|Home | About | Journals | Submit | Contact Us | Français|
The levels of hypoxia-inducible factor 1α (HIF-1α) are tightly controlled. Here, we investigated the posttranscriptional regulation of HIF-1α expression in human cervical carcinoma HeLa cells responding to the hypoxia mimetic CoCl2. Undetectable in untreated cells, HIF-1α levels increased dramatically in CoCl2-treated cells, while HIF-1α mRNA levels were unchanged. HIF-1α translation was potently elevated by CoCl2 treatment, as determined by de novo translation analysis and by monitoring the polysomal association of HIF-1α mRNA. An internal ribosome entry site in the HIF-1α 5′ untranslated region (UTR) was found to enhance translation constitutively, but it did not further induce translation in response to CoCl2 treatment. Instead, we postulated that RNA-binding proteins HuR and PTB, previously shown to bind HIF-1α mRNA, participated in its translational upregulation after CoCl2 treatment. Indeed, both RNA-binding proteins were found to bind HIF-1α mRNA in a CoCl2-inducible manner as assessed by immunoprecipitation of endogenous ribonucleoprotein complexes. Using a chimeric reporter, polypyrimidine tract-binding protein (PTB) was found to bind the HIF-1α 3′UTR, while HuR associated principally with the 5′UTR. Lowering PTB expression or HuR expression using RNA interference reduced HIF-1α translation and expression levels but not HIF-1α mRNA abundance. Conversely, HIF-1α expression and translation in response to CoCl2 were markedly elevated after HuR overexpression. We propose that HuR and PTB jointly upregulate HIF-1α translation in response to CoCl2.
The hypoxia-inducible factor (HIF) functions as a master regulator of angiogenesis, a process whereby new blood vessels are formed to ensure adequate oxygen perfusion of tissues. Under conditions of low oxygen tension (hypoxia), HIF activates the transcription of genes bearing hypoxia response elements in their promoter regions. HIF-regulated genes include many cancer-related genes, such as vascular endothelial growth factor (VEGF), the glucose transporter GLUT1, erythropoietin, β-catenin, and many other genes implicated in cellular functions, such as proliferation, angiogenesis, glucose metabolism, and survival (64). Consequently, understanding the regulation of HIF function and searching for therapies to inhibit HIF activity have been the focus of intense study in recent years (56).
Expressed in two main isoforms (HIF-1 and HIF-2), HIF is a heterodimeric complex consisting of a transiently expressed, oxygen-regulated α-subunit (HIF-1α and HIF-2α) and a constitutively expressed β subunit (HIF-1β). The HIF-1α gene has been proposed to be constitutively transcribed through the action of transcription factors such as AP-1, AP-2, NF-1, and NF-κB (10). Under normal oxygen conditions (normoxia, ranging from 2% to 14% O2, depending on the tissue ), HIF-1α protein levels are essentially undetectable, but in response to hypoxia, HIF-1α expression is rapidly and potently elevated. The posttranslational stabilization of the HIF-1α protein is the best-understood mechanism to increase its expression levels. Under normoxic conditions, HIF-1α has a very short half-life (~5 min), achieved through hydroxylation at conserved prolyl and asparagyl residues that are targeted for degradation by the von Hippel-Lindau protein (pVHL), the recognition factor of a ubiquitin E3 ligase complex (5, 30, 55). During hypoxia, these hydroxylases are quickly inhibited, in turn causing a rapid and robust stabilization of HIF-1α (37).
Additional mechanisms to regulate HIF-1α expression during hypoxia have begun to emerge. HIF-1α expression is also strongly regulated at the posttranscriptional level, primarily through the action of trans-acting factors (noncoding RNAs and RNA-binding proteins) that interact with the HIF-1α mRNA and regulate its decay and translation rates. Among the noncoding RNAs, the naturally occurring antisense HIF transcript has been shown to bind to the HIF-1α mRNA and to regulate its expression posttranscriptionally (51, 59), and microRNAs have been postulated to influence HIF-1α expression in response to hypoxia (11). The polypyrimidine tract-binding protein (PTB) was shown to bind to the internal ribosome entry site (IRES) located in the 5′ untranslated region (UTR) of the HIF-1α mRNA (and in many other viral and cellular mRNAs) (23, 46) and to promote the translation of HIF-1α during hypoxia (34, 54). In fact, the translational control of HIF-1α expression is emerging as a major regulatory step, estimated to account for ~40 to 50% of the elevation in HIF-1α protein levels by short-term hypoxia (1% O2, 3 h) (6, 22). In this regard, while hypoxia can inhibit general protein synthesis and the translational regulator mammalian target of rapamycin (mTOR), the cap-mediated translation of a subset of mRNAs, including HIF-1α mRNA, is induced; this effect was proposed to be linked to protein kinase Cα activity (22).
Under normoxic conditions, numerous other stimuli, such as cytokines, vascular hormones, iron chelators, and growth factors, have also been shown to modulate HIF-1α expression (reviewed in references 30 and 67). Nonhypoxic inducers of HIF-1α expression appear to function primarily by enhancing HIF-1α biosynthesis in a variety of cells. For example, vasoactive hormones (such as angiotensin II, thrombin, and endothelin) were found to increase HIF-1α expression through a combination of upregulated HIF-1α gene transcription and heightened HIF-1α translation (47). The latter effect was linked to the activation of phosphatidylinositol 3-kinase (PI3K) by vasoactive hormones, leading to the activation of the downstream effectors mTOR, p70 S6 kinase (S6K), and the ribosomal protein S6 (rpS6). Phosphorylated rpS6, in turn, has been widely believed to promote the translation of a specific subset of transcripts bearing 5′-oligopyrymidine tracts (5′TOP), among them HIF-1α mRNA (3, 47). This hypothesis was recently proven incorrect, as S6K knockout mice and rpS6 knockin mice (carrying mutations in all rpS6 phosphorylation sites) displayed normal translational control of TOP mRNAs (48, 52; reviewed in reference 53). In addition, mTOR-independent, Akt-dependent pathways of increased HIF-1α translation have also been reported (49). Treatment with another nonhypoxic agent, the androgen dihydrotestosterone, has also been proposed to influence HIF-1α expression posttranscriptionally, via the formation of ribonucleoprotein (RNP) complexes comprising HIF-1α mRNA and the RNA-binding protein HuR (57). A member of the Hu/Elav family, HuR has been implicated in stabilizing and regulating the translation of many target mRNAs (4, 18, 29); however, the specific posttranscriptional influence of HuR upon the expression of HIF-1α in dihydrotestosterone-treated cells was not examined directly (57).
Our own studies also identified the HIF-1α mRNA as a putative target of HuR, since it contained several hits of a motif present in HuR target transcripts (38). Given the links between HuR and cancer (19, 39), and between cancer and HIF-1α (50, 56, 62), we were interested in testing experimentally if HuR contributed to the regulation of HIF-1α expression. Here, we investigate the response of human cervical carcinoma HeLa cells to the hypoxia mimetic cobalt chloride (CoCl2). Like other metal ions, cobalt(II) can mimic the response to hypoxia in cells under normal oxygen tension, elevating HIF-1α levels and implementing a HIF-1-regulated gene expression program (41); however, the two stimuli differ in that hypoxia triggers an antiapoptotic response, but CoCl2 and other mimetics do not (12). Using CoCl2 in the present studies was advantageous because it enabled many experiments that could not be performed under the restrictions of hypoxia chambers. Moreover, CoCl2 is a physiologic stress in its own right, as metal exposure represents a significant environmental and occupational risk for humans today (9). Our results show that HuR and PTB bind the HIF-1α mRNA and promote HIF-1α translation under hypoxia-like conditions.
Human cervical carcinoma HeLa cells were cultured in Dulbecco's modified essential medium (Invitrogen), and human lung carcinoma A549 cells were cultured in Ham's F-12K medium (Invitrogen), each supplemented with 10% fetal bovine serum and antibiotics. Cells were treated at ~80% confluence. Hypoxia was achieved by placing cells in a hypoxia incubator (1% O2, 5% CO2, 94% N2; Thermo Forma). CoCl2 (Sigma) was used at 200 μM for 2.5 h (unless otherwise indicated) in complete medium containing 5% fetal bovine serum. Small interfering RNAs (siRNAs; all from Qiagen) were used at a final concentration of 30 nM. siRNAs targeting HuR (pooled before use) were AATCTTAAGTTTCGTAAGTTA (HuR U1), TTCCTTTAAGATATATATTAA (HuR U2), and AAGTGCAAAGGGTTTGGCTTT (HuR H4); PTB-directed siRNA was S100301490; control (Ctrl) siRNAs were AATTCTCCGAACGTGTCACGT (c1) and AllStars negative control siRNA (c2). Adenovirus AdHuR (custom-made; containing the entire HuR coding region) and control AdCtrl were generated and amplified, and titers were determined by ViraQuest, Inc.; they were used at the indicated PFU.
Plasmid pEGFP-HIF(5′) was generated by cloning the HIF-1α 5′UTR (bp 12 to 352) into the multiple cloning site of pEGFP-N1 (Clontech) at the HindIII restriction site; pEGFP-HIF(3′) was generated by cloning the HIF-1α 3′UTR (bp 2793 to 3934) into the multiple cloning site of pEGFP-C1 (Clontech) after HindIII and BamH1 digestion. The HIF-1α 5′UTR was amplified by PCR using cDNA prepared from HeLa cells as a template and the oligonucleotide primers GCCGGCTAGCTCGTCTGAGGGGACAGGA (NheI site underlined) and GCCGCTCGAGGGTGAATCGGTCCCCGCGATGT (XhoI site underlined). The resulting product was subsequently cloned into the NheI and XhoI sites of the bicistronic reporter plasmid pBIC (20) to generate pBIC-HIF-1α. The promoterless pBIC-HIF-1α(−CMV) plasmid was generated by restriction digestion of pBIC-HIF-1α with BamHI and BglII to remove the cytomegalovirus (CMV) promoter, followed by religation of the plasmid backbone.
Whole-cell lysates (WCE) were prepared using radioimmunoprecipitation (RIPA) buffer. Cytoplasmic and nuclear protein fractions (CE and NE, respectively) were isolated with NE-PER nuclear and cytoplasmic extraction reagents (Pierce) following the manufacturer's specifications. Proteins were resolved by 8%, 10%, or 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene difluoride membranes (Invitrogen). Incubations with primary mouse monoclonal antibodies recognizing HuR (Santa Cruz), β-actin (Abcam), α-tubulin (Santa Cruz), or HIF-1α (BD Biosciences Pharmingen), as well as with goat polyclonal anti-PTB antibody (Abcam), were followed by incubations with the appropriate secondary antibodies (Amersham). Protein bands were detected by ECL-Plus (Amersham) or SuperSignal West Femto maximum sensitivity substrate (Pierce). For coimmunoprecipitation (co-IP) analysis, whole-cell lysates (5 mg/ml) prepared in RIPA buffer were either left untreated or incubated with RNase A (100 μg/ml) and RNase T1 (1,000 U/ml) for 10 min at 30°C. IP was performed using 20 μg antibody (mouse immunoglobulin G [IgG] from BD Biosciences Pharmingen, anti-HuR antibody from Santa Cruz, or anti-PTB from Abcam) for 16 h at 4°C; IP samples were then assayed by Western blotting.
Transiently transfected cells were washed in 1 ml phosphate-buffered saline and harvested in 300 μl chloramphenicol acetyltransferase (CAT) enzyme-linked immunosorbent assay (ELISA) kit lysis buffer according to the manufacturer's protocol (Roche Molecular Biochemicals). β-Galactosidase (β-Gal) enzymatic activity was determined by spectrophotometric assay using o-nitrophenyl-β-d-galactopyranoside as previously described (40). CAT levels were determined using the CAT ELISA kit according to the protocol provided by the manufacturer (Roche Molecular Biochemicals). The relative IRES activity was calculated as the CAT/β-Gal ratio. The levels of neomycin phosphotransferase (NEO; NPTII) were determined using the NPTII ELISA kit following the manufacturer's protocol (Agdia).
Total RNA was isolated from transfected cells using the RNAspin Mini RNA isolation kit according to the manufacturer's instructions (GE Healthcare). cDNA was generated using an oligo(dT)18 primer and the Bulk 1st-Strand synthesis kit according to the protocol provided by the manufacturer (GE Healthcare). The synthesized cDNA was used as the template for quantitative PCR using the QuantiTect SYBR green PCR kit (Qiagen) and analyzed on an ABI Prism 7000 detection system using the ABI Prism 7000 SDS software. Quantitative PCRs were carried out using primer pairs (forward and reverse in each case) ACTATCCCGACCGCCTTACT and CTGTAGCGCTGATGTTGAA to detect the β-Gal mRNA and GCGTGTTACGGTGAAAACCT and GGGCGAAGAACTTGTCCATA to detect the CAT mRNA.
For the detection of HIF-1α transcriptional targets VEGF and GLUT1, total RNA was prepared from HeLa cells and mRNA levels were measured by RT-qPCR analysis. The primer pair CCTAAGGATCTCTCAGGAGCACAG and TCAGGTTTGGAAGTCTCATCCAG was used for GLUT1. Transcript levels were normalized to those of the housekeeping glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA; GAPDH and VEGF primer sequences are listed above.
HeLa cells were either left untreated or treated with CoCl2 for 2.5 h. Translation was inhibited with 0.1 mg/ml cycloheximide (10 min), and then cells were scraped and lysed in 300 μl PEB lysis buffer (0.3 M NaCl, 15 mM MgCl2, 15 mM Tris-HCl, pH 7.6, 1% Triton X-100) on ice for 10 min. Nuclei were pelleted at 10,000 × g for 10 min, and 250 μg of protein of the resulting supernatant was fractionated through a 10 to 50% linear sucrose gradient, as described elsewhere (15). Eleven fractions were obtained using a fraction collector (Brandel), and their quality was monitored at 254 nm using a UV-6 detector (ISCO). RNA in each fraction was extracted with TRIzol (Invitrogen); following RT, real-time qPCR was performed using SYBR green (Applied Biosystems) and the following primer pairs: AACCCATTCCTCACCCATCA and TCCACCTCTTTTGGCAAGCA for HIF-1α and ACATCAAGAAGGTGGTGAAGCAGG and CCAGCAAGGATACTGAGAGCAAGAG for GAPDH.
Endogenous mRNA-protein complexes were precipitated as previously described (32). Briefly, HeLa cytoplasmic lysates were prepared in polysome lysis buffer (100 mM KCl, 5 mM MgCl2, 10 mM HEPES, pH 7.0, 0.5% Nonidet P-40 [NP-40], 1 mM dithiothreitol) containing 100 units/ml RNase Out (Invitrogen) and a protease inhibitor cocktail (Roche). One mg of lysate was incubated (1 h, 4°C) with 100 μl of a 50% (vol/vol) suspension of protein A-Sepharose beads precoated with 30 μg each of mouse anti-HuR (Santa Cruz), goat anti-PTB (Abcam), goat anti-TIAR or anti-TIA-1 (Santa Cruz), G3BP (BD Transduction Laboratories), and mouse IgG1 or goat IgG1 (BD Pharmingen). The beads were washed five times with NT2 buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM MgCl2, and 0.05% NP-40). For RNA analysis, the beads were incubated with 100 μl NT2 buffer containing 20 units of RNase-free DNase I (15 min, 30°C), washed twice with 1 ml NT2 buffer and further incubated in 100 μl NT2 buffer containing 0.1% SDS and 0.5 mg/ml proteinase K (15 min, 55°C) to digest the proteins bound to the beads. RNA was extracted using phenol and chloroform, precipitated in the presence of glycoblue, and used in RT reactions in the presence of random hexamers, oligo(dT) primer, and SuperScript II reverse transcriptase (Invitrogen). Conventional PCR (28 cycles) and qPCR amplification were performed using the following specific primer pairs: AACCCATTCCTCACCCATCA and TCCACCTCTTTTGGCAAGCA for HIF-1α, TCTACCTCCACCATGCCAAGT and GATGATTCTGCCCTCCTCCTT for VEGF, ACGTAAACGGCCACAAGTTC and AAGTCGTGCTGCTTCATGTG for enhanced green fluorescent protein (EGFP), and ACATCAAGAAGGTGGTGAAGCAGG and CCAGCAAGGATACTGAGAGCAAGAG for GAPDH.
Nascent translation of HIF-1α and GAPDH was studied by incubating HeLa cells with 1 mCi l-[35S]methionine and l-[35S]cysteine (Easy Tag Express; NEN/Perkin-Elmer, Boston, MA) per 60-mm plate for 15 min. Cells were lysed in RIPA buffer (10 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1% NP-40, 1 mM EDTA, 0.1% SDS, and 1 mM dithiothreitol), and IP reactions were carried out in 1 ml TNN buffer (50 mM Tris-HCl [pH 7.5], 250 mM NaCl, 5 mM EDTA, 0.5% NP-40) for 16 h at 4°C using anti-HIF-1α (BD Pharmingen), IgG1 (BD Pharmingen), or anti-GAPDH (Santa Cruz Biotechnology) antibodies. Following extensive washes in TNN buffer, the IP samples were resolved by SDS-PAGE, transferred onto polyvinylidene difluoride filters, and visualized with a PhosphorImager (Molecular Dynamics). The signals were quantified and presented as a percentage of the signals in CoCl2-treated cells; the values of radiolabeled signals were adjusted to the amount of label in that sample.
In keeping with earlier literature (36), exposure of HeLa cells to either hypoxia (1% O2) or to the widely used hypoxia mimetic CoCl2 (200 or 500 μM) for 2.5 h did not change HIF-1α mRNA levels significantly, as assessed after RT of total RNA followed by conventional PCR or real-time qPCR amplification (Fig. 1A and B). By contrast, HIF-1α protein levels increased dramatically in cells treated with CoCl2 (200 μM; the dose used in all subsequent experiments) and showed a much less pronounced induction (~20-fold lower) after incubation in 1% O2 (Fig. (Fig.1C1C).
Given the robust increase in HIF-1α protein levels by CoCl2 treatment, we postulated that in addition to possibly stabilizing the HIF-1α protein (like hypoxia does), CoCl2 treatment also elevated HIF-1α translation. To test this hypothesis, we monitored the rate of nascent HIF-1α translation by performing a brief (15-min) incubation with l-[35S]methionine and l-[35S]cysteine in HeLa cells that had been treated with 1% O2 or CoCl2 for 2.5 h or had been left untreated. Following cell lysis, nascent HIF-1α was visualized by IP of radiolabeled material using either anti-HIF-1α or IgG antibodies. This assay revealed a markedly elevated de novo translation of HIF-1α in CoCl2-treated cells compared with untreated cells; by contrast, hypoxia (1% O2, 2.5 h) had only a modest stimulatory effect on HIF-1α translation (~2% of what was seen after CoCl2 treatment) (Fig. (Fig.2A).2A). De novo translation of the housekeeping protein GAPDH, tested in parallel, remained unchanged among treatment groups (Fig. (Fig.2A2A).
Further evidence that HIF-1α was translated more actively in the presence of CoCl2 was obtained from experiments that examined the association of HIF-1α mRNA with the translation machinery. The relative distribution of the HIF-1α mRNA on sucrose gradients used to fractionate polysomes according to their size was used as a measure of the HIF-1α mRNA engagement with the translational apparatus. Compared with the distribution of HIF-1α mRNA in untreated cells, CoCl2 treatment shifted the distribution of HIF-1α mRNA towards heavier fractions of the gradient, revealing an enhanced association of HIF-1α mRNA with larger polysomes (Fig. (Fig.2B)2B) that was consistent with the elevation in HIF-1α de novo protein biosynthesis (Fig. (Fig.2A).2A). A control housekeeping mRNA (GAPDH) was routinely found to shift towards heavier polysomes. This observation contrasted with the prediction that global translation would decrease after CoCl2 treatment. However, in this study system, we did not observe a loss in overall protein translation rates after 2.5 h in 1% O2 or CoCl2 (see Fig. S1 in the supplemental material), although translational elongation has been reported to slow down during hypoxia due to alteration of eEF2 activity (8).
It is worth noting that the GAPDH coding region is significantly shorter (1.0 kb) than the HIF-1α coding region (2.5 kb). Accordingly, an actively translated HIF-1α mRNA would be predicted to accommodate many more ribosomes within its coding region and, hence, form heavier polysomes than those seen with GAPDH mRNA. Instead, HIF-1α polysomes peaked at a size that cosedimented with GAPDH polysomes (approximately fraction 7). It was only after CoCl2 treatment that HIF-1α polysomes showed the expected heavier size (approximately fraction 10), revealing a robust enhancement in HIF-1α translation, while GAPDH polysomes showed a comparatively less pronounced shift (approximately fraction 8) after CoCl2 treatment (Fig. (Fig.2B).2B). Taken together, these results indicate that exposure to CoCl2 promotes the translation of the HIF-1α mRNA, linked to the increase in HIF-1α protein levels.
To begin to study the mechanisms mediating the CoCl2-induced translation of HIF-1α mRNA, we treated cells with pateamine A, a small molecule that inhibits cap-dependent translation. As shown in Fig. Fig.3A,3A, pateamine A treatment potently suppressed HIF-1α expression, suggesting that CoCl2 induced HIF-1α translation in a cap-dependent manner. However, since pateamine A could also function by inhibiting translation driven by IRES, and given earlier reports that an IRES within the 5′UTR of the mouse HIF-1α promoted HIF-1α translation (34, 46, 54), we sought to study if the human HIF-1α 5′UTR also harbored IRES activity and, if so, if HIF-1α IRES activity was enhanced in response to CoCl2 treatment. The human HIF-1α 5′UTR was cloned into the β-Gal-CAT bicistronic reporter system (pBIC) (Fig. (Fig.3B),3B), in which expression of the 5′ cistron (β-Gal) is driven by cap-dependent translation, whereas expression of the 3′ cistron (CAT) is driven by IRES-dependent translation. HEK 293 cells were transfected with either the empty reporter plasmid pBIC or the pBIC-HIF-1α construct. We found that the HIF-1α 5′UTR could enhance expression of the 3′ CAT cistron (Fig. (Fig.3C);3C); importantly, expression of the 3′ CAT cistron mediated by the insertion of the HIF-1α 5′UTR sequence was not due to either cryptic promoter activity (Fig. (Fig.3D)3D) or splicing of the bicistronic mRNA (Fig. (Fig.3E).3E). These data indicate that the human HIF-1α 5′UTR exhibits IRES activity.
To determine if HIF-1α IRES activity was enhanced during CoCl2 treatment, 18 h after transfection of the pBIC-HIF-1α bicistronic reporter, HeLa cells were treated with CoCl2 (200 μM, 2.5 h). Surprisingly, CoCl2 treatment did not enhance HIF-1α IRES activity above basal levels, despite the strong induction of endogenous HIF-1α (Fig. (Fig.3G).3G). These findings suggest that increased translation of HIF-1α during CoCl2 treatment is not mediated by an enhancement of HIF-1α IRES activity.
Besides the reported IRES-dependent increase in mouse HIF-1α translation after hypoxia, RNA-binding proteins have also been shown to enhance HIF-1α translation. Therefore, we next examined the possibility that RNA-binding proteins elicit the CoCl2-mediated increase in HIF-1α translation. The ~4-kb HIF-1α mRNA contains seven computationally predicted hits of a previously identified HuR motif (38) in the HIF-1α coding region and 3′UTR, in agreement with earlier findings that HuR forms a complex with the HIF-1α mRNA in leukemia and prostate carcinoma cell extracts (57). However, using biotinylated partial transcripts spanning the entire HIF-1α mRNA, we observed that cytoplasmic HuR present in lysates from CoCl2-treated cells bound preferentially the HIF-1α 5′UTR and was also capable of binding the coding region (see Fig. S2 in the supplemental material). Unexpectedly, PTB associated with all of the biotinylated fragments tested in vitro, indicating that PTB had a general affinity for each of the transcripts (see Fig. S2 in the supplemental material), all of which contained polypyrimidine tracts of different lengths, and suggesting that the specificity of binding of PTB to RNAs could not be studied adequately by this in vitro assay. Therefore, we decided to narrow down the region to which HuR and PTB bound the HIF-1α mRNA by constructing and assaying chimeric reporters containing the EGFP coding region along with either the 5′UTR or the 3′UTR of HIF-1α (Fig. (Fig.4).4). Following transfection of these constructs into HeLa cells, we performed RNP-IP assays using anti-HuR, anti-PTB, or (control) IgG, and tested the abundance of EGFP, EGFP-HIF(5′), or EGFP-HIF(3′) mRNAs in the IP materials by using RT followed by qPCR. This analysis revealed that HuR preferentially bound the 5′UTR-containing EGFP chimeric transcript, while PTB preferentially bound the 3′UTR-containing EGFP chimeric transcript (Fig. (Fig.4,4, graph).
Given the evidence that HuR and PTB associated with chimeric RNAs carrying the HIF-1α 5′UTR and 3′UTR, respectively, we sought to test if HuR and PTB also bound the endogenous HIF-1α mRNA. Employing lysates from untreated HeLa cells, RNP-IP assays revealed a ~10-fold enrichment in HIF-1α mRNA when using an anti-HuR antibody compared with control IP reactions using IgG (Fig. (Fig.5A).5A). When using lysates from CoCl2-treated cells, the association of HIF-1α mRNA with HuR was a striking >150-fold higher than that seen in control IgG IP reactions (Fig. (Fig.5A).5A). A substantial enrichment in PTB-bound HIF-1α mRNA was also observed, particularly after CoCl2 treatment (~80-fold more abundant in PTB IP than in IgG IP). As a positive control, the binding of VEGF mRNA to HuR was tested, revealing ~60-fold-increased binding after exposure to CoCl2; PTB also associated with the VEGF mRNA (as previously reported ), showing an enrichment of ~14-fold (Fig. (Fig.5A).5A). As a negative control, the abundance of the housekeeping (nontarget) GAPDH mRNA in the HuR IP or PTB IP was very low (comparable to the IgG IP) and was unchanged by CoCl2 treatment (Fig. (Fig.5A).5A). The relative enrichment of these mRNAs in each RNP-IP reaction was also tested by RT followed by conventional PCR amplification and was visualized on agarose gels (Fig. (Fig.5B).5B). These RNP complexes were also readily detected in another cell type, the human lung carcinoma cell line A549, after hypoxia for 2.5 h, indicating that these associations were not specific to HeLa cells (Fig. (Fig.5C).5C). Other RNA-binding proteins tested showed little or no enrichment compared with IgG (like G3BP or TIA-1 [Fig. [Fig.5C])5C]) or showed no hypoxia-inducible binding (TIAR [Fig. [Fig.5C]).5C]). Collectively, our findings indicate that PTB and HuR can bind to the 3′UTR and 5′UTR of HIF-1α mRNA, and their association with the endogenous HIF-1α mRNA increases following CoCl2 treatment.
Since earlier studies indicated that PTB stimulated the translation of HIF-1α in hypoxic cells (54), we hypothesized that PTB performed this function as part of a multiprotein RNP complex with the HIF-1α mRNA, possibly involving HuR. First, we examined if HuR formed direct protein-protein associations with PTB in an RNA-independent manner. Indeed, HuR was found to coimmunoprecipitate with PTB (Fig. (Fig.6A)6A) and, conversely, PTB was found to coimmunoprecipitate with HuR (Fig. (Fig.6B).6B). However, digestion with RNases A and T1 inhibited these associations, indicating that HuR and PTB formed complexes with shared RNAs but the two proteins did not interact directly.
To assess if HuR and PTB were functionally linked as HIF-1α translational inducers, we began by testing the influence of PTB upon HIF-1α translation. We reduced PTB expression levels in HeLa cells by using RNA interference, achieved by transfection of siRNA targeting the PTB mRNA (see Materials and Methods). As shown in Fig. Fig.7A,7A, PTB was effectively silenced, particularly in cytoplasmic extracts; in keeping with previous studies using hypoxia (54), the silencing of PTB strongly inhibited HIF-1α expression in CoCl2-treated HeLa cells (Fig. (Fig.7A).7A). To test if the levels of PTB affected HuR function, we examined the abundance of HuR-HIF-1α mRNA complex RNPs in cultures with normal or silenced PTB levels. PTB indeed appeared to be necessary for maximal binding of HuR in CoCl2-treated cells, since in PTB siRNA populations, HuR-HIF-1α mRNA complex RNPs were diminished to approximately one-half of the RNP levels seen in control siRNA populations (Fig. (Fig.7B).7B). These differences in HuR-HIF-1α mRNA complex abundance were assessed by RT followed by both real-time qPCR (Fig. (Fig.7B,7B, top) and conventional PCR analyses (Fig. (Fig.7B,7B, bottom). No changes in the steady-state levels of HIF-1α mRNA were observed, as these remained unaltered among the transfection groups (Fig. (Fig.7C).7C). Together, these findings indicate that binding of HuR to the HIF-1α mRNA decreases when PTB levels are reduced and support the view that PTB and HuR cooperate in promoting binding to HIF-1α mRNA.
Next, we tested if modulating HuR levels influenced HIF-1α expression. First, we overexpressed HuR by transfecting cells with a vector that expressed HuR linked to a tandem affinity purification (TAP) tag (HuR-TAP) (33). By 72 h after transfection, HuR overexpression in untreated cells had a negligible effect on HIF-1α production; by contrast, the CoCl2-triggered rise in HIF-1α expression was enhanced in the presence of elevated HuR levels (Fig. (Fig.8A).8A). Again, the positive effect of HuR overexpression on HIF-1α abundance was most clearly observed in CE, likely because changes in translation are reflected first in the cytoplasm. The positive influence of HuR upon HIF-1α expression levels in CoCl2-treated cultures was due, at least in part, to the enhanced association of HIF-1α mRNA with actively translating polysomes. In CoCl2-treated cultures, the distribution of HIF-1α mRNA along sucrose gradients shifted towards heavier polysome fractions in the presence of elevated HuR levels, both when considering individual fractions (Fig. (Fig.8B,8B, inset) and when pooling fractions by translational activity into untranslated (nonpolysomal [NP]) fractions, moderately translated (low-molecular-weight polysomes [LMWP]) fractions, or actively translated (high-molecular-weight polysomes [HMWP]) fractions (Fig. (Fig.8B).8B). In additional experiments, a more accentuated overexpression of HuR was achieved by infection with an adenoviral vector (AdHuR). This intervention also elevated HIF-1α protein expression levels. HIF-1α expression was modestly elevated when using low PFU of AdHuR (Fig. (Fig.8C,8C, left) but was strongly elevated at high AdHuR PFU (Fig. (Fig.8C,8C, right).
Second, we reduced HuR expression by RNA interference. As shown, silencing of HuR by transfecting HeLa cells with HuR-directed siRNAs (see Materials and Methods) lowered HuR abundance and strongly diminished the CoCl2-induced HIF-1α levels (Fig. (Fig.9A);9A); as with PTB silencing (Fig. (Fig.7A),7A), these effects were observed more prominently in the cytoplasm (Fig. (Fig.9A,9A, CE). This intervention also reduced the binding of PTB to the HIF-1α mRNA (Fig. (Fig.9B)9B) but did not affect HIF-1α mRNA levels (Fig. (Fig.9C).9C). Together with the effect of PTB silencing (Fig. (Fig.7),7), our results support a cooperative association of HuR and PTB with the HIF-1α mRNA, since reducing HuR levels diminishes the binding of PTB to HIF-1α mRNA, while reducing PTB levels lowers HuR binding to the HIF-1α mRNA (Fig. (Fig.7B7B and and9B9B).
The decreases in HIF-1α levels after PTB or HuR silencing (Fig. (Fig.77 and and9,9, respectively) were due, at least in part, to the reduced association of HIF-1α mRNA with actively translating fractions of the cytoplasm in the silenced cultures. As shown, the global translational profiles of CoCl2-treated cells did not differ significantly among the silenced groups (Fig. 10A). However, the HIF-1α mRNA profile curves shifted leftward, albeit modestly, indicating a reduction in translational activity; pooled populations again showed that the HIF-1α mRNA was more abundant in LMWP fractions and less abundant in HMWP fractions (Fig. 10B). Importantly, nascent HIF-1α translation was significantly reduced in HeLa cells in which HuR or PTB was silenced; the incorporation of 35S-labeled amino acids into de novo-synthesized HIF-1α was ~15% and 22%, respectively, relative to the incorporation measured in the control siRNA-transfected cultures (Fig. 10C). Incorporation of 35S-labeled amino acids into the housekeeping protein GAPDH was not dependent on PTB or HuR levels (see Fig. S3 in the supplemental material).
The reduction in HIF-1α expression levels in cells with silenced HuR or PTB (Fig. (Fig.7,7, ,9,9, and 10C) led to a diminution in the levels of the HIF-1α transcriptional targets VEGF and GLUT1 (Fig. 10D). Together with the HuR overexpression experiments, our findings support a model whereby HuR and PTB positively influence the translation of HIF-1α in response to CoCl2 treatment. These regulatory processes in turn influence HIF-1α-directed gene expression programs.
We report here that HIF-1α translation increases after cells are challenged with the hypoxia mimetic CoCl2, and we show that this effect depends, at least in part, on the association of HIF-1α mRNA with the RNA-binding proteins HuR and PTB. PTB was previously shown to bind to the mouse HIF-1α IRES, thereby increasing HIF-1α translation (54). While we also found IRES activity within the human HIF-1α 5′UTR (Fig. (Fig.3),3), this IRES was not responsible for promoting HIF-1α translation after CoCl2 treatment. Instead, modulation of the levels of RNA-binding proteins HuR and PTB did influence HIF-1α translation after CoCl2 treatment, with overexpression of HuR elevating HIF-1α biosynthesis and silencing of PTB or HuR reducing it. Moreover, HuR-HIF-1α mRNA complexes were diminished in PTB-depleted cells and, likewise, PTB-HIF-1α mRNA complexes were decreased in HuR-depleted cells, further revealing a functional cooperation between PTB and HuR during the CoCl2-induced translation of HIF-1α.
In agreement with earlier literature, neither hypoxia nor CoCl2 seemed to enhance HIF-1α gene transcription or elevate HIF-1α mRNA stability, since HIF-1α mRNA levels remained unchanged after these treatments (Fig. (Fig.1).1). Formal measurement of the HIF-1α mRNA half-life using actinomycin D further confirmed that there were no CoCl2-dependent changes (not shown). Instead, the potent induction of HIF-1α protein expression likely results from a combination of two distinct processes: increased HIF-1α translation and decreased HIF-1α proteolysis. CoCl2 treatment can induce HIF-1α protein levels through its inhibitory effect on prolyl hydroxylases and thereby block the proteasome-dependent degradation of HIF-1α (7, 13, 61). In addition, CoCl2 treatment can also promote HIF-1α translation, as reported here. While bona fide hypoxia is arguably a more physiologic stress, gaining the molecular understanding of these processes and elucidating the function of HIF-1α RNPs using hypoxia would have been hampered by space and methodological limitations imposed by the use of hypoxic chambers/incubators. However, considering the similarities between the mechanisms of action of CoCl2 and hypoxia, we anticipate that hypoxia-regulated HIF-1α expression will also prove to have a significant translational component, likely modulated by PTB and HuR.
The mechanisms whereby CoCl2 treatment increased HuR binding to HIF-1α mRNA are unclear, but neither the whole-cell abundance of HuR nor its cytoplasmic levels (Fig. (Fig.8A8A and data not shown) were elevated in response to CoCl2. These findings were in contrast to what we and others have observed regarding this shuttling protein, as its cytoplasmic concentration increases markedly following exposure to numerous other stress agents, likely a result of HuR increased cytoplasmic export or reduced nuclear import. In addition, a variety of damaging agents induced the aggregation of cytoplasmic, HuR-containing stress granules (reviewed in reference 28). By contrast, CoCl2 treatment neither elevated the cytoplasmic HuR levels nor induced HuR aggregation into stress granules (see Fig. S4 in the supplemental material). These observations support the notion that HuR binding to the HIF-1α mRNA is influenced by posttranslational processes that occur in the absence of net changes in nucleocytoplasmic distribution. Several regulatory mechanisms have been described which influence, directly or indirectly, the abundance of HuR-containing RNP complexes. For example, the translational suppression imposed by microRNA miR-122 upon the CAT-1 mRNA was relieved following displacement by HuR in response to stress conditions (2), the activity of the checkpoint kinase Chk2 influenced HuR binding to the SIRT1 mRNA after oxidative stress (1), and HuR methylation by coactivator-associated arginine methyltransferase 1 affected HuR binding to tumor necrosis factor alpha mRNA in lipopolysaccharide-treated cells (35). In addition, PI3K and the mitogen-activated protein kinase p38 influenced HuR function (44, 58), while the activity of mitogen-activated protein kinase extracellular signal-regulated kinase (ERK) promoted HuR binding to target p21 mRNA in cells treated with prostaglandin A2, as ERK inhibition diminished [HuR-p21 mRNA] complexes but not the cytoplasmic abundance of HuR (66). The latter effect might be implicated in the CoCl2-triggered increase in HuR binding to HIF-1α mRNA, since CoCl2 treatment activates ERK (65). A more in-depth analysis of the mechanisms responsible for the CoCl2-induced HuR-HIF-1α mRNA complexes is warranted.
Similar questions arise regarding the influence of CoCl2 upon PTB binding to HIF-1α mRNA. What are the cellular effectors of the CoCl2-enhanced formation of PTB-HIF-1α mRNA complexes? The signaling pathways that are activated in response to CoCl2 treatment have not been fully delineated, but exposure to cobalt activates the production of reactive oxygen species and induces PI3K activity linked to increased HIF-1α expression; the latter events are significant because PI3K/mTOR/p70 S6 kinase signaling has been linked to increased HIF-1α translation (6, 10). However, whether CoCl2 regulates the function of PTB via any of the reported kinase pathways that control HIF-1α translation, dependently or independently of mTOR (3, 22, 47, 49), remains to be investigated. It is also possible that cobalt itself directly elicits structural changes in the HIF-1α mRNA, as reported for other metal ions, such as iron and magnesium, which can affect mRNA tertiary structure (17, 27, 63). Thus, the possibility remains that cobalt directly alters the accessibility of RNA-binding proteins to the HIF-1α mRNA and thereby facilitates translation.
Although HuR was found to bind the HIF-1α 5′UTR and PTB was found to bind the HIF-1α 3′UTR, the specific region(s) of association of HuR and PTB on the human HIF-1α mRNA also awaits more detailed study. It was surprising to discover that the HIF-1α 3′UTR did not associate with HuR (Fig. (Fig.4),4), despite the predicted existence of three HuR motif hits (38) and a previous report of HuR-HIF-1α 3′UTR complexes (57). HuR was shown to enhance the translation of numerous target mRNAs, including those that encode prothymosin α, p53, cytochrome c, GLUT1, and CAT-1 (2, 16, 26, 33, 42). Paradoxically, HuR was also shown to inhibit the translation of other target mRNAs, such as those encoding cytokines, p27, and IGF-IR (25, 31, 43). In the cases in which these effects were characterized in molecular detail, HuR was found to associate via the 5′UTR of target mRNAs which had IRESs (31, 43). In this regard, given our findings described here that HuR bound the IRES-containing 5′UTR of HIF-1α, we anticipated that HuR would trigger the inhibition, not the promotion, of HIF-1α translation. Finally, the functional interaction between HuR and PTB upon the translation of HIF-1α is reminiscent of those described for Apaf-1 and the human rhinovirus RNA, whose translation is potentiated through binding of PTB to the RNA-binding protein Unr (upstream of N-ras) (24, 45).
In sum, these studies have identified HuR and PTB as two RNA-binding proteins that function jointly to stimulate HIF-1α translation in response to CoCl2 treatment. The translational control of HIF-1α expression complements the previously reported influence of CoCl2 on HIF-1α protein stability (7, 13, 61). The concomitant increase in HIF-1α biosynthesis and inhibition of HIF-1α degradation underscore the need for rapid and efficient increases in HIF-1α abundance following environmental changes. The expression of a growing number of additional genes has been found to be regulated simultaneously at multiple levels, transcriptional, posttranscriptional, translational, and posttranslational. Such extensive regulatory networks afford the appropriate degree of redundancy, checks, and safeguards to ensure that critical proteins such as HIF-1α are expressed in the necessary abundance, time, and location to maintain cellular homeostasis.
We thank S. D. Baird and B. Frank for their assistance with this study.
This research was supported by the Intramural Research Program of the National Institute on Aging, National Institutes of Health.
Published ahead of print on 29 October 2007.
†Supplemental material for this article may be found at http://mcb.asm.org/.