PRDX2 and PRDX4 are negative regulators of hypoxia-inducible factors under conditions of prolonged hypoxia

Hypoxia-inducible factors (HIFs) control the transcription of genes that are crucial for the pathogenesis of cancer and other human diseases. The transcriptional activity of HIFs is rapidly increased upon exposure to hypoxia, but expression of some HIF target genes decreases during prolonged hypoxia. However, the underlying mechanism for feedback inhibition is not completely understood. Here, we report that peroxiredoxin 2 (PRDX2) and PRDX4 interact with HIF-1α and HIF-2α in vitro and in hypoxic HeLa cells. Prolonged hypoxia increases the nuclear translocation of PRDX2 and PRDX4. As a result, PRDX2 and PRDX4 impair HIF-1 and HIF-2 binding to the hypoxia response elements of a subset of HIF target genes, thereby inhibiting gene transcription in cells exposed to prolonged hypoxia. PRDX2 and PRDX4 have no effect on the recruitment of p300 and RNA polymerase II to HIF target genes and the enzymatic activity of PRDX2 and PRDX4 is not required for inhibition of HIF-1 and HIF-2. We also demonstrate that PRDX2 is a direct HIF target gene and that PRDX2 expression is induced by prolonged hypoxia. These findings uncover a novel feedback mechanism for inhibition of HIF transcriptional activity under conditions of prolonged hypoxia.

HIF-1α protein stability is also regulated by oxygen-independent mechanisms. The ubiquitin E3 ligase CHIP cooperates with HSP70 to induce HIF-1α protein degradation in the 26S proteasome during prolonged hypoxia [18]. HAF is another ubiquitin E3 ligase that mediates proteasome-dependent HIF-1α protein degradation and decreases HIF-1 activity [19]. BHLHE41 (also known as SHARP1) binds to, and promotes VHL-independent proteasomal degradation of HIF-1α and HIF-2α [20]. HSP90 inhibitors increase the ubiquitination and proteasomal degradation of HIF-1α that is triggered by binding of RACK1 at the site vacated by HSP90 [21]. SSAT1 binds to both HIF-1α and RACK1 to promote ubiquitination of HIF-1α [22]. The tumor suppressor p53 also binds to HIF-1α and induces MDM2dependent ubiquitination and proteasomal degradation of HIF-1α [23]. Finally, HIF-1α is also subject to lysosomal degradation through chaperone-mediated autophagy, which is mediated by binding of HSC70 and LAMP2A [24].
The peroxiredoxin (PRDX) family of peroxidases is abundantly expressed in cells and metabolizes intracellular H 2 O 2 through the thioredoxin system [36]. In mammals, there are six family members (PRDX1-6), which are divided into three subgroups according to their catalytic mechanism: typical 2-cysteine PRDX (PRDX1-4), atypical 2-cysteine PRDX (PRDX5), and 1-cysteine PRDX (PRDX6) [36]. Hypoxia induced PRDX1 expression in oral squamous carcinoma SCC15 cells [37], whereas HIF-1 suppressed PRDX3 expression in VHLdeficient clear cell renal carcinoma cells [38]. PRDX1 functioned as a ligand for Toll-like receptor 4 to enhance HIF-1α expression and HIF-1 binding to the promoter of the VEGF gene in endothelial cells, thereby potentiating VEGF expression [39]. Expression of PRDX5 targeted to the mitochondrial intermembrane space decreased hypoxia-induced reactive oxygen species and attenuated HIF-1α protein levels and HIF-1 target gene expression in rat pulmonary artery smooth muscle cells [40]. Regulation of HIF activity by PRDX2 or PRDX4 has not been reported.
In the present study, we demonstrate that several PRDX family members directly interact with HIF-1α and HIF-2α in hypoxic human HeLa cells. PRDX2 and PRDX4 suppress transcription of a subset of HIF-1 and HIF-2 target genes under conditions of prolonged hypoxia. PRDX2 is a novel HIF target gene and hypoxia-induced PRDX2 expression results in feedback inhibition of HIF activity in HeLa cells subjected to prolonged hypoxia.

Identification of PRDX family members as novel HIF-1α-and HIF-2α-interacting proteins
Our previous proteomic screening identified several positive and negative regulators that directly control the activity of HIF-1 and HIF-2 [17,41,42]. PRDX2 was also identified as a candidate HIF-1α-interacting protein in the screen. To validate the screening data, HeLa cells were transfected with an expression vector encoding V5-epitope-tagged PRDX2 and exposed to 1% O 2 for 24 h. Anti-HIF-1α antibody co-immunoprecipitated PRDX2-V5 protein from hypoxic HeLa cell lysates ( Figure 1A). Conversely, endogenous HIF-1α was coimmunoprecipitated from hypoxic cell lysates by anti-V5 antibody ( Figure 1B). These data indicate that PRDX2 interacts with HIF-1α in HeLa cells.
To further determine whether PRDX2 and PRDX4 have a direct effect on HIF-1α transactivation function, HeLa cells were co-transfected with: pGalA, which encodes the GAL4 DNA-binding domain fused to the HIF-1α transactivation domain (531-826); reporter plasmid pG5E1bLuc, which contains five GAL4-binding sites and a TATA box upstream of Fluc coding sequences [43]; pSV-Renilla; and PRDX expression vector or EV. Transfected cells were exposed to 20% or 1% O 2 for 24 h. Expression of PRDX2-V5 or PRDX4-V5, but not other PRDX family members, significantly inhibited HIF-1α transactivation domain function in HeLa cells ( Figure 3F).
We next investigated whether PRDX2 or PRDX4 regulates HIF-1α or HIF-2α protein levels. Overexpression of PRDX2-V5 ( Figure 4A) or PRDX4-V5 ( Figure 4B) did not alter HIF-1α or HIF-2α protein levels in non-hypoxic or hypoxic HeLa cells. Double knockdown of PRDX2 and PRDX4 also had no effect on expression of HIF-1α or HIF-2α levels in HeLa cells ( Figure 4C). These data rule out decreased HIF-1α or HIF-2α protein stability as the cause of PRDX2-and PRDX4-mediated inhibition of HIF transcriptional activity.
To determine whether PRDX2 or PRDX4 regulates
of cytosolic and nuclear fractions, respectively ( Figure  5). Analysis of the subcellular fractions revealed that PRDX4-V5 was present in the nucleus and the cytosol, whereas PRDX2-V5 was localized to the cytosol of non-hypoxic HeLa cells ( Figure 5). Prolonged hypoxia dramatically increased the nuclear translocation of PRDX2-V5 and PRDX4-V5 in HeLa cells ( Figure 5). However, the presence of PRDX2-V5 or PRDX4-V5 did not alter the nuclear translocation of HIF-1α or HIF-2α ( Figure 5). These data rule out impaired nuclear translocation of HIF-α subunits as the mechanism by which PRDX2 and PRDX4 interfere with HIF transcriptional activity.

Catalytic activity of PRDX2 and PRDX4 is not required for inhibition of HIF transcriptional activity
We next investigated whether the catalytic activity of PRDX2 or PRDX4 is required for HIF suppression. HeLa cells were co-transfected with: p2.1; pSV-Renilla; and expression vector encoding wild-type (WT) PRDX2-V5 or PRDX4-V5, or catalytically inactive PRDX2(C51S)-V5 or PRDX4(C124S)-V5, or EV. PRDX2(C51S)-V5 decreased hypoxia-induced p2.1 luciferase activity, similar to the effect of WT PRDX2-V5 ( Figure 7A). The effect of catalytically inactive PRDX4(C124S)-V5 was also similar to WT PRDX4-mediated suppression of HIF activity ( Figure 7B). We further confirmed these data by GalA reporter assays, which demonstrated similar effects of WT and catalytically inactive forms of PRDX2 ( Figure  7C) and PRDX4 ( Figure 7D). These data indicate that the catalytic activity of PRDX2 and PRDX4 is not required for suppression of HIF transcriptional activity.

PRDX2 and PRDX4 block HIF binding to the HREs of a subset of target genes
To study the molecular mechanisms of PRDX2-and PRDX4-mediated inhibition of HIF-1α transactivation domain function, we first tested whether PRDX2 and PRDX4 are recruited to the HREs of HIF target genes in response to prolonged hypoxia using chromatin immunoprecipitation (ChIP) assays. HeLa subclones that were stably transduced with an empty lentivirus vector (EV) or lentivirus encoding PRDX2-V5 or PRDX4-V5 were exposed to 20% or 1% O 2 for 72 h. Recruitment of PRDX2-V5 or PRDX4-V5 to the SLC2A3 gene HRE was significantly increased in HeLa cells exposed to prolonged hypoxia, whereas recruitment to the PGK1 gene HRE was not significantly increased in hypoxic cells ( Figure 10A). Thus, PRDX recruitment to SLC2A3, but not to PGK1 ( Figure 10A), was associated with negative regulation of SLC2A3, but not PGK1, gene expression ( Figure 9). We further analyzed whether PRDX2 or PRDX4 affects HIF binding. Overexpression of PRDX2-V5 significantly decreased occupancy of the SLC2A3 HRE by HIF-1α and HIF-1β, but HIF-2α occupancy of the SLC2A3 HRE was not affected by PRDX2-V5 overexpression ( Figure  10B). PRDX4-V5 overexpression significantly decreased occupancy by HIF-2α and HIF-1β, but not HIF-1α, of the SLC2A3 HRE ( Figure 10B). In contrast, overexpression of PRDX2-V5 or PRDX4-V5 did not inhibit HIF binding to the HRE of the PGK1 gene ( Figure 10C). Taken together, these data indicate that PRDX2 and PRDX4 may act in part by selectively decreasing HIF binding to a subset of target genes, leading to reduced gene transcription under prolonged hypoxia.

PRDX2 and PRDX4 do not influence RNA polymerase II binding to HREs of HIF target genes
Serine 5 phosphorylation of RNA polymerase II is necessary for gene transcription [44]. To determine whether PRDX2 and PRDX4 regulate the recruitment of phosphorylated RNA polymerase II, we performed ChIP assays using anti-RNA polymerase II (pSer5) antibody in HeLa cells exposed to 20% or 1% O 2 for 72 h. Hypoxia significantly increased RNA polymerase II (pSer5) binding to the HRE of the SLC2A3 and CA9 genes in HeLa cells ( Figure 11). Overexpression of PRDX2-V5 or PRDX4-V5 did not influence RNA polymerase II (pSer5) binding ( Figure 11). Thus, altered phosphorylation or recruitment of RNA polymerase II does not represent the mechanism by which PRDX2 and PRDX4 inhibit HIF transcriptional activity.

PRDX2 and PRDX4 do not inhibit the interaction of p300 with HIF-1α
To determine whether PRDX2 and PRDX4 regulate the recruitment of p300 to HIF-1α, we performed co-IP assays. HeLa cells were transfected with PRDX2-V5 or PRDX4-V5 expression vector, or EV, and exposed to 1% O 2 for 24 h. As shown in Figure 12, forced expression of PRDX2-V5 or PRDX4-V5 did not alter HIF-1α-p300 interaction in hypoxic HeLa cells. Thus, PRDX2 and PRDX4 do not inhibit the recruitment of p300 to HIF-1α.

PRDX2 expression is regulated by HIFs
To determine whether HIFs control PRDX expression, we exposed HeLa cells to 20% or 1% O 2 for 24 or 72 h. Reverse transcription and real-time quantitative PCR (RT-qPCR) assays demonstrated that PRDX2 mRNA levels were significantly increased after 24 h of hypoxia and then decreased to baseline levels at 72 h of hypoxia ( Figure 9G and Figure 13A), whereas PRDX4 mRNA levels were not significantly increased after 24 h of hypoxia ( Figure 13A). Consistent with mRNA induction, PRDX2 protein expression was induced by hypoxia in a time-dependent manner (Figure 8 and Figure  13B). Knockdown of HIF-1α or HIF-2α alone slightly decreased PRDX2 protein levels in hypoxic HeLa cells, but double knockdown of HIF-1α and HIF-2α prevented hypoxia-induced PRDX2 expression ( Figure 13C). These data indicate that both HIF-1 and HIF-2 induce PRDX2 expression in hypoxic HeLa cells.
To determine whether PRDX2 is a direct HIF target gene, we analyzed the genomic DNA sequence and identified the HIF binding site sequence 5'-ACGTG-3' on the antisense strand in the 5'-flanking region of the human PRDX2 gene ( Figure 14A). To determine whether HIF binds to this sequence, HeLa cells were exposed to 20% or 1% O 2 for 24 h and the chromatin was extracted, sheared, and precipitated by antibodies against HIF-1α, HIF-2α, HIF-1β, or IgG. qPCR assays using primers spanning the putative HIF binding site revealed that hypoxia significantly increased occupancy by HIF-1α, HIF-2α, and HIF-1β ( Figure 14B-14C). These data indicate that PRDX2 is a direct HIF-1 and HIF-2 target gene. These results are consistent with the effect of HIF-1α and HIF-2α Figure 12: Effect of PRDX2 and PRDX4 on HIF-1α-p300 interaction. HeLa cells were transfected with empty vector (EV) or vector encoding PRDX2-V5 or PRDX4-V5, and exposed to 1% O 2 for 24 h. WCL was subject to IP with anti-p300 antibody, followed by immunoblot assays using antibodies against HIF-1α, V5, and p300. knockdown on PRDX2 expression ( Figure 13C).
Next, we introduced a 54-bp PRDX2 DNA fragment encompassing the HIF binding site into the pGL2promoter reporter plasmid upstream of SV40 promoter and Fluc coding sequences and designated the reporter construct pGL-WT-HRE. In HeLa cells co-transfected with pSV-Renilla and pGL-WT-HRE, hypoxia dramatically increased the ratio of Fluc:Rluc activity ( Figure 14D). Mutation of the HIF binding site sequence from 5'-ACGTG-3' to 5'-AAAAG-3' significantly decreased Fluc:Rluc activity ( Figure 14D). Overexpression of HIF-1α or HIF-2α significantly increased pGL2-WT-HRE reporter activity in non-hypoxic and hypoxic HeLa cells ( Figure 14E). Taken together, these data indicate that both HIF-1 and HIF-2 bind to an HRE in the 5' flanking region of the PRDX2 gene to enhance PRDX2 expression under prolonged hypoxia.

DISCUSSION
In the present study, we analyzed the role of PRDX family members in regulating HIF transcriptional activity. We found that PRDX2 and PRDX4 interact with HIF-1α and HIF-2α, and inhibit the transcriptional activity of HIF-1 and HIF-2 in multiple cell lines. PRDX2 is localized to the cytosol of non-hypoxic HeLa cells, whereas PRDX4 is present in both the nucleus and the cytosol. Prolonged hypoxia increases the nuclear localization of PRDX2 and PRDX4. As a result, PRDX2 and PRDX4 are recruited to the HREs of a subset of HIF target genes and inhibit their transcription. PRDX2 transcription is controlled by both HIF-1 and HIF-2 in hypoxic HeLa cells. Thus, prolonged hypoxia regulates PRDX2 at both transcriptional and posttranslational levels. Recent studies also demonstrated the nuclear localization of PRDX2, with different subcellular fractions of PRDX2 having distinct functions in regulation of the androgen receptor [45,46]. Nuclear PRDX2 also Figure 13: PRDX2 expression is regulated by HIF-1 and HIF-2. A. HeLa cells were exposed to 20% or 1% O 2 for 24 h. RT-qPCR assays were performed using primers specific for the indicated mRNAs. Data are shown as mean ± SEM, n = 3. *** p < 0.001 versus 20% O 2 . B. HeLa cells were exposed to 20% or 1% O 2 for the indicated time. WCLs were subject to immunoblot assays with antibodies against PRDX2, HIF-1α, HIF-2α, and actin. The PRDX2 and actin bands were quantified by densitometry and normalized to 0 h (20% O 2 ). Normalized data are shown as mean ± SEM, n = 3. * p < 0.05, *** p < 0.001 versus 20% O2. C. HeLa-shSC (SC), HeLa-shHIF-1α(1α), HeLa-shHIF-2α(2α), and HeLa-sh1α+2α (DKD) cells were exposed to 20% or 1% O 2 for 72 h. WCLs were subject to immunoblot assays with antibodies against PRDX2, HIF-1α, HIF-2α, and actin. inhibits STAT3 transcriptional activity through redox effects [47]. Mutation of the PRDX2 catalytic site abolishes PRDX2 regulation of androgen receptor and STAT3 [47]. In contrast, the enzymatic activity of PRDX2 was not required for inhibition of HIF-1 and HIF-2 transcriptional activity. Our data indicate that the physical interaction of PRDX2 or PRDX4 with the inhibitory domain of HIF-1α is crucial for suppression of HIF-1 transcriptional activity.
There are three types of hypoxia, i.e. acute, intermittent, and prolonged hypoxia, with each type of hypoxia having distinct effects on HIF-1α and HIF-2α [9,49,50]. We have previously shown that prolonged hypoxia causes the selective decay of HIF-1α protein, which is mediated at least in part by HSP70-and CHIP-dependent ubiquitination and proteasomal degradation [18]. In the current study, we have demonstrated that PRDX2 and PRDX4 knockdown increases the expression of HIF target genes during prolonged hypoxia. However, only a subset of HIF target genes was controlled by PRDX2 and PRDX4, as previously reported for Reptin [30]. Reduced recruitment of HIFs to the HREs of selected HIF target genes represents a mechanism by which PRDX2 or PRDX4 inhibits HIF-mediated transactivation. Interestingly, PRDX2 and PRDX4 differentially inhibit recruitment of HIF-1α and HIF-2α to the HRE of the SLC2A3 gene. The mRNA levels of three out of six genes we tested (SLC2A3, GPI, and PDK3) were selectively decreased by PRDX2 and PRDX4 during prolonged hypoxia, suggesting that PRDX2 and PRDX4 may be negative regulators of glucose reprogramming in cancer cells exposed to prolonged hypoxia. Cancer cells exposed to prolonged hypoxia are more aggressive and resistant to radiation and chemotherapy, and differential gene expression is likely to contribute to this phenotype [51]. It will be interesting to determine if PRDX2 and PRDX4 regulate the aggressive phenotype of cancer cells under prolonged hypoxia. As a further measure of complexity, we demonstrate that PRDX2 itself is encoded by a HIF target gene, providing a mechanism for feedback inhibition of genes whose high-level expression is only required during the acute phase of hypoxic exposure.

Cell culture and transfection
HeLa cells, mouse embryo fibroblasts, and HEK293T cells were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum at 37ºC in a 5% CO 2 /95% air incubator. Cells were transfected using PolyJet DNA, according to the manufacturer's protocol (SignaGen). HeLa cells stably transfected with Teton-shPRDX2 and Teton-shPRDX4 vectors were treated with doxycycline (0.5 µg/ml) and sodium pyruvate (10 mM). All cells were verified as Mycoplasma free by PCR.

Hypoxia
Cells were placed in a modular incubator chamber (Billups-Rothenberg) flushed with a gas mixture containing 1% O 2 , 5% CO 2 , and balance N 2 and incubated at 37 o C.
Equal amounts of GST and GST-HIF-1α fusion proteins immobilized on glutathione-Sepharose beads were incubated overnight with whole cell lysates. After washing three times, the bound proteins were fractionated by SDS-PAGE, followed by immunoblot assays.

Subcellular fractionation assays
HeLa cells were lysed in hypotonic buffer [10 mM HEPES/KOH (pH 7.5), 10 mM KCl, 1.5 mM MgCl 2 , 1 mM K 2 EDTA, 1 mM EGTA, 0.1% Igepal, 1 mM DTT, and protease inhibitor cocktail] in a Dounce homogenizer (30 strokes). Intact cells were removed by centrifugation at 50 g for 10 min. The nuclei were collected by centrifugation at 800 g for 10 min, washed, and lysed in isotonic buffer (hypotonic buffer plus 250 mM sucrose) by sonication to prepare the nuclear fraction. The supernatant was centrifuged at 13,000 g for 10 min and the resulting supernatant was taken as the cytosolic fraction [41].

RT-qPCR assays
Total RNA was isolated using Trizol (Invitrogen) and treated with DNase I (Ambion). RT-qPCR assays were performed as described [41]. Primer sequences are shown in Table 1.