Comprehensive in silico and functional studies for classification of EPAS1/HIF2A genetic variants identified in patients with erythrocytosis

Abstract

Gain-of-function mutations in the EPAS1/HIF2A gene have been identified in patients with hereditary erythrocytosis that can be associated with the development of paraganglioma, pheochromocytoma and somatostatinoma. In the present study, we describe a unique European collection of 41 patients and 28 relatives diagnosed with an erythrocytosis associated with a germline genetic variant in EPAS1. In addition we identified two infants with severe erythrocytosis associated with a mosaic mutation present in less than 2% of the blood, one of whom later developed a paraganglioma. The aim of this study was to determine the causal role of these genetic variants, to establish pathogenicity, and to identify potential candidates eligible for the new hypoxia-inducible factor-2 α (HIF-2α) inhibitor treatment. Pathogenicity was predicted with in silico tools and the impact of 13 HIF-2b variants has been studied by using canonical and real-time reporter luciferase assays. These functional assays consisted of a novel edited vector containing an expanded region of the erythropoietin promoter combined with distal regulatory elements which substantially enhanced the HIF-2α-dependent induction. Altogether, our studies allowed the classification of 11 mutations as pathogenic in 17 patients and 23 relatives. We described four new mutations (D525G, L526F, G527K, A530S) close to the key proline P531, which broadens the spectrum of mutations involved in erythrocytosis. Notably, we identified patients with only erythrocytosis associated with germline mutations A530S and Y532C previously identified at somatic state in tumors, thereby raising the complexity of the genotype/phenotype correlations. Altogether, this study allows accurate clinical follow-up of patients and opens the possibility of benefiting from HIF-2α inhibitor treatment, so far the only targeted treatment in hypoxia-related erythrocytosis disease.

Introduction

Erythrocytoses are characterized by an elevated red cell mass of more than 125% of the predicted value for the age and body mass of the subject, usually reflected by increased levels of hemoglobin (Hb) and/or hematocrit (Ht) values.¹ Erythrocytoses can be acquired as in the primary polycythemia vera, a myeloproliferative neoplasm as a consequence of a gain of function mutation in the JAK2 gene (p.Val617Phe), or secondary to diverse pathological situations (pulmonary or heart disease, kidney cancer, carbon monoxide poisoning, etc.) due to an increased secretion of erythropoietin (EPO). On the other hand, erythrocytosis can be observed in a context of inherited disease, which can be primary when there is an intrinsic defect in the progenitor cells of the bone marrow (EPOR mutations), or secondary when the oxygen-sensing pathway is dysregulated and EPO is produced at a high level (VHL, EGLN1/PHD2, EPAS1/HIF2A mutations).

This study focuses on the EPAS1 gene that encodes the hypoxia-inducible factor-2 α (HIF-2α), a major player in the oxygen-sensing pathway, also known as the hypoxia pathway.

HIF is a transcription factor that is stabilized when the oxygen concentration is reduced.² HIF is a α/β hetero-dimer consisting of a tightly regulated oxygen-labile α-subunit and a constitutive β-subunit. The HIF-α subunits (HIF-1α, 2α, 3α) contain an oxygen-dependent degradation (ODD) domain, and two independent transcriptional activation domains. In the presence of oxygen, the HIF-α subunits are hydroxylated by prolyl hydroxylases (PHD1-4)³ that regulate their stability and an asparaginyl hydroxylase factor inhibiting HIF (FIH) that regulates their transcriptional activity.⁴ The PHD hydroxylate proline residues are located within the HIF-α ODD (P402 and P564 for HIF-1α; P405 and P531 for HIF-2α). This hydroxylation allows the binding of the von Hippel-Lindau (VHL) protein, a recognition subunit of an E3 ubiquitin ligase multiprotein complex. Binding of VHL to HIF-α subunits induces ubiquitination, which targets them for degradation by the proteasome. Under hypoxic conditions, when the co-factor oxygen is limiting, hydroxylation of HIF-α subunits slows down and results in its stabilization. HIF-α then translocates to the nucleus, associates with the HIF-1α subunit and, upon recruiting appropriate co-activators, the HIF-α/b heterodimer binds to hypoxia response elements (HRE) within DNA and activates expression of HIF target genes.⁵ HIF regulate the transcription of more than 200 genes involved in many pathways, notably erythropoiesis via the synthesis of erythropoietin (EPO)⁶ and iron metabolism regulation (tranferrin, transfer-rin receptor, divalent metal transporter 1 [DMT1], ferroportin). HIF-2 is the main isoform that controls EPO expression which regulates the proliferation and differentiation of erythroid progenitors, thereby linking decreased tissue oxygenation to an adequate erythropoietic response.⁷ HIF-2α is an 870 amino acid protein encoded by the EPAS1 gene located in 2p21, that contains 16 exons. The first mutation in the EPAS1 gene has been described in 2008 associated with polycythemia, also called erythocytosis (ECYT) developed by patients over three generations.⁸ Since then, more than 40 other cases have been published (Online Supplementary Table S1). These are always germline missense mutations in the heterozygous state. Patients described with a mutation in the EPAS1 gene developed erythrocytosis frequently associated with high EPO levels and are at increased risk of pulmonary hypertension and thrombotic events (thrombosis, infarction, pulmonary embolism).^8-11 This disease has been classified as ECYT4 (OMIM#603349, MIM#611783). Importantly, mutations in EPAS1/HIF2A gene are also responsible for the development of pheochromocytomas, paragangliomas (PPGL), somatostatinomas and ocular lesions¹² sometimes associated with erythrocytosis and rare cases of cyanotic congenital heart disease and hemangioblastomas (Online Supplementary Table S2).^13,14,12,15 In these cases, the mutations have been found at the somatic level within the tumor, but further investigations sometimes revealed that the mutation was acquired during development and was actually present in a mosaic state.¹⁶

The majority of these missense mutations are located in exon 12 and cluster close to the proline residues Pro531, hydroxylation of which regulates HIF-2α stability. Combined biochemical and cellular assays showed that the majority of these mutations may reduce both hydroxy-lation of HIF-2α by the PHD, and subsequent recognition of HIF-2α by pVHL. Functional studies showed that mutations associated with tumor development have a more severe gain-of-function than mutations associated with erythrocytosis alone.¹⁷ These most deleterious mutations (amino acids 529-532) affect residues close to the key proline 531.

In the present collaborative study, we describe a unique series of 43 patients diagnosed with an erythrocytosis in whom we identified 33 different genetic variants in EPAS1. The aim of this study is to determine the causal role of these genetic variants in the pathogenesis of erythrocytosis and the potential risk for developing tumors. In order to study the functional consequences of the different HIF-2α mutations, we performed luciferase reporter assays. We used a vector encoding the firefly luciferase driven by multiple proximal and distal regulatory elements of the HIF-2 target gene EPO, the key target gene linked to the development of erythrocytosis. A recent publication suggested that the EPO gene may contain complex regulatory elements in its proximal promoter based on the identification of a mutation located in a region (c.-136, upstream the ATG codon¹⁸) not included in previously existing luciferase constructs.^19,20 We therefore generated a luciferase vector driven by a substantially extended region of the EPO promoter and distal enhancer regions. In order to accurately quantify subtle changes in HIF-2α activity real-time periodical luminescence measurements were performed.

Methods

Sequencing

Samples were obtained from laboratories specialized in diagnosis and research of idiopathic erythrocytosis after exclusion of classical causes of erythrocytosis (polycythemia vera or secondary erythrocytosis associated with particular renal, cardiac or pulmonary disorders). After receiving approval from the Ethics Committee and obtaining written informed consent from the proband and family members, blood samples were collected and DNA was extracted for genetic analysis. Molecular screening of genes associated with erythrocytosis was performed by high throughput sequencing with different technologies, depending on the sequencing center (see the Online Supplementary Appendix).

In silico analysis

Genetic variants located in the EPAS1 gene with a global frequency lower than 5.10^-4 in the gnomAD v3 database were selected. The MetaDome²¹ analysis was performed on the EPAS1 gene by using the GENCODE: ENST00000263734.3, RefSeq: NM_001430.4 and UniProt: Q99814. The effect of each non-synonymous variant was assessed using Protein Variation Effect Analyzer (PROVEAN)21 Protein Batch v1.1.3 prediction integrative webtool. ENSEMBL ID HIF-2α wild-type amino acid sequence (ENSP00000263734) was used as a reference. Cutoff of -2.5 and -4.1 were applied to PRO-VEAN predictions. EPAS1 variants were studied by using MobiDetails,²² an annotation platform for DNA variants. Values obtained by single-and meta-predictor tools were normalized (0-1), 0 being the less damaging and 1 the most damaging for each predictor. For the final interpretation and classification of the variants, we used the ACMG (American College of Medical Genetics and Genomics) criteria and guidelines.²³ The ACMG uses the following classification to describe variants identified in Mendelian disorders: class 1: benign; class 2: likely benign; class 3: variant of uncertain significance (VUS); class 4: likely pathogenic; class 5: pathogenic. Criteria used in our study are detailed in the Online Supplementary Appendix.

Full erythropoietin promoter plasmid generation

The full human core EPO promoter sequence was recovered from ENSEMBL (7 dna:chromosome:GRCh38:7: 100720400:100721001:1, ENSR00000833692). Briefly, the pGL3-5’HRE290-FullProm-3’HRE126 EPO promoter-driven luciferase plasmid was generated after polymerase chain reaction (PCR) amplification of a 604 bp fragment of genomic DNA extracted from HeLa cells that was inserted into a pGL3-luciferase vector (details are provided in the Online Supplementary Appendix).

End point luciferase reporter assays

Cells were cultured as detailed in the Online Supplementary Appendix. Briefly, 4x10⁵, 3.5x10⁵ and 6.5x10⁵ cells were transiently transfected with 500 ng or 100 ng reporter plasmid in a 6-well format using CaCl₂ or JetOptimus (Polyplus), respectively for HEK293T and Hep3B or Kelly cells, and 1 mg of YFP-HIF-1α, YFP-HIF-2α or pcDNA3-HA-HIF-2α (Add-gene) constructions. In order to control for differences in transfection efficiency and extract preparation, 50 ng or 75 ng pRL-SV40 Renilla luciferase reporter vector (Promega) was co-transfected, respectively for HEK293T and Hep3B or Kelly cells. The next day, cultures were evenly split onto 6-well plates, incubated for an additional 24 hours, under normoxic or hypoxic conditions (0.2% O₂, 5% CO₂ and 37°C). Cells were lysed with passive lysis buffer and luciferase activities of duplicated wells were determined using the Dual Luciferase Reporter Assay System (Promega) as described before.²⁴ Reporter activities were expressed as relative Firefly/Renilla luciferase activities normalized to control under hypoxic conditions. All reporter gene assays were performed at least three times independently. Proteins were extracted and immunoblotted to quantify the HIF-2α proteins as described before²⁵ (see the Online Supplementary Appendix).

Real-time luciferase reporter assays

Luciferase assays were performed on HEK293T cells seeded in 24-well Black Visiplate Perkin Elmer) (1×10⁵ cells per well), 24 hours before transfection. Cells were transfected by using jetPRIME^® (Ozyme Polyplus). The expression vectors pcDNA3-HA-HIF-2α (25 ng) were co-transfected with the pGL3-5’HRE-FullProm-3’HRE-EPO promoter-driven luciferase plasmid (100 ng), and empty vector for a total amount of 500 ng transfected DNA. Luciferase activity was monitored over 48 hours using the bioluminometer WSL-1565 Kronos HT^® (ATTO). Cells were harvested and lysed in extraction buffer (Macherey Nagel) for quantification of transfected plasmids by PCR (for details see the Online Supplementary Appendix).

Statistical analysis

Values in the figures of the end point luciferase assays are presented as mean ± standard error of the mean (SEM). For the real-time luciferase reporter assay, differences in means among multiple groups were analyzed by using one-way ANOVA of Kruskal-Wallis and Dunn's post hoc tests. All statistics were performed with GraphPad Prism software 7.05. Values of P≤0.05 were considered statistically significant.

Results

Diagnosis and genetic screening

Patients with erythrocytosis were selected after exclusion of classical causes of erythrocytosis (polycythemia vera or secondary erythrocytosis associated with particular renal, cardiac or pulmonary disorders). Genetic screening was performed by using next-generation sequencing (NGS) panels dedicated to erythrocytosis, and a total of 1,450 patients were sequenced at all centers. Results are presented in Table 1 and show 33 missense genetic variants identified in a total of 43 patients (41 patients with heterozygous germline variant and 2 patients with mosaicism) and 28 relatives from four different European countries. Ten variants have already been described.^15,26-43 Twenty-four variants are located in exon 12 and extend from amino acids 525 to 658 surrounding the key proline in position 531. Family history of erythrocytosis is present in 14 families. The pedigrees are of variable size (Figure 1; Online Supplementary Figure S1) and can show a multigenerational history of erythrocytosis (family [F] 14, F22), but sometimes also display a family history related to the consequences of undiagnosed erythrocytosis, such as ischemic stroke (F13). Clinical manifestations are in general variable, ranging from mild to severe erythrocytosis (Ht up to 77.5% in patient 19) (Table 1 for probands; Online Supplementary Table S3 for relatives). The serum EPO level is rarely elevated except in two patients (up to 7,500 UI/mL in patient 20). However, we were unable to obtain full information on the patient's phlebothomy time frame that may influence these factors.

Figure 1.Pedigree of patients carrying a mutation in EPAS1. Roman numerals indicate generations. Squares indicate men and circles women; black filling indicate the development of confirmed erythrocytosis; gray filling indicate a suspected erythrocytosis; arrows indicate probands; +, indicates a positive genetic screen.

Associated symptoms characteristic to EPAS1 mutations have been observed in a limited number of patients. Pulmonary arterial hypertension (PAH) was developed by two patients and ocular lesions have been detected in three patients as previously reported in association with EPAS1 mutations.¹² A history of thrombosis was reported in six patients and six relatives which mainly were of the stroke type. Close examination of the medical records of family 13 (variant D525H identified in the 76-year-old proband and her son with no history of thrombosis) found a stroke in five sisters at an age of onset between 60 and 80 years, however, no genetic test has been done in these relatives to indicate that the variant is associated with thrombosis. We also found a stroke in a child (patient 20, variant Y532C) with a Moya-Moya type cerebral vascular malformation. Unfortunately, no information could be obtained regarding the existence of any other cardiovascular risk factor in the relatives.

Table 1.Clinical data of probands with a genetic variant identified in EPAS1.

Of major importance, a paraganglioma has been detected in one mosaic patient (patient 19). Patient 19 presented at 16 months old with a strongly elevated Ht (77.5%), Hb (24.2 g/dL) and EPO (573 UI/mL). A first screening by NGS reads did not reveal germline mutations. In the presence of such a severe diagnosis, whole-exome and subsequently whole-genome sequencing were performed on the patient and parent’s DNA, but no significant genetic abnormality was identified. At the age of 10 years, due to severe hypertension highly suggestive of the presence of a catecholamine secreting tumor, a deeper re-analysis of NGS reads allowed the identification of the c.1591C>T, p.Pro531Ser variant at a maximum rate of 1.9% of the reads that was confirmed by using droplet PCR (Online Supplementary Figure S2). Indeed, two abdominal paragangliomas were identified and resected. Sequencing of the tumor using droplet PCR showed a variant allele frequency of 60%, confirming the role of EPAS1-mutated cells in oncogenesis (Online Supplementary Figure S2B). After surgical resection of the paragangliomas, disappearance of the hypertrophic cardiomyopathy and the non-compaction syndrome initially found was noted. A second patient presented at a very young age (patient 18, 9 months old) with severe erythrocytosis. The mutation c.1589C>A, p.Ala530Glu was detected at 1.5% of reads by NGS and confirmed by droplet PCR. The risk of tumor development in this patient is carefully monitored.

In silico analysis of the genetic variants

Subsequently, we employed the MetaDome web server to analyze the mutation tolerance at each position of the HIF-2α protein. The amino acids targeted by the missense genetic variants identified in our study were localized on the HIF-2α protein map (Figure 2). The resulted tolerance is reported in Table 2 for each variant, and the detailed score are indicated in the Online Supplementary Table S4. The majority of the variants target amino acids are located in the oxygen-dependent degradation (ODD) domain, an intolerant zone surrounding the key residue proline P531. The location of this hydroxylation site is shown in the 3D structure of HIF-2α (upper left panel of Figure 2). We then used the Mobidetails online platform which gathers many sources of data for the interpretation of DNA variants in the context of molecular diagnosis. For each variant, we analyzed the scores obtained with single- and meta-predictors and classified the variants as benign when both scores were <0.5, and as deleterious when both scores were >0.7 (Figure 3A). These scores are illustrated with a Radar view presenting the prediction results obtained from the different in silico tools (Online Supplementary Figure S3). The analysis was completed by using the PROVEAN in silico tool. We classified the variants depending on their score regarding the cutoff: eutral >-2.5, possibly deleterious <-2.5 and >-4.1, deleterious <-4.1 (Figure 3B). The final classification of the variants was based on a global analysis of in silico and functional studies (see below) with the ACMG (American College of Medical Genetics and Genomics) criteria which previously developed guidance for the interpretation of sequence variants (see the last column of Table 2, and details of scores in the Online Supplementary Table S4)²³.

Generation of an erythropoietin promoter-driven reporter vector

As HIF-2α-dependent regulation of EPO expression represents the key pathophysiological mechanism involved in the occurrence of erythrocytosis, we focused our functional studies on reporter assays using a luciferase gene driven by EPO regulatory elements. Previously existing luciferase constructs under the control of the EPO promoter contain the proximal region located between position -194 and -341 upstream the ATG codon (termed minimal promoter, labeled in black; Online Supplementary Figure S4).^19,20 A mutation located at position -136 in the EPO promoter has been recently described and linked to the development of erythrocytosis in two families.¹⁸ We therefore introduced a larger promoter region into the reporter constructs from position -17 to -564 of the EPO promoter (construct termed Full EPO promoter, labeled in gray; Online Supplementary Figure S4).

Functional studies by using end point and real-time reporter luciferase assay

We performed luciferase reporter assays using this novel EPO promoter-driven construct in the absence or presence of the distal 5’ and 3’ hypoxia-responsive elements (HRE) responsible for tissue-specific and hypoxia-inducible regulation of EPO gene expression in the kidney and the liver, respectively. Transient transfection of these constructs showed higher luciferase activity of the full promoter under hypoxic conditions compared to the minimal promoter. Furthermore, transient co-transfection of vectors encoding wild-type HIF-1α or HIF-2α displayed an increased luciferase signal of both the minimal and enlarged promoter, but resulted in preferential response to HIF-2α only when the enlarged promoter was present under both normoxic and hypoxic conditions. A similar preferential HIF-2α-dependent increase in luciferase activity was obtained with reporter genes in the presence or absence of distal HREs (Figure 4; Online Supplementary Figure S5). The same experiments were performed with co-transfection of plasmids encoding the mutated HIF-2α proteins, including P531A as positive control in HEK293, Hep3B and Kelly cells (1 clone for each variant in triplicates). We observed a significant gain of function for the P531S mutant in all three cell lines and for A530S, G537R and F540L in Hep3B cells (Figure 5; Figure 6B; Online Supplementary Figure S7A). We also performed these experiments under hypoxic conditions confirming the results under normoxic conditions, with a significant gain of function for the P531S mutant in three cell lines and for the E538K in Hep3B (Online Supplementary Figures S6 and S7B).

Figure 2.HIF-2α three-dimensional structure and HIF-2α protein tolerance landscape. The main domains of the HIF-2α protein are presented in purple in the middle of the figure. The severity of the impact induced by sequence variation targeting each amino acid was determined by using the MetaDome in silico analysis tool. It is presented with a color code, from blue (tolerant) to red (intolerance) on the right. The variants identified in our series are listed. PAS (Per-ARNT-Sim) domain: dimerization domain, ODD: oxygen dependent degradation domain, c-TAD: c-terminal transactivation domain. In the upper left panel, the three-dimensional structure of HIF-2α is presented (from AlphaFold Protein Structure Database). The HIF-2α inhibitor (MK-6482/Belzutifan/Welireg) binds to a region that causes a conformational change of the amino acids H293 and M252 (red stars in the figure), while the amino acids found mutated in patients target the close vicinity of the hydroxylation proline P531 located in the ODD domain.

Table 2.Results of in silico analysis and functional studies of a genetic variant identified in EPAS1.

Figure 3.In silico prediction of the impact of an amino acid substitution on the HIF-2α protein. (A) Representation score obtained by the predictors analyzed by the Mobidetails annotation platform. Values are normalized (0-1), 0 being the less damaging and 1 the most damaging for each predictor. The graph indicates the mean normalized score obtained by single predictors (SIFT, Polyphen 2 HumDiv an0d HumVar) and meta-predictors (Fathmm, REVEL, ClinPred, Meta SVM, Meta LR, Mistic). (B) Representation of protein variation effect analyzer (PROVEAN) analysis performed on each non-synonymous HIF-2α variant. Variants with a score above -2.5 were predicted to be neutral. A threshold score below -2.5 was predicted to be deleterious and a more stringent threshold score of -4.1 is associated with increased specificity.

Figure 4.End point reporter gene assays demonstrate increased erythropoietin promoter-driven luciferase activity with a full erythropoietin promoter compared to minimal construct. HEK293 cells were co-transfected with the erythropoietin (EPO) minimal or full promoter-driven constructs containing the 5’HRE and 3’HRE and with HIF-1α or HIF-2α isoform overexpression plasmids, as indicated. The luciferase activity is reported as the induction compared to the control (Ctrl) under hypoxic condition and represents the ratio of firefly (FF) to Renilla (RL) relative light units (RLU). Each column represents the mean ± standard error of the mean of 4 to 14 different experiments performed in duplicate. One-way ANOVA (*P≤0.05; **P≤0.01; ****P≤0.0001).

In order to increase the sensitivity of the reporter assay we reduced the amount of transfected HIF-2α expression vectors. We used HEK293 cells in order to avoid additional induction of the reporter vector by endogenous HIF-2α (expressed in Hep3B and Kelly cells). In order to determine the optimal time frame to quantify the luciferase activity after transfection, we followed it using a real-time luciferase measurement system for 48 h (Figure 6A). We set the 100% activity to the relative light unit (RLU) values obtained at the plateau of the wild-type protein after 30 h of expression from the beginning of the reaction (i.e., 34 h after transfection). We observed a large scale of activity from wild-type (around 100%) to elevated gain-of-function (150-300%). The mean of results derived from repeated experiments obtained at the plateau are shown in Figure 6B. We calculated the slope of the different curves to obtain an indicator of the velocity of the reaction (Figure 6C). The use of three independent clones for each variant shows consistent results with strong reliability of this test within an experiment, but some variants display variable behavior between experiments, demonstrating the importance to replicate assays. Altogether, a significant increase of the activity and/or the velocity of the reaction was found for the tested variants identified in patients targeting amino acids from D525 to E538. One variant presents an activity very close to the wild-type protein and was classified as benign (Q557H), the other variants have been classified as variant of unknown signification (VUS).

Discussion

We report here the largest collection of EPAS1/HIF2A gene variants from a large cohort of patients with idiopathic erythrocytosis. Variants were characterized using detailed in silico and new functional studies.

The use of a wide range of in silico prediction tools can be very useful for classifying variants in genetic diagnosis. However, the pathogenicity of a genetic variant can be difficult to classify as pathogenic when the disease is associated with a gain-of-function mutation. For this reason, functional studies need to be included in the diagnostic tools. In the case of erythrocytosis, the functional tests must be very sensitive because it is known that many mutations may be hypomorphic,^44–46 especially because the genes of the hypoxia pathway play a major role in physiology and a germline alteration that is too severe and would likely be incompatible with life. The construction of an improved, more sensitive EPO promoter-driven vector in combination with the use of a real-time luciferase reporter assay measurements allowed us to accurately measure gain of function of new HIF-2α mutants. Altogether, combined in silico, segregation and functional studies performed in this study allowed the classification of 11 variants as likely pathogenic or pathogenic (including 4 never previously described) in 17 patients and 23 relatives. Our study further expanded the region of HIF-2α associated with erythrocytosis. Thus far, this region spanned from amino acids 529 to 537, but the classification of pathogenic variants of amino acids 525, 526 and 527 prompt an expansion of this area. Additional investigations will be necessary to definitely classify the E538K and F540L variants. Our results will allow an appropriate follow-up of families at the clinical level, especially with regard to complications and associated diseases. It is indeed intriguing that so few patients carrying pathogenic variants present with additional symptoms associated with EPAS1 mutations:

PAH was only described in two patients (P#20 and 27), ocular lesions in two patients (P#19, 20) and thrombosis and ischemic accidents in five families (P#13, 15, 20, 22, 29). Interestingly, patient 20 carrying the variant p.Tyr532Cys also suffers from the Moya-Moya disease. This is a rare and chronic disease of mostly unknown causes that affects the blood vessels in the brain. No link has ever been found with the hypoxia pathway genes and additional investigations should be performed on the potential role of EPAS1 in the etiology of Moya-Moya disease. Some families are currently being closely monitored due to the high risk of evolution of tumor development. In the literature, somatic or mosaic pathogenic variants described in EPAS1 target the amino acid from position 529 to 532. In our study, four families carry mutations that target these amino acids. Among them, only a single patient (P#19), carrying the p.Pro531Ser pathogenic variant at a mosaic state, developed multiple paragangliomas at the age of 10 years. Two other young patients at high risk carry EPAS1 variants already described in tumors and will need close follow-up: patient 18 (mosaic p.Ala530Glu, 6 years old) and patient 20 (germline p.Tyr532Cys, 9 years old). Noteworthy, it would be important to extend the follow-up of tumor occurrence to the family carrying the variant p.Leu526Phe. This variant indeed targets the consensus hydroxylation motif “LXXLAP” (L₅₂₆XXL₅₂₉A₅₃₀P₅₃₁) that plays a major role in tumor occurrence when mutated. It is interesting to note the complexity of genotype/phenotype correlations in this disease¹⁷ with the example of family 17 who carries the p.Ala530Ser variant (a substitution never described before this study) and do not present a tumor history despite a highly significant gain of function. A similar case has been described in a family who developed only erythrocytosis associated with the germline mutation variant p.Tyr532His.⁴⁰ The precise substitution targeting key amino acids may therefore be of major importance for the severity of the disease. Of note, our study reintroduces the debate about the variant F374Y that was initially described as causal in patients with erythrocytosis and paraganglioma^15,26-28 and found to be deleterious in one functional study.⁴⁷ Here this variant, that could be considered as a polymorphism due to its high frequency in the population, has been clearly classified as benign. For patients in for whom variants have been classified as VUS with predictions in favor of a benign effect (i.e., patients 38 to 42), additional investigations should be conducted to identify the cause of the carrier's pathology. The search for associated variants in the same gene was carried out and only one was found: the c.1218C>G, p.Thr406Thr variant (P#41), whose high frequency provides little support for a gain of function. The association with other variants in the hypoxia pathway genes associated with erythrocytosis (EGLN1, VHL) should also be explored, including in intronic regions that may impact splicing.⁴⁵ As a matter of fact, substantial research efforts remain to be made as approximately 70% of erythrocytosis cases are still of unknown genetic cause. The identification of two young cases carrying a variant at a mosaic state at a rate up to 1.9% of reads shows the potential failure to diagnose a number of patients with erythrocytosis. The development of a paraganglioma 8 years after the discovery of a major erythrocytosis, associated with very high EPO levels, is instructive: mosaicism may be present at a very low level that is not automatically detectable by usual NGS analysis. It is therefore necessary to lower the variant detection thresholds or to manually analyze the NGS results for the EPAS1 gene, notably in young children with severe erythrocytosis. Interestingly, our study opens up new research avenues related to HIF-2α-dependent EPO gene regulation. Indeed, expansion of the DNA sequence surrounding the minimal promoter previously described in the reporter vector not only resulted in substantially increased luciferase activity under both normoxic and hypoxic conditions, but also conferred more specificity towards HIF-2α (vs. HIF-1α) under overexpression conditions. This additional sequence does not contain any consensus HRE (G/ACGTG)⁵ and strongly suggests the presence of additional regulatory sequences within the EPO promoter. It will be of interest to analyze which factors bind to this region and if these proximal elements further contribute to tissue-specific and conditional EPO gene regulation in co-operation with distal enhancers.

Figure 5.End point reporter gene assays demonstrate a significant gain of function of the P531S HIF-2α mutant. HEK293 cells (A) and Hep3B cells (B) were transfected with 5’HRE and 3’HRE-dependent erythropoietin (EPO) full promoter-driven construct and different HIF-2α mutants identified from patients, as well as with wild-type (WT) and positive control P531A HIF-2α constructs, as indicated. Luciferase activity is reported as the induction compared to the control (Ctrl) under normoxic conditions and represents the ratio of firefly (FF) to Renilla (RL) relative light units (RLU). Each column represents the mean ± standard error of the mean of 3 to 5 different experiments performed in duplicate. (C) Expression levels of the different HIF-2α proteins were assessed with a HIF-2α antibody. Actin and α-tubulin were used as loading controls. One-way ANOVA, compared to HIF-2α WT (*P≤0.05; ***P≤0.001; ****P≤0.0001).

Figure 6.Real-time reporter gene assays demonstrate a significant gain of function for additional HIF-2α variants. (A) HEK293 cells were transfected with 3’HRE erythropoietin (EPO) full promoter luciferase reporter construct and different HIF-2α mutants. Relative light unit (RLU) was measured every 30 minutes for 48 hours (h). A plateau was obtained 30 h after the beginning of reaction (i.e., 34 h after transfection). The 100% was set on the activity of the wild-type (WT) protein at this plateau. Each mutant curve represents the mean of 3 independent clones. (B) Representation of luciferase activity obtained with the real-time luciferase assay, at the plateau obtained 30 h after the beginning of the reaction (i.e., 34 h after transfection); The 100% was set on the activity of the wild-type protein at this plateau. Each column represents the mean ± standard error of the mean of experiments performed in triplicate and each chip represents 1 clone, 3 different clones was tested each time. The quantity of transfected HA-plasmid was verified by polymerase chain reaction targeting the HA tag. The results obtained on agarose gel are shown. One-way ANOVA, Kruskal–Wallis tests with Dunn’s post hoc analysis for multiple comparisons with HIF-2α WT was performed (*P≤0.05; **P≤0.01; ***P≤0.001). (C) Representation of curve slopes obtained from the realtime luciferase measurement between 10 h and 15 h after the beginning of the reaction (see grey triangle in Figure 6A), normalized to the WT protein. Each column represents the mean ± standard error of the mean of an experiment performed in quadruplicate, each color represents an experiment and each chip represents 1 clone, 3 different clones being were tested each time. One-way ANOVA, Kruskal–Wallis tests with Dunn’s post hoc analysis for multiple comparisons with HIF-2α WT was performed (*P≤0.05; **P≤0.01; ***P≤0.001).

Overall, this study allowed the classification of EPAS1 variants classified as causal mutations in 40 individuals (17 patients and 23 relatives). This classification is of major importance giving the new therapeutics that specifically target and inhibit the HIF-2α protein. Indeed, a clinical trial with the HIF-2α inhibitor (MK-6482/Belzutifan/Welireg) on a single patient with a HIF-2α mutation has been recently published.⁴⁸ This patient carried the p.Ala530Glu pathogenic variant at a mosaic state and the treatment led to a rapid resolution of the erythrocytosis, hypertension, headaches in addition to paraganglioma response.⁴⁸ This remarkable effect opens up the possibility to treat more patients carrying mutations in EPAS1 which may potentially modify the drug binding capacity. Indeed, the chemical drug binds to a region that causes a conformational change of the amino acids H293 and M252 located in the HIF-2α-PAS-B domain. This drug-induced shift of the residue position, with a move of the side chains towards the binding surface, weakened the binding capacity of HIF-2α with HIF-β (ARNT-PASB) to form an active transcription factor.⁴⁹ Importantly, none of the mutations associated with erythrocytosis are located in the drug-targeted region (Figure 2, upper left panel). Therefore, all the patients carrying a causal mutation in EPAS1, located between amino acids 525 to 537, are theoretically eligible for HIF-2α inhibitor treatment.

Conclusions

Our collaborative study showed that enhanced functional assays in combination with in silico methods can improve the diagnosis in patients with erythrocytosis with an unclassified mutation in EPAS1. We also demonstrated the advantage of federating all diagnostic laboratories working on this rare pathology. This allows to increase the number of cases and the power of the analyses that can render genetic variants informative. It is also necessary, for a pathology linked to frequently hypomorphic mutations, to multiply the in silico analysis and to refine the functional approaches. A precise classification of mutations is indeed essential for a better diagnosis, clinical follow-up and access to a targeted treatment.

Footnotes

• Received September 1, 2022
• Accepted January 17, 2023

Correspondence

^*VK and DM contributed equally as co-first authors.

^#FG, DH and BG contributed equally as co-senior authors.

°An appendix with all ECYT-4 consortium members can be found at the end of the manuscript

B. Gardie betty.gardie@inserm.fr

Disclosures

No conflicts of interest to disclose.

Contributions

VK, DM, AB, LS, VA, ALR, SC, VL, MD, DH and BG performed experiments. AR, FL, performed bioinformatics analyses. CG, NM, FA, BA, LM, MR, SB, BC, FG, AP GR, NB, HC, RvW, CB, FG and the consortium ECYT4 conducted the medical and diagnostic studies. BG, DH and FG wrote the manuscripts. BG, FG and DH designed the study. BG directed the study. All authors contributed to the research and approved the final manuscript.

Data-sharing statement

Data and detailed information related to the study are available from the corresponding author upon request.

Funding

This study was supported by grants from the Agence Nationale de la Recherche (ANR; PRTS 2015 “GenRED”; AAPG2020 "SplicHypoxia"), the Labex GR-Ex, reference ANR-11-LABX-0051, the Fondation Maladies Rares (FMR) and Kiwanis project FONDATION-GenOmics 2017, and the associations VHL Alliance USA, VHL France and Génavie. This work was also supported as a part of NCCR Kidney.CH, a National Center of Competence in Research, funded by the Swiss National Science Foundation (grant number 183774) and by Swiss National Science Foundation project grant 310030_207460.