of both TADs, although the detailed mechanism of action of the ID is not clearly understood (22).
In this study, we used the yeast two-hybrid system to identify novel proteins that interact with the ID of HIF-1␣ and might affect HIF-1␣ transcriptional activity. We identified the HIF-1␣-interacting clone histone deacetylase (HDAC) 7 , which is a transcriptional repressor belonging to the mammalian class II HDAC family. Class II HDAC family members include HDAC4, HDAC5, HDAC6, HDAC7, HDAC9, and HDAC10. HDAC4, HDAC5, and HDAC7 share a high degree of homology with one another, whereas HDAC6, HDAC9, and HDAC10 share only the conserved HDAC domain (23,24). HDAC7 plays a role in repression of gene transcription by associating with HDAC3 in vivo via recruitment of co-repressors N-CoR and SMRT in the nucleus. HDAC7, like other HDAC family members HDAC4 and HDAC5, requires HDAC3 for enzymatic activity (25)(26)(27)(28). HDAC4, HDAC5, and HDAC7 contain a high homologous conserved domain (HDAC domain) containing catalytic domain in the C-terminal region. The N-terminal region and the C-terminal tail of HDAC7 are less homologous to the corresponding regions of HDAC4 and HDAC5. HDAC4, HDAC5, and HDAC7 also contain N-terminal nuclear localization signal sequences and C-terminal signal-responsive nuclear export sequences (29,30). In addition, HDAC4, HDAC5, and HDAC7 are known to shuttle between the cytoplasm and the nucleus in a process regulated by calcium/calmodulin-dependent protein kinase (31,32).
In this study, we show that HDAC7 forms a complex with HIF-1␣ and p300 in the nucleus under hypoxic conditions, resulting in increased levels of HIF-1␣ target genes (VEGF and Glut-1). Conversely, HDAC4 and HDAC5 did not bind HIF-1␣. Moreover, HDAC7 translocates from the cytoplasm to the nucleus under hypoxic conditions. We therefore propose a novel role for HDAC7 in regulating the transcriptional activity of HIF-1␣.

MATERIALS AND METHODS
Plasmids and Clones-The N-terminal Myc-, HA-, and FLAG-tagged pcDNA3.1(ϩ) vectors (Invitrogen) were from Dr. Toshiaki Suzuki, and human HIF-2␣ and rat HIF-3␣ full-length cDNAs were kindly provided by Dr. Thomas Kietzman. Human HIF-1␣ cDNA was purchased from Novus Biologicals. The full-length HDAC4 and HDAC5 cDNAs were cloned from a human brain library (Takara) and mouse heart, respectively. All mutants were generated by PCR-based site-directed mutagenesis, carried out by mismatch amplification using two sequential PCRs (33) by the Expand High Fidelity PCR System (Roche Applied Science). All constructs were sequenced to confirm no undesired mutations. pEFHAp300 was provided from Dr. T. Fujita (34).
Cloning of Full-length HDAC7 cDNA-A full-length cDNA of HDAC7 was cloned using 5Ј-rapid amplification of cDNA ends system, version 2 (Invitrogen) and Expand High Fidelity PCR System (Roche Applied Science). 5Ј-CCTTTCGGAGCAGTGGATTC-3Ј primer was used for the first-strand cDNA synthesis from human brain poly(A) ϩ RNA (Clontech). The amplified cDNA fragments were subcloned into Myc-and HA-tagged pcDNA3.1(ϩ) as described above.
Yeast Three-hybrid Assay-Yeast three-hybrid assays were performed using pGADT7 and the pBridge vector (Clontech). In this assay, the effects of a third protein on specific protein-protein interactions can be measured. The pBridge vector allows constitutive expression of a cloned protein (DBD fusion protein) through the alcohol dehydrogenase promoter. In addition, a third protein is conditionally expressed from the MET25 promoter in pBridge in response to changes in methionine concentrations in the media. In the presence of 1 mM methionine, protein expression is repressed, and in the absence of methionine, the protein is expressed. cDNAs of HIF-1␣ comprising amino acids 601-826 and 735-785 were subcloned into pGADT7, and the resulting vector constructs were transformed into strain Y187. The cDNAs of HDAC7 (aa 669 -952) and the N-terminal domain of p300 (aa 1-437) or CBP (aa 1-452) were subcloned into the two multiple cloning sites in the pBridge vector and transformed into Y187. After mating of these colonies, the diploid yeast were grown on media with or without methionine. The effects of the third protein (HDAC7, p300, or CBP) on the protein-protein interaction between HIF-1␣ and HDAC7 or p300 or CBP were measured by ␤-galactosidase filter assay.
Western Blot Analysis-Whole cell lysates were prepared using a Nuclear Extraction Kit (Active Motif). Protein was boiled at 95°C in SDS sample buffer for 5 min, run on 7.5% polyacrylamide gels, and transferred to polyvinylidene difluoride membranes. Membranes were incubated overnight at 4°C with blocking buffer, followed by exposure to primary antibodies for 1 h. Anti-HIF-1␣ antibody was obtained from Transduction Laboratories. Anti-Myc antibody and anti-HA antibody were purified from 9E10 and 12CA5 hybridomas, respectively (MBL). Monoclonal anti-FLAG antibody (M2-FLAG) and polyclonal anti-p300 antibody were from Sigma and Santa Cruz Biotechnology. The membranes were incubated with horseradish peroxidase-conjugated anti-mouse IgG (Chemicon) or horseradish peroxidase-conjugated anti-rabbit IgG (Chemicon) and washed six times with phosphate-buffered saline containing 0.1% Tween. West-Dura (Pierce) was used to detect chemiluminescence.
Immunofluorescence Staining-HEK293 cells were plated on 18-mm micro-cover glasses (Matsunami) and transfected with 2 g of the appropriate plasmid using the Ca/phosphate method. After 5 h, cells were washed with Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and incubated at normal or hypoxic (2% O 2 ) oxygen concentrations for 12 h. Cells were fixed in 3.7% paraformaldehyde in phosphate-buffered saline and washed with phosphate-buffered saline containing 0.1% Nonidet P-40. For immunostaining, fixed cells were incubated with antibodies for 1 h, washed, and incubated with Cy3conjugated secondary antibody for 30 min. Cells were washed and mounted with Vectashield (Vector Laboratories) mounting medium for fluorescence with 4Ј,6-diamidino-2-phenylindole. Images were visualized with an Olympus 1x70 inverted system microscope equipped with charge-coupled device. The resulting images were stored using Metamorph computer software.
Immunoprecipitation-After 24 h of transfection, HEK293 cells were washed with phosphate-buffered saline and lysed with lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40, 0.5 mM dithiothreitol, and protease inhibitor mixture (Roche Applied Science)). The cell lysate was centrifuged at 14,000 ϫ g for 30 min, and the supernatant was removed. Anti-Myc-Sepharose (Santa Cruz Biotechnology) or anti-FLAG M2-agarose (Sigma) was added to the supernatant and incubated for 2 h. After washing with the lysis buffer five times, SDS sample buffer was added and boiled for Western blot analysis.
Real-time PCR-Total RNA was isolated from HEK293 cells (plated on 6-well dishes) using Isogen (Nippon Gene). First-strand DNA was synthesized by SuperScript First-Strand Synthesis System for reverse transcription-PCR (Invitrogen). Real-time quantitative PCR was performed by TaqMan PCR using 7000 Sequence Detection System (Applied Biosystems). A standard curve for serial dilutions of 18S rRNA was generated. The relative standard curve method (Applied Biosystems) was used to calculate the amounts of VEGF and Glut-1 RNA.

HIF-1␣
Associates with HDAC7 in Yeast-HIF-1␣ has been shown to interact with several proteins such as CBP/p300 and factor inhibiting HIF/asparaginyl hydroxylase. Regulation of HIF-1␣ transcriptional activity, however, is not well understood. To clarify the transcriptional activation mechanisms of HIF-1␣, we set out to identify proteins that interact with HIF-1␣ using the yeast two-hybrid system. Because both the N-terminal transactivation domain (N-TAD) and C-terminal transactivation domain (C-TAD) of HIF-1␣ showed high back- ground in yeast two-hybrid screenings, we used the ID (aa 601-785) located between the N-TAD and C-TAD as bait. 1.2 ϫ 10 6 clones from a human brain library were screened. Two of 13 ␤-galactosidase-positive clones contained the same ϳ1.8-kb fragment encoding the C-terminal domain of HDAC7 (aa 669 -952), a member of the class II family of mammalian HDACs. The cloned HDAC7 contained the C-terminal conserved domain (HDAC domain), the catalytic domain, and the C-terminal unique tail.
We further characterized the interaction between HDAC7 and HIF-1␣ to map the essential binding domains of HDAC7 and HIF-1␣. Fig. 1A shows the relative strength of the interactions of the proteins in yeast determined by a quantitative ␤-galactosidase assay using ONPG. Full-length HIF-1␣ and constructs containing amino acids 401-600, 601-826, 601-785, 681-826, 735-826, 735-785, and 786 -826 of HIF-1␣ retained ␤-galactosidase activities. Constructs with amino acids 1-481, 601-734, and 601-680 had no activity. Amino acids 401-600 and 786 -826 of HIF-1␣ (the N-TAD and the C-TAD) alone showed relatively strong ␤-galactosidase activity without the presence of AD-HDAC7. These activities were due to the autonomous ␤-galactosidase activity as reported in the manufacturer's protocol. From this assay, the region containing amino acids 735-785 (within the ID of HIF-1␣) seemed to be the minimum region required for binding to HDAC7.
To determine the region within HDAC7 essential for binding to HIF-1␣, we constructed vectors containing the AD fused to HDAC7 fragments and introduced them into yeast along with a vector expressing the DBD fused to a fragment of HIF-1␣ (aa 735-785). Fig. 1B shows that the C-terminal fragment of HDAC7 (aa 669 -952), which contains the catalytic domain, C-terminal conserved domain (HDAC domain), and C-terminal tail, bound to HIF-1␣. Constructs containing other C-terminal fragments of HDAC7 (Fig. 1B)  HDAC7 (data not shown) showed no ␤-galactosidase activity. These data demonstrate that HIF-1␣ binds to HDAC7 in yeast and suggest that the minimum contact sites included amino acids 735-785 of HIF-1␣ and amino acids 669 -952 of HDAC7.
HIF-1␣ and HDAC7 belong to their respective families, the HIF-␣ family and the class II HDAC family. To identify whether other members of each family also associated in yeast, we performed the ONPG assay using DBD fusions with the C-terminal domains of HIF-2␣ (aa 571-870 or 571-828) and HIF-3␣ (aa 506 -662 or 506 -635). Because other class II HDAC family members (HDAC4 and HDAC5) were structurally similar to HDAC7, we also constructed AD fusions with the C-terminal domains of HDAC4 (aa 802-1084) and HDAC5 (aa 823-1113), corresponding to the HIF-1␣-binding domain of HDAC7 (aa 669 -952).
HIF-1␣ and HDAC7 Co-localize to the Nucleus-To examine the subcellular localization of HDAC7, HEK293 cells were transfected with a plasmid expressing HDAC7 and visualized by immunofluorescence staining. As shown in Fig. 2, HDAC7 was mainly expressed in the cytoplasm under normal oxygen concentrations; expression of HDAC7 was observed in the cytoplasm of 65.5 Ϯ 0.67% of the transfected cells. Other class II HDAC family members, HDAC4 and HDAC5, were also visualized, and the results are shown in Fig. 2. HDAC4 was localized to the cytoplasm, whereas HDAC5 was localized to the nucleus. HDAC7 changed expression pattern from the cytoplasm to the nucleus when transfected cells were incubated under hypoxic conditions (2% O 2 , 12 h); 94.4 Ϯ 0.37% of the cells expressed HDAC7 in the nucleus. On the other hand, localization of HDAC4 and HDAC5 was not altered by hypoxic treatment. During hypoxia, HIF-1␣ was expressed solely in the nucleus, suggesting that HDAC7 and HIF-1␣ co-localize to the nucleus. Similar changes in HDAC7 localization induced by hypoxia were observed in HeLa cells (data not shown).
To confirm the interaction between HIF-1␣ and HDAC7 in HEK293 cells, immunoprecipitation experiments were conducted. Cells were co-transfected with Myc-HIF-1␣ and HA-HDACs (Fig. 3A) or Myc-HDAC7 and HA-HIF-␣ family members (Fig. 3B). Cell lysates were prepared and immunoprecipitated with anti-Myc antibodies conjugated to Sepharose. Co-immunoprecipitated HDACs or HIF-␣ family members were detected by anti-HA antibody. In support of our results in yeast, only HDAC7 was co-precipitated by HIF-1␣, and HIF-1␣ was co-precipitated by HDAC7 under normoxic and hypoxic conditions. These results indicated specific interaction between HIF-1␣ and HDAC7 because other HIF-␣ family members and class II HDACs did not interact with each other.
HIF-1␣ Brings HDAC7 to the Nucleus-HIF-1␣ and HDAC7 contain nuclear localization signal sequences at amino acids 718 -721 and 157-192, respectively. To determine whether the signal sequence of each protein was important for localization to the nucleus, we constructed nuclear localization signal sequence (NLS) deletion mutants of HIF-1␣ and HDAC7 (Myc-HIF-1␣-⌬NLS and HA-HDAC7-⌬NLS). The wild-type or each mutant construct was transfected into HEK293 cells and incubated under normoxic or hypoxic conditions. The transfected cells were subjected to immunofluorescence staining. As shown in Fig. 4, A and B, overexpressed Myc-HIF-1␣-⌬NLS and HA-HDAC7-⌬NLS localized to the cytoplasm in cells incubated under normoxic or hypoxic conditions, whereas wild-type Myc-HIF-1␣ (Myc-HIF-1␣-WT) and HA-HDAC7 (HA-HDAC7-WT) localized to the nucleus under hypoxic conditions. We conducted co-transfection experiments using different combinations of the full-length and ⌬NLS mutant expression vectors under hypoxic conditions. When expressed together, Myc-HIF-1␣-WT and HA-HDAC7-WT were co-localized to the nucleus. Interestingly, when Myc-HIF-1␣-WT and HA-HDAC7-⌬NLS were co-expressed in the cells, HA-HDAC7-⌬NLS was localized with Myc-HIF-1␣-WT in the nucleus (Fig. 4C). Conversely, when Myc-HIF-1␣-⌬NLS and HA-HDAC7-WT were co-expressed, both proteins were localized to the cytosplasm (Fig.  4C). The binding of ⌬NLS mutants and the wild-type constructs of Myc-HIF-1␣ and HA-HDAC7 were confirmed by immunoprecipitation (data not shown). These results suggest that translocation of HDAC7 from the cytoplasm to the nucleus under hypoxic conditions requires association of HDAC7 with HIF-1␣ and that NLS of HIF-1␣ plays an important role in this translocation of HDAC7.
HDAC7 Increases Transcriptional Activity of HIF-1␣-We next investigated the role of HDAC7 association with HIF-1␣ in the nucleus under hypoxic conditions. To determine whether HDAC7 could affect transcriptional activity of HIF-1␣, we used real-time PCR to quantify expression of two HIF-1␣ target genes, VEGF and Glut-1. HEK293 cells were transfected with HIF-1␣ and HDAC7 and incubated under normoxic or hypoxic conditions. 18S rRNA was used as an internal control because glyceraldehyde-3-phosphate dehydrogenase, the most commonly used for assay control, was one of HIF-1␣ target genes. The relative amount of VEGF and Glut-1 RNAs was quantified. Fig. 5 shows that overexpression of HIF-1␣ causes an increase in the amount of VEGF and Glut-1 RNAs (157 Ϯ 5.1% and 202.8 Ϯ 5.1%, compared with hypoxia-treated mock cells, respectively), whereas HDAC7 did not change the amounts of both RNAs. Interestingly, co-transfection of HIF-1␣ with HDAC7 further enhanced the amount of VEGF and Glut-1 RNAs (200.0 Ϯ 4.5% and 188.9 Ϯ 6.7%, compared with cells transfected with HIF-1␣ alone). This enhancement of the transcriptional activity of HIF-1␣ by HDAC7 was not observed under normal O 2 conditions. In addition, when the HDAC7binding domain (aa 735-785) was deleted from HIF-1␣ (HIF-1␣-⌬), the enhancement of HIF-1␣ transcriptional activity by HDAC7 was diminished. These results demonstrate that association of HDAC7 with HIF-1␣ enhances transcriptional activity of HIF-1␣ in hypoxia-stressed cells (2% O 2 ).
HIF-1␣, HDAC7, and p300 Form a Complex-The C-TAD of HIF-1␣ has been shown to bind with transcriptional co-activators via the CH1 domain in the N-terminal domain of CBP and p300 (6,12). To determine whether HDAC7, HIF-1␣, and p300/ CBP form a complex or whether there is competition for binding between these proteins, yeast three-hybrid experiments were performed. We utilized an AD vector and the pBridge vector, a three-hybrid vector that expresses two proteins, a DBD fusion protein and an additional protein (the third pro-tein). The third protein, under control of the Met25 promoter, was expressed in methionine-free media and repressed in the presence of 1 mM methionine. We constructed pBridge vectors expressing HDAC7, CBP, or p300 as the DBD fusion proteins or as the methionine-regulated third proteins. We tested the effects of the third protein on binding between the DBD-HIF-1␣ fragment (aa 601-826 or 735-785) and AD-HDAC7 (aa 669 -952) or the AD-N-terminal domain of p300 (aa 1-437) or the AD-N-terminal domain of CBP (aa 1-452). The N-terminal domains of CBP and p300 containing the CH1 domain bind the C-TAD of HIF-1␣ (aa 786 -826). Therefore, HIF-1␣ fragments (aa 601-826) containing the binding sites of HDAC7 and CBP/ p300 were used. As summarized in Table I, the ␤-galactosidase filter assays show that HIF-1␣ fragment (aa 601-826) binds to HDAC7, CBP, or p300. CBP and p300 do not inhibit the binding between HIF-1␣ and HDAC7, whereas HDAC7 does not inhibit the binding between HIF-1␣ and CBP or HIF-1␣ and p300. On the other hand, the minimal binding fragment of HIF-1␣ (aa 735-785) binds only to HDAC7. CBP and p300 do not inhibit the binding between this HIF-1␣ fragment and HDAC7. These data indicate that CBP and p300 do not appear to interfere with the interaction between HDAC7 and HIF-1␣. In addition, HDAC7 does not replace CBP and p300 on HIF-1␣. To further confirm these results, immunoprecipitation was performed using HEK293 cells. Myc-HIF-1␣, FLAG-HDAC7, and HA-p300 were co-transfected into the cells and incubated under normoxic and hypoxic conditions. Cell lysates were immunoprecipitated with anti-Myc-Sepharose or anti-FLAG M2agarose. As shown in Fig. 6A, HDAC7 was immunoprecipitated with HIF-1␣ under normoxic or hypoxic conditions, whereas p300 was immunoprecipitated with HIF-1␣ only in cells incubated under hypoxic conditions. The expression levels of HDAC7 and p300 were the same under normoxic and hypoxic conditions. At normal O 2 levels, endogenous factor inhibiting HIF-1/asparaginyl hydroxylase seemed to modify the binding site of CBP/p300 in HIF-1␣ and inhibit the binding of CBP/ p300 to HIF-1␣, as described previously (19,20). As shown in Fig 6B, HDAC7 did not associate with p300 in the absence of HIF-1␣ under normoxic and hypoxic conditions. However, p300 was immunoprecipitated with HDAC7 under hypoxia when cells were transfected with HDAC7, p300, and HIF-1␣. Taken together, these results indicated that HIF-1␣ forms a complex with HDAC7 and p300 under hypoxic conditions. DISCUSSION In this study, we used the yeast two-hybrid system to identify novel proteins that interact with HIF-1␣ and might affect HIF-1␣ transcriptional activity. We identified the HIF-1␣-interacting clone HDAC7, which belongs to the class II HDAC family. We found that HDAC7 translocates to the nucleus along with HIF-1␣ and enhances transcription of HIF-1␣ target genes by forming a complex with HIF-1␣ and p300 under hypoxic conditions (2% O 2 ). These results suggest that HDAC7 is a new transcriptional regulatory partner of HIF-1␣.
Here we show that HDAC7 interacts exclusively with HIF-1␣ and does not bind either HIF-2␣ or HIF-3␣. HIF-2␣ and HIF-3␣ do not contain the ID found in HIF-1␣. It is therefore likely that this ID domain plays an important role in regulating the transcriptional activity of HIF-1␣ through interaction with HDAC7.
Within their HDAC domains, class II HDAC4 and HDAC5 share ϳ80% amino acid sequence identity with HDAC7. These proteins also share ϳ50 -60% amino acid sequence identity within their C-terminal tail. Our yeast two-hybrid assays show that HDAC4 and HDAC7 interact with HIF-1␣, whereas HDAC5 does not bind HIF-1␣. The binding activity of HDAC7 to HIF-1␣ was four times as high as that of HDAC4.
HDAC4, HDAC5, and HDAC7 have been shown to shuttle between the cytoplasm and the nucleus in mammalian cells when they respond to extracellular signals (31,37) or are transformed by oncogenic Ras (38). In our study, HDAC7 localized to the cytoplasm under normoxic conditions and translocated to the nucleus under hypoxia, suggesting that HDAC7 shuttles between the cytoplasm and the nucleus depending on    in yeast using yeast three-hybrid system The cDNAs of HIF-1␣ comprising amino acids 601-826 and 735-785 were subcloned into the AD-vector and transformed into yeast strain Y187. The cDNAs of HDAC7 (aa 669 -952) or the N-terminal domain of p300 (aa 1-437) or CBP (aa 1-452) were subcloned into two multiple cloning sites (MCS-I and MCS-II) in pBridge vector, a three-hybrid vector. pBridge constitutively expresses an AD-protein (MCS-I) and the additional third protein (MCS-II) under control of the MET25 promoter. The AD-vector-HIF-1␣ fragments and pBridge vectors harboring HDAC7 or p300 and CBP were transformed into strain AH109. After mating of these colonies, the diploid cells were grown on media with methionine (not inducing the third protein) or without methionine (inducing the third protein). The effects of the third protein (HDAC7 or p300 or CBP) on protein-protein interaction between HIF-1␣ and HDAC7 or p300 or CBP were measured by ␤-galacotosidase filter assay.
the O 2 tension surrounding the cells. This shuttling of HDAC7 was more clearly observed when HDAC7 and HIF-1␣ were overexpressed in cells, indicating that HDAC7 associates with HIF-1␣ and translocates to the nucleus under hypoxia. Although HIF-1␣ is rarely expressed under normoxic conditions due to oxygen-dependent degradation through the ubiquitinproteasome system, several cancer cell lines have been reported to express HIF-1␣ under normoxic conditions (39,40). It will be intriguing to see whether HDAC7 is localized to the nucleus in these cancer cell lines, resulting in the increased levels of VEGF expression reported in numerous cancers.
Our study also showed that HDAC4 localized to the cytoplasm and HDAC5 localized to the nucleus under normoxic conditions. These proteins did not translocate to the nucleus or cytoplasm under hypoxia. One feasible explanation why HDAC4 did not associate with HIF-1␣ in mammalian cells is that the two proteins are localized to different areas of the cell.
HDAC4, HDAC5, and HDAC7 possess the NLS and the nuclear export sequence in their N-terminal domain and Cterminal domain, respectively. The subcellular localizations of the HDACs, however, are reported to be regulated by a mechanism involving calcium/calmodulin-dependent protein kinase and an intracellular chaperone protein, 14-3-3. 14-3-3 binds to calcium/calmodulin-dependent protein kinase-phosphorylated serine residues in the N-terminal domain of the HDACs, resulting in the export of HDACs with a cellular export factor, CRM1 (30,31,41). Our data show that HDAC7 localizes to the cytoplasm under normoxic conditions and translocates to the nucleus under hypoxia. Co-transfection experiments using HIF-1␣-⌬NLS and HDAC7-⌬NLS mutants demonstrated that the NLS in HIF-1␣ is involved in translocation of the HIF-1␣⅐HDAC7 complex into the nucleus. Deletion of the HDAC7 NLS, however, did not affect subcellular localization of HIF-1␣. Therefore, it is less likely that phosphorylation of HDAC7 plays a role in translocation of HDAC7 under hypoxia. Because hypoxia causes an increase in intracellular calcium concentration in epithelial cells (42), hypoxia-induced phosphorylation of HDAC7 might be involved in stabilization of HIF-1␣⅐HDAC7 complex or export of HIF-1␣ from the nucleus to the cytoplasm.
Measurement of the relative RNA contents of HIF-1␣ target genes VEGF and Glut-1 revealed that HDAC7 enhanced HIF-1␣ transcriptional activity, leading to increases in the amount of both RNAs under hypoxia. This enhancement of activity required direct binding of HDAC7 to HIF-1␣. Currently, it is reported that HDAC4, HDAC5, and HDAC7 are not functionally active alone but acquire deacetylase activity through formation of a complex with the transcriptional corepressors SMRT/N-CoR and HDAC3 (26). It is reported that the entire HDAC domain and the C-terminal domain of HDAC7 are necessary for the binding of HDAC7 to SMRT (23). Our yeast two-hybrid data mapped part of the SMRT-binding domain of HDAC7 as a HIF-1␣-associating region. Because there is no amino acid sequence homology between SMRT and HIF-1␣, HDAC7 binds SMRT/N-CoR and HIF-1␣ by a different manner.
Recent work with mammalian cells demonstrated that HIF-1␣ binds cofactors (CBP/p300) and starts transcription of HIF-1␣ target genes under hypoxia (14). CBP/p300 binds the C-TAD (aa 786 -826) of HIF-1␣. This region is in close proximity to the HDAC7-binding domain (aa 735-785) of HIF-1␣ (16,17). Using a yeast three-hybrid system, we found that CBP/ p300 and HDAC7 did not inhibit or replace their binding to HIF-1␣, with respect to each other. These results suggest that CBP/p300 and HDAC7 form a complex with HIF-1␣ via their different binding domains within HIF-1␣. Indeed, immunoprecipitation experiments suggested that HIF-1␣, HDAC7, and p300 formed a complex in the nucleus. How does HDAC7 promote transcriptional activity of HIF-1␣? We propose that the HDAC domain of HDAC7 may be masked through its interaction with HIF-1␣, resulting in interference of HIF-1␣ binding to SMRT/N-CoR under hypoxia. Alternately, the ID of HIF-1␣ may act as a regulatory domain of TADs. Binding of HDAC7 to HIF-1␣ may lead to a conformational change within the ID of HIF-1␣ resulting in facilitated binding of co-activators such as CBP/p300 and an increase in transcriptional activity under hypoxia. Additional molecular structural and biochemical studies are needed to understand the detailed mechanism of transcriptional regulation of HIF-1␣ by HDAC7.