Erythrocytosis-associated HIF-2α Mutations Demonstrate a Critical Role for Residues C-terminal to the Hydroxylacceptor Proline*

A classic physiologic response to hypoxia in humans is the up-regulation of the ERYTHROPOIETIN (EPO) gene, which is the central regulator of red blood cell mass. The EPO gene, in turn, is activated by hypoxia inducible factor (HIF). HIF is a transcription factor consisting of an α subunit (HIF-α) and a β subunit (HIF-β). Under normoxic conditions, prolyl hydroxylase domain protein (PHD, also known as HIF prolyl hydroxylase and egg laying-defective nine protein) site specifically hydroxylates HIF-α in a conserved LXXLAP motif (where underlining indicates the hydroxylacceptor proline). This provides a recognition motif for the von Hippel Lindau protein, a component of an E3 ubiquitin ligase complex that targets hydroxylated HIF-α for degradation. Under hypoxic conditions, this inherently oxygen-dependent modification is arrested, thereby stabilizing HIF-α and allowing it to activate the EPO gene. We previously identified and characterized an erythrocytosis-associated HIF2A mutation, G537W. More recently, we reported two additional erythrocytosis-associated HIF2A mutations, G537R and M535V. Here, we describe the functional characterization of these two mutants as well as a third novel erythrocytosis-associated mutation, P534L. These mutations affect residues C-terminal to the LXXLAP motif. We find that all result in impaired degradation and thus aberrant stabilization of HIF-2α. However, each exhibits a distinct profile with respect to their effects on PHD2 binding and von Hippel Lindau interaction. These findings reinforce the importance of HIF-2α in human EPO regulation, demonstrate heterogeneity of functional defects arising from these mutations, and point to a critical role for residues C-terminal to the LXXLAP motif in HIF-α.

HIF binds to the promoters and enhancers of a broad range of genes involved in cellular, local, and systemic responses to hypoxia. They include genes that facilitate glucose uptake, enhance glycolysis, inhibit the Krebs cycle, promote angiogenesis, and augment red blood cell production (4 -6). The latter is mediated by EPO, a glycoprotein that is the product of the EPO gene, which is considered a paradigm of oxygen-regulated gene expression because of its high dynamic range of inducibility and the central importance of this gene for oxygen delivery to tissues (7,8).
Aside from addressing the fundamental biologic interest in oxygen-dependent regulation of EPO, this pathway has also provided a framework for understanding certain rare forms of erythrocytosis in which there is abnormally increased red blood cell production due to aberrant up-regulation of EPO (9 -12). The first gene of the oxygen-sensing pathway in which erythrocytosis-associated mutations were identified was the VHL gene (13), followed by mutations in the PHD2 gene (14). Subsequently, we described a mutation of the HIF2A gene in a family with hereditary erythrocytosis (15). This mutation, G537W, affects a residue in close proximity to the primary prolyl hydroxylation site in HIF-2␣, Pro-531. Functional studies indicated that it impairs both the ability of PHD2 to bind to and hydroxylate HIF-2␣, as well as the subsequent capacity of VHL to recognize it.
We more recently reported two additional erythrocytosisassociated HIF2A mutations, G537R and M535V (16). One of these mutations, G537R, has subsequently been independently identified in other patients with erythrocytosis (17,18). The existence of these new mutations raises questions as to whether they lead to stabilization of HIF-2␣, and if so, whether this is due to altered interactions with PHD2 and/or VHL. Such altered interactions might also raise the possibility of dominant negative mechanisms. Our analysis of these mutations as well as a third novel erythrocytosis-associated HIF2A mutation, P534L (19), reveals that whereas all mutations result in the aberrant stabilization of HIF-2␣, they do so in ways that are not entirely TCA GTA CCC AGA CGG ATT TCA ATG AGC TGG ACT  TGG AGA CAC TGG CAC CCT ATA TCC CCG TGG ACG  GGG AAG ACT TCC AAT TGA GCC CCA TCT GCC CCG  AGG AGT AG-3Ј and 5Ј-CTA GCT ACT CCT CGG GGC  AGA TGG GGC TCA ATT GGA AGT CTT CCC CGT CCA  CGG GGA TAT AGG GTG CCA GTG TCT CCA AGT CCA  GCT CAT TGA AAT CCG TCT GGG TAC TG-3Ј; G537R,  5Ј-AAT TCA GTA CCC AGA CGG ATT TCA ATG AGC  TGG ACT TGG AGA CAC TGG CAC CCT ATA TCC CCA  TGG ACA GGG AAG ACT TCC AAT TGA GCC CCA TCT  GCC CCG AGG AGT AG-3Ј and 5Ј-CTA GCT ACT CCT  CGG GGC AGA TGG GGC TCA ATT GGA AGT CTT CCC  TGT CCA TGG GGA TAT AGG GTG CCA GTG TCT CCA  AGT CCA GCT CAT TGA AAT CCG TCT GGG TAC TG-3Ј. Additional pcDNA-GAL4-HIF constructs were prepared in an analogous manner using duplexes comprised of the following sequences: HIF-2␣ Sec (residues 390 -423): 5Ј-AAT TCT  TCA CCA AGC TAA AGG AGG AGC CCG AGG AGC TAG  CCC AGC TGG CTC CCA CCC CCG GGG ACG CCA TCA  TCT CTC TGG ATT TCG GGA ATC AGA ACT TCG AGG  AGT AG-3Ј and 5Ј-CTA GCT ACT CCT CGA AGT TCT GAT  TCC CGA AAT CCA GAG AGA TGA TGG CGT CCC CGG  GGG TGG GAG CCA GCT GGG CTA GCT CCT CGG GCT  CCT CCT TTA GCT TGG TGA AG-3Ј; Prim/Sec, 5Ј-AAT  TCA GTA CCC AGA CGG ATT TCA ATG AGC TCG ACT  TGG AGA CAC TGG CAC CCA CCC CCG GGG ACG CCA  TCA TCT CTC TGG ATT TCG GGA ATC AGA ACT TCG  AGG AGT AG-3Ј and 5Ј-CTA GCT ACT CCT CGA AGT TCT  GAT TCC CGA AAT CCA GAG AGA TGA TGG CGT CCC  CGG GGG TGG GTG CCA GTG TCT CCA AGT CGA GCT  CAT TGA AAT CCG TCT GGG TAC TG-3Ј; Sec/Prim,  5Ј-AAT TCT TCA CCA AGC TAA AGG AGG AGC CCG  AGG AGC TAG CCC AGC TGG CTC CCT ATA TCC CCA  TGG ACG GGG AAG ACT TCC AAT TGA GCC CCA TCT  GCC CCG AGG AGT AG-3Ј and 5Ј-CTA GCT ACT CCT  CGG GGC AGA TGG GGC TCA ATT GGA AGT CTT CCC  CGT CCA TGG GGA TAT AGG GAG CCA GCT GGG CTA  GCT CCT CGG GCT CCT CCT TTA GCT TGG TGA AG-3Ј. All constructs were verified by DNA sequencing.
PHD2 Binding Assays-(His) 6 FlagPHD2 immobilized on 10 l of M2-agarose was incubated with 35 S-labeled, in vitro translated GAL4-HIF fusion proteins as described (21). The resins were then washed, eluted with 2ϫ SDS loading buffer, and the eluates subjected to SDS-PAGE and phosphorimager analysis on an ABI Storm 860 PhosphorImager. 6 FlagPHD2 immobilized on 10 l of M2-agarose was washed and assayed for activity essentially by the method of Kivirikko and Myllyla (23) as modified by Linke et al. (24). Reactions were performed in a total volume of 45 l of 6 mM Hepes, pH 7, 45 mM NaCl, 7.5 mM ␤-glycophosphate, 0.6 mM sodium pyrophosphate, 3% glycerol, 0.3% Triton X-100, 0.3 mM dithiothreitol, with 2.4 mM ascorbate, 120 M FeCl 2 , and 0.5 l of [ 14 C]2-oxoglutarate (50 mCi/mmol; 0.1 mCi/ml) in the absence or presence of 25 M wild type or mutant HIF-2␣-(527-542) peptides. Reactions were conducted for 1 h at 37°C in 1.5-ml screw top tubes fitted with filter papers saturated with Ca(OH) 2 . The 14 CO 2 liberated and captured on the filter paper was quantitated by scintillation counting. Uncoupled decarboxylation measurements performed in the absence of HIF peptide were subtracted from that in its presence to obtain HIF hydroxylase activity.

2-Oxoglutarate Decarboxylation Assays-(His)
Mass Spectrometry Assays-(His) 6 FlagPHD2 immobilized on 10 l of M2-agarose was washed with 50 mM NH 4 HCO 3 containing 0.5 mM dithiothreitol and incubated at 37°C in a volume of 40 l of the same buffer supplemented with 1 mM ascorbate and 0.5 mM 2-oxoglutarate in the absence or presence of 10 M HIF-2␣-(527-542) peptides. At various times, aliquots of 5 l were suspended in 50:50 MeOH:H 2 O with 0.1% formic acid and subjected to electrospray ionization with an Advion TriVersa NanoMate spray source. All mass spectrometry analysis was performed on a Thermo LTQ Orbitrap hybrid mass spectrometer. Scans accumulated at high resolution (100,000) with 10 microscan accumulation. Typical total accumulations were 15 to 20 s.
VHL Binding Assays-Prolyl-hydroxylated GST-HIF-1␣-(531-575) was prepared by incubating GST-HIF-1␣-(531-575) immobilized on GSH-agarose with (His) 6 FlagPHD2 in the presence of ascorbate and 2-oxoglutarate, and then washing the resin. Alternatively, 0.12 g of biotinylated Hyp-564 HIF-1␣-(556 -574) peptide was incubated with 10 l of streptavidin-agarose (Sigma). The resin was then incubated with 35 S-labeled, in vitro translated FlagVHL as described previously in the absence or presence of Hypcontaining peptide. The resins were washed, eluted with 2ϫ SDS loading buffer, and the eluates subjected to SDS-PAGE and phosphorimager analysis. Cell Culture and Generation of Stable Transfectants-Stably transfected HEK293 Flp-In TRex cell lines (Invitrogen), in which the pcDNA5/FRT/TO construct is integrated into a single defined locus, were generated using Flp recombinase according to the manufacturer's instructions. Those expressing wild type and P531A 3xFLAG-HIF-2␣ have been previously described (15). These cells were maintained in Dulbecco's modified Eagles medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. HIF-2␣ expression was induced by treating cells with 0.1 g/ml doxycycline for 16 h. For protein stability assays, cycloheximide was employed at a concentration of 20 g/ml. Hypoxic treatment was performed in an In Vivo 200 Hypoxia Work station (Ruskinn Technologies).
Real Time PCR-Total RNA was isolated from stably transfected HEK293 Flp-In T-REx cell lines induced with doxycy- cline for 16 h using TRIzol (Invitrogen) according to the manufacturer's protocol. cDNA was prepared using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real time PCR analysis was performed using SYBR Green Master mix (ABI) and an Applied Biosystems 7300 Real Time PCR System. Target gene levels were normalized to that of ␤-actin. SYBR Green primers, designed with the use of Primer Express software, version 3.0 (Applied Biosystems), were as follows: 3xFLAG, 5Ј-GCT TGG TAC CAT GGA CTA CAA AGA-3Ј and 5Ј-CGT CAT CCT TGT AAT CGA TGT CA-3Ј; hVEGF, 5Ј-AGA CTC CGG CGG AAG CAT-3Ј and 5Ј-AAT GGC GAA TCC AAT TCC AA-3Ј; hNDRG1, 5Ј-AGT GCG GCT GCC AGG TT-3Ј and 5Ј-CAC GGT GAG CCA AAA TGA AA-3Ј; hADM, 5Ј-ACC GCC AGA GCA TGA ACA A-3Ј and 5Ј-AAG CGG CAG CCA AAG CT-3Ј. Dissociation curve analysis for each set of probes revealed single peaks. The sequences of the ␤-actin primers have been previously described (25). Relative quantification was performed using the ⌬⌬C T method and ␤-actin as the endogenous control.
Luciferase Assays-The HEK293 Flp-In TRex cell lines were cotransfected with (eHRE) 3 -luciferase (26) and the Renilla luciferase internal control pRL-TK using FuGENE 6 (Roche), and then treated with 0.1 g/ml doxycycline for 16 h. In some cases, cells were exposed to hypoxia (0.2% O 2 ). Cells were harvested with Passive Lysis Buffer (Promega) and then clarified by centrifugation at 13,000 ϫ g for 15 min at 4°C. Luciferase activities were measured using a Dual Luciferase Reporter Assay System (Promega) and a Berthold Lumat LB9507 luminometer.
Statistical Analysis-Student's t test and analysis of variance were used for statistical analysis. Differences were considered significant when p Ͻ 0.05.
The [ 14 C]2-oxoglutarate decarboxylation assay indirectly measures hydroxylation. As an independent means of assessing whether the mutations impair PHD2-induced hydroxylation, we incubated recombinant PHD2 with wild type, P534L, M535V, or G537R HIF-2␣-(527-542) peptide, and then assessed the extent of hydroxylation by electrospray ionization mass spectrometry (Fig. 2). Under conditions in which the wild type peptide was very efficiently hydroxylated ( Fig. 2A), all three mutant peptides were hydroxylated substantially less so (Fig. 2, B-D), evident at both the 30and 60-min time points.
We next generated stably transfected, Flp-In TRex HEK293 cells that express doxycycline-inducible FLAG-tagged P534L, M535V, and G537 HIF-2␣ (Fig. 4A). The constructs encoding for the HIF-2␣ mutants integrate into a single FRT site in this cell line so as to generate isogenic cell lines. As a reflection of this, two independent clones for each of these three mutants revealed comparable levels of steady state, doxycycline-induced protein levels, as assessed by Western blotting with anti-FLAG  Two independent clones for each of the P534L, M535V, and G537R mutants were examined and are shown. HIF2-␣ P534L, M535V, and G537R clones depicted in lanes 4, 5, and 7, respectively, were chosen for further analysis. Western blot for ␤-tubulin is also shown. B, real time PCR of messenger RNA transcripts from the 3xFlagHIF-2␣ gene in wild type or mutant 3xFlagHIF-2␣ cell lines. p values for each mutant (as compared with wild type (WT)) were Ͼ0.05. antibodies (Fig. 4A). When compared with wild type FLAGtagged HIF-2␣ expressed from the same system, we find that the protein levels for each mutant is higher. Another cell line expressing a hydroxylation-defective P531A HIF-2␣ also displays higher steady state levels of protein than wild type, as reported previously (15). For subsequent experiments, we selected a single clone from each pair of P534L, M535V, and G537R mutants. Real time PCR employing primers specific for the nucleotide sequence encoding the FLAG tag show that the wild type, P534L, M535V, and G537R mRNA levels are all comparable (Fig. 4B).
To assess the stability of these proteins in more detail, we arrested protein synthesis by treatment of cells with cycloheximide, and then the monitored protein level by Western blotting using anti-FLAG antibodies (Fig. 5). As a control, wild type HIF-2␣ is rapidly degraded (left-hand side of each panel). We find that P534L, M535V, and G537R mutants of HIF-2␣ mutants are each degraded more slowly than wild type HIF-2␣ (top three sets of panels), although perhaps not as slowly as the P531A hydroxylation defective control (bottom set of panels).
We next examined, by real time PCR, the expression of the HIF target genes adrenomedullin (ADM), N-myc downstream regulated gene 1 (NDRG1), and vascular endothelial growth factor (VEGF) (27) in the wild type and mutant HIF-2␣ expressing HEK293 cells (Fig. 6). The expression of M535V and G537R HIF-2␣, as well as P531A HIF-2␣, induced significantly increased mRNA transcript levels from the ADM gene (Fig. 6A) as compared with wild type HIF-2␣. The expression of P534L and G537R HIF-2␣, as well as P531A HIF-2␣, induced significantly increased mRNA transcript levels from the NDRG1 and VEGF genes as compared with wild type (Fig. 6, B and C). The baseline levels of a given target gene showed slight variation among the different cell lines. We therefore also calculated the induction ratios of each HIF target gene (Fig. 6, D-F). When analyzed in this manner, the induction ratios of all three genes are higher with all of the HIF-2␣ mutants than that obtained with wild type HIF-2␣.
The higher levels of HIF target genes induced by the HIF-2␣ mutants is consistent with increased transcriptional activity. To examine this independently, we transfected the HIF-2␣expressing HEK293 cell lines with a hypoxia response element reporter gene (Fig. 7). Under normoxic conditions, and consistent with the previous data (Fig. 6), we find that reporter gene activity is increased with all three mutants as compared with wild type. In contrast, hypoxic conditions, which would be expected to stabilize wild type HIF-2␣, leads to levels of reporter gene activity that are similar to that of wild type.
All three of the mutations examined here affect residues C-terminal to the primary site of hydroxylation in HIF-2␣, Pro-531, suggesting a critically important role for these residues. The secondary site of hydroxylation in HIF-2␣, Pro-405, is dissimilar in this region (Fig. 8A, top, compare HIF-2␣ Prim and HIF-2␣ Sec). Indeed, it differs at all three of the residues that are mutated in HIF2A-associated erythrocytosis patients (indicated by asterisks). To explore this further, we prepared 35 Slabeled, in vitro translated, HIF-2␣ Sec (residues 390 -423), corresponding to the secondary site of hydroxylation. We find that it binds more weakly to PHD2 than the corresponding sequence (residues 516 -549) from the primary site of hydroxylation (Fig. 8A, bottom, compare lanes 3 and 12). We next prepared a chimera containing the N-terminal residues from the primary site of hydroxylation and C-terminal residues from the secondary site of hydroxylation (Prim/Sec) and find that it too also binds more weakly (lane 6). Conversely, and in contrast, a peptide with the N-terminal residues from the secondary site of hydroxylation and C-terminal residues from the primary site of hydroxylation (Sec/Prim) binds with substantial affinity to PHD2 (lane 9).
In other experiments, we also compared the ability of Hypcontaining peptides corresponding to the primary and secondary sites of hydroxylation to bind to VHL (Fig. 8B, top). In this case, we used the previously described competition assay. As shown in Fig. 8B (bottom), a hydroxylated peptide containing the primary site of HIF-2␣ hydroxylation binds with high affinity to VHL, as reflected by its capacity to inhibit VHL binding to hydroxylated HIF-1␣-(556 -574) (compare lane 4 with lane 3). The secondary site of hydroxylation binds with substantially lower affinity to VHL (lane 7). Notably, a chimera containing the C-terminal residues from the primary site of hydroxylation and N-terminal residues from the secondary site binds with substantial affinity to VHL (lane 6), and more tightly than that of the chimera with the converse switch (lane 5). These observations therefore reinforce the importance of residues C-terminal to the hydroxylacceptor proline in HIF-2␣, both in the interaction with PHD2 and in the interaction with VHL.

DISCUSSION
After the initial identification and characterization of an erythocytosis-associated HIF2A mutation (15), we have subse- quently identified three additional distinct HIF2A mutations (16,19). The present characterization of these mutations provides new observations and insights into this disease.
First, all three mutations lead to decreased degradation and hence, stabilization of HIF-2␣. Consequently, all display a gain of function phenotype. Dominant negative mechanisms, such as, for example, increased mutant binding to PHD2 leading to impaired hydroxylation of the wild type HIF-2␣, or alternatively increased mutant binding to VHL leading to impaired degradation of wild type HIF-2␣, are not supported by the present data. None of the mutants was found to have increased affinity for either PHD2 or VHL, and indeed, in most cases there were significant decreases in affinity.
Second, whereas all mutations act to stabilize HIF-2␣, they do so in ways that are not identical. Thus, whereas the P534L and G537R mutations impair both binding to PHD2 and VHL, we find that the M535V mutation impairs only binding to PHD2. In the threedimensional structure of a Hyp HIF-1␣ peptide bound to VHL, HIF-1␣ Met-568 (which corresponds to HIF-2␣ Met-535) is not a contact residue with VHL, and its side chain points away from VHL (28,29). In contrast, Pro-567 (which corresponds to HIF-2␣ Pro-534) forms a van der Waals contact with VHL. Furthermore, Gly-537, which is not conserved in HIF-1␣, is just C-terminal to a residue that corresponds to HIF-1␣ Asp-569, which also makes a van der Waals contact with VHL (28,29).
This therefore suggests that the M535V mutation is pathogenic mainly through effects on PHD2 binding and PHD2-catalyzed hydroxylation. This in turn suggests that impairment of the PHD2/HIF-2␣ interaction alone is sufficient to induce erythrocytosis, and is consistent with the fact that mutations in PHD2 alone comprise a distinct cause of erythrocytosis (10,12,14,30,31). That impairment of the HIF-␣/VHL interaction alone is sufficient to induce erythrocytosis is independently supported by the existence of erythrocytosis-associated VHL mutations (10,13,20,(32)(33)(34). This, in turn, raises the possibility that there might exist yet to be identified HIF2A mutations that selectively impair the interaction of HIF-2␣ with VHL, but not with PHD2.
A third notable result of these studies is that all erythrocytosis-associated HIF2A mutations identified to date affect residues C-terminal to the hydroxylacceptor proline. This is somewhat of an unexpected finding, because much of the attention on sequence determinants in HIF-␣ initially focused on residues in a highly conserved LXXLAP motif (where underlining  indicates the hydroxylacceptor proline) that lie N-terminal to the hydroxylacceptor proline (35)(36)(37). Residues C-terminal to the hydroxyacceptor proline are substantially less well conserved, but these naturally occurring missense mutations now highlight the fact that they are indeed functionally critical. This is reinforced by the observations that chimeras between the primary and secondary sites of hydroxylation of HIF-2␣ bind with high affinity to PHD2 and to VHL in a manner that primarily tracks with the C-terminal, as opposed to N-terminal, residues (Fig. 8). It will in this regard be of interest to examine the three-dimensional structure of HIF-2␣ bound to PHD2.
Collectively, these studies reinforce the importance of prolyl hydroxylation as the key post-translational modification regulating the HIF pathway in response to changes in oxygen tension, reaffirm the role of the primary site of hydroxylation in HIF-2␣ in regulating its protein stability, and support the importance of HIF-2␣ in the control of EPO and hence erythropoiesis in humans.