Introduction
The hypoxia‐inducible factor (HIF) transcriptional system plays a central role in physiological responses to oxygen availability, and regulates genes involved in processes such as angiogenesis, erythropoiesis, vasomotor control, energy metabolism, carbon dioxide metabolism and cell survival decisions (
Semenza, 2000;
Wenger, 2000). Although oxygen availability can influence multiple steps in HIF activation (
Jiang et al., 1997;
Pugh et al., 1997;
Huang et al., 1998;
Kallio et al., 1998;
Bhattacharya et al., 1999), the primary mode of regulation occurs through oxygen‐dependent proteolysis of HIF‐α subunits (
Huang et al., 1996,
1998;
Pugh et al., 1997;
Salceda and Caro, 1997). In normoxic cells, HIF‐α subunits have an exceptionally short half‐life (
Jewell et al., 2001), and steady‐state levels are very low. Increasing severity of hypoxia retards degradation of HIF‐α subunits in a graded manner (
Jiang et al., 1996), allowing nuclear localization, dimerization with HIF‐β and formation of a DNA‐binding HIF complex.
HIF‐α subunits are therefore part of a large set of cellular regulators whose activity is determined by tightly controlled proteolysis. Many of these molecules are known to contain transferable destruction domains that can confer instability on heterologous proteins. In some cases, phosphorylation of particular residues provides a specific recognition signal that targets the substrate to ubiquitin ligase complexes, though in many cases the determinants that regulate proteolysis remain unknown (
Jackson et al., 2000).
In the case of HIF, two isoforms of the α‐subunit, HIF‐1α and HIF‐2α, have been shown to be regulated in a similar manner. Each possesses an extensive, transferable, oxygen‐dependent degradation domain (ODDD) encompassing >200 residues in the central region of the molecule (
Huang et al., 1998;
Ema et al., 1999;
O‘Rourke et al., 1999;
Sutter et al., 2000). Interestingly studies of isolated sequences have demonstrated that partial instability can be conveyed by subdomains within the ODDD (
Huang et al., 1998;
O'Rourke et al., 1999;
Yu et al., 2001). Since these domains must interact with the oxygen‐sensitive signal, their sequences have been analysed intensively in an effort to gain an understanding of the sensing/transduction process. Important insights have been gained from studies of interactions of HIF‐α with the von Hippel–Lindau tumour suppressor protein (pVHL) (
Maxwell et al., 1999). pVHL is part of a multiprotein ubiquitin E3 ligase complex (VHLE3), homologous to the SCF (Skp‐1‐Cdc53/Cullin‐F‐box) class of E3 ligases (
Lisztwan et al., 1999;
Stebbins et al., 1999). pVHL itself plays a role analagous to the F‐box substrate recognition component, and can interact directly with HIF‐α subunits and target them for VHLE3‐dependent ubiquitylation
in vitro (
Cockman et al., 2000;
Kamura et al., 2000;
Ohh et al., 2000). Protein interaction and ubiquitylation assays have defined a subdomain in the C‐terminal portion of the HIF‐1α ODDD that is necessary and sufficient for VHLE3‐dependent ubiquitylation under the conditions of assay, and shown that residues 556–574 of HIF‐1α constitute a minimal pVHL‐binding domain within this region (
Cockman et al., 2000;
Ohh et al., 2000;
Tanimoto et al., 2000). Interaction of this region with pVHL is promoted by enzymatic hydroxylation of Pro564 (
Ivan et al., 2001;
Jaakkola et al., 2001). Although the HIF‐ prolyl hydroxylase(s) remain to be characterized, these findings are of interest in relation to the mechanism of oxygen sensing. Since known enzymes of this class are dioxygenases that utilize molecular oxygen as co‐substrate (
Kivirikko and Myllyharju, 1998), these findings provide a direct link between the availability of oxygen and the regulation of HIF. In the simplest model, the availability of oxygen would affect the rate of modification of Pro564 and hence degradation of HIF‐α by the VHLE3–ubiquitin–proteasome pathway. However, such a model provides little insight into how the complex physiological response to oxygen availability is controlled with such precision, does not define a role for the majority of the HIF‐α ODDD sequences and cannot explain the partial instability conveyed by other regions of the ODDD.
To investigate this, we have analysed interactions between HIF‐α ODDDs and the VHLE3 complex using a series of ubiquitylation and interaction assays based on crude cell lysates or purified components of the ubiquitylation system. We show that two independent regions of the HIF‐1α ODDD are targeted for ubiquitylation by VHLE3 in a manner that is dependent upon hydroxylation of specific proline residues. Although both proline residues are located in a motif that is conserved between HIF‐1α and HIF‐2α at both sites, the target sites differ in overall sequence and requirements for interaction with pVHL.
Discussion
In this work, we demonstrate that HIF‐1α contains two independent regions in the ODDD that are targeted by the VHLE3 ubiquitin ligase. In vitro assays demonstrated that both sites can function independently, supporting interactions with the VHLE3 complex and VHLE3‐dependent ubiquitylation. Furthermore, mutational analysis demonstrated the functional importance of each site in regulating transcriptional responses in vivo.
At the C‐terminal site, residues 556–572 previously have been defined as a minimal VHLE3 interaction domain (
Tanimoto et al., 2000). The current study has defined a minimal VHLE3 target site within HIF‐1α residues 380–417. The definition of this second site provides additional evidence for the critical role played by pVHL in the regulation of the HIF system, and also explains previously puzzling observations regarding the function of isolated portions of the ODDD. Whilst previous studies have indicated that interaction with the VHLE3 complex appears to be limited to the C‐terminal site, both N‐ and C‐terminal sequences within the ODDD are able to mediate partial levels of protein instability. This has been reported both in the context of the native HIF‐1α molecule and when isolated subsequences from HIF‐1α and HIF‐2α have been analysed as fusions to heterologous proteins (
Huang et al., 1998;
Ema et al., 1999;
O'Rourke et al., 1999;
Sutter et al., 2000;
Yu et al., 2001). The current work indicates that the observed partial instability may be accounted for by the operation of one, but not both VHLE3 target sites.
Importantly, both VHLE3 target sites in HIF‐1α appear to be regulated by enzymatic hydroxylation of specific prolyl residues. Hydroxylation of Pro564 recently has been identified as the key modification controlling activity of the C‐terminal target site (
Ivan et al., 2001;
Jaakkola et al., 2001). Analysis of the N‐terminal VHLE3 interaction site now implicates hydroxylation of Pro402 in the regulation of targeting at this site, and demonstrates the presence of a common motif, LXXLAP, at the two sites. Thus, the new findings provide further evidence for the importance of prolyl hydroxylation in the regulation of HIF, and for the importance of prolyl hydroxylation as a mechanism of protein recognition by the VHLE3 complex.
However, despite the similarities in the operation of the two VHLE3 target sites, important differences were defined. First, reticulocyte lysate contains an activity that is able to hydroxylate the C‐terminal site effectively and to promote interaction with the VHLE3 complex. In contrast, exposure to RCC4 or other tissue culture cell cytoplasmic extract was required to promote the VHLE3 interaction with the N‐terminal site, whereas reticulocyte lysate had no such activity. Secondly, although both VHLE3 target sites contain a common motif, the sequences required for efficient enzymatic modification appear to extend well beyond this motif and are quite different at the two sites. For instance, the synthetic peptide B28HYP corresponding to HIF‐1α residues 390–417 and containing hydroxyproline at residue 402 could bind to VHLE3 and therefore must contain the VHLE3 interaction determinants. Nevertheless, neither the corresponding peptide containing proline, B28PRO, nor the GAL‐HIF‐1α fusion GAL390–417, could be modified by cytoplasmic extract so as to capture VHLE3, indicating that more extensive sequences were required to direct the modification step. Taken together, these findings suggest either that additional factors present in RCC4 cytoplasmic extract but not reticulocyte extract are required for sequence recognition and hydroxylation at the N‐terminal site, or that a different enzyme or enzyme isoform is involved.
The HIF‐1α sites also differed in their mode of interaction with the VHLE3 complex. Whereas the hydroxylated C‐terminal site interacts readily with recombinant pVHL produced in a variety of expression systems, and has been shown to interact directly with pVHL, the N‐terminal site could not interact with recombinant pVHL expressed in vitro in reticulocyte lysate. In contrast, a robust interaction of the N‐terminal target site was obtained with VHLE3 derived from lysates of 786‐0 HA·VHL cells. This difference could indicate the need for a modification of pVHL that occurs in vivo but not in vitro in reticulocyte lysate, or the operation of an additional factor that cooperates in a more complex interaction with VHLE3 at the N‐terminal target site. So far, however, we have not been able to reconstitute the interaction in vitro with known components of this complex.
Whatever the precise reasons for these findings, the existence of more than one destruction domain targeted by VHLE3 is itself of interest. Similar analyses in other systems have indicated that multiple destruction domains are not uncommon among proteins that are regulated by degradation. For instance, there are several examples of cell cycle proteins that contain two destruction boxes (D boxes) targeted by the anaphase‐promoting complex, another multicomponent E3 ligase. In these cases, the D boxes also differ in overall sequence and potency in some assays. The reasons why these proteins have evolved multiple destruction boxes are not well understood, but they presumably provide for increased combinatorial interactions that lend specificity to the destruction process.
Modification of the destruction domains in HIF‐α subunits by enzymatic prolyl hydroxylation is of particular interest in relation to the underlying physiology of oxygen sensing. Recent evidence has indicated that hydroxylation at Pro564 is performed by one or more members of the 2‐oxoglutarate‐dependent dioxygenase superfamily (
Jaakkola et al., 2001). The use of molecular oxygen as co‐substrate by such enzymes provides a direct link between HIF regulation and the availability of molecular oxygen. However, it is difficult to envisage how modification of a particular peptide substrate by a single enzyme could account for the precisely shaped physiological responses of the system. Many other systems of oxygen sensing have been proposed and might impact at different points in the pathway (
Semenza, 1999;
Zhu and Bunn, 2001). Equally, other ubiquitin ligase systems such as mdm‐2 have been proposed to impact on HIF regulation (
Ravi et al., 2000), although the site and mode of targeting of the HIF system have not been defined, and might involve different oxygen‐sensitive processes. Neverthe less, the recognition of two sites of modification by prolyl hydroxylation with different properties with respect to the modifying activity and VHLE3 interaction indicates a potential for more complex responses to oxygen availability to be mediated through enzymatic prolyl hydroxylation. It will now be of interest to identify the prolyl hydroxylase(s) operating at the different VHLE3 target sites and compare their oxygen‐dependent characteristics for modification of different degradation domains.
Materials and methods
Plasmids
His
6‐tagged mouse E1 cDNA in pRSET was kindly donated by T.Hunt. GAL fusion proteins were encoded by plasmids based on pcDNA4 that contain a truncated GAL4 gene encoding amino acids 1–147 followed by a polylinker bearing
SacII and
AscI sites into which the HIF‐1α or HIF‐2α sequences generated by PCR were cloned. All PCRs were performed using
Pfu DNA polymerase (Stratagene). pcDNA3‐HIF‐1α, pVHLHA, pGAL/VP16, pGAL/α344–417/VP16, pUAS‐tk‐Luc and pCMVβGal have been described previously (
Pugh et al., 1997;
O‘Rourke et al., 1999;
Cockman et al., 2000). pGL3PGK6TKp contained six copies of the HRE from the mouse phosphoglycerate kinase‐1 gene linked to a luciferase reporter gene. pGAL/344–400/VP16 was constructed by insertion of HIF‐1α sequence encoding amino acids 344–400 into
SacII–
AscI‐digested pGAL/VP16. Mutations were generated using a site‐directed mutagenesis kit (QuickChange; Stratagene) and mutagenic oligonucleotides designed according to the manufacturer's recommendations. The integrity of all plasmids was confirmed by DNA sequencing.
Cell culture and transient transfection
RCC4 cells stably transfected with pcDNA3‐VHL (RCC4/VHL) or empty vector (RCC4) (
Cockman et al., 2000) and the HIF‐1α‐deficient CHO cell line (Ka13) (
Wood et al., 1998) have been described previously. 786‐0 HA·VHL cells (786‐0 cells stably transfected with plasmid pRC‐HA‐VHL) were a gift from W.G.Kaelin, and U2OS cells were a gift from S.Geley. All cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, glutamine (2 mM), penicillin (50 IU/ml) and streptomycin sulfate (50 μg/ml). For RCC4, RCC4VHL and 786‐0 HA·VHL, G418 (0.5 mg/ml) was added to the growth medium.
Transient transfections were performed using Fugene 6 (Roche Molecular Biochemicals). For luciferase assay in U20S cells, 10 ng of GAL/HIF‐1α/VP16 activator plasmid, 100 ng of pUAS‐tk‐Luc luciferase reporter and 500 ng of pCMVβgal (to enable correction for variation in transfection efficiency) were used per well of a 6‐well plate. For luciferase assay in Ka13 cells, 2 μg of pcDNA3‐HIF‐1α plasmid, 0.1 μg of pGL3PGK6TKp luciferase reporter and 0.5 μg of pCMVβgal were used. Hypoxic incubation was in an atmosphere of 0.1% oxygen, 5% CO2, balance nitrogen in a Napco 7001 incubator (Jouan). For immunoblotting experiments, Ka13 cells were transfected with 2 μg of pcDNA3‐HIF‐1α wild‐type and mutant plasmids.
Luciferase and β‐galactosidase assays
Luciferase activities were determined in extracts made from transfected cells maintained for 48 h, either entirely in normoxia or with hypoxic stimulation for the final 16 h. Luciferase activities were determined using a commercially available luciferase assay system (Promega) and a TD‐20e luminometer (Turner Designs). Relative β‐galactosidase activity in extracts was measured using o‐nitrophenyl‐β‐d‐galactopyranoside (0.67 mg/ml) as substrate in a 0.1 M phosphate buffer pH 7.0 containing 10 mM KCl, 1 mM MgSO4 and 30 mM β‐mercaptoethanol incubated at 30°C for 15–45 min. The A420 was determined after stopping the reaction by the addition of sodium carbonate to a final concentration of 0.4 M.
Substrate and cell extract preparation
[35S]methionine‐labelled HIF‐1α and GAL‐HIF‐1α substrates were prepared by coupled IVTT using TnT7 rabbit reticulocyte (Promega). GAL–HIF‐1α proteins appear as two bands, with the faster migrating form arising from aberrant translational initiation within the GAL sequence. In all assays, the faster migrating form is recognized less efficiently by VHLE3.
Ka13 whole‐cell extracts for immunoblotting were prepared 24 h after transfection by lysis in 8 M urea, 10% glycerol, 1% SDS, 5 mM dithiothreitol (DTT), 10 mM Tris pH 6.8, followed by disruption using a hand‐held homogenizer (Ultra‐Turrax T8 with 5G dispersing tool; Janke & Kunkel GmbH). Cytoplasmic extract for ubiquitylation assays and for modification of HIF‐1α substrates
in vitro was prepared as previously described (
Cockman et al., 2000). Briefly, cells were washed twice with cold hypotonic extraction buffer (HEB: 20 mM Tris pH 7.5, 5 mM KCl, 1.5 mM MgCl
2, 1 mM DTT). After removal of excess buffer, cells were lysed in a Dounce homogenizer. Following lysis, crude cytoplasmic extract was centrifuged at 10 000
g for 5 min at 4°C to remove cell debris and nuclei, and stored in aliquots at −70°C. Heat‐treated cytoplasmic extract was prepared by incubation at 60°C for 3 min, followed by centrifugation at 10 000
g for 5 min to remove precipitated material. S100 extract was obtained by an additional ultracentrifugation step at 100 000
g at 4°C for 4 h. Cytoplasmic extract was also incubated at 4°C for 4 h to ensure that effects seen with S100 extract were due to the 100 000
g spin. ATP‐depleted cytoplasmic extract was prepared by hexokinase/glucose treatment as follows. To each 100 μl of cytoplasmic extract, 5 μl of 1 M glucose and 5 U of 1 U/μl hexokinase (Sigma) were added and the sample incubated at 30°C for 30 min to allow ATP depletion. Nuclear extract was obtained by extraction of the crude nuclear pellet that remains following cytoplasmic extract preparation with three volumes of buffer C [20 mM Tris pH 8.0, 25% glycerol (v/v), 0.42 M NaCl, 1.5 mM MgCl
2, 0.2 mM EDTA], followed by 3‐fold dilution in 20 mM Tris pH 8.0 and storage at −70°C. 786‐0 HA·VHL cell extract, used as a source of VHLE3, was prepared by lysis in NP‐40 lysis buffer (10 mM Tris pH 7.5, 0.25 M NaCl, 0.5% NP‐40). NP‐40 lysis buffer does not support the HIF‐1α‐modifying activity (data not shown).
Modification of substrate was achieved by incubation of GAL–HIF‐1α translate (4 μl) with 70 μl of RCC4 cell lysate in HEB, or control HEB alone, at 30°C for 20 min prior to anti‐GAL immunoprecipitation. For modification in the presence of Fe (II), ferrous chloride (100 μM) was added to the reaction. For modification in the presence of prolyl hydroxylase inhibitor, N‐oxalylglycine (1 mM) was used either alone or in combination with 2‐oxoglutarate (5 mM). Modification reactions using hexokinase‐treated cytoplasmic extract and nuclear extract were as above.
Antibodies, immunoblotting and peptides
Mouse anti‐HA antibody (12CA5) used for immunoprecipitation and rat anti‐HA antibody (3F10) used for immunoblotting were from Roche Molecular Biochemicals. Anti‐GAL4 (RK5C1) agarose conjugate was from Santa Cruz Biotechnology. Anti‐HIF‐1α antibody (clone54) was from Transduction Laboratories. Following SDS–PAGE, proteins were transferred on to Immobilon‐P membrane (Millipore) and processed for immunoblotting using the indicated antibody. Biotinylated HIF‐1α peptides (Biopeptide Co.) were retrieved using streptavidin Dynabeads M‐280 (Dynal ASA).
Ubiquitylation enzymes and assays
The E1‐activating enzyme used in ubiquitylation assays was either obtained from Affiniti Research or purified from BL21 (DE3) Escherichia coli transfected with plasmid expressing His6‐tagged mouse E1. His6‐E1 was purified by Ni2+–agarose affinity chromatography. After dialysis against phosphate‐buffered saline, glycerol was added to 10% (v/v) and 25 ng/μl aliquots stored at −80°C. Human CDC34 recombinant E2 enzyme was from Affiniti Research. VHLE3 was obtained by anti‐HA immunoprecipitation from stably transfected 786‐0 HA·VHL cell lysates. Briefly, 1 ml of 786‐0 HA·VHL cell lysate (∼107 cells) and 5 μg of anti‐HA antibody were incubated at 4°C for 1 h. A 12 μl aliquot of protein G–Sepharose beads was added and incubation continued at 4°C with mixing. Beads were then washed four times in IP wash buffer (125 mM NaCl, 25 mM Tris pH 7.5, 0.1% NP‐40), with a final wash in HEB. GAL–HIF‐1α substrate was prepared by anti‐GAL immunoprecipitation from [35S]methionine‐labelled TnT7 rabbit reticulocyte (Promega) translates.
The purified component ubiquitylation reaction (40 μl) consisted of 4 μl of 5 mg/ml ubiquitin, 4 μl of 10× ATP‐regenerating system (20 mM Tris pH 7.5, 10 mM ATP, 10 mM magnesium acetate, 300 mM creatine phosphate, 0.5 mg/ml creatine phosphokinase), 2 μl of E1, 3 μl of E2, 6 μl of VHLE3 immunopurified on protein G–Sepharose (∼5 × 106 cells), 6 μl of GAL‐HIF‐1α substrate immunopurified on agarose beads (derived from 4 μl of translate) and 15 μl of HEB. Reactions were incubated at 30°C for 2 h with occasional mixing, stopped by the addition of SDS sample buffer and analysed by SDS–PAGE and autoradiography. For ubiquitylation reactions in the presence of peptide, peptides were added at a final concentration of 18.5 μM.
The cytoplasmic extract‐based ubiquitylation assays have been described previously (
Cockman et al., 2000).
In vitro interaction assays
For pVHL interaction assays, HEB‐treated or cytoplasmic extract‐modified GAL–HIF‐1α substrates (4 μl, non‐radiolabelled) were immunopurified using anti‐GAL antibody‐conjugated Sepharose beads (10 μl of beads). [35S]methionine‐labelled VHLHA translate (4 μl) was added together with 60 μl of HEB and samples incubated at 30°C for 1 h with mixing. Beads were then washed four times with IP wash buffer and once with IP wash buffer lacking NP‐40. Co‐precipitating VHLHA proteins were analysed by SDS–PAGE and fluorography.
For VHLE3 interaction assays, cytoplasmic extract‐modified GAL–HIF‐1α substrates were incubated together with 786‐0 HA·VHL cell lysate and anti‐HA antibody (2.5 μg) at 4°C for 1 h. Protein G–Sepharose beads were added and samples mixed at 4°C for 30 min. Beads were washed as above. Co‐precipitating GAL–HIF‐1α proteins were analysed by SDS–PAGE and fluorography.
For VHLE3 capture assays using biotinylated peptides, HEB‐ or cytoplasmic extract‐treated peptides (750 pmol) were incubated with 786‐0 HA·VHL cell lysate at 4°C for 1 h. Streptavidin dynabeads (7 × 107 beads) were added and samples mixed at 4°C for a further 30 min. Beads were washed four times with IP wash buffer and once with IP wash buffer lacking NP‐40. Co‐precipitating HA·VHL was analysed by anti‐HA immunoblotting.