Eukaryotic cells sense oxygen and adapt to hypoxia by regulating a number of genes. Hypoxia‐inducible factor 1 (HIF‐1) is the ‘master’ in this pleiotypic response. HIF‐1 comprises two members of the basic helix–loop–helix transcription factor family, HIF‐1α and HIF‐1β. The HIF‐1α protein is subject to drastic O2‐dependent proteasomal control. However, the signalling components regulating the ‘switch’ for ‘escaping’ proteasomal degradation under hypoxia are still largely unknown. The rapid nuclear translocation of HIF‐1α could represent an efficient way to escape from this degradation. We therefore asked, where in the cell is HIF‐1α degraded? To address this question, we trapped HIF‐1α either in the cytoplasm, by fusing HIF‐1α to the cytoplasmic domain of the Na+‐H+ exchanger (NHE‐1), or in the nucleus, by treatment with leptomycin B. Surprisingly, we found that HIF‐1α is stabilized by hypoxia and undergoes O2‐dependent proteasomal degradation with an identical half‐life (5–8 min) in both cellular compartments. Therefore, HIF‐1α entry into the nucleus is not, as proposed, a key event that controls its stability. This result markedly contrasts with the mechanism that controls p53 degradation via MDM2.
Hypoxia‐inducible factor 1 (HIF‐1) plays a central role in oxygen homeostasis by inducing the expression of a broad range of genes including VEGF, VEGFR1, and almost every gene of the glycolytic pathway in response to a decrease in pO2 (Guillemin and Krasnow, 1997; Semenza, 1998). HIF‐1 is a heterodimer composed of the constitutively expressed HIF‐1β, and the rate limiting factor HIF‐1α (Wang and Semenza, 1995). HIF‐1β is the already characterized aryl hydrocarbon receptor nuclear translocator (ARNT), previously shown to heterodimerize with the aryl hydrocarbon receptor (AHR) (Hoffman et al., 1991). In contrast, HIF‐1α specifically mediates hypoxic responses. HIF‐1α is a short‐lived protein, the levels of which are tightly regulated by oxygen concentration via the ubiquitin–proteasome system (Salceda and Caro, 1997; Huang et al., 1998; Kallio et al., 1999). In normoxia, HIF‐1α is maintained at low and often undetectable levels, whereas hypoxia rapidly and strongly increases the amounts of HIF‐1α by relaxing its proteasome‐dependent degradation. It has been demonstrated previously that the product of the von Hippel‐Lindau tumour suppressor gene (pVHL) directly interacts with HIF‐1α as a component of an E3 ubiquitin–protein ligase complex responsible for the degradation of HIF‐1α (Maxwell et al., 1999; Cockman et al., 2000; Ohh et al., 2000; Tanimoto et al., 2000). pVHL/HIF‐1α association has been reported to take place even in hypoxic conditions. This finding however, unexpected for a ligase/substrate complex, was in fact invalidated by two recent reports demonstrating that pVHL/HIF‐1α interaction is indeed regulated by the intracellular O2 levels. The key enzyme controlling this O2‐dependent step is a specific HIF‐1α‐proline hydroxylase (Ivan et al., 2001; Jaakkola et al., 2001). In addition to stabilizing HIF‐1α, hypoxia also induces the nuclear translocation of the protein (Kallio et al., 1998). This relocalization of HIF‐1α could be a way by which the protein escapes from proteasomal degradation. In fact, HIF‐1α nuclear localization has been recently proposed as a regulatory step involved in its stabilization (Tanimoto et al., 2000). To test this hypothesis, and get more insight into the mechanism of HIF‐1α degradation, we impaired the cytoplasmic–nuclear shuttling of HIF‐1α. We first trapped HIF‐1α in the cytoplasmic compartment by fusing HIF‐1α to the Na+‐H+ exchanger (NHE‐1) and secondly we blocked nuclear export by using leptomycin B (Fornerod et al., 1997; Ossareh‐Nazari et al., 1997). Interestingly, and in marked contrast to the previously proposed model (Tanimoto et al., 2000), our data unequivocally demonstrate that: (i) nuclear translocation is not necessary for HIF‐1α stabilization, and (ii) both ‘nuclear’ and ‘cytoplasmic’ proteasomes are fully competent for HIF‐1α degradation in an oxygen‐dependent manner. This result also contrasts with the mechanism regulating the stability of p53, a transcription factor that is required to exit the nucleus for degradation via MDM2.
Results and Discussion
As schematized in Figure 1, we fused the entire human HIF‐1α protein to the cytoplasmic domain of the NHE‐1. NHE‐1 is an integral plasma membrane protein that exchanges intracellular protons at the expense of the inwardly‐directed Na+ gradient (Sardet et al., 1989). The protein (815 aa) is composed of two domains, the transporter per se (500 aa) which spans the plasma membrane 10 times followed by a cytoplasmic domain of about 315 residues, dispensable for transport activity but essential for conveying extracellular signals to the H+ modifier site (Counillon and Pouysségur, 2000; for review). The NHE‐1 sequence used has been engineered by point mutations to confer resistance to amiloride analogues allowing the direct selection of the transgene by H+ suicide‐selection (Pouysségur and Roux, 1999). Therefore, transfectants resistant to H+ suicide, in the presence of the amiloride analogue, must express HIF‐1α fused in‐frame with a functional, amiloride‐resistant NHE‐1. Because NHE‐1 is exclusively targeted to the plasma membrane or membrane vesicles, this approach ensures the sequestration of HIF‐1α to the cytoplasmic compartment. Hamster lung fibroblasts were transfected with either NHE‐1 alone (control) or the NHE‐1≫HIF‐1α chimera and populations resistant to H+ suicide were selected and independent clones isolated for further study.
Figure 2A and C show no HIF‐1α immunoexpression in normoxic conditions, whereas hypoxia (1–2% O2 for 3 h) is sufficient to result in increased expression of HIF‐1α in both clones. As expected, HIF‐1α accumulated in the nucleus of control cells (Figure 2B) and chimerical HIF‐1α was restricted to the cytoplasmic compartment (Figure 2D). The cytoplasmic localization of HIF‐1α seen in Figure 2D is identical to that of NHE‐1 in cells that overexpress the transporter. Confocal analysis indicates a localization in both the plasma membrane and the cytosolic network of membrane vesicles (data not shown). At this stage we could say that the chimerical HIF‐1α protein appears to be regulated as endogenous HIF‐1α: degraded in normoxia and stabilized at low O2 tension. CoCl2 is well known to mimic hypoxia. Indeed Figure 2E shows, by immunoblotting of total cellular extracts, a stabilization of both, endogenous (115 kDa) and chimerical (160–180 kDa) HIF‐1α proteins.
As we could identify in total extracts both the endogenous and chimerical HIF‐1α proteins by their different apparent molecular weight, we next investigated their respective half‐life following reoxygenation. Figure 3, left panel, shows that following hypoxia (1–2% O2 for 4 h), reoxygenation leads rapidly to dramatic destruction of the protein in control cells. Interestingly, the same experiment conducted in cells expressing the chimerical HIF‐1α protein demonstrated that reoxygenation, following hypoxia, rapidly leads to degradation of both proteins with parallel kinetics (Figure 3, middle panel). We obtained a half‐life of 5–8 min for both HIF‐1α proteins. It is well established that this rapid oxygen‐driven degradation is mediated via the ubiquitin‐dependent proteasomal system, an event which is prevented by the proteasomal inhibitor, lactacystin (Fenteany et al., 1995). Figure 3, right panel, shows a drastically increased half‐life for both HIF‐1α proteins in cells pre‐treated with this proteasome inhibitor. Note, in the presence of lactacystin, the increased level in normoxic conditions and the steady‐state level of both HIF‐1α proteins upon reoxygenation (compare middle and right panels of Figure 3). Altogether, these results clearly establish that cytoplasmic proteasomes are fully competent for O2‐dependent degradation of HIF‐1α.
The next step was to investigate whether HIF‐1α entry and retention in the nucleus facilitates in some way its escape from proteasomal degradation upon reoxygenation. HIF‐1α and p53 levels were followed in parallel since p53, another transcription factor stabilized by a variety of stresses, must exit the nucleus to undergo MDM2‐dependent proteasomal degradation (Freedman and Levine, 1998). Thus, we prevented HIF‐1α nuclear export by treatment with leptomycin B. Figure 4 shows that addition of leptomycin B has no effect on the basal level (normoxic) of HIF‐1α nor on the rate of HIF‐1α degradation upon reoxygenation. Leptomycin B, however strongly stabilizes p53 either in basal conditions or after hypoxic stress. This result, which validates the action of leptomycin, strongly indicates that the pVHL‐dependent degradation of HIF‐1α is totally independent of its subcellular localization. Leptomycin B was used here as a ‘forcing device’ to prevent possible HIF‐1α nuclear export. Upon reoxygenation of cells treated with leptomycin B, and lactacystin to abolish degradation, HIF‐1α remained in the nucleus (see Supplementary data).
In this communication, we have firmly established that HIF‐1α escape from proteasomal degradation, in response to hypoxia, could not be explained by the rapid entry of HIF‐1α into the nucleus. A previous study has eliminated the possible role of heterodimerization of HIF‐1α/HIF‐1β in this process (Chilov et al., 1999). The authors reported that the HIF‐1α protein translocated and accumulated in the nucleus of cells lacking HIF‐1β We fully agree with this finding. Indeed the cytoplasmic chimerical HIF‐1α protein when stabilized by hypoxia was not found to be associated with the HIF‐1β partner (data not shown), therefore excluding this association as a mechanism for proteasomal escape. Our findings however, contradict a recent report from Poellinger's group (Tanimoto et al., 2000). These authors propose that hypoxia‐induced protection of HIF‐1α against regulation by pVHL involves two distinct and successive steps: nuclear translocation of HIF‐1α and a hypoxic signal leading to escape from pVHL‐induced proteolysis. We agree on the need for a hypoxic signal but not for the nuclear translocation of HIF‐1α. Now one could argue that the HIF‐1α proteasomal degradation that we reported here with the NHE‐1≫HIF‐1α chimera is not regulated by pVHL. It is important to clarify this point because it has been suggested that HIF‐1α could be sent for proteasomal degradation via at least two E3‐ubiquitin ligases, pVHL and MDM2 (Maxwell et al., 1999; Ravi et al., 2000). The experiment conducted with leptomycin B (Figure 4) shows that sequestration of the MDM2–p53 complex in the nucleus, away from cytoplasmic proteasomes, has no impact on HIF‐1α degradation. This result suggests that the contribution of MDM2 to O2‐induced HIF‐1α degradation, if it exists, remains minimal. Our findings also indicate that the E3‐ligase pVHL is fully competent in targeting HIF‐1α to both cytoplasmic and nuclear proteasomes, a result in marked contrast with the E3 ligase, MDM2. Although there are many striking similarities between p53 and HIF‐1α, two transcription factors auto‐regulated post‐transcriptionally via proteasomal degradation and maintained at very low levels in normal and unstressed cells (Berra et al., 2001), p53 must exit the nucleus for degradation, while HIF‐1α is destroyed in both compartments. This ‘attack’ in both subcellular locations could explain the extraordinary short half‐life of HIF‐1α protein. Although this area of research has rapidly progressed, many steps in the adaptative response to hypoxia remain to be deciphered, in particular the hypoxic signal leading to HIF‐1α stabilization. Now with the implication of the HIF‐1α‐proline hydroxylase as a key enzyme controlling the O2‐dependent ubiquitylation (Ivan et al., 2001; Jaakkola et al., 2001), our results imply that proline hydroxylation of HIF‐1α and subsequent ubiquitylation must take place in both cellular compartments. We are currently testing this hypothesis.
Materials and plasmids.
For the cloning of NHE‐1≫HIF‐1α, human HIF‐1α (pcDNA3‐HA‐HIF‐1α) was initially digested with HindIII and XbaI and subcloned into pBluescript KS (Stratagene). The N‐terminal region of HIF‐1α was amplified by PCR using the primers: 5′‐CCCGGGGGCCCAGGGCGCCGGCGGCGCG‐3′ (sense) and 5′‐GTTCATAGTTCTTCCTCGGC‐3′ (antisense), digested with ApaI and HindIII and subcloned into the same vector. The NHE‐1≫HIF‐1α chimera was finally generated by subcloning of the ApaI–XbaI fragment of HIF‐1α into pECE‐NHE‐1 previously digested with the same enzymes (as shown in Figure 1, HIF‐1α is linked to the residue 698 of NHE‐1). Details for the two plasmids expressing either the amiloride‐resistant NHE‐1 or human HIF‐1α have been described previously (see references Pouysségur and Roux, 1999 and Richard et al., 1999; Gothié et al., 2000, respectively). The anti‐HIF‐1α antibody (antiserum 2087) was raised in our laboratory in rabbits immunized against the last 20 amino acids C‐terminal of the human HIF‐1α (Richard et al., 1999). The monoclonal antibody against human p53 was from DAKO (DO‐7).
Cell culture and transfections.
CCL39, NHE‐1 and NHE‐1≫HIF‐1α cell lines were cultured in Dulbeco's modified Eagle's medium (DMEM) supplemented with 7.5% fetal calf serum (FCS), penicillin G (50 U/ml), and streptomycin (50 μg/ml) (Gibco‐BRL) in an incubator (5% CO2) at 37°C. HeLa cells were incubated in the same medium but supplemented with inactivated FCS. Hypoxic conditions were performed by incubation of cells in a sealed ‘Bug‐Box’ anaerobic workstation (Ruskinn Technologies, Jouan). The oxygen in this workstation was maintained at 1–2% with a residual gas mixture being 93–94% nitrogen, and 5% carbon dioxide.
Clones expressing the amiloride‐resistant form of NHE‐1 or NHE‐1≫HIF‐1α were obtained by stable transfection of CCL39 cells with pECE‐NHE‐1 or pECE‐NHE‐1‐HIF‐1α plasmids, respectively. Transfection was performed by the calcium phosphate method. Three H+‐suicide selections were applied as previously described (Pouysségur and Roux, 1999) to obtain a stable cellular population. From the population expressing the chimera, independent clones were isolated and after selection for HIF‐1α expression, three different clones were selected for further study. We notice that high levels of NHE‐1≫HIF‐1α had a tendency to decline with passage because this expression slows down the growth rate.
Cells grown on glass coverslips were analysed using the anti‐HIF‐1α antiserum as previously described (Richard et al., 2000). Immunofluorescence was then analysed with a Leica DM‐R microscope equipped with a DC‐100 digital camera.
Western blot analysis.
Subconfluent cells were lysed in Laemmli buffer. The protein concentration was determined using the Lowry assay and 50 μg of whole cellular extracts were resolved by SDS–PAGE (7.5%) and electrophoretically transferred onto a polyvinylidene difluoride membrane (Immobilon‐P). HIF‐1α and p53 were revealed with specific antibodies. The immunoreactive bands were visualized with the ECL system (Amersham Pharmacia Biotech).
Supplementary Figure 1
Supplementary Figure 2
We thank Dr C. Dargemont for kindly providing leptomycin B and Dr C. Brahimi‐Horn for critical review of the manuscript. This work was supported by grants from the Centre National de la Recherche Scientifique (CNRS), Le Ministère de l'Education, de la Recherche et de la Technologie, La Ligue Nationale Contre le Cancer (équipe labellisée), and the GIP HMR (contract No. 1/9743B‐A3). E.B was the recipient of fellowships from the Human Frontiers Science Program and is presently supported by the GIP HMR. D.E.R. is a recipient of a fellowship from Canada (FRSQ).
- Copyright © 2001 European Molecular Biology Organization