Oxygen-dependent Regulation of Erythropoietin Receptor Turnover and Signaling*

von Hippel-Lindau (VHL) disease is a rare familial cancer predisposition syndrome caused by a loss or mutation in a single gene, VHL, but it exhibits a wide phenotypic variability that can be categorized into distinct subtypes. The phenotypic variability has been largely argued to be attributable to the extent of deregulation of the α subunit of hypoxia-inducible factor α, a well established target of VHL E3 ubiquitin ligase, ECV (Elongins/Cul2/VHL). Here, we show that erythropoietin receptor (EPOR) is hydroxylated on proline 419 and 426 via prolyl hydroxylase 3. EPOR hydroxylation is required for binding to the β domain of VHL and polyubiquitylation via ECV, leading to increased EPOR turnover. In addition, several type-specific VHL disease-causing mutants, including those that have retained proper binding and regulation of hypoxia-inducible factor α, showed a severe defect in binding prolyl hydroxylated EPOR peptides. These results identify EPOR as the second bona fide hydroxylation-dependent substrate of VHL that potentially influences oxygen homeostasis and contributes to the complex genotype-phenotype correlation in VHL disease.

von Hippel-Lindau (VHL) 2 disease is a rare familial cancer predisposition syndrome characterized by the development of tumors in several organ systems, including the central nervous system and retinal hemangioblastoma (HAB), which represent the two cardinal features of VHL disease, as well as endolymphatic sac tumor of the inner ear, pancreatic cystadenoma, benign cystadenoma of epididymis, adnexal papillary cystadenoma of probable mesonephric origin, clear-cell renal cell carcinoma (RCC), paragangliomas, and pheochromocytoma (PHEO) (1). Some VHL patients also develop polycythemia with or without the other stigmata of VHL disease. VHL disease is subcategorized into type 1, characterized by HAB and RCC without PHEO; type 2A, characterized by PHEO and HAB without RCC; type 2B, characterized by PHEO and HAB with RCC; type 2C, characterized by exclusive development of PHEO; and type 3, which is characterized by the development of polycythemia without an increased cancer predisposition (2). Notably, despite the phenotypic heterogeneity, VHL disease is caused by a mutation or loss of a single gene, VHL.
VHL is the substrate recognizing subunit of a multiprotein E3 ubiquitin ligase ECV (Elongins BC/Cul2/VHL) that targets the ␣ subunit of hypoxia-inducible factor (HIF␣) for ubiquitylation under normal oxygen tension (3). HIF␣ is hydroxylated on conserved prolines within the oxygen-dependent degradation domain (ODD) by a family of prolyl hydroxylase enzymes (PHD1-3) in the presence of 2-oxoglutarate, iron, and oxygen (4 -7). VHL binds exclusively to prolyl hydroxylated HIF␣. Thus, under hypoxia, the unmodified HIF␣ escapes the recognition by VHL and heterodimerizes with the constitutively expressed HIF␤ to form an active transcription factor, which transactivates numerous hypoxia-inducible genes involved in various adaptive processes to hypoxia such as anaerobic metabolism, angiogenesis, and erythropoiesis (8).
Considering the wide phenotypic variability in VHL disease, a prediction was that ECV had more than one substrate. Over the past few years, several proteins have been reported to be substrates of ECV or VHL-associated proteins, such as RBP1 (9), aPKC (10), SPRY2 (11), and ␤2AR (12). However, the biological or pathophysiological significance in the context of VHL disease of such interactions has remained largely unclear. In contrast, studies have suggested that HIF␣ deregulation upon VHL mutation or loss contributes to the development of VHL disease in particular HAB and RCC but did not seem to be involved in the development of PHEO (13,14). Intriguingly, binding to extracellular matrix protein fibronectin was lost in all disease-causing VHL mutants tested to date (15). However, the biological significance of this lost interaction or its contribution, if any, to genotype-phenotype correlation in VHL disease remains unanswered.
Erythropoietin (EPO) is a glycoprotein hormone whose gene is transactivated by HIF2␣ (16 -21). EPO binds to EPO receptor (EPOR) expressed on erythroid progenitor cells to initiate JAK2-STAT5 signal transduction cascade that promotes a surge of proliferation followed by terminal differentiation into mature oxygen-carrying red blood cells (22,23). Truncation mutations in EPOR that lead to a failure in negative regulation or JAK2 mutations that cause increased activity are often found in primary form of polycythemia, which are instigated by an exaggerated response to EPO stimulus (24 -26). In this regard, it is noteworthy that primary polycythemia patients often have a reduced serum EPO level.
Mutations in HIF2␣ have been identified in the vicinity of its conserved prolyl residue, Pro 531 , in several polycythemia patients (27). Some of these mutations cause a reduction in binding to and catalysis by PHD enzymes, as well as a defect in binding to VHL (28). Mutations in PHD2 have also been identified, which cause erythrocytosis in a HIF2␣-dependent manner (27,29). These patients have elevated or higher than the expected serum EPO levels. Consistent with these observations, PHD2 knock-out mice develop secondary polycythemia with high serum EPO level (30,31). In contrast, PHD1 and PHD3 double knock-out mice develop polycythemia but curiously without an increased serum EPO level (30), suggesting a yet undefined HIF2␣-independent mechanism. Notably, anemia in rats has been shown to increase the number of cell surface EPOR (32).
Here, we show for the first time that EPOR turnover is regulated by oxygen tension. The conserved prolyl residues within the cytoplasmic region of EPOR become hydroxylated by PHD3, a process that is strictly dependent on oxygen, and targeted for ubiquitylation via ECV. Molecular knockdown or knock-out of PHD3 or VHL promoted EPOR expression level and EPO-dependent downstream signaling. Moreover, myeloid colony formation and signal transduction were accentuated by hypoxia. Notably, several type-specific VHL mutants, some of which retained proper binding and regulation of HIF␣, showed a profound defect in binding synthetically hydroxylated EPOR peptide. These findings reveal EPOR as a potential substrate of VHL tumor suppressor complex that may contribute to the phenotypic spectrum of VHL disease.
Chemicals and Reagents-Dimethyloxalylglycine was purchased from Frontier Scientific. Cycloheximide (C4859) and cobalt chloride were obtained from Sigma. MG132 (IZL-3175-v) was obtained from Peptides International. Streptavidin-agarose resin was obtained from Thermo Scientific. HAubiquitin was purchased from Boston Biochem. The cell surface biotinylation was performed using EZ-Link Sulfo-NHS-LC-Biotin (Thermo) in accordance with the manufacturer's instructions.
Lentiviral Production and Infection of Cell Lines-HEK293T cells were transfected with psPAX2, pMDG1.vsvg, and either a pGIPZ, pLenti-CMV-GFP-Hygro, or pLentiCRISPR transfer vector. Lentivirus containing supernatant was collected at 48 and 72 h post-transfection. Lentiviral supernatant was filtered and applied to the indicated cell lines. UT-7 cells required the addition of 4 g/ml Polybrene (Millipore) and GM-CSF at a final concentration of 5 ng/ml. Selection was started 24 h after infection using puromycin (2 g/ml) or hygromycin (250 g/ml) (both from Wisent). Note that polyclonal populations were generated and used for experiments, except for knock-out cells for which monoclonal cell lines were generated.
Protein Purification-pGEX-4T-1-PHD3 (or empty pGEX-4T-1 for GST control protein) was expressed in BL21 Escherichia coli and induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside (Sigma) overnight. Bacterial pellets were lysed in PBS supplemented with protease inhibitors (Roche), sonicated, cleared, and bound to glutathione-Sepharose (GE Healthcare). The beads were washed and GST-PHD3 was eluted using GST elution buffer (50 mM Tris, pH 8, 30 mM glutathione). pET15-huCytoEPOR was expressed in Rosetta-2 and induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside overnight. Bacterial pellets were lysed in 50 mM Tris, pH 8, 300 mM NaCl, 20 mM imidazole supplemented with protease inhibitors; sonicated; cleared; and bound to Ni-NTA-agarose (Invitrogen). Beads were washed and eluted using an imidazole gradient. VHL 19 was co-expressed with EloB/C in E. coli BL21 (DE3). The cell cultures were grown to A 600 ϭ 0.6 and induced with a final concentration of 1 mM isopropyl-␤-D-thiogalactopyranoside for 5 h at 37°C. The cells were subsequently resuspended in Buffer A (5 mM HEPES, pH 7.4, 200 mM NaCl, and 1 mM DTT) supplemented with 1ϫ Complete EDTA-free protease inhibitor (Roche) and lysed using a Constant Systems cell disruption system (30 kpsi). The lysate was then briefly sonicated on ice for 30 s, and the cellular debris was removed by centrifugation at 34,000 ϫ g for 40 min prior to loading onto a glutathione-Sepharose column (GE Healthcare). The VHL-EloB/C complex was eluted in Buffer A with 10 mM reduced glutathione and treated with thrombin (Millipore; 1 unit/mg of protein) at 4°C overnight to remove the GST fusion protein. The complex was dialyzed overnight in buffer A prior to purification on a preparation grade Superdex 200 10/300 column. The peak corresponding to the 1:1:1 VHL-EloB/C complex (44.1 kDa) was collected.
Biolayer Interferometry-Biotinylated EPOR and HIF1␣ ODD peptides were synthesized (Genscript), diluted to 100 g/ml in BLI kinetics buffer (buffer A, 0.5% (w/v) BSA and 0.02% (v/v) Tween 20), and immobilized over 120 s onto a streptavidin-coated BLI biosensor (Pall Corp). The purified 1:1:1 VHL-EloB/C ternary complex was diluted in BLI kinetics buffer over various protein concentrations, and its association to the immobilized peptide was measured over 120 s. Subsequently, the biosensor was immersed into BLI kinetics buffer over 180 s to dissociate the VHL-EloB/C-peptide complex. The kinetic parameters (K d , k a , and k d ) were calculated from the sensorgrams using the BLItz Pro software (version 1.1.0.16).
In Vitro Binding Assays-Purified proteins were incubated in a final reaction volume of 30 l of EBC at 30°C. After 1 h of incubation, 1 ml of EBC buffer was added along with Ni-NTA, and the mixture was rocked at 4°C for 1 h. Bound proteins were washed five times prior to elution by boiling in sample buffer. Proteins were visualized by staining with GelCode (Thermo). Binding assays using biotinylated peptides were performed as previously described (37) with the following modifications: methionine was not radiolabeled, so immunoblotting was used to detect rabbit reticulocyte in vitro translated (Promega) proteins; EBC buffer was supplemented with 1 mM DTT; washes were conducted in NETN containing 300 mM NaCl instead of 100 mM NaCl.
In Vitro Hydroxylation-In vitro hydroxylation assays were performed essentially as previously described (37). 1 g of the indicated peptide was attached onto 30 l of streptavidin-agarose slurry. The final reaction volume was 200 l, consisting of 1ϫ PHA (40 mM HEPES, pH 7.4, 80 mM KCl), 100 M FeCl 2 , 2 mM ascorbate, 5 mM 2-oxoglutarate, and 10 l of programmed reticulocyte lysate. The reaction was incubated at room temperature for 2 h. Where purified components were used, 2 g of His-CytoEPOR was used and 1 g of GST-PHD3.
Competition Assay-HEK293 cells were transfected with VHL and substrate. The complex was affinity-purified and separated into four tubes each with an equivalent quantity of beads, the indicated amount of peptide (0 -6 g), and EBC lysis buffer up to 100 l. The tubes were incubated at 30°C for 30 min with mixing every 5 min. Next, 1 ml of EBC was added, and the tubes were rocked at 4°C for 1 h. The beads were washed in NETN, and proteins were eluted by boiling in sample buffer.
Ubiquitylation Assays-In vitro ubiquitylation assay was performed using S100 fractions as described previously (3). For in vivo ubiquitylation, UT-7 cells were treated with 10 M MG132 for 4 h. The cells were lysed in EBC lysis buffer supplemented with 2% SDS and boiled for 15 min to denature proteins and eliminate protein interactions. The supernatant was collected and diluted so the final concentration of SDS would be less than 0.1%, and immunoprecipitation was performed.
Cytokine Stimulation of Cells-Prior to experiments involving cytokine treatment, cells were washed extensively in plain medium and then resuspended in base medium supplemented with 10% fetal bovine serum. The cells were maintained in cytokine free medium overnight. The cells were treated with the indicated cytokine for the indicated period of time. In experiments involving a short dose of EPO treatment to UT-7 cells, plain ␣-MEM was added to a Falcon tube containing EPO and an equivalent volume of medium as the cells that were starved. The cells were poured into the Falcon tube and inverted to eliminate the slow rate at which diffusion would have occurred.
In Vitro Differentiation of UT-7 Cells-The cells were washed extensively in plain ␣-MEM and starved overnight in ␣-MEM supplemented with 20% fetal bovine serum and puromycin and without GM-CSF. The cells were then grown at 7% oxygen tension for 9 days in ␣-MEM supplemented with 20% fetal bovine serum, puromycin and 1 unit/ml EPO. On the last day, cells were washed in PBS twice and then lysed by vortexing in 3 volumes of lysis solution H (50 mM Tris, pH 7, 25 mM KCl, 5 mM MgCl 2 , 1 mM ␤-mercaptoethanol, 0.3% Triton X-100) similar to as previously described (38) and centrifuged on a tabletop centrifuge at 4°C for 10 min at maximum speed. The supernatant was collected and placed on ice. The total protein concentration was determined using a Bradford assay. The quantity of hemoglobin was determined using a hemoglobin assay kit (Abnova). 50 l of the above sample was incubated with 200 l of the reagent provided for 5 min, and the absorbance at 400 nm was read. After determining the concentration of hemoglobin in the sample, the values were normalized to total protein level. The experiment was performed in triplicate or quadruplicate three independent times. UT-7-pGIPZ control cells were arbitrarily set to 10.
Colony Forming Assay-UT-7 cells were cultured in triplicate in Methocult H4230 (Stemcell Technologies) supplemented with 10% heat-inactivated premium FBS (Wisent), ␣-MEM and EPO where indicated. The cells were maintained in normoxia (21% oxygen) or hypoxia (1% oxygen) for 10 days. Colonies of greater than 30 cells were scored. The experiment was performed three independent times.
Mass Spectrometry-In vitro hydroxylated His-huCytoEPOR was reduced, alkylated, and subjected to proteolysis with LysC for 16 h, followed by V8 for an additional 8 h. The resulting peptides were lyophilyzed and resuspended in 0.1% formic acid. The samples were analyzed on an Orbitrap Velos mass spectrometer in data-dependent mode, where each full MS scan (at 60,000 resolution in the Orbitrap) was followed by up to 20 collision-induced dissociations in the ion trap. The hydroxylated peptides were identified by multiple database search algorithms (Peaks score 40.67, Sequest XCorr 1.27, deltaCN 0.11, MS-Amanda score 33.18, PeptideProphet confidence value of 0.94), with a parent mass error of Ϫ1.4 ppm.
Immunofluorescence Microscopy-Cells were grown on coverslips, and 48 h post-transfection they were washed and fixed in cold 100% methanol at Ϫ20°C for 30 min. After an hour of blocking with 1% normal goat serum in PBS, the cells were then incubated for 1 h (at room temperature) with antibodies recognizing calreticulin (Abcam) and FLAG. The cultures were washed and then incubated for an additional 1 h with fluorescein-conjugated goat anti-rabbit FITC (green) and goat antimouse rhodamine (red) secondary antibodies. The nuclei were stained with DAPI (blue). Coverslips were mounted on slides with VectaShield mounting medium (Vector Labs). Immunostained cells were visualized with an Axioplan2 imaging microscope (Carl Zeiss) and imaged with an Axiocam HRM digital camera.
Statistical Analysis-The Results are presented as the mean values Ϯ the standard deviation. A Student's t test was used to determine the level of statistical significance achieved. Statistical analysis was performed using GraphPad Prism 5.0 software.

Results
Oxygen Promotes EPOR Turnover-We asked whether oxygen influences the EPO-dependent growth of human erythroleukemia UT-7 cells that express endogenous EPOR. The cells were plated for a colony forming assay in semisolid medium containing increasing amounts of EPO and maintained in normoxia or hypoxia. Increased colony formation was observed under hypoxia relative to normoxia (Fig. 1A). After validating and optimizing the anti-EPOR antibody (Fig. 1B), we asked whether oxygen influences EPOR expression and downstream signaling. UT-7 cells were cytokine-starved and maintained in a gradient of oxygen tension, and then the cells were briefly treated with or without constant concentration of exogenous EPO. Interestingly, oxygen tension was inversely correlated to total EPOR level, whereas the level of total JAK2 remained unchanged (Fig. 1C) (note that EPOR presents as variable multiple bands because of glycosylation). As expected, EPO treatment promoted EPOR and downstream JAK2-Tyr 1007 phosphorylation, but notably, the extent of phosphorylation was markedly elevated under reduced oxygen tension (Fig. 1C). However, the level of EPOR mRNA remained unchanged irrespective of oxygen tension (Fig. 1, D and E). The maintenance of mouse embryonic fibroblasts (MEFs) that ectopically express murine EPOR-FLAG and murine pro-B Ba/F3 cells that stably express ectopic EPOR under hypoxia likewise increased EPOR level (Fig. 1, F and G).
EPOR has previously been reported to undergo endoplasmic reticulum-associated degradation (39). We asked whether the oxygen-dependent regulation of EPOR expression was direct or indirect via endoplasmic reticulum-associated degradation by measuring the steady-state expression level of human (hu) and murine (mu) cytoplasmic EPOR lacking the N-terminal segment containing the ER-targeting sequences in the presence or absence of hypoxia treatment or mimetics. As expected, hu/muCytoEPOR was excluded from the ER as noted by nonoverlapping staining pattern with calreticulin and localized in the cytosol (Fig. 1H), and its level was increased under hypoxia or in the presence of hypoxia mimetics, cobalt chloride (CoCl 2 ), and dimethyloxalylglycine (Fig. 1, I and J). These results suggest that the cytoplasmic region contains an element(s) that imparts oxygen-dependent stability of EPOR.
VHL Negatively Regulates EPOR via Ubiquitylation-Considering the pivotal role of VHL in oxygen homeostasis and in the oxygen-dependent recognition of proly hydroxylated substrate HIF␣, we asked whether VHL binds to EPOR in an oxygen-dependent manner. Ectopic VHL expressed in HEK293 FIGURE 1. Oxygen influences EPOR stability. A, an equal number of UT-7 cells were plated in triplicate in semisolid medium containing the indicated dosage of EPO. The cells were maintained in normoxia (21% O 2 ) or hypoxia (1% O 2 ) for 10 days. Colonies consisting of greater than 30 cells were scored. A Student's t test was used to assess the significance. Graphed is a representative result of three independent experiments. B, HEK293 cells were transfected with human (hu) or murine (mu) EPOR. The cells were lysed, resolved by SDS-PAGE, and immunoblotted using the indicated antibodies. C, UT-7 cells were maintained in varying levels of hypoxia and deprived of cytokines overnight. The cells were treated with 1 unit/ml EPO for 90 s and lysed, and the indicated proteins were detected by immunoblotting. Presented is a representative of three independent experiments. D and E, UT-7 cells were cytokine-starved and maintained at ambient oxygen tension or 1% oxygen tension overnight (D) or the indicated number of hours (E). Total RNA was isolated from three biological replicates, and quantitative real time PCR was performed. Normoxic expression was arbitrarily set to 1. A Student's t test determined that the results were not significantly different. F, MEF cells were infected with lentivirus encoding GFP or EPOR. The cells were maintained at the indicated oxygen tensions overnight. The cells were lysed, and the indicated proteins were detected by immunoblotting using the indicated antibodies. G, Ba/F3 cells were treated with 10 M MG132 and/or less than 1% oxygen for 5 h. The indicated proteins were detected by immunoblotting. H, HEK293 cells were transfected with FLAG-muCytoEPOR. The indicated proteins were visualized by immunofluorescence. Scale bar, 10 m. I and J, HEK293 cells were transfected with human or mouse CytoEPOR plasmids and plated into a large dish. The cells were split into equivalent smaller plates and treated with 200 M CoCl 2 , 1 mM dimethyloxalylglycine (DMOG), or hypoxia overnight. Proteins were detected by immunoblotting using the indicated antibodies. Densitometry represents the average value from three independent experiments. * denotes a nonspecific band. cells bound to EPOR, but this interaction was diminished under hypoxia ( Fig. 2A). Similar to HIF␣, EPOR bound to the ␤ domain of VHL (residues 63-155), whereas the ␣ domain of VHL (residues 156 -213) bound to Cul2, as expected (Fig. 2B) (3). In addition, EPOR co-precipitated VHL and Cul2 in VHLnull 786-O renal cell carcinoma cells reconstituted with HA-VHL and infected with lentivirus-EPOR-FLAG (Fig. 2C), because 786-O cells do not contain detectable endogenous EPOR (Fig. 2D).
␤-TrCP and SOCS3 containing cullin-RING E3 ubiquitin ligases have also been shown to associate with EPOR but following EPO stimulation (40 -42). We asked whether VHL interacts with EPOR in a cytokine-dependent manner. Ba/F3-EPOR cells stably expressing HA-VHL were cytokine-starved The cells were lysed, and anti-EPOR or control immunoprecipitation was performed. The proteins were resolved by SDS-PAGE, and immunoblotting was used to detect the indicated proteins. E, Ba/F3 cell lines were cytokine-starved overnight. The cells were treated with 10 M MG132 for 5 h prior to lysis. The cells were treated with the indicated cytokines. The cells were lysed, and immunoprecipitation was performed with the indicated antibodies. Bound proteins were resolved by SDS-PAGE and detected using the indicated antibodies. F, Ba/F3 cell lines were treated with 10 M MG132 for 5 h. The cells were lysed and immunoprecipitated using the indicated antibodies. The bound proteins were resolved by SDS-PAGE and detected using the indicated antibodies. G, ␥2A cells were transfected with the indicated constructs. The cells were lysed and immunoprecipitated with the indicated antibodies. The bound proteins were resolved by SDS-PAGE and detected using the indicated antibodies. WCE, whole cell extract; IP, immunoprecipitated. * denotes a nonspecific band. All experiments were performed at least three times. and then treated with MG132 and EPO. VHL bound to EPOR, but not IL-3 receptor, irrespective of cytokine stimulation (Fig.  2, E and F). JAK2 has been reported previously to assist in EPOR trafficking to the plasma membrane (43). We asked whether JAK2 is necessary for the interaction between EPOR and VHL and observed that ectopic VHL bound to EPOR in ␥2A cells, which are devoid of JAK2 (Fig. 2G). These results suggest that JAK2 and EPO stimulation is not required for the binding between VHL and EPOR.
EPOR half-life increased from ϳ30 to 50 min in 786-O cells devoid of VHL in comparison to isogenically matched VHLreconstituted cells (p ϭ 0.0209; Fig. 3A). We next asked whether the increased rate of EPOR turnover was due, at least in part, to VHL-mediated ubiquitylation. CytoEPOR polyubiq-uitylation was observed using VHL-reconstituted S100 cellular extracts, which are devoid of 26S proteasome, whereas negligible polyubiquitylation pattern was produced using VHL-null S100 extracts (Fig. 3B). Furthermore, robust EPOR ubiquitylation was observed in Ba/F3 cells under normoxia, but not under hypoxia (Fig. 3C). CRISPR/Cas9 was used to generate UT-7 cells devoid of VHL (Fig. 3D). Loss of VHL was associated with decreased level of in vivo ubiquitylation of endogenous EPOR (Fig. 3E). These results suggest that VHL binds to and promotes the ubiquitylation of EPOR, leading to its increased turnover in an oxygen-dependent manner.
PHD3 Hydroxylates EPOR on Proline 419 and 426 -Considering the oxygen-dependent nature of EPOR turnover and binding to VHL, reminiscent of the interaction between HIF␣ The cells were treated with 50 g/ml cycloheximide for up to 60 min. EPOR protein levels were detected through anti-FLAG immunoblotting and quantified. The experiment was performed in triplicate, and the mean was plotted. Student's t test at 60 min was calculated as p ϭ 0.0209. A one-phase decay curve was used to model the amount of EPOR remaining as a function of time. B, in vitro translated SBP-CytoEPOR was incubated with the indicated S100 fractions in the presence or absence of ubiquitin. A streptavidin pulldown was performed, beads were washed, and bound protein was eluted by boiling in sample buffer. Ubiquitylated species were detected by immunoblotting. C, Ba/F3 cells were treated with MG132 and the indicated oxygen tension for 5 h. Anti-EPOR immunoprecipitation was performed, and the indicated proteins were detected by immunoblotting. D, UT-7 cells were lysed, and the indicated proteins were detected by immunoblotting to assess for CRISPR-Cas9 mediated gene editing. E, parental UT-7 cells or VHL-null UT-7 cells were lysed and boiled in SDS-containing buffer. Endogenous EPOR or control was immunoprecipitated, and ubiquitylated EPOR was detected by immunoblotting. WCE, whole cell extract; IP, immunoprecipitation; Strep-PD, streptavidin pulldown. All experiments were performed at least three times. and VHL, we asked whether PHD enzymes bind EPOR and, if so, whether they promote oxygen-dependent modification of EPOR. Interestingly, EPOR-FLAG preferentially co-precipitated PHD3 and negligibly PHD1 and PHD2 (Fig. 4, A and B). In addition, recombinant GST-PHD3 bound specifically to purified His-CytoEPOR (Fig. 4C). These results suggest that the binding between EPOR and PHD3 is direct. An in vitro hydroxylation reaction was then performed using purified GST-PHD3 and His-CytoEPOR followed by mass spectrometry, which identified hydroxylation of Pro 419 and Pro 426 (Fig. 5A). Furthermore, hydroxylation is a chemical reaction that absolutely requires oxygen; thus, hydroxylation does not occur in the absence of oxygen. Interestingly, Pro 426 is highly conserved in several organisms examined, and although Pro 419 appears to be well conserved in the alignment, Pro 419 , unlike Pro 426 , is not part of a diprolyl motif (Fig. 5B). However, oxygen-dependent prolyl-hydroxylation on endogenous EPOR could not be formally determined via mass spectrometry, at least in part because less than 1% of the total cellular endogenous EPORs are on the cell surface (39) and because of the limitation of the available anti-EPOR antibodies. Notably, prolyhydroxylation of endogenous HIF1␣, as well as HIF2␣ or HIF3␣, has not been formally confirmed via mass spectrometry.
EPOR Hydroxylation Is a Prerequisite for VHL Binding-We next asked whether EPOR prolyl hydroxylation is required for binding to VHL. An in vitro hydroxylation reaction was first performed on biotinylated 15-amino acid PEP6 peptides spanning Pro 419 and Pro 426 in the presence of in vitro translated PHD3(WT), catalytically inactive PHD3(H196A), or empty plasmid. PEP6 peptide preincubated with PHD3(WT), but not with PHD3(H196A), bound to in vitro translated 3ϫFLAG-VHL (Fig. 6, A and B). Biotinylated HIF1␣ ODD peptide was used to validate that the enzyme was indeed functional, and synthetically hydroxylated ODD peptide (ODD-OH) served as a positive control for binding to VHL (Fig. 6A). Similar results were observed using bacterially purified GST-PHD3 in promoting ODD(WT) and PEP6 binding to 3ϫFLAG-VHL (Fig.  6C). We next generated PEP6 peptides with a single synthetic hydroxyl group on Pro 419 , Pro 425 , or Pro 426 and performed an in vitro binding assay with HA-VHL. Pro 419 -OH peptide bound most robustly to HA-VHL, followed by Pro 426 -OH, whereas Pro 425 -OH and the unhydroxylated PEP6(WT) showed negligible interaction with HA-VHL (Fig. 6D). Notably, Pro 425 was not identified as a hydroxylation site via MS (Fig. 3C). Pro 419 -OH peptide did not bind to SOCS3, which recognizes phosphorylated EPOR (Fig. 6E). These results suggest that Pro 419 is the major site on CytoEPOR for PHD3-dependent hydroxylation, which is required for binding to VHL.
The bacterially purified GST-VBC complex bound to Pro 419 -OH, but not to the unhydroxylated Pro 419 peptide (Fig. 6F). Pro 419 -OH, but not Pro 419 peptide, co-precipitated VHL and Cul2 in Ba/F3-EPOR cells (Fig. 6G). Furthermore, Pro 419 -OH co-precipitated Cul2 and elongin B only in the presence of VHL (Fig. 6H), which suggests that the recruitment of ECV ubiquitin ligase components by Pro 419 -OH is via binding to VHL.
We showed that, similar to HIF␣, EPOR binds to the ␤ domain of VHL (Fig. 2B). Consistent with this notion, increased concentration of exogenous ODD-OH attenuated the interaction between HA-VHL and EPOR-FLAG in HEK293 cells (Fig.  6I). Unhydroxylatable ODD(P564A) peptide failed to block the interaction between HA-VHL and EPOR-FLAG (Fig. 6I). These results further support the notion that EPOR and HIF␣ share an overlapping binding site on VHL. GAL4-HA-PEP6(PPPPPP) (amino acids 412-446 of EPOR) was able to bind to 3ϫFLAG-VHL (Fig. 6J). Notably, mutation of every proline within PEP6 to alanine markedly decreased its interaction to 3ϫFLAG-VHL (Fig. 6J).
We next performed biolayer interferometry analysis to determine the binding kinetics of the interaction between single (Pro 419 or Pro 426 ) or double (Pro 419 and Pro 426 ) hydroxylated or unhydroxylated EPOR peptide and bacterially purified ternary complex comprised of VHL, elongin B, and elongin C (VBC) ( Table 1 and Fig. 7). The single Pro 419 -OH peptide had a K d of 41 nM, and the double Pro 419 -OH/Pro 426 -OH peptide had a K d of 38 nM. However, single Pro 426 -OH had very weak binding. Notably, ODD-Pro 564 -OH peptide had a K d of 31 nM, which is tightly comparable with Pro 419 -OH. As expected, both unhydroxylated ODD and the unhydroxylated EPOR peptide failed to bind the VBC complex. These results indicate that the hydroxylation of Pro 419 is critical for promoting the high affinity interaction to VHL and that hydroxylation of Pro 426 has a minor or negligible influence on the binding between EPOR and VHL.  Fig. 4C were used to perform an in vitro hydroxylation reaction. His-huCytoEPOR was digested, and the peptides were analyzed by mass spectrometry. Red denotes hydroxylated Pro, whereas blue represents alkylated Cys. B, an alignment of PEP6 derived from humans was performed against the indicated organisms using NCBI Homologene. * denotes a hydroxylated Pro residue identified by mass spectrometry in human EPOR. Red represents PP motifs, whereas yellow highlights LCP. Note that for Danio rerio the signal peptide length was unknown so it was included in the numbering.

PHD3 and VHL Negatively Regulate EPOR Signaling-
Diminishing oxygen tension was associated with increasing EPOR expression level and downstream signaling in the presence of constant EPO concentration (Fig. 1C). We next asked whether the loss of PHD3 or VHL directly influences EPOR expression level and JAK2-STAT5 signaling. ShRNA-mediated knockdown of PHD3 in MEFs increased the expression level of EPOR-FLAG (Fig. 8A). Furthermore, UT-7 polyclonal stable cell lines expressing shRNA targeting PHD3, VHL, or nontargeting control (pGIPZ) were maintained in EPO for in vitro differentiation analysis (Fig. 8B). UT-7 cells are one of a handful of cell lines that are able to initiate hemoglobinization in FIGURE 6. VHL binds prolyl hydroxylated EPOR. A and B, an in vitro hydroxylation reaction was performed on the indicated biotinylated peptides using in vitro translated PHD3 enzyme. Afterward, an in vitro binding assay was performed using in vitro translated 3ϫFLAG-VHL. A streptavidin pulldown was performed, and bound VHL was detected by immunoblotting. C, an in vitro hydroxylation reaction was performed on biotinylated-PEP6 peptide or biotinylated HIF1␣(ODD) peptide using purified GST-PHD3 followed by binding to in vitro translated 3ϫFLAG-VHL. D, PEP6 that was chemically synthesized to contain hydroxyproline at the indicated residues was attached to streptavidin-agarose. Binding to in vitro translated HA-VHL was assessed. Dot blotting was performed to assess the quantity of peptide. E, in vitro translated 3ϫFLAG-VHL and 3ϫFLAG-SOCS3 were incubated with biotinylated Pro 419 -OH peptide, which was prebound to streptavidin-agarose. Efficiency of binding to Pro 419 -OH peptide was assessed by anti-FLAG immunoblotting. F, the indicated peptides were bound to streptavidin-agarose. GST or GST-VBC was incubated with the peptides, the beads were washed, and proteins were eluted in sample buffer. Protein was resolved by SDS-PAGE and visualized by GelCode staining. G, Ba/F3-EPOR cells were lysed. The cell lysate was incubated with Pro 419 or Pro 419 -OH peptide that was bound to streptavidin-agarose. In parallel, an anti-VHL or control immunoprecipitation was performed. The beads were washed, and bound proteins were eluted by boiling in sample buffer. The indicated proteins were detected by immunoblotting. H, 786-O-Mock and 786-O-HA-VHL cell lysate was incubated with Pro 419 or Pro 419 -OH peptide that was prebound to streptavidin-agarose. Interacting proteins were detected through immunoblotting with the indicated antibodies. I, HEK293 cells were transfected with HA-VHL and EPOR. The complex was isolated onto beads using HA-agarose and split into four identical tubes that were then incubated with the indicated peptides. Amount of EPOR that remained bound to HA-VHL was detected through immunoblotting. J, HEK293 cells were transfected with the indicated plasmids. The cells were treated with 10 M MG132 for 4 h, lysed, and immunoprecipitated using the indicated antibody. The bound proteins were resolved by SDS-PAGE and detected by immunoblotting with the indicated antibodies. IB, immunoblot; DB, dot blot; IP, immunoprecipitation; Strep-PD, streptavidin-agarose pulldown; IVBA, in vitro binding assay; IVOH, in vitro hydroxylation. All experiments were performed at least three times.
response to EPO stimulus. Differentiation of these UT-7 cells in the presence of EPO, but not GM-CSF, was also visibly noticeable with increased redness (Fig. 8B) (44). Notably, spectrophotometric quantification revealed a statistically significant increase in hemoglobinization in UT-7-shPHD3 and UT-7-sh-VHL cell lines as compared with control UT-7-pGIPZ cells (Fig.  8B). CRISPR/Cas9 system (36) was used to knock out the endogenous PHD3 or VHL in UT-7 cells (Figs. 8C and 3D, respectively). The cell surface of knock-out UT-7 cells and parental UT-7 cells was biotinylated. Subsequent streptavidin pulldown assay showed that VHL-null and PHD3-null cells had higher amounts of cell surface EPOR (Fig. 9A). In comparison with wild-type UT-7 cells, UT-7(PHD3 Ϫ/Ϫ ) and UT-7 (VHL Ϫ/Ϫ ) cells had greater amplitude in signaling as measured by increased levels of phosphorylated EPOR, JAK2, and STAT5 upon EPO stimulation following cytokine starvation (Figs. 8D and 9B, respectively). Reconstitution of UT-7(VHL Ϫ/Ϫ ) cells with HA-VHL(WT) restored the reduced EPOR signaling compared with GFP expressing control cells (Fig. 9C). These results support the notion that PHD3 and VHL negatively regulate EPOR signaling and potentially influence EPOR-dependent physiologic response.
VHL disease has several phenotypic manifestations. Using several known type 2-associated mutants (13,14) (type 2A is associated with the development of PHEO and HAB with a low risk of RCC, type 2B is associated with PHEO and HAB with a high risk of RCC, and type 2C is associated exclusively with PHEO), we asked whether any of the VHL mutants exhibited a defect in binding to EPOR Pro 419 -OH peptide as well as to HIF1␣ODD-OH. In comparison with in vitro translated VHL(WT), type 2A and type 2B in vitro translated VHL mutants displayed reduced binding to ODD-OH as previously described (13,14), but these deficiencies were more pronounced for Pro 419 -OH (Fig. 9D). As expected, VHL(WT) bound to the hydroxylated Pro 419 -OH or ODD-OH peptides but not to the unhydroxylated peptides (Fig. 9D). Unlike type 2A/B, type 2C mutants have been generally reported to retain the binding to and regulation of HIF1␣ (13,14). Consistent with these findings, three of five type 2C mutants tested, namely VHL(R64P), VHL(V84L), and VHL(L188V), showed relatively robust binding to ODD-OH (Fig. 9D). Notably, all type 2C mutants tested showed a marked defect in binding to Pro 419 -OH (Fig. 9D). Consistent with these observations, both ODD-OH and Pro 419 -OH peptides bound robustly to ectopically expressed HA-VHL(WT), whereas Pro 419 -OH peptide, but not ODD-OH, showed markedly reduced binding to ectopically expressed HA-VHL(R64P) in HEK293 cells (Fig. 9E). Moreover, HA-VHL(R64P) showed attenuated binding to fulllength EPOR-FLAG compared with HA-VHL(WT) (Fig. 9F).

Discussion
Red blood cell production is a key physiologic response to hypoxia that allows the body to increase its oxygen carrying capacity. Erythropoiesis is strictly dependent upon the cytokine EPO with the output signaling being governed by erythroid cell intrinsic proteins (23). EPO production is regulated through an oxygen-sensitive extrinsic PHD2-HIF2␣-VHL axis (17)(18)(19)(20)(21)29).
Here we identify an additional oxygen-sensitive pathway comprised of an intrinsic PHD3-EPOR-VHL axis that can modulate the duration and amplitude of EPO-based signaling by regulating the cell surface expression level of EPOR.
We show here that EPOR is hydroxylated on prolines at positions 419 and 426 within the cytoplasmic region by PHD3. However, hydroxylation of Pro 419 appears to be a key event that FIGURE 8. EPOR signaling can be modulated by PHD3. A, MEF cells expressing EPOR-FLAG were infected using pGIPZ control or shPHD3 lentivirus. The cells were maintained at the indicated oxygen tension overnight. The cells were lysed and resolved by SDS-PAGE, and immunoblotting was performed using the indicated antibodies. B, in vitro differentiation assay was performed three independent times on UT-7 cells stably expressing the indicated shRNA constructs. A Student's t test was used to assess the statistical significance. C, parental UT-7 cells or the indicated clones were assessed for CRISPR-Cas9-mediated gene editing. The cells were maintained in hypoxia overnight where indicated. The cells were lysed, and the indicated proteins were detected by immunoblotting. D, UT-7 parental cells or the indicated PHD3 Ϫ/Ϫ clones were cytokine-starved overnight. The cells were treated with 1 unit/ml EPO for the indicated amount of time. The proteins were visualized by immunoblotting with the indicated antibodies. Densitometry values were normalized to loading control. * denotes a nonspecific band.
marks EPOR for VHL recognition and subsequent ECV-mediated ubiquitylation under normoxia. Hypoxia or the molecular inhibition of PHD3 or VHL attenuates EPOR turnover, resulting in an increased JAK2-STAT5 signaling in the presence of EPO. Notably, Pro 419 is within a region of EPOR that contains numerous negative regulatory signals. For example, adjacent are Tyr 430 and Tyr 432 (Tyr 429 and Tyr 431 in mice), which upon EPO engagement and JAK2 phosphorylation promote EPOR endocytosis (45). EPO stimulation also promotes binding to and negative regulation by ␤-TrCP in a Ser 438 -dependent manner (40). EPOR truncations are commonly found in patients with primary familial and congenital polycythemia. Interest- FIGURE 9. EPOR signaling can be modulated by VHL. A, an equivalent number of cells from the indicated cell lines was cell surface-biotinylated. A streptavidin pulldown was performed to enrich the biotinylated proteins, which were eluted and resolved by SDS-PAGE. The indicated proteins were detected by immunoblotting. EPOR signals were measured by densitometry and normalized to ␤c. B, UT-7 parental cells or UT-7 VHL knock-out cells were cytokine-starved overnight. The cells were treated with 1 unit/ml EPO for the indicated amount of time. The proteins were visualized by immunoblotting with the indicated antibodies. The experiments were performed in triplicate. Densitometry values were normalized to ␤-actin. C, VHL-null UT-7 cells were infected with lentivirus containing GFP or HA-VHL(WT). The cells were cytokine-starved overnight. The cells were treated with 1 unit/ml EPO for the indicated amount of time. Proteins were visualized by immunoblotting with the indicated antibodies. D, the indicated VHL constructs were in vitro translated using rabbit reticulocyte lysate that was incubated with Pro 419 -OH or ODD-OH peptides (or unhydroxylated control peptides) prebound to streptavidin-agarose. VHL binding was assessed by anti-HA immunoblotting. E, the indicated peptides were attached onto streptavidin-agarose. The beads were incubated with the indicated HEK293 transfected lysates. The beads were washed, and interacting proteins were resolved by SDS-PAGE and detected by immunoblotting. F, HEK293 cells were transfected with the indicated plasmids. The cells were lysed, and anti-HA immunoprecipitation was performed. The proteins were resolved by SDS-PAGE and detected by immunoblotting using the indicated antibodies. WCE, whole cell extract; IP, immunoprecipitate; IVBA, in vitro binding assay; Strep-PD, streptavidin-agarose pulldown; Ctrl, control.
ingly, these mutations have been shown to lead to the accumulation of truncated EPOR at the plasma membrane, which is thought to underlie increased downstream signaling (24,25). Perhaps most notably, many of the disease-associated truncations eliminate Pro 419 (and Pro 426 ), which is predicted to cause a failure in the negative regulation of EPOR via VHL and a consequential prolongation of cell surface EPOR expression. Future experiments would be needed to address the factors and processes controlling hydroxylation of endogenous EPOR and the specific role of Pro 419 .
At present it is unclear why a loss of both PHD1 and PHD3 is required to produce a hematologic phenotype in mice (30). Recently, PHD1 was shown to hydroxylate FOXO3a to promote its degradation through a mechanism that does not seemingly involve VHL (46). FOXO3a has been reported to promote the expression of antioxidant genes that through their regulation of ROS determine the lifespan of a circulating red blood cell (47). We show here that a loss of PHD3 leads to exaggerated EPO-mediated signaling via EPOR up-regulation. Thus, the aforementioned effects when combined upon the loss of both PHD1 and PHD3 might have contributed to the development of polycythemia in these double knock-out mice despite exhibiting reduced serum EPO level.
It is also unclear at present what determines the genotypephenotype correlation observed in VHL disease. We show here that ODD-OH and Pro 419 -OH peptides share an overlapping binding region on VHL within the ␤ domain. It is perhaps not surprising then that type 2A and 2B disease-associated VHL mutants, which are known to have a defect in binding to HIF␣, show a similar defect in binding to Pro 419 -OH. Type 2C mutations, which cause PHEO without other manifestation of VHL disease, have been generally regarded to retain the capacity to negatively regulate HIF␣ (13,14). In contrast, type 2C mutants showed a marked defect in binding to Pro 419 -OH. For example, R64P, V84L, and L188V mutants, which have comparable binding to ODD-OH as wild-type VHL, showed negligible to modest binding to Pro 419 -OH. These results suggest the possibility and significance of a loss of EPOR regulation via VHL in the pathogenesis of PHEO.
Several studies have suggested a potential involvement of PHD3 in PHEO. At a cellular level, PHEO arises from the increased survival or decreased apoptosis of sympathetic neuronal progenitor cells. Kaelin and co-workers (35) have shown that the overexpression of catalytically active PHD3 in PC12 PHEO cells leads to apoptosis. The target of PHD3 in PHEO is unknown. Although it is formally unknown whether EPOR is a relevant target of PHD3 in PHEO, EPO treatment of PC12 and neuroblastoma SH-SY5Y cells was shown to be protective against apoptosis (48,49), which strongly suggests EPOR expression in these cell types. Furthermore, Phd3 Ϫ/Ϫ mice have increased numbers of sympathetic neurons (50), and PHD3 (R8S) mutation was identified in a patient with PHEO (51). Whether these phenomena are attributable at least in part to elevated EPOR dependent pro-survival signaling remains an outstanding question.
Furthermore, although individuals with VHL R200W/WT genotype do not display features of polycythemia, two polycythemia patients with VHL R200W/L188V genotype have been report-ed; one of them had a normal serum EPO level (52,53). These findings, although limited, suggest that an impaired VHL-mediated EPOR degradation could be a contributing factor in the development of polycythemia. Consistent with this notion, PHD3 was identified in an shRNA screen on patient-derived hematopoietic progenitor cells that when lost would promote longer stem cell maintenance or increased growth (54).
HAB of the retina and central nervous system (cerebellum, brainstem and spinal cord) is a cardinal feature of VHL disease. A hemangioblast is a cell that is capable of differentiating into endothelial and hematopoietic cells. Consistent with this notion, islands of blood cells are often found within HAB (55)(56)(57). In addition, studies have suggested that HAB potentially express EPOR (56). These observations suggest that a failure in the proper negative regulation of EPOR in addition to the predicted up-regulation of EPO production upon the loss or mutation of VHL plays a role in the development of HAB.
These results, taken together, identify EPOR as a substrate of PHD3, which upon oxygen-dependent prolyl hydroxylation is targeted for VHL-mediated ubiquitylation and subsequent degradation. It remains an outstanding question whether the oxygen-dependent negative regulation of EPOR via VHL plays a critical role in normal erythropoiesis, as well as in the pathogenesis of VHL disease upon the loss of this newly defined function of VHL.