|Home | About | Journals | Submit | Contact Us | Français|
Hypoxia-inducible factors (HIFs) are transcriptional regulators that mediate the cellular response to low oxygen. Although HIF-1 is usually considered as the principal mediator of hypoxic adaptation, several tissues and different cell types express both HIF-1 and HIF-2 isoforms under hypoxia or when treated with hypoxia mimetic chemicals such as cobalt. However, the similarities or differences between HIF-1 and HIF-2, in terms of their tissue- and inducer-specific activation and function, are not adequately characterized. To address this issue, we investigated the effects of true hypoxia and hypoxia mimetics on HIF-1 and HIF-2 induction and specific gene transcriptional activity in two hepatic cancer cell lines, Huh7 and HepG2. Both hypoxia and cobalt caused rapid induction of both HIF-1α and HIF-2α proteins. Hypoxia induced erythropoietin (EPO) expression and secretion in a HIF-2-dependent way. Surprisingly, however, EPO expression was not induced when cells were treated with cobalt. In agreement, both HIF-1- and HIF-2-dependent promoters (of PGK and SOD2 genes, respectively) were activated by hypoxia while cobalt only activated the HIF-1-dependent PGK promoter. Unlike cobalt, other hypoxia mimetics such as DFO and DMOG activated both types of promoters. Furthermore, cobalt impaired the hypoxic stimulation of HIF-2, but not HIF-1, activity and cobalt-induced HIF-2α interacted poorly with USF-2, a HIF-2-specific co-activator. These data show that, despite similar induction of HIF-1α and HIF-2α protein expression, HIF-1 and HIF-2 specific gene activating functions respond differently to different stimuli and suggest the operation of oxygen-independent and gene- or tissue-specific regulatory mechanisms involving additional transcription factors or co-activators.
Hypoxia-inducible factors (HIFs) play a central role in hepatocellular carcinoma (HCC) (Mylonis and Simos, 2012). HIFs are frequently up-regulated in HCC and seem to control tumor progression and sensitivity to radiation therapy (Nath and Szabo, 2012). HIFs are heterodimers of the protein subunits HIF-α, which are induced by hypoxia and HIF-β or aryl hydrocarbon receptor nuclear translocator (ARNT), which is constitutively expressed (Keith et al., 2012). The HIF-α subunits are rapidly degraded in normoxia by a process involving prolyl hydroxylation, interaction with the von Hippel-Lindau protein (pVHL) E3 ubiquitin ligase complex and proteasomal degradation. Hydroxylation is catalyzed by a family of Fe(II) and 2-oxoglutarate-dependent prolyl hydroxylases (PHDs) whose absolute requirement for molecular oxygen confers sensitivity to hypoxia, under which HIF-α members rapidly accumulate, translocate inside the nucleus, heterodimerize with ARNT and bind to hypoxia-response elements (HREs) DNA elements in the promoters or enhancers of their target genes. The activity of PHDs and subsequent stabilization of HIF-1α is also affected by chemical agents (“hypoxia mimetics”) such as the iron chelator desferrioxamine (DFO) (Wang and Semenza, 1993), flavonoids such as quercetin (Triantafyllou et al., 2007), 2-oxoglutarate-dependent oxygenase inhibitors such as DMOG (Elvidge et al., 2006) and transition metals such as cobalt (Kaelin and Ratcliffe, 2008). Treatment with these agents (“chemical hypoxia”) is often used to simulate hypoxic conditions and induce HIF-1. However, the action of certain of these agents (e.g. cobalt and flavonoids) may be mediated by signaling pathways not necessarily shared by the “true” hypoxic response and may cause, therefore, different and oxygen-independent biological effects (Chachami et al., 2004; Triantafyllou et al., 2006, 2008).
HIF-1α was the first isoform of HIF-α to be identified by its capacity to bind to the hypoxia-responsive element (HRE) present in the erythropoietin (EPO) enhancer (Wang et al., 1995). HIF-1α is ubiquitously expressed under hypoxia and is responsible for the regulation of a wide range of cellular adaptation responses and more preferentially metabolic enzymes (Majmundar et al., 2010; Mylonis et al., 2012). A second isoform (HIF-2α), is encoded by the EPAS1 gene, its expression is restricted to specific cell types, including hepatocytes, it appears to be more involved in angiogenesis and erythropoiesis and its regulation is considerably less investigated (Keith et al., 2012). Like HIF-1α, HIF-2α is degraded under normal conditions via the same PHD-VHL-proteasome-dependent system and is stabilized under hypoxia due to PHD inhibition. However, it is not yet known whether oxygen-independent mechanisms that regulate HIF-1 activity, such as for example nuclear transport and phosphorylation (Mylonis et al., 2008; Kalousi et al., 2010), also apply to HIF-2. In the same line, the effects of “hypoxia mimetics” on different cell types is usually studied by monitoring HIF-1α expression and whether HIF-2α activity is also affected remains relatively unknown. This has important biological relevance given that HIF-α stabilizers are used both experimentally and clinically as erythropoiesis-stimulating agents.
HIF-1α and HIF-2α share certain overlapping functions by regulating common hypoxia-inducible genes. However, it has recently become clear that, through independent regulation of distinct target genes or unique protein-protein interactions, HIF-1α and HIF-2α can also mediate divergent functions when expressed in the same cell type and especially in the context of cancer (Keith et al., 2012). The predominant role of HIF-2 in erythropoiesis has been established by studies in HIF-2α-deficient (knock-out) mice documenting HIF-2α as the main regulator of hepatic EPO production and essential for the maintenance of systemic EPO and iron homeostasis (Scortegagna et al., 2005; Rankin et al., 2007; Kapitsinou et al., 2010). Other prominent HIF-2-specific targets are genes for major antioxidant enzymes such as superoxide dismutase (SOD2) (Scortegagna et al., 2003). However, these animal studies have not determined the exact degree to which individual HIF-α subunits contribute to EPO production in cancer cells. Many subtle differences on the regulation of the expression and transcriptional activity between HIF-1α and HIF-2α need to be elucidated, especially in HCC as inhibition of HIFs may have an important role in targeted cancer therapy (Nath and Szabo, 2012).
In this report, we investigate the expression and transcriptional activity of HIF-1α and HIF-2α as well as expression and secretion of EPO in response to hypoxia and the “hypoxia mimetic” cobalt in hepatic cancer cell lines expressing both HIF-α isoforms. Our findings show that cobalt has opposing effects on HIF-2α expression and activity, rendering it ineffective in EPO gene activation. This specific effect of cobalt on HIF-2 signaling was accompanied by reduced interaction of the HIF-2α with USF2. This suggests that HIF-2α, but not HIF-1α, requires the assistance of a cobalt-sensitive factor in order to activate transcription of its target genes in liver cancer cells.
pGL3-SOD2 promoter and pGL3-PGK promoter were kindly provided by Joseph A. Garcia (Department of Medicine, University of Texas) and M. Celeste Simon (Abramson Family Cancer Research Institute, University of Pennsylvania) respectively (Scortegagna et al., 2003; Hu et al., 2007). Plasmid pEGFP-HIF-2α was constructed by inserting the full length HIF-2α cDNA into the BamHI position of the multicloning site of the pEGFP-C1 plasmid (Clontech). pcDNAHIF-2α, kindly provided from Dr. S. L. McKnight (Department of Biochemistry, University of Texas) (Tian et al., 1997). pEGFP-HIF-1α plasmid was previously described (Mylonis et al., 2006). Plasmid pCDNA-Flag-USF2 full-length was previously described (Pawlus et al., 2012).
Huh7 and HepG2 cells were cultured in DMEM (Gibco) containing 10% FCS and 100 U/ml penicillin–streptomycin (Gibco). All cultures were maintained at 37°C in a humidified atmosphere containing 5% CO2. When required, cells were treated for 2–48 h with 150 μM cobalt chloride (CoCl2). The concentration of CoCl2 (150 μM) was determined in preliminary dose–response experiments (using 50–200 μM) to produce maximal induction of HIF-α at 4–48 h, without being toxic to cells (results not shown). For hypoxic treatment, cells were exposed for 2–48 h to 1% O2, 95% N2 and 5% CO2 in an IN VIVO2 200 hypoxia workstation (RUSKINN Life Sciences). All other incubation protocols were as stated in figure legends.
Antibodies used for western blotting were: affinity purified rabbit polyclonal (pAbs) antibody against HIF-1α (Lyberopoulou et al., 2007), anti-HIF-2α pAb (Abcam and Novus Biologicals for Fig. 5), anti-Flag pAb (Sigma), anti-USF2 pAb (Santa Cruz) and anti-actin monoclonal antibody (Millipore). Analysis by immunoblotting and immunoprecipitation was carried out as previously described (Lyberopoulou et al., 2007; Mylonis et al., 2008). All experiments were performed in triplicate and representative results are shown.
To examine the transcriptional activity of HIF-1 and HIF-2, cells were co-transfected with the firefly luciferase reporter plasmid pGL3-PGK or pGL3-SOD2 and the Renilla luciferase expressing plasmid PCIRenilla, under the control of an autologous promoter (pGL3) as previously described (Lyberopoulou et al., 2007; Mylonis et al., 2008). Luciferase activity was measured using the Dual-luciferase assay system (Promega, WI, USA) with a luminometer (TD20/20, Turner Designs).
Huh7 cells were incubated in serum-free DMEM for 4 h with siRNA (10 nM) against HIF-1α or HIF-2α (Qiagen) or in the presence of LipofectamineTM2000 (Invitrogen). All Stars siRNA (Qiagen) was used as negative control.
Total RNA from Huh7 cells was isolated using the Trizol reagent (Invitrogen) and cDNA was synthesized with the High Capacity Reverse Transcription kit (Applied Biosystems). Real-time PCR was performed with KAPA SYBR® FAST One-Step qRT-PCR Universal (Kapa Biosystems) in a MiniOpticon instrument (Bio-Rad). The mRNAs encoding EPO, PGK and actin were amplified using primers listed in Supplementary Table 1. Each sample was assayed in triplicate for both target and internal control. Relative quantitative gene expression was calculated using the ΔΔCT method and presented as relative units.
Medium was collected from 105 cells in 6-well culture plates and centrifuged at 800 r.p.m. for 4 min at 4°C to remove cellular debris and then stored at −70°C. EPO in the medium was measured by using the EPO-TracTM RIA kit (Diasorin) according to the manufacturer’s instruction.
The Graph Pad Instat Statistical package for Windows was used. Data are expressed as ±mean standard deviation (SD). The oneway analysis of variance (ANOVA) with the Bonferroni post-test was used for the comparison of data and the statistical significance limit was set at p < 0.05.
The expression of HIF-2α in Huh7 hepatoma cells under different “hypoxic” conditions, such as true hypoxia (1% O2) or in the presence 150 μM CoCl2 (chemical hypoxia) was compared to that of HIF-1α by western blotting analysis shown in Fig. 1A. When cells were exposed to 1% O2, HIF-2α as well as HIF-1α protein levels increased rapidly and remained elevated for up to 16 h. Then, both HIF-2α and HIF-1α expression gradually declined and returned to control levels by 48 h. When cells were treated with CoCl2, HIF-2α and HIF-1α protein levels also increased with similar kinetics. However unlike hypoxia, they both remained high up to 48 h of treatment. Therefore, expression of both HIF-2α and HIF-1α is induced by either hypoxia or CoCl2 but it is differentially affected by prolonged exposure to low oxygen or cobalt, indicating dissimilar induction mechanisms. It has been previously suggested that HIF-1α protein induction in neuroblastoma cells may be more sensitive to oxygen decline (Holmquist-Mengelbier et al., 2006). To test this in Huh7 cells, HIF-α protein levels were analyzed in cells exposed to different oxygen concentrations (1–10% O2). HIF-2α and HIF-1α proteins were barely detectable at 10% O2 but the expression of both was induced at 6%O2 and increased with similar sensitivity at less oxygen (Fig. 2B) indicating no major differences in the hypoxic stabilization of HIF-1α and HIF-2α in Huh7 cells.
It is generally assumed that induction of HIF-2α is caused by protein stabilization through inhibition of proteolysis. However, recent data indicate that control of HIF-2α expression can also take place at the levels of transcription, translation and protein stability (Keith et al., 2012). To investigate the mechanisms behind the induction of HIF-2α accumulation in Huh7 cells, we performed experiments with the inhibitor of transcription actinomycin D or the translation inhibitor cycloheximide. Our data show that induction of HIF-2α by low oxygen or CoCl2 was not dependent on new RNA synthesis but required active protein synthesis (Supplementary Fig. S1A and B). The involvement of translation was further confirmed using LY294002, an inhibitor of the PI-3 K and its downstream AKT-mTOR pathway that has been shown before to control HIF-1α protein synthesis (Chachami et al., 2004, Triantafyllou et al., 2006). Treatment with LY98059 indeed reduced the induction of HIF-2α by both low oxygen and cobalt (Supplementary Fig. S1C) suggesting the involvement the PI-3K pathway under both conditions. On the other hand, treatment with PD98059, an inhibitor of the MAPK pathway leading to activation of ERK1/2, attenuated the expression of HIF-2α only under cobalt treatment, suggesting the participation of ERK1/2 in the cobalt but not hypoxia-triggered induction (Supplementary Fig. 1SC). As the PI3K/Akt-dependent pathway can be activated by the generation of ROS (Kietzmann and Gorlach, 2005), known radical scavengers, such as NAC, glutathione and Trolox were used to treat cells simultaneously exposed to either hypoxia or cobalt. In all cases HIF-2α expression was reduced (Supplementary Fig. S1D) suggesting the involvement of ROS in the induction of HIF-2α in Huh7 cells irrespective of the type of “hypoxic” stimulus.
In order to define the relative contributions of HIF-1 and HIF-2 in the regulation of EPO or PGK expression by hepatoma cells, we transiently knocked-down HIF-1α and HIF-2α expression using siRNA-mediated silencing in Huh7 cells grown under hypoxia (Fig. 2A) Hypoxia triggered a robust increase in EPO mRNA in control cells and in cells with silenced HIF-1α while up-regulation of PGK mRNA by hypoxia was observed also in control cells and in cells with silenced HIF-2α (Fig. 2B and D). In contrast, knock-down of HIF-1α decreased hypoxic PGK mRNA levels, whereas knock-down of HIF-2α decreased hypoxic EPO mRNA levels (Fig. 2B and D). Quantification of hypoxia-induced EPO secretion in the culture medium by RIA analysis gave similar results as determination of EPO mRNA (Fig. 3A) indicating that transcriptional activation of the EPO gene in Huh7 cells and subsequent stimulation of EPO production and secretion by low oxygen is predominantly, if not only, mediated by HIF-2α. On the other hand, the transcriptional activation of the PGK gene in Huh7 cells by hypoxia is mediated only by HIF-1α.
Surprisingly, treatment of the cells with CoCl2, that unregulated HIF-1α and HIF-2α expression (Fig. 1A), did not increase in EPO mRNA levels, which remained similar to that under normoxic conditions (Fig. 2C). On the other hand, CoCl2 induced upregulation of PGK mRNA levels in Huh7 cells to similar extent as hypoxia (Fig. 2E). To further compare the effects of low oxygen and cobalt on EPO production, we measured EPO secretion during a time course of 8–24 h. Only true hypoxia increased EPO secretion from Huh7 cells in a time-dependent manner, whereas the presence of CoCl2 produced no effect (Fig. 3B). We, therefore, conclude that CoCl2, unlike hypoxia, cannot activate HIF-2-dependent EPO mRNA and protein production in Huh7 cells suggesting that HIF-2α induced by cobalt treatment is somehow unable to activate transcription of the EPO gene.
The inability of cobalt to stimulate production of EPO indicates that HIF-2α that is induced by this treatment lacks transcriptional activity, preconditions for which are translocation of HIF-2α inside the nucleus and binding to HRE-containing DNA elements. To test whether induction of HIF-2α by CoCl2 is followed by accumulation of the protein inside the nucleus, Huh7 cells exposed to hypoxia or CoCl2 were analyzed by immunofluorescence microscopy. A strong nuclear signal of HIF-2α was readily visible under both hypoxia and CoCl2 treatment (Supplementary Fig. S2A). These data were confirmed by biochemical fractionation, which showed the exclusive presence of HIF-2α in the nuclear fraction under both conditions (Supplementary Fig. S2B). Therefore, cobalt treatment, like hypoxia, results in both induction and nuclear accumulation of HIF-2α.
To test whether cobalt-induced HIF-2α as well as HIF-1α are active in transcription, we analyzed the effect of their induction on known HRE-containing promoters such as those of the SOD2 and PGK genes, which have been previously suggested to represent specific targets of HIF-2α and HIF-1α, respectively (Scortegagna et al., 2003; Hu et al., 2007). To first test target specificity, SOD2 or PGK promoter reporter plasmids were co-transfected with HIF-1α or HIF-2α expression plasmids into Huh7 cells. Over-expression of HIF-2α, but not HIF-1α, activated the SOD2 promoter, whereas over-expression HIF-1α stimulated the PGK promoter to a much larger extent than HIF-2α (Fig. 4A). In the inverse experiment, knocking down of HIF-2α, but not HIF-1α, impaired the hypoxic activation of the SOD2 promoter, whereas hypoxic stimulation of the PGK promoter was repressed by knocking down HIF-1α but not HIF-2α (Fig. 4B). Therefore, the SOD2 or PGK promoters are reliable indicators of HIF-2 and HIF-1 unique activities, respectively, in Huh7 cells. As expected, treatment of Huh7 cells harboring the aforementioned reporter constructs with hypoxia activated both SOD2 and PGK promoter-driven transcription (Fig. 4C). In contrast, treatment with cobalt activated the PGK promoter to the same extent as hypoxia but left the SOD2 promoter completely inactive (Fig. 4C), suggesting that the presence of cobalt does not generally affect HRE-dependent transcription but rather specifically inhibits the activity of HIF-2α.
To test whether this effect is specific to CoCl2 treatment, or whether other hypoxia mimetics such as DFO or DMOG produce the same result, we analyzed their effect on the activation of the SOD2 or PGK promoter in Huh7 cells. Both DFO and DMOG activated both SOD2 and PGK promoters (Fig. 4D), suggesting that inactivity of HIF-2 is a phenomenon specific to CoCl2 treatment of these cells.
As inhibition of HIF-2α activity, when cells are treated with CoCl2, cannot be attributed to its altered expression, stability or nuclear localization, the affected step in HIF-2 activation must lie downstream of its nuclear translocation and may involve its ability to interact with HIF-2-specific co-activators such as USF2 (Pawlus et al., 2012). This is also suggested by the fact that CoCl2 affects negatively only HIF-2 but not HIF-1.
To investigate this possibility, we first examined the effect of hypoxia or CoCl2 on the expression of USF2 in Huh7 cells. USF2 was expressed at similar levels under all conditions tested (Fig. 5A). We then tested the interaction of USF2 with HIF-2α. Huh7 cells expressing Flag-tagged USF2 were incubated under normoxia or hypoxia or in the presence of CoCl2, lysed and subjected to immunoprecipitation using an anti-Flag antibody. Analysis of the immunoprecipitates revealed that association of HIF-2α with USF2 was reduced in cobalt-treated cells (by approximately 40%) in comparison to hypoxia-treated cells despite similar levels of HIF-2α protein expression (Fig. 5B). This suggests that CoCl2 impairs HIF-2 activity, at least partly, by weakening the HIF-2α interaction with USF2.
To further examine the effect of CoCl2 on HIF-2 activity, we tested combined treatment of Huh7 cells with both CoCl2 and hypoxia. CoCl2 reduced the hypoxic stimulation of EPO mRNA synthesis whereas it augmented the hypoxic induction of PGK mRNA synthesis (Fig. 5C). These results suggest that CoCl2 actively impairs the hypoxic stimulation of HIF-2 transcriptional activity probably by interfering with the formation of a competent USF2-containing transcription complex.
To investigate whether this negative, in terms of HIF-2 activity, response to CoCl2 is cell-line specific or a general characteristic of liver cancer cells, we tested a different hepatic cell line, HepG2. Firstly, both HIF-1α and HIF-2α proteins were shown to be expressed in this cell line under hypoxia or in the presence of CoCl2 (Fig. 6A). Using HIF-1α or HIF-2α silencing, we also confirmed that SOD2 is a HIF-2α specific gene and PGK a HIF-1α specific gene (Fig. 6B). Exposure to hypoxia activated both SOD2 and PGK promoters but addition of CoCl2 only stimulated the PGK promoter (Fig. 6C). Finally hypoxia increased EPO secretion from HepG2 cells in a time dependent manner but incubation with CoCl2 had no effect (Fig. 6D), suggesting that the inhibitory effect of CoCl2 on HIF-2 transcriptional activity characterizes hepatic derived cancer cell lines.
Many human cancer cell types exhibit increased expression of either one or both of the main HIF-α isoforms, HIF-1α and HIF-2α, and in many cases this overexpression is associated with poor prognosis (Keith et al., 2012). This has made HIFs major targets of anti-cancer therapy and many HIF inhibitors are currently under clinical evaluation (Semenza, 2010). However, there are certain cases of tumors in which one of the two HIF-α isoforms, but not the other, may actually confer favorable prognosis (Lidgren et al., 2005; Noguera et al., 2009). This adds to the increasingly important and acceptable notion that HIF-1α and HIF-2α, despite their similarities and possibly overlapping functions, can be subject to differential regulation so as to mediate unique and even opposing cellular responses often in a cell-type specific context. It has, therefore, become important to individually study each HIF-α iso-form, especially in cell types that express both. To our knowledge, this report is the first systematic study of parallel HIF-1α and HIF-2α induction and activation in liver cancer cells. Our analysis has revealed similarities and differences both between HIF-1α and HIF-2α and between the types of hypoxic stimulus, i.e. low oxygen (true hypoxia) and cobalt (chemical or mimetic hypoxia), which are physiologically important at least in the context of hepatic EPO production.
It is widely reported in the literature that cancer cell HIF-1α and HIF-2α exhibit diverse sensitivities to different oxygen concentrations and to acute (short-term) or prolonged hypoxia (Keith et al., 2012). This stems from studies with lung adenocarcinoma, neuroblastoma and glioma cells in which HIF-1α was induced at lower oxygen concentrations than HIF-2α and HIF-1α expression was transient while HIF-2α long-lasting (Uchida et al., 2004; Nilsson et al., 2005; Holmquist-Mengelbier et al., 2006; Dioum et al., 2009). Our results clearly show that this is not a universal but rather a cell-type specific rule as in liver cancer cells we detected no isoform-specific differences when HIF-2α and HIF-1α expression was examined over a range of different oxygen concentrations and hypoxic exposure times. Expression of both proteins became detectable at 6% oxygen and continued to increase with declining oxygen concentrations reaching maximum at 2%. Concerning kinetics, both proteins appeared 2 h after onset of hypoxia, reached peak levels at 16 h and gradually declined to basal levels after 48 h. The down-regulation of both HIF-α isoforms after prolonged hypoxic incubation probably represents a homeostatic adaptation to the continuing reduced availability of oxygen and could be mediated by one of several mechanisms such as HIF-driven up-regulation of prolyl hydroxylases (Berra et al., 2003; Aprelikova et al., 2004; Nakayama et al., 2004), induction of antisense HIF-α (Rossignol et al., 2002) or production of dominant negative HIF-α isoforms (Lee et al., 2004). Significantly, this negative feedback loop is not established when HIF-α is induced by cobalt since expression of both isoforms remains high even after 48 h of treatment. The same difference between hypoxia and cobalt was also observed for HIF-1α in HeLa cells (Triantafyllou et al., 2006) and pinpoints the “non-physiological” effects of chemical hypoxia mimetics.
It is clear that stabilization of HIF-α subunits by inhibition of PHD activity is a key and universal mechanism of HIF induction by low oxygen. However, additional, often cell-type specific, regulatory mechanisms have been identified that fine-tune HIF activity to cues from the extracellular environment and the particular cellular needs (Keith et al., 2012). As we have recently shown for HIF-1α in prostate cancer cells (Befani et al., 2012), hypoxic induction of HIF-2α in hepatoma cells involves the PI-3K signaling pathway, activation of which is most likely required to sustain protein synthesis that also seems to be a prerequisite for HIF-α elevated expression. Cobalt treatment also requires the PI-3K signaling pathway but in addition an active MAPK pathway to induce HIF-2α. We have made similar observations for HIF-1a concerning both the PI-3K (Chachami et al., 2004; Triantafyllou et al., 2006) and the MAPK (Triantafyllou et al., 2006) pathways suggesting that cobalt induces expression of the two HIF-α isoforms through similar mechanisms, which are overlapping but distinct from those utilized by cells exposed to low oxygen.
However, the most striking difference between cobalt and low oxygen that this report reveals is their effect on EPO regulation as EPO gene induction and protein secretion were not observed in liver cancer cells treated by, the widely used as “hypoxia mimetic”, cobalt. This has physiological importance, as EPO is a hematopoietic growth factor that, by regulating production of red blood cells, controls a key determinant of oxygen homeostasis, blood oxygen-carrying capacity. During fetal development, EPO is mainly produced in liver; although its primary site of production then changes to the kidney during late gestation, it continues to be produced to a lesser extent in liver. Hepatic EPO contributes about 10% of the plasma EPO in adults. Hypoxia is the primary physiological stimulus for EPO production, which, depending on the hypoxic condition increases serum EPO levels up to several hundred-fold (Haase, 2010; Jelkmann, 2011).
As already mentioned, animal studies have shown than the main regulator of hepatic EPO production is HIF-2α and not HIF-1α (Scortegagna et al., 2005; Rankin et al., 2007; Kapitsinou et al., 2010). We confirmed the HIF-2-dependency of human EPO gene hypoxic activation at the cellular level in Huh7 and HepG2 cells in agreement with previous studies using other cell types (Warnecke et al., 2004; Chavez et al., 2006). Therefore, the inability of cobalt to induce EPO production suggested a defect in HIF-2-dependent transcriptional activation. Indeed, we show that this was the case in both cell lines by using specific HIF targets: cobalt activated the HIF-1-dependent PGK promoter but not the HIF-2-dependent SOD2 promoter, while both promoters could be stimulated by low oxygen.
This came as a surprise, as cobalt is extensively used as a hypoxia mimetic to study HIF responsiveness. Previous reports (Vengellur et al., 2005; Borcar et al., 2013) had already indicated that cobalt can elicit unique physiological responses, influencing transcription of distinct set of genes compared to hypoxia. However, these studies did not examine the role of HIF isoforms in the cobalt-specific effects. Our study is the first to compare both HIF-1 and HIF-2 functions between hypoxia and cobalt and clearly shows that, in contrast to HIF-1α, HIF-2α produced in cells treated with cobalt is transcriptionally inactive. This effect is specific for cobalt and not other hypoxia mimetic chemicals such as DFO and DMOG, which apparently share with low oxygen the common ability to inhibit the activity of PHDs (Wang and Semenza, 1993; Elvidge et al., 2006).
This unanticipated observation, i.e. presence of HIF-2α protein but lack of HIF-2 activity, is not without precedent, albeit in non-cancer cells. Two previous reports have demonstrated the expression of inactive HIF-2α in mouse embryo fibroblasts (MEFs) and embryonic stem (mES) cells (Park et al., 2003; Aprelikova et al., 2006). In the former case, the lack of HIF-2 activity was attributed to its HIF-2α cytoplasmic mislocalization (Park et al., 2003). This does not appear to be the case in cobalt-treated hepatic cancer cells, as HIF-2α accumulated inside the nucleus but still remained inactive. In mES cells, inhibition of HIF-2 function occurred at the level of transcription cofactor recruitment (Aprelikova et al., 2006). A very recent report indeed identified USF2 as a factor required for hypoxic induction of HIF-2 target genes in several cell lines including mES and Hep3B cells (Pawlus et al., 2012). We tested the interaction of USF2 with HIF-2α protein in cells treated with CoCl2 and we could demonstrate weaker association of HIF-2α with USF2 as compared with cells treated with hypoxia. This finding explains, at least partly, the low transcriptional activity of cobalt-induced HIF-2α. Cobalt not only induces an inactive HIF-2α but also inhibits the activity of hypoxia-induced HIF-2 suggesting that cobalt plays an energetic negative role and does not just fail to induce a step required for HIF-2 activation. Apart of the USF2/HIF-2α interaction, cobalt may also interfere with additional cofactors involved in HIF-2 target gene activation. There is a growing list of transcription factors the cooperation of which is required for maximal and cell-type-specific upregulation of HIF-2-dependent transcription. HIF-2α interacts with multiple ETS family members such as ETS-1, ETS-2, GABP and ELK that may confer target specific gene selection in numerous cell types (Elvert et al., 2003; Aprelikova et al., 2006; Hu et al., 2007). The NF-kB essential modulator (NEMO) also physically interacts with HIF-2α, but not with HIF-1α, and enhances HIF-2α transcriptional activity (Bracken et al., 2005). Finally, the stress-responsive deacetylase Sirt1 selectively augments HIF-2 signaling and EPO gene expression by deacetylating HIF-2α during hypoxia or in the presence of other environmental stresses (Dioum et al., 2009). It is, therefore, conceivable that cobalt down-regulates the transactivation of HIF-2-specific target genes by interfering with any of these factors that may be expressed in liver cancer cells.
The lesson learnt from our study is that the widely used hypoxia mimetic cobalt is not at all equivalent to true, low oxygen hypoxia: despite the fact that both induce expression of HIF-α isoforms, their biological end-effects may be completely different. Second, the mere over-expression of a HIF-α subunit does not at all mean that a full-blown response to hypoxia has been mounted: one has to also confirm activation of target genes and corresponding biological phenotype. Finally, behavior of one HIF-α isoform tells nothing about the behavior of the second and, given their distinct functions, both should be systematically monitored in terms of expression as well as activity. These should be taken seriously into consideration by research and clinical oncologists, especially as expression of HIF-α isoforms in tumor specimens is often linked to clinical outcome and HIF-inhibitors, with poorly characterized isoform selectivity, are increasingly more often tested as targeted therapeutics against different types of cancer, including HCC, or represent promising future interventions in various other hepatic diseases (Nath and Szabo, 2012).
We thank Drs. J. A. Garcia (University of Texas), M. C. Simon (University of Pennsylvania) and M. U. Muckenthaler (University of Heidelberg) for providing plasmids.
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.biocel.2013.07.025.