Defective Tibetan PHD2 Binding to p23 Links High Altitude Adaption to Altered Oxygen Sensing*

Background: Tibetans have a genetic signature in the coding region of their PHD2 gene. Results: Tibetan PHD2 variant displays markedly impaired binding to the HSP90 cochaperone p23. Conclusion: Because p23 couples PHD2 to HIF-α hydroxylation, Tibetans possess a loss of function PHD2 allele. Significance: This study uncovers a mechanism for Tibetan adaptation to high altitude. The Tibetan population has adapted to the chronic hypoxia of high altitude. Tibetans bear a genetic signature in the prolyl hydroxylase domain protein 2 (PHD2/EGLN1) gene, which encodes for the central oxygen sensor of the hypoxia-inducible factor (HIF) pathway. Recent studies have focused attention on two nonsynonymous coding region substitutions, D4E and C127S, both of which are markedly enriched in the Tibetan population. These amino acids reside in a region of PHD2 that harbors a zinc finger, which we have previously discovered binds to a Pro-Xaa-Leu-Glu (PXLE) motif in the HSP90 cochaperone p23, thereby recruiting PHD2 to the HSP90 pathway to facilitate HIF-α hydroxylation. We herein report that the Tibetan PHD2 haplotype (D4E/C127S) strikingly diminishes the interaction of PHD2 with p23, resulting in impaired PHD2 down-regulation of the HIF pathway. The defective binding to p23 depends on both the D4E and C127S substitutions. We also identify a PXLE motif in HSP90 itself that can mediate binding to PHD2 but find that this interaction is maintained with the D4E/C127S PHD2 haplotype. We propose that the Tibetan PHD2 variant is a loss of function (hypomorphic) allele, leading to augmented HIF activation to facilitate adaptation to high altitude.

. These studies consistently identified genetic signatures in the EGLN1 (also known as prolyl hydroxylase domain protein 2, or PHD2) 2 and EPAS1 (also known as hypoxia-inducible factor-2␣, or HIF2A) genes in the Tibetan population (10).
PHD2 and HIF2A are compelling candidate genes in this population because of the key positions that they occupy in the HIF pathway, which is the central pathway for transducing changes in oxygen tension to changes in gene expression (11)(12)(13). Under normoxic conditions, PHD2 site-specifically prolyl hydroxylates the ␣ subunit of HIF (of which there are three paralogues: HIF1-␣, HIF-2␣, and HIF-3␣). In HIF-1␣, the primary site of hydroxylation is Pro-564; in HIF-2␣, it is Pro-531 (14). This post-translational modification allows recognition by the von Hippel-Lindau tumor suppressor protein (VHL), a component of an E3 ubiquitin ligase complex that then targets HIF-␣ for degradation by the ubiquitin-proteasome pathway. Under hypoxic conditions, prolyl hydroxylation is attenuated, leading to the stabilization of HIF-␣, its dimerization with HIF-␤, and activation of HIF target genes. These genes participate in a broad range of processes that promote systemic and cellular adaptation to hypoxia. HIF-1␣ and HIF-2␣ have overlapping as well as distinct gene targets. For example, HIF-1␣ activates many of the genes that encode for glycolytic enzymes, whereas HIF-2␣ is the principal regulator of the ERYTHRO-POIETIN (EPO) gene (13,14).
The current challenge is to identify the molecular basis for how Tibetan-associated haplotypes in the PHD2 and HIF2A genes promote adaptation to high altitude. With regard to the PHD2 gene, resequencing of this gene in 46 Tibetan individuals ruled out a classic de novo mutation sweep (15). Instead, this study reported on selection from standing variation. In particular, there is significant enrichment in the Tibetan population of two nonsynonymous coding region single-nucleotide polymorphisms, originally identified by others (16), that reside in exon 1 of the PHD2 gene: rs186996510 and rs12097901. The single-nucleotide polymorphisms are predicted to produce D4E and C127S amino acid substitutions, respectively, in the PHD2 protein.
Amino acid residues that are essential for the catalytic activity of PHD2 are encoded for largely by exons 2-5 of the PHD2 gene. Exon 1 (amino acids , in contrast, is most notable for a zinc finger of the MYND (myloid, Nervy, DEAF-1) type, which encompasses amino acids 21-58. This zinc finger shows strong phylogenetic conservation (17,18) and in fact is present in the PHD2 orthologue in the simplest metazoan, Trichoplax adhaerens (17). We have recently discovered that the function of this zinc finger is to couple PHD2 to the HSP90 pathway (19). Specifically, it binds to a Pro-Xaa-Leu-Glu (PXLE) motif that is present in select HSP90 cochaperones such as p23, and this allows recruitment of PHD2 to HSP90 to facilitate prolyl hydroxylation of HIF-1␣, which itself is an HSP90 client protein (20 -22).
The presence of the Tibetan-associated rs186996510 (D4E) and rs12097901 (C127S) single-nucleotide polymorphisms in exon 1 of the PHD2 gene led us to explore the possibility that this haplotype might affect PHD2 function, potentially through its zinc finger. Here, we provide evidence that the Tibetan D4E/ C127S haplotype impairs PHD2 function and that mechanistically, it leads to a dramatically decreased ability of the PHD2 zinc finger to associate with the HSP90 cochaperone p23. Remarkably, this impairment depends on both the D4E and C127S amino acid substitutions. On the basis of these results, we conclude that the Tibetan PHD2 D4E/C127S haplotype leads to a loss of function (hypomorphic) PHD2 allele, and we propose that this is a key mechanism by which the HIF pathway has been reconfigured for adaptation to chronic hypoxia in this population.
pcDNA3-HA-PHD2 (1-131) C127S was constructed by amplifying by PCR a 0.4-kb band using pcDNA3-HA-PHD2 as a template and the following primers: CMV 5Ј, 5Ј-GCG TGT ACG GTG GGA GGT C-3Ј; and PHD2 C127S 3Ј (see above). The product was digested with BamHI and NotI and subcloned into the BamHI/NotI site of pcDNA3-HA to give pcDNA3-HA-PHD2 pcDNA3-Flag-p23 W106A was constructed by overlapping PCR. In the first round of PCR, we performed two PCRs. We amplified a 0.32-kb band using pcDNA5/FRT/TO-3ϫFlag-p23 as a template and the following primers: CMV 5Ј, 5Ј-GCG TGT ACG GTG GGA GGT C-3Ј; and P23 W106A 3Ј, 5Ј-GAA TCA TCT TCC CAG TCT TTC GCA TTA TTG AAA TCG ACA CTA AGC CAA TTA AG-3Ј. We amplified a 0.2-kb band using pcDNA5/FRT/TO-3ϫFlag-p23 as a template and the following primers: P23 W106A 5Ј, 5Ј-CTT AAT TGG CTT AGT GTC GAT TTC AAT AAT GCG AAA GAC TGG GAA GAT GAT TC-3Ј; and BGH rev, 5Ј-TAG AAG GCA CAG TCG AGG-3Ј. The two PCR products were mixed and employed as template in a second round of PCR using the CMV 5Ј and BGH rev primers. The product was digested with BamHI and XhoI and subcloned into the BamHI/XhoI site of pcDNA5/FRT/ TO-3ϫFlag.
Mass Spectrometry-WT and D4E/C127S HT-FlagPHD2 purified from baculovirus-infected Sf9 cells were subjected to filter-aided sample preparation (27). Mass spectrometry was then performed as described (19) with the following modifications: fragmentation spectra (MS2s) were selected from within a reduced parent mass scan range (600 -800 m/z) and using a targeted parent mass list designed to selectively acquire spectra for the indicated peptides. Masses selected for fragmentation included both fully and semitryptic peptides in addition to mass shifts associated with zero, one, or two serine phosphorylation events (ϩ79.966 Da). The data were analyzed using the Max-Quant analytical package (28) version 1.2.2.5 and included N-terminal acetylation, oxidation of methionine, and phosphorylation of serine (with potential accompanying neutral losses) as variable modifications and carbamidomethylation of cysteine as a fixed modification.
Prolyl Hydroxylase Assays-Recombinant WT or D4E/ C127S HT-FlagPHD2 was incubated with substrate in a buffer consisting of 20 mM Hepes, pH 8.0, 100 mM KCl, 10 M FeCl 2 , 1 mM 2-oxoglutarate, 5 mM ascorbate, 1 mM DTT, and 1 M ZnCl 2 at 37°C. For hypoxic conditions, the reactions were conducted in either a Ruskinn In Vivo 200 hypoxia work station or a Billups Rothenberg modular incubator perfused with a gas mixture containing the appropriate oxygen concentration. The products were subjected to SDS-PAGE, transferred to Immobilon membranes, blocked with PBS containing 0.5% Tween 20, and 1% bovine serum albumin for 30 min, and then incubated with recombinant HT-HA-VHL (in complex with ElonginB and ElonginC) for 1 h in the same buffer at 4°C. Following washing, HT-HA-VHL was visualized using anti-HA antibodies conjugated to alkaline phosphatase.
Peptides (0.5 g) were prebound to 15-l aliquots of streptavidin-agarose (Sigma). The resins were incubated with Sf9 lysates containing baculovirus-expressed HT-FlagPHD2 (23) for 1 h with rocking at 4°C in buffer A (20 mM Tris, pH 7.6, 150 mM NaCl, 10% glycerol, 1% Triton X-100). The resins were washed four times with buffer A and eluted, and the eluates were subjected to SDS-PAGE and Western blotting.
Cell Culture-HEK293 FT cells were maintained and transfected with plasmids as described (19). Luciferase assays were performed as described (19). Hypoxia experiments were performed in an In Vivo 200 hypoxia work station (Ruskinn Technologies).
Mouse Embryonic Fibroblast (MEF) Generation-Phd2 ϩ/Ϫ mice (29) were crossed, and MEFs were obtained from embryonic day 11.5 Phd2 Ϫ/Ϫ embryos by the primary explant technique (30). MEFs were immortalized by transfection with pSG5-T, which contains the coding sequence for the SV40 large T antigen (31), and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml of streptomycin. Wild type MEFs that were immortalized in the same manner were a gift from Mr. Scott Sutherland.
Immunoprecipitations-For purified recombinant proteins, proteins were mixed in buffer B (50 mM Tris, pH 7.5, 100 mM NaCl, 0.5% Triton X-100, and 1 M ZnCl 2 ) supplemented with 1% bovine serum albumin for 1 h at 4°C. In some cases lysates from 293 FT cells were included. The reactions were then added to 15-l aliquots of anti-Flag (M2)-agarose and rocked for 1 h at 4°C. The resins were washed three times with buffer B and eluted, and the eluates were subjected to SDS-PAGE and Western blotting.
For proteins expressed by transient transfection, cells were lysed in buffer B. The lysates were clarified by centrifugation at 15,800 ϫ g for 10 min at 4°C. The lysates were then added to 15-l aliquots of anti-Flag (M2)-agarose and rocked for 1 h at 4°C. The resins were washed three times with buffer B and eluted, and the eluates were subjected to SDS-PAGE and Western blotting. In the experiments in which FlagPHD2 was immunoprecipitated from stably transfected MEFs, the lysates were precleared by incubation with Sepharose CL-4B (Sigma) prior to incubation with anti-Flag agarose. The results presented are representative of two to four independent experiments.
Western Blotting-The sources of antibodies to p23, FKBP38, Flag tag, HA tag, and mouse Hif-1␣, as well as the procedure for Western blotting, have been described (19,29). Rabbit mAb D31E11 to PHD2 was obtained from Cell Signaling Technology. The E7 monoclonal antibody against ␤-tubulin developed by Dr. Michael Klymkowsky was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD, National Institutes of Health and maintained by the University of Iowa Department of Biology (Iowa City, IA). Protein concentrations of extracts were determined by the Bio-Rad DC protein assay.
Statistical Analysis-Unpaired Student's t test was employed for statistical analysis. Differences were considered significant when p Ͻ 0.05.

RESULTS
We first transfected HEK293 FT cells with constructs for WT or D4E/C127S PHD2 along with a luciferase reporter gene driven by three copies of a hypoxia response element and maintained cells under normoxia or hypoxia (1% O 2 ). As shown in Fig. 1A, we find that hypoxia activates the reporter gene and that overexpression of WT PHD2 inhibits this activation (seventh and ninth columns). Importantly, we observe that D4E/ C127S PHD2 is weaker in its ability to inhibit hypoxia response element activity (eighth and tenth columns).
We generated Phd2 Ϫ/Ϫ MEFs, then stably transfected them with constructs for WT or D4E/C127S PHD2, and then selected two independent clones for each with comparable PHD2 expression levels (Fig. 1B). When exposed to either normoxia or hypoxia (Fig. 1, C and D, respectively), we observe that Hif-1␣ levels are higher in the D4E/C127S PHD2 expressing MEFs as compared with those expressing WT PHD2 (Fig. 1, C  and D top panel, compare lanes 2 and 3 with lanes 4 and 5).
These experiments raise the possibility that D4E/C127S PHD2 might have intrinsically weaker catalytic activity than WT PHD2. To examine this, we prepared recombinant proteins and conducted in vitro prolyl hydroxylase assays. We employed a VHL probe-based far Western analysis, because VHL binds with extremely high specificity to the prolyl hydroxylated form of HIF-␣ (32,33). Under a range of oxygen concentrations, including 1, 3, 5, and 21%, we have not seen any consistent differences between WT and D4E/C127S PHD2 in their capacity to hydroxylate GST-HIF-1␣ (531-575), GST-HIF-2␣ (516 -549), or full-length HIF-1␣ (data not shown).
Both Asp-4 and Cys-127 reside in exon 1, the most prominent feature of which is a MYND-type zinc finger. This zinc finger binds to a PXLE motif that is found in select HSP90 cochaperones, including p23 and FKBP38 (19). We therefore examined WT and D4E/C127S PHD2 for their capacity to interact with each protein. We first incubated recombinant PHD2 with recombinant p23 or FKBP38, immunoprecipitated the PHD2, and examined the immunoprecipitates. We find that WT PHD2 interacts directly with both p23 and FKBP38; however, we do not observe any significant difference between WT and D4E/C127S PHD2 in their ability to interact with either protein (data not shown).
Recognizing that the interaction between purified proteins may not recapitulate their interactions within the cell, we next performed a series of transfection and coimmunoprecipitation studies. We coexpressed either HA-tagged WT or D4E/C127S PHD2 with Flag-p23, immunoprecipitated the latter, and examined the immunoprecipitates for the absence or presence of PHD2. In striking contrast to the results with purified recombinant proteins, we find that D4E/C127S PHD2 is defective in its ability to coimmunoprecipitate with p23 ( Fig. 2A, top panel,  compare lanes 2 and 4).
When parallel transfection experiments were performed with FKBP38, we find that D4E/C127S PHD2 is only marginally decreased in its interaction as compared with WT (Fig. 2B). We examined the D4E and C127S substitutions individually. Remarkably, we find that both of the individual substitutions preserve the ability of PHD2 to coimmunoprecipitate with p23 ( Fig. 2C, top panel, lanes 2, 4, and 6) and that only the combined D4E/C127S substitution abolishes it (Fig. 2C, top panel, lane 8).
It is conceivable that the difference in the results of Fig. 2A and those using purified recombinant proteins may be due to the fact that p23 can exist in a complex with HSP90 within the cell and that the p23⅐HSP90 complex might interact differently with WT as compared with D4E/C127S PHD2. Previous studies showed that a W106A mutant of p23 is defective in its interaction with HSP90 (34). We therefore tested the role of HSP90mediated binding by examining the ability of W106A p23 to coimmunoprecipitate PHD2. We find that this mutant p23 can coimmunoprecipitate PHD2 in a manner comparable with WT p23 (Fig. 2D, top panel, lanes 2 and 5). Significantly, the interaction is abolished by the D4E/C127S substitution (Fig. 2D, top  panel, lanes 3 and 6). Therefore, the differential interaction of WT versus D4E/C127S PHD2 with p23 is not dependent on the association of p23 with HSP90.
These results further support the notion that the D4E/C127S substitution impairs function of the zinc finger of PHD2. We considered whether the D4E/C127S substitution might also affect interaction with a different PXLE-containing protein. Toward this end, we searched for such a motif and identified one in HSP90 itself (both ␣ and ␤ isoforms; Fig. 3A). As in p23 and FKBP38 (19), the HSP90 PXLE motif is immediately preceded by a hydrophobic residue and is in close vicinity to acidic amino acids that reside N-terminal to the motif. The PXLE motif is conserved in HSP90 from mammalian, bird, amphibian, and fish species, as well as that from the simple metazoan T. adhaerens. In a structure for human HSP90␣ (293-732) determined by x-ray crystallography, the C-terminal 33 amino acids were not visualized, likely because of high flexibility in this region (35). The last five amino acids of HSP90, MEEVD, are well known to bind to tetratricopeptide repeat domain containing cochaparones such as FKBP51, FKBP52, and FKBP38 (36). To our knowledge, a function for the PXLE motif immediately N-terminal to this MEEVD motif has not yet been assigned.
To determine whether this motif in HSP90␤ (hereafter referred to as HSP90) can mediate binding to PHD2 in the context of full-length HSP90, we prepared recombinant HSP90 from baculovirus-infected Sf9 cells and incubated it with recombinant FlagPHD2. The latter was then immunoprecipitated with anti-Flag antibodies, and the eluates were examined. As shown in Fig. 3C, we observe evidence for a specific interaction between the proteins (compare lanes 2 and 3). Importantly, a mutation of the PXLE motif to AXAA abolishes it (lane 6). We then incubated recombinant PHD2 with recombinant HSP90, immunoprecipitated the PHD2, and examined the immunoprecipitates. As expected, WT PHD2 interacts directly with HSP90 (Fig. 3D, top panel, lane 2). However, we do not observe any appreciable difference between WT and D4E/C127S PHD2 in their ability to interact with HSP90 (lane 3).
We next incubated recombinant Flag-tagged WT or D4E/ C127S with HEK293 FT cell lysates, immunoprecipitated the PHD2 with anti-Flag antibodies, and then examined the immunoprecipitates for the absence or presence of endogenous p23 from the lysates. Interestingly, recombinant D4E/C127S PHD2 is substantially weakened in its ability to coimmunoprecipitate p23 (Fig. 3E, top panel, compare lane 2 and 3). We extended these studies by incubating purified Flag-tagged WT or D4E/ C127S PHD2 with purified HA-tagged p23 in the presence of HEK293 FT cell lysates, immunoprecipitating the PHD2 with anti-Flag antibodies, and then examining the immunoprecipitates for the recombinant HA-tagged p23. We find that the PHD2-p23 interaction is recapitulated, and importantly, this interaction is diminished by the D4E/C127S haplotype (Fig. 3F,  top panel, lanes 2 and 3). Thus, the defective interaction between recombinant D4E/C127S PHD2 and recombinant p23 is observed in the presence of cellular lysate (Fig. 3F), but not in its absence (data not shown).
The aforementioned studies also provide an opportunity to reexamine whether the D4E/C127S substitution affects the ability of PHD2 to interact with the PXLE motif of HSP90. We incubated purified Flag-tagged WT or D4E/C127S PHD2 with purified HA-tagged HSP90 in the presence of HEK293 FT cell lysates, immunoprecipitated the PHD2 with anti-Flag antibodies, and then examined the immunoprecipitates for recombinant HSP90. We observed evidence for the PHD2-HSP90 interaction but, in contrast to results with p23, this interaction is not appreciably affected by the D4E/C127S haplotype (data not shown). From these experiments, we conclude that the D4E/ C127S PHD2 haplotype produces a selectively detrimental defect in p23 interaction.

DISCUSSION
We have previously shown that p23 knockdown augments hypoxia-induced HIF-␣ protein levels and HIF target genes (19), indicating that p23 recruitment to the HSP90 pathway facilitates HIF-␣ hydroxylation. We now report that the Tibetan PHD2 haplotype disrupts this linkage to p23, leading directly to a model in which the Tibetan variant is impaired in its ability to prolyl hydroxylate HIF-␣ and hence is a loss of function (hypomorphic) allele (Fig. 4D). This might seem counterintuitive. Low altitude inhabitants who ascend to high altitude (Fig. 4D, middle panel) are predisposed to developing both pulmonary hypertension and erythrocytosis (Monge's disease). A loss of function PHD2 allele would be expected to potentiate HIF-␣ signaling and thereby exacerbate both.
However, genetic adaptation can result from selection at more than one loci, and multiple studies indicate that Tibetan genetic selection has occurred not only at the PHD2 locus, but also at the HIF2A locus (3,(5)(6)(7)(8)(9). Importantly, two of these studies (3,5) demonstrated both a strong association between the Tibetan HIF2A haplotype and low hemoglobin concentration (which is controlled by the HIF-2␣ target gene, EPO (37)(38)(39)(40)), thereby providing compelling evidence that the HIF2A allele is a loss of function allele. Because HIF-2␣ also plays a central role in pulmonary hypertension (41)(42)(43)(44), the loss of function HIF2A allele would be predicted to blunt the erythrocytosis and pulmonary hypertension that might otherwise result from loss of PHD2 function (Fig. 4D, right panel).
Our model could also allow for potentially beneficial effects of augmented HIF-1␣ activation, such as increased respiration and nitric oxide production, because HIF-1␣ plays a role in promoting both (45)(46)(47)(48). We therefore propose that the Tibetan population has enriched specific haplotypes of the PHD2 and HIF2A genes to reconfigure the HIF pathway to facilitate high altitude adaptation.  were incubated with the indicated biotinylated peptides immobilized on streptavidin-agarose and washed, and the eluates were examined for the presence of PHD2 using anti-Flag antibodies. Input, 1% of total. C, recombinant WT or P709A/L711A/E712A (AAA) HT-HA-HSP90␤ (2 g) was incubated with or without HT-FlagPHD2 (2 g), and the latter was then immunoprecipitated using anti-Flag antibodies. The immunoprecipitates were examined for the presence of HSP90 as shown. Input represents 2% of total. D, recombinant HT-HA-HSP90 (1 g) was incubated without or with WT or D4E/C127S HT-FlagPHD2 (1 g), and the latter was then immunoprecipitated using anti-Flag antibodies. The immunoprecipitates were examined for the presence of HSP90 as shown. E, HEK293 FT lysates were incubated without or with recombinant WT or D4E/C127S HT-FlagPHD2 (1 g), and the latter was then immunoprecipitated using anti-Flag antibodies. The immunoprecipitates and aliquots of lysate were subjected to Western blotting as indicated. F, recombinant HT-HA-p23 (1 g) was incubated without or with either WT or D4E/C127S HT-FlagPHD2 (1 g) in the presence of HEK293 FT lysate (100 g), and the HT-FlagPHD2 was then immunoprecipitated using anti-Flag antibodies. The immunoprecipitates and aliquots of lysate were examined for the presence of HA-p23 as shown. IP, immunoprecipitation; WB, Western blotting.
We find a dramatic effect of the D4E/C127S haplotype on PHD2 interaction with p23. Although it was far more significant than the effect seen on PHD2 interaction with either FKBP38 or HSP90, we cannot at present rule out the possibility that subtle alterations in these other interactions may also contribute to the phenotype. Interestingly, the impaired interaction with p23 is seen in the presence of cellular lysates, whether the proteins are transiently expressed, stably expressed, or prepared as recombinant proteins (Figs. 2; 3, E and F; and 4). The impaired interaction is not seen in the absence of lysate (data not shown). This suggests that there may be a conformational change in D4E/C127S versus WT PHD2 in the presence of lysate, a cofactor(s) in the lysate that modulates the PHD2-p23 interaction, or a post-translational modification of PHD2 that is differentially present in D4E/C127S versus WT PHD2. With regard to the latter, the C127S substitution provides a potential phosphoacceptor residue; however, we have thus far not observed phosphorylation at this residue by mass spectrometry (data not shown). The various possibilities mentioned above will require further investigation, as will the possibility that p23 may play a role in the maturation of HIF-␣ itself. Regardless, the present studies provide genetic evidence for the functional importance of the PHD2-p23 interaction. Moreover, they raise the possibility that therapeutic targeting of this interaction may recapitulate key aspects of adaptation to chronic hypoxia, which is a central feature of common diseases that include ischemic heart, cerebrovascular, and chronic obstructive pulmonary disease.  Right panel, the Tibetan PHD2 is impaired relative to WT PHD2 in its ability to hydroxylate HIF-␣ (indicated by one arrow), leading to an exaggerated HIF-1␣ response. However, the Tibetan HIF2A allele likely leads to a blunted HIF-2␣ response, thereby leading to differential activation of HIF-1␣ versus HIF-2␣.