Introduction

About half of blood volume is made up of erythrocytes (red blood cells) whose role is to deliver oxygen from the lungs to the tissues in the body¹. Low concentrations of red blood cells result in insufficient oxygen delivery to tissues while high concentrations can induce a propensity towards thrombotic events rendering red blood cell mass regulation crucial. Erythropoietin (EPO) is a hormone produced primarily in the kidneys that stimulates erythropoiesis^2,3. Through binding to EPO receptors on erythroid progenitors, EPO can activate a signaling cascade that induces differentiation into erythrocytes⁴. EPO production is regulated in response to physiological oxygen concentrations through hypoxia response elements (HRE). HREs are enhancers that transcription factors called hypoxia-inducible factors (HIFs) bind to and subsequently upregulate the expression of numerous hypoxic response genes including EPO⁵. Disrupted EPO regulation can be pathogenic and lead to an autonomous overproduction of erythrocytes causing erythrocytosis⁶. Due to EPO’s inherent link to the oxygen response, mutations in the metazoan oxygen-sensing pathway have frequently been found to underlie erythrocytosis.

The oxygen-sensing pathway has been conserved in all animals and is essential for the maintenance of cellular activities in low oxygen settings⁷. Under hypoxia, oxygen-labile HIF1-3α subunits complex with its constitutively expressed HIFβ subunits to translocate to the nucleus as active heterodimeric transcription factors⁸. Following translocation, HIF binds to HREs across the genome to trigger the transactivation of numerous genes such as GLUT1, EPO and VEGF to promote anaerobic metabolism, erythropoiesis and angiogenesis, respectively, to induce adaptation to low oxygen availability⁹. In normal oxygen conditions, HIF prolyl hydroxylases (PHDs) hydroxylate HIFα subunit at one or both of the two conserved proline sites located in the oxygen-dependent degradation domain (ODD), which initiates its ubiquitylation via the von Hippel-Lindau tumor suppressor protein (pVHL)-containing E3 ubiquitin ligase (Elongin BC/Cullin 2-Rbx1/pVHL) and subsequent destruction via the 26S proteasome. Notably, only one proline is required to be hydroxylated to trigger its rapid destruction and whether HIFα is singly or doubly hydroxylated does not appear to change the rate of its ubiquitylation or proteasome-mediated destruction¹⁰.

Moreover, the C-terminally located proline within ODD or CODD (e.g., P564 on HIF1α) has been shown to be the predominant residue that becomes hydroxylated prior to the modification of proline in the N-terminal ODD or NODD (e.g., P402 on HIF1α)^10–12. The reason for this preferential prolyl-hydroxylation is incompletely understood, and the biological significance or the mechanistic purpose of the N-terminal proline remains unclear.

Clinically observed mutations have been identified in the major components of the oxygen-sensing pathway including VHL (pVHL)^13,14, EPAS1 (HIF2α)^15–17, EGLN1 (PHD2)^18,19, and EGLN2 (PHD1)²⁰. These have been associated with the development of diseases with overlapping phenotypes with erythrocytosis being the most common. Collectively, these diseases are referred to as pseudohypoxic diseases as they involve an inappropriate hypoxic response. Erythrocytosis caused by mutations in EGLN1 is the most recently discovered pseudohypoxic disease. Since its first reported case in 2006²¹, 152 cases with 96 distinct mutations have been reported. EGLN1 mutation-driven erythrocytosis is referred to as erythrocytosis, familial, 3 (ECYT3) by the Online Mendelian Inheritance in Man (OMIM). However, three individuals with distinct EGLN1 mutations (W51X, A228S and H374R) also developed recurrent pheochromocytoma and paraganglioma (PPGL)^20,22,23, which are neuroendocrine tumors that develop from chromaffin cells either within (pheochromocytoma) or outside (paraganglioma) the adrenal gland²⁴.

Of three isozymes of PHD, PHD2 is thought to be the most critical with high expression seen in a variety of cell types ²⁵ and the only PHD able to effectively target both prolines in HIF1α and 2α ODDs^11,26,27. PHD2 belongs to the 2-oxoglutarate-dependent-dioxygenase superfamily²⁸ in which enzymes contain a double-stranded β-helix fold with a highly conserved HXE/D…H motif that coordinates Fe²⁺ at its catalytic site²⁹. In PHD2, the iron coordinating residues are His313, Asp315 and His374³⁰. Erythrocytosis-causing mutations at (H374R) and near these conserved catalytic residues have been shown to be defective at regulating HIFα in a cellular context^18,21,23, which appears to be a common mechanism observed in the other pseudohypoxic diseases. For example, numerous studies have revealed that pVHL mutants have varied abilities to recognize and regulate HIFα leading to the development of PPGL, erythrocytosis and other cancers such as clear-cell renal cell carcinoma and hemangioblastoma^31–33. Recently, we showed that phenotypic severity of Pacak-Zhuang syndrome, caused by EPAS1 mutations, is correlated to the ability of HIF2α mutants to escape degradation mediated by pVHL and PHD2^15,34.

Together, these observations suggest that disproportionately high levels of HIFα are responsible for pseudohypoxic disease phenotypes such as erythrocytosis and PPGL. In keeping with this notion, disease-causing mutants of PHD2 are likely defective in catalyzing HIFα hydroxylation to some extent, resulting in less HIFα degradation and correspondingly increased hypoxia response signaling. However, the precise mechanism(s) underlying such defect in PHD2 function is unclear. Recently, we devised a time-resolved nuclear magnetic resonance (NMR) assay to directly measure the rate of hydroxylation on the two conserved proline residues simultaneously in HIF1α ODD¹². Here, we examined 7 disease-associated PHD2 mutants via biophysical analyses and show that all PHD2 mutants have structural and/or catalytic activity defects. While all mutants had the anticipated defect in binding and hydroxylating the predominant CODD proline, we identified PHD2 P317R mutant that retained comparably wild-type level of hydroxylating CODD proline (i.e., HIF1α P564) but failed to hydroxylate NODD proline (i.e., HIF1α P402) as measured by NMR. Unexpectedly, PHD2 P317R was capable of binding to both NODD and CODD peptides with similar affinity. These results provide a biophysical mechanism that, at least in part, underlies PHD2-driven erythrocytosis, and serendipitously reveal a direct biochemical evidence to suggest that NODD prolyl-hydroxylation is necessary for normal physiological response to hypoxia.

Results

Mutations across the PHD2 enzyme induce erythrocytosis

We compiled and analyzed all reported cases of PHD2-driven erythrocytosis as of January 2025. 96 distinct variants have been reported, including two missense variants commonly found in Tibetan populations that are hypothesized to be adaptations to high altitude conditions (D4E and C127S)^35,36. Consistent with erythrocytosis in the general population, most patients with PHD2-driven erythrocytosis were male (∼70%). About half of the case reports provided information on family history of PHD2-driven erythrocytosis, of which ∼75% were reported to be familial while ∼25% were de novo mutations; a similar rate of de novo mutations is observed with VHL disease. PHD2-driven erythrocytosis is typically classified as secondary erythrocytosis, which is defined as being caused by an EPO-inducing mechanism downstream of the primary defect, ultimately resulting in normal to high EPO levels that drive erythropoiesis. This classification is aligned with the suspected mechanism of PHD2-driven disease in which mutated PHD2 dysregulates HIF leading to increased EPO production. Curiously however, EPO was reported to be largely normal in patients with PHD2-driven erythrocytosis with few exceptions (Supplementary Data), which is consistent with the loss of PHD2 in mice leading to EPO hypersensitivity in erythroid progenitors, enabling a ‘normal’ amount of EPO to generate a pathogenic response³⁷.

All cases of PHD2-driven erythrocytosis were reported as germline heterozygous mutations of PHD2, suggesting haploinsufficiency of PHD2 is enough to induce erythrocytosis. This is consistent with previous reports that HIFα regulation is dose dependent on PHD2 knockdown via siRNA²⁵. As stated previously, in three exceptional cases, individuals with PHD2 mutations developed PPGL (W51X, A228S and H374R). Biopsies of each tumor revealed a loss of heterozygosity for PHD2, implying a complete loss of wild-type PHD2 in chromaffin cells is necessary to induce tumorigenesis.

The majority of the reported mutations were missense (∼74%) (Figure 1A). Interestingly, mutations that should theoretically severely disrupt PHD2 structure (nonsense, frameshift and deletions) caused the same phenotype (erythrocytosis) as less disruptive missense mutations, suggesting an unclear or lack of correlation between mutation type and phenotypic severity. Thus, a potential phenotypic mechanism of PHD2-driven disease can be hypothesized where erythrocytosis will develop upon deleterious mutation in one PHD2 allele while PPGL will be induced upon a loss of heterozygosity in chromaffin cells.

Figure 1.

Mutations are distributed across all exons of PHD2 with some clustering across the catalytic core (Figure 1B). When mapped onto the three-dimensional crystal structure of the PHD2 catalytic core, which we and others have previously solved^34,38,39, these mutations do not seem to display a specific pattern (Figure 1C). Thus, we selected 7 PHD2 mutants with distinct residue changes in the catalytic core of PHD2 for study.

Some PHD2 mutants are less stable and defective in regulating HRE-driven transcription

We asked whether the enzymatic activity of PHD2 mutants can be inferred from downstream HIF activity using a dual luciferase reporter assay under control of HRE. Notably, we first generated CRISPR/Cas9-mediated PHD2 knockout (-/-) HEK293A cells to minimize the impact of endogenous PHD2 on the prolyl-hydroxylation of HIFα. As expected, wild-type (WT) ectopic PHD2 markedly reduced the HRE-luciferase activity driven by ectopic HIF1α compared to cells without ectopic PHD2 expression (Figure 2A). Two disease-associated PHD2 mutants, P317R and H374R, displayed significantly reduced ability to knockdown luciferase activity (∼2-fold and ∼2.5-fold, respectively) compared to PHD2 WT. Notably, the other mutants showed comparable activity to PHD2 WT in down-regulating HRE-driven luciferase signal under these in cellulo experimental conditions (Figure 2A).

Figure 2.

We next performed cycloheximide (CHX) chase assay to determine the protein stability of PHD2 mutants. We observed that mutants G206C, W334R, F366L, R371H, and H374R showed noticeable instability by 24 hours (Figure 2B and C). These results suggest that while most PHD mutants showed diminished protein stability in comparison to PHD2 WT, the stability of PHD2 mutants cannot alone explain the impact of these disease-associated mutations on HIF1α-driven HRE-luciferase activity in cells.

PHD2 mutants are thermally unstable and some are prone to aggregation

We asked if the disease-causing mutations compromised other biophysical characteristics of PHD2. We purified catalytic cores of PHD2 harboring the selected mutations for the subsequent biophysical studies. Size exclusion chromatography (SEC) was performed as a polishing step and chromatograms were acquired with monomeric PHD2 eluting from the column around 17 mL (Figure 3A). Mutants were eluted from the column at different volumes ranging from 16.9 mL (G206C) to 17.9 mL (P317R), while WT PHD2 was eluted consistently at 17 mL. These results suggest that mutations may affect the hydrodynamic size of PHD2, hinting to mutation-induced structural alteration. Moreover, PHD2 G206C, W334R and H374R aggregated severely in SEC and throughout the purification process and could not be purified.

Figure 3.

We next subjected the remaining four mutants (A228S, P317R, F366L, and R371H) in comparison to WT PHD2 to circular dichroism (CD) to determine structural abnormalities. The CD spectra of the mutant PHD2s did not display appreciable differences from WT PHD2 (Figure 3B), indicating that these PHD2 mutants’ overall structure is mostly unaffected.

To further probe the stability of PHD2 mutants, molar ellipticity was measured at 220 nm from 25 to 95 °C to determine melting temperature (T_m). Alpha helices produce a strong minimum at 220 nm so this wavelength was selected to measure thermal protein unfolding⁴⁰. Melting curves were created by setting the peak value of each spectrum at 25 °C to 1.0. The data were then fitted with a sigmoidal curve to produce melting curves and calculate T_m’s (Figure 3C and Table 1). WT PHD2’s T_m, estimated as the temperature at which half of the proteins are unfolded, was determined to be 52.3°C. All four mutants had lower T_m’s than WT, suggesting a degree of thermal instability. Although it is reasonable to assume that instability would result in less PHD2, it is unclear if the unstable enzyme is active prior to degradation and whether the remaining portion of mutant PHD2s would be sufficient to bind and hydroxylate HIFα.

Table 1.

PHD2 mutants have weaker binding to HIFα than WT PHD2

Assuming intact catalytic function, the strength of PHD2 binding affinity to HIFα should translate to the extent of prolyl-hydroxylation of HIFα via PHD2. We performed microscale thermophoresis (MST), which utilizes the motion of molecules in a temperature gradient to determine dissociation constants (K_d), to measure the binding deficiencies, if any, between PHD2 mutants and HIFα. In addition to being highly sensitive, MST is also advantageous for measuring protein-ligand binding in an untethered setting. We showed previously that this technique was effective at distinguishing between disease classes of HIF2α mutants in Pacak-Zhuang syndrome by measuring their binding affinities to WT PHD2³⁴. We performed a similar protocol with PHD2 mutants to identify potential binding defects with HIF1α and 2α peptides of the C-terminal oxygen-dependent degradation domain (CODD) that contains single target proline, which is thought to be predominant over the N-terminal oxygen-dependent degradation domain (NODD)¹⁰. PHD2 WT bound tightly to HIF1αCODD and HIF2αCODD peptides (7.5 and 15 µM, respectively). Notably, all but one tested PHD2 mutants (A228S, F366L and R371H) displayed mild binding defects, up to approximately 2-fold, to HIF1αCODD in comparison to WT PHD2 while P317R showed a severe binding defect with high K_d value (320 µM) (Figure 4A and Table 2). Similar pattern of binding deficiencies was noted for PHD2 mutants to HIF2αCODD (Figure 4B and Table 2), suggesting no significant preferential binding defects to either HIFα paralog.

Figure 4.

Table 2.

NMR reveals major defect in PHD2 P317R-mediated hydroxylation of NODD proline

Previously, we developed an assay utilizing NMR to directly measure PHD2 hydroxylation of the critical two prolines located within the intrinsically disordered ODD of HIF1α¹². Under the catalysis of PHD2, movement of resonances corresponding to the target prolines as well as adjacent residues are observed, reflecting local conformational changes due to the post-translational modification of HIF1αODD. The time-dependent shift in peak intensity from unhydroxylated to hydroxylated species can be interpreted as a function of the reaction progress. Notably, the enzymatic kinetics of the hydroxylation of both target prolines (P402 in NODD, P564 in CODD) can be measured simultaneously over time. Since binding experiments with PHD2 mutants did not implicate marked preferential defects for either HIF1α or HIF2α, these assays were performed with HIF1αODD of which more than 95% of amino acids have been assigned to their corresponding 3D NMR backbone cross-peak¹².

We selected PHD2 A228S for NMR study to determine if this mutant, which showed negligible to minute defect in the above studies from HRE-driven luciferase assay to MST based binding experiments, exhibited any appreciable enzymatic defect in hydroxylating P564 or P402. We included as controls PHD2 WT and P317R that showed major defect in binding HIF1αCODD peptide and transcriptional activation of HRE-luciferase. As shown previously¹², PHD2 WT showed robust and rapid hydroxylation of P564 in CODD and a markedly slower hydroxylation of P402 in NODD (Figure 5A). PHD2 A228S mutant effectively hydroxylated P564 faster than P402 with negligible differences in the enzymatic rate to PHD2 WT (Figure 5A). Unexpectedly, P317R displayed a near wild-type hydroxylation of P564 but failed to hydroxylate P402 (Figure 5A).

Figure 5.

Superimposition of ¹H-¹⁵N HSQC spectra showed that cross-peak of P564 shifts with PHD2 WT, A228S, and P317R mutant, but P402 only shifts when incubated with PHD2 WT and A228S, not PHD2 P317R (Figure 5B). These results are inconsistent with observations made via MST that showed P317R having a severe binding defect for HIF1αCODD peptide. Furthermore, it is of particular interest to investigate whether the lack of activity of PHD2 P317R on NODD is due to a loss of binding interaction.

PHD2 P317R binds effectively to both CODD and NODD via BLI

We performed bio-layer interferometry (BLI) using immobilized biotinylated HIF1αCODD and NODD peptides with purified PHD2 WT and P317R, which showed P317R binding to CODD peptide only slightly weaker than PHD2 WT (Table 3). The discrepancy between the MST and BLI results was likely due to the amine-linked fluorescent labelling of PHD2 P317R for MST where the fluorescent label bound to the newly available amine residue on the mutant arginine, thereby disrupting the binding to HIF1αCODD peptide. The removal of this label allowed for proper, unimpeded interaction to occur as observed with BLI. These results are consistent with the near-WT, efficient hydroxylation of HIF1αCODD by PHD2 P317R as revealed by NMR (See Figure 5).

Table 3.

The binding affinity between PHD2 P317R and HIF1αNODD was also measured to examine whether the catalytic defect observed via NMR would manifest as a binding defect. However, PHD2 P317R bound to NODD similarly to PHD2 WT suggesting that its failure to hydroxylate P402 is unlikely attributable to binding.

Discussion

Mutations occurring in the core components of the hypoxia sensing pathway have been shown to cause various phenotypes ranging from erythrocytosis and neuroendocrine tumors to hemangioblastoma and renal cell carcinoma. Collectively, these diseases are referred to as pseudohypoxic diseases due to the inappropriate hypoxic response induced by the mutations. It is presumed that the abnormal response is the reason for overlapping phenotypes caused by mutations in different proteins of the same pathway (pVHL, HIF2α and PHD2). Specifically, we have proposed that the common phenotypes of pseudohypoxic diseases are caused by the gain-of-function ability of HIFα, in particular HIF2α, to escape degradation in the presence of the disease-causing mutants³¹. It has been shown that VHL disease-causing mutants have varying ability to target HIF1/2α for ubiquitin-mediated destruction^32,41 and Pacak-Zhuang syndrome-causing HIF2α mutants are able to escape recognition to a varying degree from both pVHL and PHD2^15,34. It follows then that in the case of PHD2-driven erythrocytosis, PHD2 mutants would have a hindered ability to promote proper oxygen-dependent hydroxylation of HIF1α and 2α.

Complete deletion of PHD2 in mice causes embryonic lethality due venous congestion and cardiomyopathy, theorized to be directly related to high blood volume induced upon PHD2 loss⁴². Conditional inactivation of PHD2 whether globally⁴³ or in renal erythropoietin-producing cells⁴⁴ has been observed to cause erythrocytosis in mice. Moreover, while inactivation of PHD2 alone was sufficient to induce erythrocytosis, inactivating PHD1 or PHD3 alone did not result in erythrocytosis. However, when PHD1 or PHD3 were inactivated in addition to PHD2, erythrocytosis phenotype worsened, suggesting a supporting role for the PHD paralogs⁴⁵.

Here, we examined 7 disease-causing PHD2 mutations across the catalytic core, the most frequently mutated region within PHD2. G206C, W334R and H374R exhibited gross protein stability defects and were prone to severe aggregation, which precluded their purification necessary for further biophysical analyses. One of these mutants, H374R, is located at the catalytic site integral to the function of PHD2, which combined with the stability defect likely abrogated its ability to promote the oxygen-dependent hydroxylation of ectopic HIF1α, leading to increased HRE-luciferase transcriptional activity in PHD2-/- HEK293A cells. Previous work has shown mutating analogous His iron-coordinating residues to a charged residue (Arg or Glu) in the prolyl-4-hydroxylase responsible for hydroxylating prolines of collagen abolishes enzymatic activity⁴⁶. Given the high structural conservation in the Fe/2OG-oxygenase enzyme family²⁹, it can be predicted that a comparable mutation on PHD2 would also abolish activity. However, PHD2 G206C and W334R mutants showed similar capacity as the PHD2 WT to inhibit the transcriptional activity of ectopic HIF1α. This may have been due to the nature of transfected PHD2 in which mutant PHD2s that are prone to aggregation and instability are being translated throughout the luciferase assay, which may generate a small pool of monomeric PHDs that is sufficient to inhibit the transcriptional activity of HIF1α comparable to PHD2 WT.

The remaining four mutants (A228S, P317R, F366L, and R371H) not prone to aggregation displayed an overall structure similar to PHD2 WT, as determined by CD, and with the exception of PHD2 P317R (discussed later), were able to effectively suppress the ectopic HIF1α transcriptional activity. These mutants also exhibited slight thermal instability with mildly weaker binding affinity to HIF1αCODD peptide, which may be sufficient to cause the disease phenotype in vivo. PHD2 A228S exhibited the most minute of defects amongst all mutants tested and concordantly showed marginal reduction in enzymatic activity as measured via time-resolved NMR. Notably, despite these defects, albeit subtle, it is evident that the HRE-luciferase reporter assay is in general not reliable to reveal the disease-causing mutations on the functional output of PHD2 mutants, including those prone to aggregation (G206C and W334R).

We recently developed a time-resolved NMR approach to directly and simultaneously measure PHD2-mediated hydroxylation of P564 within CODD and P402 within NODD in the context of full-length HIF1αODD. We showed that P564 is the predominant oxygen-dependent hydroxylation site over P402. Considering that pVHL only requires one hydroxylated proline for binding and triggering rapid ubiquitin-mediated proteasomal degradation of HIF1α^10–12,47, the biological utility of P402 was thought to be inconsequential in normal hypoxic response. Here, our results revealed a previously unappreciated intricacy in the enzymatic selectivity of PHD2 P317R between CODD and NODD, challenging the current notion of NODD insignificance. NMR results showed that the disease-causing PHD2 P317R retained near-WT activity on CODD P564 but had negligible activity on NODD P402. Furthermore, despite effective activity on P564, ectopic PHD2 P317R generated abnormally high HRE-driven luciferase reporter signal, suggesting the negative impact of defective P402 hydroxylation on an otherwise rapid oxygen-dependent degradation of HIF1α. This is congruous with Schofield’s earlier work that showed, using an indirect metabolic method with individual peptides, no activity on NODD by PHD2 P317R while effectively hydroxylating CODD³⁸. Given the pathogenic nature of P317R mutation, it appears likely that this selective defect on HIF1αNODD contributes to the development of erythrocytosis and therefore, the involvement of NODD in normal oxygen-sensing pathway is not insignificant.

The molecular basis for PHD2 P317R’s selective defect for NODD is currently unknown. Intriguingly, BLI results showed that PHD2 P317R binds to both CODD and NODD peptides comparable to PHD2 WT. Substrates of Fe/2OG-oxygenases typically bind the enzymes non-covalently with a prime separation distance of 4-5 Å between the target atom and the catalytic iron⁴⁸. One possibility is that the Pro to Arg substitution at position 317 may increase the distance between HIF1α P402 and the active site within the catalytic core without compromising the interaction between the substrate and enzyme. However, such mutation did not markedly impact the efficient turnover of P564, suggesting that the varied residues between NODD and CODD may influence the substrate selectivity of PHD2.

PHD2-driven erythrocytosis is a relatively newly discovered genetic disease with the first reported case in 2006. Since then, there has been a large increase in reports owing to the addition of EGLN1 mutant screening in recent years for cases of idiopathic erythrocytosis³². As further awareness of the disease grows, screening will increase, leading to the likely identification of more disease-associated mutations. This study, therefore, will add to the groundwork knowledge of how mutant PHD2 induces disease in the context of the hypoxic response.

Materials and methods

Plasmids, antibodies, and peptides

Previously described plasmids were used as follows: pET-46-HIS₆-PHD2¹⁵, pGL3-VEGFA-HRE³⁴, pCDF-Ek/LIC-StrepII-HIS₆-HIF1α ODD¹², pcDNA3-HA-HIF1α ¹⁰. pRL-SV40 was procured from Promega. pcDNA3-3XFLAG-EGLN1 was generated by amplifying the EGLN1 gene from HA-EglN1-pcDNA3 (a gift from William Kaelin; Addgene plasmid # 18963) using primers 5’ – GCG GCG GGA TCC ATG GCC AAT GAC AGC G – 3’ and 5’ – GCG GCG TCT AGA CTA GAA GAC GTC TTT ACC GAC – 3’. The product was digested with BamHI/XbaI and subcloned into pcDNA3 vector containing an N-terminal 3XFLAG tag. pX330-U6-Chimeric_CBh-hSpCas9 was gifted to the lab from Feng Zhang (Addgene plasmid #42230). PHD2 mutants of pET-46-HIS₆-PHD2 and pcDNA3-3XFLAG-PHD2 were generated through QuikChange site directed mutagenesis (Agilent). Mutant sequences were confirmed via DNA sequencing. The following antibodies were used: α-HA (C29F4) from Cell Signaling, α-FLAG M2 (F1804) from Sigma-Aldrich, and α-PHD2/EGLN1 (D31E11) from Cell Signaling. Primary antibodies were diluted in TBS-T (20 mM Tris pH 7.6, 137 mM NaCl, 0.05% (v/v) Tween 20) with 0.02% (w/v) sodium azide for immunoblotting. Peptides were N-terminally acetylated for microscale thermophoresis (MST) and biotinylated for bio-layer interferometry (BLI). All peptides were amidated at the C-terminus. For microscale thermophoresis, peptides were reconstituted in 50mM Tris (tris(hydroxymethyl)aminomethane) pH 8.0 buffer. For bio-layer interferometry, peptides were reconstituted in 100% (v/v) DMSO.

PHD2 -/- HEK293A Cells

Wild-type (WT) human epithelial kidney (HEK293A) cells were subjected to the clustered regularly interspaces short palindromic repeats (CRISPR) method to generate homozygous PHD2 knockout (-/-) cells. The CRISPOR design tool was used to generate CRISPR guide RNAs (gRNA) targeting exon 1 in EGLN1⁴⁹. Forward and reverse gRNA oligonucleotides were phosphorylated and annealed before ligation into BbsI-digested pX330-U6-Chimeric_CBh-hSpCas9. WT HEK293A cells were transfected with pX330-EGLN1 gRNA using Lipofectamine 2000 according to manufacturer instructions. Single cell colonies were seeded into 96-well plates from the previously transfected cells at a density of 1.25 cells/mL. 200 μl of cells were seeded into each well resulting in a density of ∼0.25 cells/well. Single cell clones were screened for PHD2 knockout via immunoblotting for PHD2.

Dual Luciferase Assay

A dual luciferase assay was performed according to a modified previously reported protocol³⁴. PHD2 -/- HEK293A cells were grown and maintained in Dulbecco’s Modified Eagle Medium (DMEM, Wisent) with 10% (v/v) Fetal Bovine Serum (FBS, Wisent) and 100 IU/mL penicillin and streptomycin in a humidified incubator at 37°C and 5% CO₂. Transfection complexes consisted of 2 μg total DNA: 0.4 μg pcDNA3-HA, 0.9 μg pGL3-VEGFa-HRE, 0.1 μg pRL-SV40, 0.4 μg pcDNA3-HA-HIF1α, and 0.2 μg pcDNA3-FLAG3X-PHD2 (WT or mutant). Complexes were incubated for 15 minutes at room temperature with 8 μg polyethleneimine pH 7.2 (Polysciences) in 400 ml of OptiMEM^TM Reduced Serum Media (Gibco). Following incubation, transfection complexes were added to 5 x 10⁵ PHD2-/- HEK293A cells suspended in 400 μl OptiMEM^TM and incubated for 5 minutes. Cells were transferred to six well plates and incubated at 37°C, 5% CO₂ for 24 hours. Cells were lysed and processed according to the Promega Dual-Luciferase Reporter Assay System (#E1960) instructions. Firefly luciferase and Renilla luciferase activity were measured on a Varioskan Lux microplate reader (Thermo Fisher Scientific). Luciferase activity was normalized by dividing Firefly RLU values by the constitutively expressed, control Renilla RLU values. Normalized RLU values were transformed by setting the RLU value of the WT PHD2 sample to 1 and dividing the other sample values by the same transformation factor. A two-tailed t-test was performed to determine statistically significant differences between samples. A p-value below 0.0332 was considered significant. Protein levels of PHD2 were determined by immunoblotting for FLAG3X-PHD2. Vinculin was probed as a loading control.

Cycloheximide Chase Assay

HEK293A cells were cultured as described above. Transfection complexes consisting of 1.5 μg total DNA were incubated with 6 μg Polyethyleneimine pH 7.2 in 400 μl of OptiMEM^TM for 15 min at room temperature. Transfection complexes consisted of either 0.5 μg (WT, A228S, P317R, F366L, R371H) or 1.5 μg (G206C, W334R, H374R) of pcDNA3-FLAG3X-PHD2. Additional pcDNA3-FLAG3X-PHD2 was transfected for less stable mutants to ensure equal amounts of PHD2 expression. The total amount of DNA was brought to 1.5 μg with the addition of 1 μg of pcDNA3 if necessary.

Transfection complexes were added to 1.5 x 10⁵ cells suspended in 400 μl OptiMEM^TM and incubated for 5 minutes. Cells were plated in 60 mm plates with 5 plates for each mutant and 2 for WT. Cells were incubated for 24 hours at 37°C and 5% CO₂. Following incubation, media was replaced to remove Polyethyleneimine and fresh media with 10 μg/ml cycloheximide (CHX) was added. Cells were incubated in CHX supplemented media for 0, 3, 6, 12 and 24 hours. WT samples were incubated only for 0 and 24 hours. Cells were harvested through manual scraping and lysed via sonication in 250 μl of cold EBC lysis buffer (50 mM Tris, 120 mM NaCl, 0.5% (v/v) NP-40) supplemented with 1X protease inhibitor cocktail (BioShop PIC002.1). Protein levels were determined via Bradford assay and equal protein amounts of lysate were denatured by boiling for 5 minutes with 3X sample buffer (62.5 mM Tris, 10% (v/v) glycerol, 2% (w/v) SDS, 0.01% (w/v) bromophenol blue). Samples were then resolved on a 10% acrylamide SDS-PAGE gel for 90 minutes at 120 V. Proteins were transferred to polyvinylidene fluoride (PVDF) membrane via wet electrophoretic transfer for 70 minutes at 110 V. Membranes were blocked for 60 minutes at room temperature in 5% (w/v) skim milk in TBS-T. Membranes were then incubated with α-FLAG-M2 (1:1500) and α-vinculin (1:1500) at 4°C overnight. Following three washes with TBS-T, membranes were incubated with HRP-conjugated secondary antibodies (1:10,000 in 5% (v/v) skim milk in TBS-T) for 60 minutes at room temperature. Proteins were detected using a chemiluminescence solution and imaged with a ChemiDoc^TM MP Imaging System (BioRad). Immunoblot band density was quantified using BioRad Image Lab densitometry software. FLAG3X-PHD2 levels were normalized according to corresponding vinculin levels. Each 0-hour sample was set to 1 and each other time point was divided by this value to reflect relative protein levels. A two-tailed t-test was performed to identify significant differences between 0 hour and 24-hour timepoints. A p-value below 0.0332 was considered significant.

PHD2 (181-426) Expression and Purification

BL21 (DE3) E. coli cells (Novagen, #69450) were transformed with pET-46-HIS₆-PHD2 (wild type or mutant). 1 liter Luria Broth (LB) cell cultures were grown at 37°C until OD₆₀₀ reached ∼0.8. Cultures were then induced with a final concentration of 1 mM Isopropyl ß-D-1-thiogalactopyranoside (IPTG) for 16 hours at 16°C. Cultures were harvested via centrifugation and stored at −80°C for future use. His₆-PHD2 (181-426) was purified according to a previously described protocol⁵⁰. Bacterial pellets were resuspended in lysis buffer (50 mM Tris-HCl pH 7.9, 500 mM NaCl, 5 mM imidazole) supplemented with 1X SigmaFast protease inhibitor cocktail (Sigma-Aldrich). Cells were lysed with 3 passes through a cell disruptor at 30 kpsi. Cell lysate was cleared via centrifugation at 30,000 g for 45 minutes. Cleared lysates were applied to a 2 mL Ni-NTA resin (ThermoFisher). The resin was washed with 10 mL of lysis buffer followed by 2 separate 30 mL washes with Wash buffer (50 mM Tris-HCl pH 7.9, 500 mM NaCl, 30 mM imidazole). His₆-PHD2 was eluted from the resin with 10 mL of elution buffer (50 mM Tris-HCl pH 7.9, 500 mM NaCl, 1 M imidazole). The eluate was dialyzed against 2 liters of dialysis buffer (50 mM Tris-HCl pH 7.5) using a 10 kDa cut-off dialysis membrane. After 4 hours, the dialysis buffer was replaced with 2 fresh liters and 1.5 units of thrombin was added per mg of His₆-PHD2. His₆-PHD2 was dialyzed overnight with thrombin at 4°C. The dialyzed PHD2 was applied to Ni-NTA resin equilibrated with dialysis buffer to separate thrombin-cleaved protein from uncleaved protein and His₆ tags. The resin was washed with dialysis buffer supplemented with 5 mM imidazole. The flowthrough and wash were combined and concentrated to ∼600 μl using an Amicon 10 kDa cut-off centrifugal concentrator (Millipore). The concentrated PHD2 was cleared at 13,000 rpm for 10 minutes then applied to a Superdex^TM 200 Increase 10/300 (Cytiva) column equilibrated with dialysis buffer or labeling buffer (50 mM HEPES (N-2-hydroxyethylpiperazine-N’-2-ethanesulfonic acid) pH 7.5, 150 mM NaCl). Elution was monitored via UV absorbance and samples with PHD2 were collected. SDS-PAGE gels were run to confirm purity. PHD2 proteins were concentrated to > 2 mg/ml. The final concentration was measured via UV absorbance at 280 nm (ε_0.1% = 1.34). Unlabeled protein was flash frozen and stored at −80°C for use in circular dichroism and nuclear magnetic resonance.

PHD2 proteins used for microscale thermophoresis (MST) were labelled with Alexa Fluor-647 N-hydroxylsuccinimide ester dye (Thermo Scientific) following purification according to a previously reported protocol³⁴. PHD2 proteins were diluted to 1 mg/ml and Alexa Fluor-647 dye was added to 500 μl of protein in a 4:1 molar ratio (dye: PHD2). Following incubation at room temperature in the dark for 1-hour, excess dye was removed using a PD-10 Sephadex^TM G-25 M desalting column (Cytiva) equilibrated with dialysis buffer. Labelled protein subjected to a 10-fold dilution in dialysis buffer and was then concentrated using a 10 kDa cut-off Amicon centrifugal concentrator to remove excess dye. Alexa Fluor-647 labelled PHD2 was then brought to 10 mM in dialysis buffer, flash frozen and stored at −80°C for use in MST.

HIF1α ODD (394-574) Expression and Purification

DNA corresponding to the sequence of human HIF-1α ODD (residue 394-574) with a C-terminal Strep Tag II (WSHPQFEK) was cloned into a pCDF Ek/LIC vector (EMD Millipore). Thrombin cleavage sites were included after the N-terminal-His₆ tag and before the C-terminal Strep Tag II. HIF-1α ODD was expressed in Rosetta-2 (DE3) cells. To produce ¹⁵N-labeled proteins, cells were grown in 2X M9 minimal media, supplemented with a final concentration of 1 g/L ¹⁵NH₄Cl (ACP chemicals) and 4 g/L D-glucose (Bioshop). Expression and purification of HIF-1α ODD was previously described¹². Briefly, cleared lysate was applied to Ni-NTA resin (ThermoFisher) and eluted with 300 mM imidazole. The elution was then incubated with StrepTactin XT 4Flow resin (IBA Lifesciences) overnight. Protein was eluted from the StrepTactin resin with 50 mM biotin. This process was repeated with the flowthrough from the StrepTactin beads to capture remaining unbound protein. Protein was concentrated and run on a Bio-Rad SEC650 10/300 column. Fractions containing HIF-1α ODD were then run on a HiResQ 5/50 (Cytiva) anion exchange column. After tag cleavage with thrombin, purified HIF-1α ODD was concentrated to 0.9 mM and stored at - 80°C for future use.

Circular Dichroism

Purified PHD2 proteins were buffer exchanged on PD-10 Sephadex^TM G-25 M desalting columns (Cytiva) for 5 mM KH₂PO₄/K₂HPO₄ pH 7.4 buffer. Proteins were then diluted to 0.1 mg/ml. Far-UV wavelength scans were collected from 190-260 nm on a Jasco-J-1500 CD spectrophotometer with a 1-mm pathlength quartz cuvette (Helma). CD values were averaged from 20 accumulations, baseline buffer subtracted and converted into mean residue ellipticity (degrees cm² dmol^-1). Spectra were smoothed using Jasco Spectra Analysis software. Wavelength scans were run in triplicate and the average mean residue ellipticity was plotted with other PHD2 CD curves.

Melting curves were generated by monitoring CD values for each PHD2 protein from 25-95°C at 220 nm with 1°C steps. Values were baseline buffer subtracted and converted to mean residue ellipticity. Triplicate non-transformed data was plotted. Curves were transformed by setting the average peak value between 25-37°C to 1.0 and normalizing other points according to the same normalization factor. Curves were fitted with a sigmoidal curve on Prism. Corresponding R² values are recorded in Table 1. Melting temperature (T_m) for each protein were determined from the EC50 of each sigmoidal curve. Curves were plotted using GraphPad Prism version 10.2.3.

Microscale Thermophoresis

Microscale Thermophoresis (MST) binding tests and analysis were performed according to a previously recorded protocol³⁴. A Monolith NT.115 (Nanotemper Technologies) was used at room temperature to record MST measurements. 2X stock solutions of Alexa Fluor-647-labeled PHD2 with necessary hydroxylation reagents was made (50 nM Alexa Fluor-647-PHD2, 50 mM Tris-HCl pH 7.5, 1 mM N-oxalylglycine (NOG), 10 μM ferrous sulfate, 5 mg/ml BSA). ∼1 mg lyophilized HIF1α _{555- 574} CODD and HIF2α _522-542 CODD peptides were dissolved in 200 μl of 50 mM Tris HCl pH 8.0 and their concentrations were measured via UV absorbance with the following extinction coefficient: 1490 M^-1 cm^-1. Starting with the stock HIFα peptide solutions, two-fold serial dilutions were performed to generate 16 different 15 μl HIFα concentrated solutions. 15 μl of 2X Alexa-647-PHD2 solution was then added to each HIFα peptide dilution resulting in a final Alexa-647-PHD2 concentration of 25 nM. The solutions were incubated at room temperature for 10 minutes then loaded into standard MST capillaries (Nanotemper Technologies). MST measurements were recorded using 20% fluorescence excitation LED power and 40% (medium) infrared (IR) laser power. Measurements were recorded for 5 seconds prior to IR laser activation, 20 seconds during IR activation and 3 seconds following IR deactivation. The experimental data was analyzed and fit with a 1:1 K_d binding model using M.O. affinity analysis software (Nanotemper Technologies). Curves were plotted on GraphPad Prism (version 10.2.3).

Bio-layer Interferometry

The binding affinities of HIF1α peptides and WT or P317R PHD2 were measured by bio-layer interferometry (BLI) using a single channel BLItz instrument (Sartorius). Purified PHD2 was concentrated to 0.45 mM and diluted into BLI Kinetic Buffer (20 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM L-ascorbic acid (Sigma), 100 μM ferrous chloride tetrahydrate (Sigma), 1 mM N-oxalylglycine (Sigma), 0.1% (w/v) Bovine Serum Albumin and 0.1% (v/v) Tween-20). Following reconstitution to a final concentration of 2 mg/ml in DMSO, biotinylated HIF1α peptides were diluted 200-fold in the BLI Kinetic Buffer. HIF1α peptides were immobilized on a streptavidin-coated biosensor (Sartorius) for 120 seconds before measuring association to WT PHD2 or P317R PHD2 in a concentration series over 120 seconds. Subsequently, the biosensor was immersed in the BLI Kinetic Buffer for 120 seconds to measure dissociation of the analyte. The sensorgrams were referenced and step corrected, and the binding affinities and kinetic rates were calculated based on global fit of the data to a 1:1 binding model using the BLItz Pro v.1.3.0.5 software. All binding measurements were performed in technical triplicates.

Real-time Prolyl-hydroxylation Kinetic Measurement

Real-time prolyl hydroxylation of HIF1α ODD was measured according to a previously reported protocol based on Nuclear Magnetic Resonance (NMR)¹². NMR experiments were performed at the UHN Nuclear Magnetic Resonance Core Facility on a Bruker 800 MHz US2 equipped with a NEO console and a 5 mm triple-resonance ¹³C optimized TXO cryoprobe, Z-gradient. NMR samples were prepared in 50 mM Tris-HCl pH 7.5, 100 mM NaCl, and 0.5 mM Tris(2-carboxyethyl) phosphine hydrochloride (TCEP-HCl) (NMR buffer) supplemented with 5% D₂O. Prior to the experiments, NMR buffer was gassed with 99.6% oxygen (Messer) for 40 minutes and then used to dilute 0.9 mM of HIF1α ODD to a final concentration of 0.18 μM. Samples were transferred to a 5-mm NMR tube pre-purged with 99.6% oxygen for 10 minutes. The real-time hydroxylation experiments were conducted at 25°C using the ¹H-¹⁵N HSQC pulse sequence with sodium 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) as reference. For each experiment, a ¹H-¹⁵N HSQC spectrum of HIF1α ODD was acquired prior to initiating the hydroxylation reaction with the addition of purified WT or mutant PHD2 at a molar ratio of 1:20 versus HIF1α ODD with 5 mM α-ketoglutaric acid disodium salt dihydrate (Sigma), 2 mM L-ascorbic acid (Sigma), and 0.1 mM ferrous chloride tetrahydrate. Multiple ¹H-¹⁵N HSQC spectra with acquisition time of 602 seconds were acquired over the course of 20.18 hours. After 6 consecutive acquisitions of ¹H-¹⁵N HSQC spectra, the subsequent spectra were acquired with interscan delays of 2400 seconds. Kinetic measurements with WT, A228S, and P317R PHD2 were performed in technical duplicates.

Real-time hydroxylation of P402 and P564 of HIF1 α was measured by monitoring resonance shifting of A403 and I566¹² respectively, in all ¹H-¹⁵N HSQC spectra. Both resonances shifted as reflections of changes in the local chemical environments upon prolyl-hydroxylation. Reaction progress was reported as the peak intensity of the shifted resonance over the sum of peak intensities for both shifted and unshifted resonances. All spectra were processed with the NMRPipe software⁵¹ and analyzed using the CcpNmr Version-3 software⁵². Kinetic curves were plotted using GraphPad Prism (version 10.2.3).

Supplementary Information

Supplementary Table 1.

Supplemental Figure 1.

Supplemental Figure 2.

Supplemental Figure 3.

Acknowledgements

We thank the members of the Lee and Ohh labs for their critical comments and helpful discussions. This study was supported by funds from the Canadian Institutes of Health Research (PJT-191811 to M.O. and MOP-133694 to J.E.L.). The UHN NMR Core Facility was supported by the Princess Margaret Cancer Foundation. Infrastructure for biophysics and NMR was funded by the Canada Foundation for Innovation (CFI).

Additional information

Author contributions

C.C.T. designed and performed the experiments, interpreted the data, and wrote the manuscript. W. H. performed the Bio-layer interferometry. W.H. and G.M.C.G.-S. performed the real time Prolyl-hydroxylation assay. M.H. assisted in performing the CHX chase assays. C.C.T., F.G.F, and M.O. conceptualized the project. G.M.C.G.-S., M.I., and J.E.L. contributed new reagents and analytical tools. M.O. interpreted the data, wrote and edited the manuscript. All authors reviewed and edited the manuscript.

Funding

Canadian Institutes of Health Research (PJT-191811)

Canadian Institutes of Health Research (MOP-133694)

Additional files

Supplementary Data