HDAC4 Protein Regulates HIF1α Protein Lysine Acetylation and Cancer Cell Response to Hypoxia

Background: HIF1α is a target of anticancer therapy.

Results: Lysines within the HIF1α N terminus are targets of HDAC4 deacetylation. HDAC4 inhibition causes the increase of HIF1α protein acetylation and decrease of protein stability, which lead to the reduction of HIF-1-mediated target gene expressions and activities in cancer cells.

Conclusion: HDAC4 provides a novel HIF1α regulatory mechanism.

Significance: HIF-1 can be targeted by HDAC4 inhibition.

Abstract

Hypoxia-inducible factor 1 α (HIF1α) is an essential part of the HIF-1 transcriptional complex that regulates angiogenesis, cellular metabolism, and cancer development. In von Hippel-Lindau (VHL)-null kidney cancer cell lines, we reported previously that HIF1α proteins can be acetylated and inhibited by histone deacetylase (HDAC) inhibitors or specific siRNA against HDAC4. To investigate the mechanism and biological consequence of the inhibition, we have generated stable HDAC4 knockdown via shRNA in VHL-positive normal and cancer cell lines. We report that HDAC4 regulates HIF1α protein acetylation and stability. Specifically, the HIF1α protein acetylation can be increased by HDAC4 shRNA and decreased by HDAC4 overexpression. HDAC4 shRNA inhibits HIF1α protein stability. In contrast, HDAC1 or HDAC3 shRNA has no such inhibitory effect. Mutations of the first five lysine residues (lysine 10, 11, 12, 19, and 21) to arginine within the HIF1α N terminus reduce protein acetylation but render the mutant HIF1α protein resistant to HDAC4 and HDACi-mediated inhibition. Functionally, in VHL-positive cancer cell lines, stable inhibition of HDAC4 decreases both the HIF-1 transcriptional activity and a subset of HIF-1 hypoxia target gene expression. On the cellular level, HDAC4 inhibition reduces the hypoxia-related increase of glycolysis and resistance to docetaxel chemotherapy. Taken together, the novel biological relationship between HDAC4 and HIF1α presented here suggests a potential role for the deacetylase enzyme in regulating HIF-1 cancer cell response to hypoxia and presents a more specific molecular target of inhibition.

Introduction

Hypoxia-inducible factor 1 α (HIF1α)² is an essential part of the HIF-1 transcriptional complex that regulates gene expressions critical to cellular response and adaptation to hypoxia (1, 2). HIF-1 is comprised of an α and a β subunit. The transcriptional activity of HIF-1 is primarily determined by the availability of HIF1α (3, 4). In addition, transcriptional cofactors, including histone acetyltransferases such as p300 and histone deacetylases (HDACs), have been reported to interact with and influence HIF-1 activity (5–9). However, the relationship between HIF-1 and these cofactors is not fully understood.

Biologically, the HIF1α protein is constantly synthesized but rapidly degraded by oxygen under non-hypoxic conditions. The newly synthesized HIF1α protein is posttranslationally modified by hydroxylations at proline residues 402 and 564 via oxygen-dependent prolyl hydroxylases (1). The von Hippel-Lindau (VHL) protein then binds hydroxylated HIF1α and recruits an E3 ubiquitin ligase complex that targets HIF1α for 26 S proteasome-mediated degradation (10). There are also oxygen-independent mechanisms in HIF1α protein regulation. Hsp90 is a key molecular chaperone in maintaining HIF1α stability (11). In contrast, RACK1, Hsp70, and carboxyl terminus of Hsc70 interacting protein (CHIP) promote HIF1α degradation (12, 13).

Besides hydroxylation, the HIF1α protein can also be posttranslationally modified by reversible lysine acetylation (14–16), which can be pharmacologically modulated by HDAC inhibitors (HDACi) (16–18). Treating cells with HDACi also reduces HIF1α on the protein level under normoxic, hypoxic and hypoxia mimic conditions (16, 19–22). The protein inhibition is dependent on the 26 S proteasomal degradation system but can be VHL-independent (16, 19). Two of the acetylation sites within the HIF1α protein have been identified at residues Lys-532 and Lys-674 (14, 15). Functionally, the acetylation at Lys-532 can lead to VHL-dependent HIF1α protein degradation (14). In contrast, the acetylation at Lys-674 is required for HIF-1 transcriptional activity (15).

Currently, the mechanism and the functional consequence of HIF1α acetylation/deacetylation at different lysine residues are unclear. Although the identity of HDAC isozymes deacetylating Lys-532 is unknown (14), the Lys-674 deacetylation is mediated by a class III HDAC, Sirt1 (15). We have shown previously that the inhibition of class II HDAC isozymes HDAC4 and HDAC6 via siRNA inhibits HIF1α protein in VHL-null kidney cancer cell lines. The HDAC6 siRNA-mediated HIF1α inhibition is thought to be related with the acetylation of Hsp90, which disrupts the Hsp90 chaperone function toward its client proteins, including HIF1α (16, 19). However, the mechanism for HIF1α inhibition via HDAC4 siRNA is unclear. HDAC4 siRNA does not increase Hsp90 acetylation nor disrupt the interaction between HIF1α and Hsp90/Hsp70 (16). On the other hand, the inhibition of HDAC4, not HDAC6, increases HIF1α protein acetylation (16). These results suggest that HDAC4 can regulate HIF1α protein acetylation and stability.

In this study, we first recapitulated the HIF1α protein acetylation and inhibition by stable HDAC4 shRNA knockdown in VHL-positive prostate and liver cancer cell lines. We then identified that the HIF1α N terminus lysine residues are targets of HDAC4 and associated with HIF1α sensitivity toward HDAC inhibition. Further, we observed that stable HDAC4 knockdown attenuates cancer cell response and adaptation to hypoxia in terms of HIF-1-mediated gene-transcriptional up-regulation, glycolysis, and chemoresistance.

EXPERIMENTAL PROCEDURES

Cell Lines and Reagents

Hep3Bc1 was a gift from Dr. Gregg Semenza at Johns Hopkins University. It was engineered by stable transfection of plasmids for hypoxia/HIF-1 response element (HRE)-driving firefly luciferase reporter (p2.1) and constitutive renilla luciferase reporter (pSV-renilla) in the human liver cancer cell line Hep3B (23). The prostate cancer C42B cell line was a gift from Dr. John Isaacs at Johns Hopkins University. The Hek293 cells were kindly provided by the laboratory of Dr. Zhengfeng Zhou at Oregon Health & Science University. The Hek293T cells were purchased from the ATCC. All cell lines were maintained as described previously (23, 24). The cell culture hypoxic condition was defined as 1% oxygen, 5% CO₂, and 94% nitrogen. Cycloheximide (CHX), cobalt chloride (CoCl₂), HDACi-suberoylanilide hydroxamic acid, and MG132 were purchased from Sigma and Cayman.

Plasmids and Transient Transfection

Plasmids encoding for 3×FLAG-HIF1α, p2.1, and constitutive pSV-renilla were gifts from Dr. Gregg Semenza at Johns Hopkins University. Plasmids encoding HA-HIF1α, FLAG-HDAC1, and FLAG-HDAC4 were purchased from Addgene (Cambridge, MA). Transfection was performed using FuGENE6 (Roche). All transfections were equalized with empty vector (Ev) to ensure the same DNA amount.

Stable shRNA Knockdown Cell Lines

The pLKO.1-puro vector-based lentiviral transduction particles containing HDAC1, HDAC3, HDAC4, and HIF1α shRNA or scramble control were purchased from Sigma.

Immunoprecipitation (IP) and Antibodies

IP was performed by incubating whole cell lysates with the primary antibody at 4°C followed by incubation with protein A/G plus-agarose beads (Santa Cruz). For FLAG- and HA-based IP, anti-FLAG M2 affinity gel and anti-HA-agarose were purchased from Sigma. The beads were extensively washed with lysis buffer, and the associated proteins were eluted and analyzed by Western blot analyses. Key antibodies used for IP and Western blot analyses include anti-HIF1α (R&D Systems), anti-acetyl-lysine (Ace-K) (Cell Signaling), anti-FLAG (Sigma), anti-HA (Santa Cruz Biotechnology, Inc.), and anti-HDAC4 (Cell Signaling Technology, Inc.).

HIF1α Protein Half-life Measurement Using CHX

Cells were seeded in 10-cm culture dishes. HIF1α proteins were induced by CoCl₂ (150 μm) for 6 h. Then, the protein synthesis inhibitor CHX was added into the media, and proteins were harvested at the indicated time points. Western blotting for HIF1α and β-tubulin were scanned by a LI-COR Odyssey fluorescence scanner and quantified using LI-COR Odyssey infrared software.

Site-directed Mutagenesis

We used the wild-type 3×FLAG-HIF1α plasmid (HIF1α-WT) as the template and two-step PCR to generate the mutant HIF1α (HIF1α-mut) that contains the lysine-to-arginine mutation at residues Lys-10, Lys-11, Lys-12, Lys-19, and Lys-21. The resultant HIF1α-mut plasmid was sequenced to confirm to mutations.

Real-time Quantitative RT-PCR

Quantitative RT-PCR was performed as described previously by us (24). The ΔΔCt method was used to calculate the mRNA transcript level.

Reporter Gene Assay

Reporter assays using Hep3Bc1 were done as described (23). For transient assays in the Hek293 and C42B cell lines, cells were cotransfected by 1.0 μg of plasmid p2.1 and 0.1 μg of plasmid pSV-renilla. For assays in Hek293 cells, 20 ng of HIF1α-WT or HIF1α-mut plasmids were also transfected. 24 h later, cells were incubated in normal or hypoxic incubators overnight, and the firefly and renilla luciferase activities for each sample were measured as described by the dual-luciferase activity kit (Promega) (23).

Lactate Measurement

Cells were added to fresh media and cultured under normal and hypoxic condition for 48 h. The lactate concentrations in media were measured with a kit (Biovision), and all values were adjusted with the total cell number and normalized to the value of sh-Scr cell lines under the normoxic condition.

In Vitro Deacetylation Assay

Hek293T cells were transfected with the plasmid expressing HA-HIF1α for 48 h. Then, total HA-HIF1α proteins (including the acetylated and the unacetylated) were IP-purified using anti-HA antibodies. Simultaneously, FLAG-HDAC1 or FLAG-HDAC4 was overexpressed in HEK293T cells and IP-purified using anti-FLAG M2-agarose. The agarose beads containing the immunocomplex of either HDAC4 or HDAC1 were mixed with the agarose containing HA-HIF1α in deacetylation buffer (Upstate-Millipore) as described (25, 26). At various time points, the deacetylation was terminated by adding Western blot loading buffer. The total HA-HIF1α and its acetylation can be visualized by Western blot analyses using anti-HA and anti-acetyl-K antibodies.

Statistical Analysis

Differences between the means of unpaired samples were evaluated by Student's t test. p values < 0.05 were considered to be statistically significant.

RESULTS

HDAC4 Regulates HIF1α Protein Acetylation and Stability

Because most types of cancer cells have functional VHL, to determine the effect of HDAC4 inhibition on HIF1α in a VHL-positive background we transduced a human prostate cancer cell line, C42B, and a variant of human liver cancer cell line, Hep3Bc1, with pseudo-lentivirus-containing shRNA against HDAC4 (sh-HDAC4) or scramble control (sh-Scr). We incubated the cells under normal oxygen or hypoxic conditions for 6 h, and we observed that the HIF1α protein level was reduced significantly in hypoxic cells of sh-HDAC4 compared with the sh-Scr (Fig. 1A). Similar inhibitory results were observed when we used cobalt chloride (CoCl₂) as a hypoxic mimic (Fig. 1B). To measure the HIF1α acetylation, we treated cells with CoCl₂ and the proteasome inhibitor MG132. As seen previously in the VHL-null cell lines (16), the inhibited HIF1α protein, because of HDAC4 knockdown, can be rescued by MG132 (Fig. 1C; IB, HIF1α), and there was no significant change in HIF1α interaction with Hsp90/Hsp70 (data not shown). Significantly, the HIF1α acetylation level was higher in the HDAC4 knockdown cells (Fig. 1C; IB, Ace-K) than the scramble controls. To measure the kinetics of HIF1α protein degradation, we treated sh-HDAC4 and sh-Scr cells with CoCl₂ to stabilize the HIF1α protein. Then, we added CHX to arrest protein synthesis. The degradation rate of existing HIF1α protein was obtained by measuring HIF1α and tubulin in cell lysates harvested at 0, 15, 30, 60, 90, and 120 min after the CHX addition. We observed that the HIF1α protein degraded faster in the sh-HDAC4 cells than in the sh-Scr cells (Fig. 1D), which was confirmed by densitometry of multiple Western blot analysis results (E). These data confirm our ongoing hypothesis that HDAC4 shRNA knockdown causes an increase of HIF1α acetylation and a decrease of HIF1α stability.

FIGURE 1.

HDAC4 regulates HIF1α protein acetylation and stability in cancer cells. A, Hep3Bc1 cells with stable sh-Scr (-) or sh-HDAC4 (+) were cultured under normal or hypoxic condition for 6 h. B, C42B sh-Scr and sh-HDAC4 cells were cultured with or without CoCl₂ for 6 h. The HIF1α, HDAC4, and tubulin protein levels were measured using whole cell lysates by Western blot analyses. C, Hep3Bc1 sh-Scr and sh-HDAC4 cells were cotreated with proteasome inhibitor MG132 and CoCl₂ for 6 h. HIF1α was IP-purified (+) from whole cell lysates and immunoblotted (IB) for HIF1α and its acetylation (ace-K). D, Hep3Bc1 sh-Scr and sh-HDAC4 cells were treated with CoCl₂ for 6 h, followed by protein synthesis inhibitor CHX treatment for the indicated times. HIF1α and tubulin protein levels were measured by Western blot analyses. E, the decay of HIF1α protein level in D was quantified by florescence densitometry. The HIF1α protein band intensity was divided by the tubulin band intensity in each sample and normalized to the time 0 of CHX treatment. The lines represent mean ± S.D. of three experiments. *, p < 0.01, sh-Scr versus sh-HDAC4 at the indicated time points.

We reported previously that HDAC6 inhibition can also reduce HIF1α protein level. However, such inhibition was not associated with HIF1α protein acetylation (16). To further investigate the differential regulation of HIF1α by HDAC4 and other HDAC isozymes, similar stable shRNA knockdown experiments were performed in Hek293 cells with shRNA against HIF1α, HDAC1, HDAC3, and HDAC4. HDAC4 shRNA robustly inhibited HIF1α proteins under hypoxic conditions. In contrast, neither HDAC1 nor HDAC3 shRNA had a significant effect on the HIF1α protein level (Fig. 2A). We reported previously that HDAC4 interacts with HIF1α (16). To further examine the relationship between HIF1α acetylation level and HDAC4, we transiently co-overexpressed a HA-tagged HIF1α with either an empty vector or a FLAG-HDAC4 plasmid in Hek293T cells. Co-IP experiments showed that FLAG-HDAC4 interacted with HA-HIF1α (Fig. 2B). Importantly, HDAC4 overexpression reduced the HA-HIF1α protein basal acetylation (Fig. 2B). In addition, we performed an in vitro deacetylation assay (25) using overexpressed and IP-purified FLAG-HDAC1, FLAG-HDAC4, and HA-HIF1α. In the course of 2 h of deacetylation reaction, we observed that the level of HA-HIF1α acetylation gradually decreased by the presence of the HDAC4 but not by HDAC1 (Fig. 2C). These data further suggest that the HIF1α acetylation level and protein stability can be regulated by HDAC4.

FIGURE 2.

HDAC4 specifically regulates HIF1α proteins. A, Hek293 cells were transduced with lentivirus containing shRNA against scramble sequence (sh-Scr), HIF1α (sh-HIF1α), HDAC1 (sh-HDAC1), HDAC3 (sh-HDAC3), or HDAC4 (sh-HDAC4). The indicated cells were cultured under either normal or hypoxic conditions for 6 h before Western blot analysis. B, an equal amount of HA-HIF1α plasmid was co-overexpressed with either empty vector (-) or FLAG-HDAC4 (+). 48 h after transfection, whole cell lysates were used for IP with anti-HA antibodies. The level of HA-HIF1α protein and its acetylation were measured by Western blot analyses using anti-HA and anti-acetyl-lysine (Ace-K) antibodies, respectively. The FLAG-HDAC4 was measured by the anti-FLAG antibody. C, the HA-HIF1α plasmid or FLAG-HDAC4 or FLAG-HDAC1 plasmid was overexpressed in HEK293T cells. 48 h after transfection, the HA-HIF1α, FLAG-HDAC4, and FLAG-HDAC1 were IP-purified using anti-HA and anti-FLAG antibodies, respectively. The purified HA-HIF1α was mixed with purified HDAC4 or HDAC1 in deacetylase buffer for the indicated time in an in vitro deacetylation assay. The resulted proteins were analyzed by Western blot analysis using anti-HA or anti-ace-K antibodies.

HDAC4 Regulates HIF1α Protein Stability via HIF1α N-terminal Lysines

To identify the lysine residues that are deacetylated by HDAC4, we ectopically overexpressed the full-length FLAG-HIF1α in Hek293 cells with shRNA against HDAC4 (sh-HDAC4) or scramble control (sh-Scr). The overexpressed FLAG-HIF1α proteins were anti-FLAG IP-purified, resolved on SDS-PAGE gel, digested by trypsin, and subjected to LC-MS/MS at the OHSU Proteomic Core Facility. MS/MS analysis of the trypsin-digested FLAG-HIF1α peptides did not reveal a significant change in lysine acetylation between the sh-Scr and sh-HDAC4 cells. In repeated experiments, however, trypsin digestion (cleaves at the unacetylated lysine) did not generate a peptide containing the first 30 amino acids of HIF1α that was suitable for LC-MS/MS analysis. There is a cluster of five lysine residues (Lys-10, Lys-11, Lys-12, Lys-19, and Lys-21) within the HIF1α N-terminal amino acids 1–30. We speculated that this can be a region regulated by HDAC4 deacetylation. Thus, we took an alternative approach. We used the site-directed mutagenesis to replace all five lysines to arginines in the full-length FLAG-HIF1α wild-type (HIF1α-WT) construct. We observed that the acetylation level of the resulted HIF1α mutant (HIF1α-mut) was significantly lower than HIF1α-WT (Fig. 3A). This differential protein acetylation level was also associated with differential protein stability in the context of HDAC4. Co-overexpression of HDAC4 with either the HIF1α-WT or HIF1α-mut showed that HDAC4 decreased the protein acetylation of HIF1α-WT but not HIF1α-mut (Fig. 3B). On the other hand, HDAC4 overexpression increased the protein level of HIF1α-WT but not HIF1α-mut (Fig. 3B). When we overexpressed the HIF1α-WT and HIF1α-mut constructs in Hek293 cells with stable sh-Scr or sh-HDAC4 under both normal and hypoxic mimic conditions, we observed that HIF1α-WT (high in acetylation) was inhibited by HDAC4 shRNA knockdown compared with sh-Scr (Fig. 3C). This was consistent with the results in Fig. 2A. In contrast, HIF1α-mut (low in acetylation) was resistant to the HDAC4 shRNA knockdown (Fig. 3C). Further, we overexpressed the wild-type or mutant FLAG-HIF1α construct in Hek293T cells and treated cells with solvent or HDACi-suberoylanilide hydroxamic acid. Under both normal and hypoxic mimic conditions, HIF1α-WT was sensitive to HDACi. In contrast, HIF1α-mut was resistant to HDACi-induced protein degradation (Fig. 3D). These data show that the acetylation and stability of HIF1α can be regulated by HDAC4 via N-terminal lysines.

FIGURE 3.

HDAC4 regulates HIF1α protein via HIF1α N-terminal lysines. A, FLAG-HIF1α-WT or FLAG-HIF1α-mut was overexpressed in Hek293T cells for 48 h, then HIF1α proteins were IP-purified by anti-FLAG M2-agarose and analyzed by Western blot analysis using anti-HIF1α and anti-ace-K antibodies. B, FLAG-HIF1α-WT and FLAG-HIF1α-mut were co-overexpressed with Ev (-) or HDAC4 (+) in Hek293T cells. To control transfection efficiency and specific regulation by HDAC4, all cells were also cotransfected with a GFP plasmid. 48 h after transfection, whole cell lysates (WCL) were harvested and analyzed either by Western blot analysis using anti-FLAG, HDAC4, GFP, and tubulin antibodies or IP-purified by anti-FLAG M2-agarose and probed for acetylation. C, FLAG-tagged HIF1α-WT and HIF1α-mut were overexpressed in Hek293 sh-Scr cells (-) and sh-HDAC4 (+) cells for 48 h. CoCl₂ was added 4 h prior to protein harvest to mimic hypoxia. Whole cell lysates were analyzed by Western blot analyses. D, FLAG-tagged HIF1α-WT and HIF1α-mut were overexpressed in Hek293T cells for 48 h. 4 h prior to protein harvest, CoCl₂ was added to mimic hypoxia, and cells were treated with either vehicle (-) or 5 μm of HDACi-suberoylanilide hydroxamic acid (+). Whole cell lysates were analyzed by Western blot analysis.

HDAC4 Regulates HIF-1 Protein Activity via HIF1α N-terminal Lysines

Using a hypoxia/HRE-driving luciferase reporter (p2.1) assay (23), we also compared the transcriptional activity of HIF1α-WT and HIF1α-mut under non-hypoxic conditions where the endogenous HIF1α protein is inhibited by oxygen and does not confound the reporter assay result. We overexpressed either HIF1α-WT or HIF1α-mut in Hek293 cells along with a firefly luciferase reporter gene plasmid (p2.1) under the control of HRE and a constitutive renilla luciferase gene. The firefly activity can be robustly induced by HIF1α or hypoxia stimulation. The renilla activity can be used as a control for equal transfection and cell density (23). We observed that HIF1α-mut had a higher transcriptional activity than HIF1α-WT (Fig. 4A). Co-overexpression with HDAC4 significantly enhanced the HIF1α-WT activity but had no effect on HIF1α-mut (Fig. 4A). This is consistent with the HIF1α protein level regulation by HDAC4 in Fig. 3B, in which the HIF1α-WT protein level was lower than HIF1α-mut but can be up-regulated by HDAC4 overexpression. We also tested the transcriptional activity of HIF1α-WT and HIF1α-mut in HDAC4 shRNA knockdown conditions by performing a reporter gene assay in Hek293 cells with sh-Scr or sh-HDAC4. Similar to the protein results in Fig. 3C, we observed that the HIF1α-WT transcriptional activity was significantly reduced in the presence of HDAC4 shRNA knockdown (Fig. 4B). In contrast, HIF1α-mut had a higher activity than HIF1α-WT and was not significantly affected by HDAC4 shRNA (Fig. 4B).

FIGURE 4.

HDAC4 regulates HIF-1 transcriptional activity via HIF1α N-terminal lysines. A, HDAC4, HIF1α-WT, HIF1α-mut, or empty vector (Ev) was overexpressed in Hek293 cells as a single agent or in combination along with HRE-firefly and constitutive renilla luciferase. B, HIF1α-WT, HIF1α-mut, or Ev were overexpressed in Hek293 sh-Scr and sh-HDAC4 cells along with HRE-firefly and constitutive renilla luciferase plasmids. 48 h later, dual-luciferase activities were measured and expressed as values relative to the Ev. * and **, p < 0.01. All bars represent mean ± S.D. of three experiments.

The Effect of HDAC4 Knockdown on HIF-1 Activity in Cancer Cells

To measure the effect of HDAC4 shRNA knockdown on HIF-1 transcriptional activities in cancer cell lines, we performed the luciferase reporter assay using Hep3Bc1 sh-Scr and sh-HDAC4 cells. Previously, Hep3Bc1 cells had been stably cotransfected with the HRE-firefly luciferase reporter gene plasmid (p2.1) and the constitutive renilla luciferase gene (23). We observed that the sh-Scr cells exhibited very robust firefly reporter activity because of hypoxia. In contrast, the sh-HDAC4 cells did not respond to hypoxia stimulation (Fig. 5A). To confirm this, we transiently cotransfected the HRE-firefly and constitutive renilla plasmids into C42B cells containing either stable sh-HDAC4 or sh-Scr and performed similar experiments under normal and hypoxic conditions. Similar to the results in the Hep3Bc1 cell lines, hypoxia up-regulated firefly activity in sh-Scr cells. HDAC4 shRNA significantly inhibited this up-regulation (Fig. 5B). On the basis of these results, we measured the effect of HDAC4 shRNA on endogenous HIF-1 target gene up-regulation in hypoxia by quantitative RT-PCR. We found that hypoxia significantly up-regulated a subset of HIF-1 target genes, including VEGFa, lactate dehydrogenase A (LDHA), and Glut1 in sh-Scr cell lines, however, the up-regulation was significantly inhibited in the sh-HDAC4 cells (Fig. 5, C and D).

FIGURE 5.

HDAC4 regulates HIF-1 transcriptional activity and target gene expression in cancer cells. A, Hep3Bc1 cells containing stable sh-Scr or sh-HDAC4 were cultured under normal and hypoxic incubators overnight. The ratio of firefly/renilla of all samples was normalized to the value of sh-Scr under normal conditions and expressed as relative luciferase (luc). *, p < 0.001. B, C42B sh-Scr and sh-HDAC4 cells were transiently cotransfected with HRE-firefly (p2.1) and constitutive-renilla plasmids. 48 h after transfection, the transfected cells were cultured under normal or hypoxic conditions overnight. The luciferase activity was measured and calculated as above. **, p < 0.01. C and D, the indicated cells were cultured under normal or hypoxia conditions for 24 h. Selective HIF-1 target gene expressions were measured by real-time quantitative RT-PCR, and the fold change was expressed as relative values to sh-Scr under normal conditions. #, p < 0.01; &, p < 0.05. All bars represent mean ± S.D. of three experiments.

The Effect of HDAC4 Knockdown on Cellular Response to Hypoxia

On the basis of the reduced glycolytic gene (LDHA and Glut1) up-regulation by HDAC4 shRNA (Fig. 5C), we hypothesized that HDAC4 shRNA will also affect hypoxia-induced glycolysis. We measured glycolysis levels on the basis of the lactate production in the sh-Scr and sh-HDAC4 cancer cells growing under normal or hypoxic condition. Indeed, we found that the lactate levels were significantly increased in the hypoxic sh-Scr cells and that HDAC4 shRNA significantly reduced the hypoxia-induced lactate production (Fig. 6A). Next, in cell proliferation assays, we observed that HDAC4 shRNA did not change the growth rate of the cancer cell lines under normal oxygen and short-term (48-h) hypoxic conditions (Fig. 6B). When cells were cultured in long-term hypoxia (96 h), C42B cells did not survive, regardless the HDAC4 status (data not shown), but the Hep3Bc1 sh-HDAC4 cells grew significantly slower than the sh-Scr cells (Fig. 6B). Hypoxia and HIF-1 are considered key factors in causing resistance toward cytotoxic therapies in solid cancer (27, 28). We therefore tested the effect of HDAC4 shRNA on prostate cancer cell response to docetaxel, which is a microtubule-targeting agent and standard chemotherapy for late-stage metastatic prostate cancer patients. We treated C42B sh-Scr and sh-HDAC4 cells with increasing doses of docetaxel for 48 h under normal or hypoxic conditions. Cell viability at each drug concentration was calculated and adjusted to the solvent control. We observed that HDAC4 shRNA did not significantly modify docetaxel efficacy in normoxic conditions but significantly inhibited the hypoxia-induced resistance (Fig. 6C).

FIGURE 6.

HDAC4 regulates cancer cell response to hypoxia. A, lactate levels (indicative of glycolysis) were measured in sh-Scr and sh-HDAC4 cells in normoxia (n) and hypoxia (h), normalized with total cell numbers, and expressed as relative fold change to the value of sh-Scr in normoxia. *, p < 0.05. B, Hep3Bc1 sh-Scr and sh-HADC4 cells were cultured under normoxia and hypoxia for 5 days. The viable cells were quantitated at days 1, 3, and 5 of the experiments and normalized to day 1. The lines represent mean ± S.D. of three experiments. #, p < 0.05, sh-Scr (h) versus sh-HDAC4 (h) on day 5. D, C42B sh-Scr and sh-HDAC4 cells were treated with the indicated dose of docetaxel for 48 h under the normal and hypoxic conditions. For each condition, the viable cells were counted and compared with the solvent control as % of viability. **, p < 0.05, sh-Scr (h) versus sh-HDAC4 (h) at 20 nm and 30 nm of docetaxel. All experiments represent mean ± S.D. of three experiments.

DISCUSSION

Because the HIF1α protein plays a pathological role in human diseases, including cancer, understanding novel mechanisms regulating the HIF1α and HIF-1 activity is important for developing more effective and targeted therapies. In the current study, we determined that a member of the HDAC family, HDAC4, plays an important role in regulating HIF1α protein N-terminal lysine acetylation level, stability, and HIF-1 activity. These results provide a mechanistic explanation to the well observed phenomena that HDACi can induce HIF1α protein acetylation, reduce HIF1α protein level and HIF-1 target gene expression in cancer cell lines, and reduce angiogenesis in preclinical mouse xenograft models (16, 19, 29).

A likely scenario is that the N-terminal lysine acetylation level of the HIF1α protein is in part regulated by HDAC4. The inhibition of HDAC4 by either HDACi or shRNA causes an inappropriate increase of HIF1α protein acetylation (hyperacetylation), which in turn disrupts the protein stability and leads to the reduction in HIF1α protein level and HIF-1 activity. In the previous study, we have shown that the HIF1α protein inhibition because of HDAC4 inhibition requires proteasomal degradation system, but is independent of VHL and does not impact the HIF1α interaction with Hsp70/90 (16). Currently, the exact molecular detail connecting the protein acetylation and degradation is unknown and warrants further additional studies. The HIF1α N-terminal sequence is also highly similar to the HIF2α N terminus. Both are critical for mediating protein-protein and protein-DNA interaction. However, we did not observe a significant HIF2α inhibition in sh-HDAC4 cells. On the basis of the novel information regarding the HIF1α N-terminal lysines and HDAC4 presented here, we will, in additional studies, investigate in detail whether N-terminal lysine acetylation plays a role in HIF1α interaction with DNA and the additional degradation machinery such as RACK1, CHIP, and ubiquitin.

In this study, the first five lysine residues in the HIF1α N terminus collectively play a pivotal role in HIF1α protein acetylation and stability in the context of HDAC4. We have not been able to identify the exact location(s) of acetylation. Because the single lysine mutants were not significantly different from HIF1α-WT (data not shown), it is likely that a combination of these five lysines (if not all five) is involved.

On the basis of the results in this and other studies, it is apparent that multiple lysine residues within the HIF1α protein can be substrates for multiple HATs and HDACs. To date, HIF1α can be deacetylated at least at two lysine residues, Lys-532 and Lys-674, by different HDACs (14, 15). The Lys-532 deacetylation is likely to be carried out by the class I and/or II HDACs (14). The Lys-674 deacetylation requires a class III HDAC-Sirt1 (15). In our study, using HIF1α containing a point mutation at Lys-532 or Lys-674 did not indicate a clear regulation of HDAC4 toward these two lysine residues (data not shown). It is unclear whether these regulations occur to HIF1α ubiquitously or only under specific physiological conditions. This suggest that multiple HDACs are involved in HIF1α regulation. We speculate that these specific associations between HDAC isozymes and deacetylation reflect the biological complexity and redundancy on HIF-1 regulation.

Further, the acetylation and deacetylation at different lysine residues within HIF1α can lead to different biological consequences. Acetylations at the N terminus (shown in this study) and at the oxygen-dependent degradation domain (Lys-532) can promote HIF1α protein degradation and inhibition of downstream HIF-1 activity (14). In contrast, the acetylation at Lys-674 promotes the transcriptional activity of HIF-1 by recruiting the transcriptional cofactor p300 (15). Recently, it has been shown that PCAF can act as HIF1α acetylase at Lys-674 (15). Although our primary goal is to identify the deacetylase of HIF1α, we did observe that endogenous PCAFs are well expressed in all the cell lines and that PACF overexpression increased the HIF1α-WT and HIF1α-mut acetylation level. Whether the HIF1α N-terminal lysines are substrates of PCAF warrants more careful investigation in additional studies.

In this study, not all HIF-1 target genes are down-regulated by HDAC4 inhibition. This suggests that 1) the down-regulation of HIF1α protein by HDAC4 shRNA is not the only mechanism responsible for the down-regulation of HIF-1 target gene expression, 2) HDAC4 is only required for a subset of HIF1α target genes, or 3) HDAC4 directly regulates a subset of HIF-1 target genes either independently or in collaboration with HIF-1. HDAC4 is known to regulate the activity of transcriptional factors and cofactors by participating in the formation of transcriptional complexes, including HIF-1 (30–33). Therefore, we speculate that only a subset of HIF-1 target genes require HDAC4 as an HIF-1 cofactor to stabilize HIF1α and to form an effective transcriptional complex. In HDAC4 knockdown cell lines, the up-regulation of these genes in hypoxia is inhibited. In addition, the regulation of glycolysis and cytotoxic stress by HDAC4 in hypoxic cells suggests a novel role for HDAC4 in cellular adaptation toward hypoxia. Genetic studies have shown that HDAC4 regulates cellular proliferation and differentiation and is an important determinant for the development of muscle, bone, and heart (30–32, 34). Because the proper development of these tissues depends on the availability of oxygen (35), the regulatory role of HDAC4 in tissue development can potentially include the participation in responding and adapting to the change of oxygen concentrations.