Novel PHEX gene locus‐specific database: Comprehensive characterization of vast number of variants associated with X‐linked hypophosphatemia (XLH)

Abstract X‐linked hypophosphatemia (XLH), the most common form of hereditary hypophosphatemia, is caused by disrupting variants in the PHEX gene, located on the X chromosome. XLH is inherited in an X‐linked pattern with complete penetrance observed for both males and females. Patients experience lifelong symptoms resulting from chronic hypophosphatemia, including impaired bone mineralization, skeletal deformities, growth retardation, and diminished quality of life. This chronic condition requires life‐long management with disease‐specific therapies, which can improve patient outcomes especially when initiated early in life. To centralize and disseminate PHEX variant information, we have established a new PHEX gene locus‐specific database, PHEX LSDB. As of April 30, 2021, 870 unique PHEX variants, compiled from an older database of PHEX variants, a comprehensive literature search, a sponsored genetic testing program, and XLH clinical trials, are represented in the PHEX LSDB. This resource is publicly available on an interactive, searchable website (https://www.rarediseasegenes.com/), which includes a table of variants and associated data, graphical/tabular outputs of genotype‐phenotype analyses, and an online submission form for reporting new PHEX variants. The database will be updated regularly with new variants submitted on the website, identified in the published literature, or shared from genetic testing programs.


| INTRODUCTION
Hypophosphatemic rickets (HR) was originally described in 1937 as a form of childhood rickets unresponsive to vitamin D in doses that were typically effective for the treatment of nutritional rickets (Albright et al., 1937). In contrast to transient nutritional deficiency, patients identified with this disorder were found to occur in families and experienced lifelong chronic hypophosphatemia due to excessive renal phosphate loss. Subsequent investigations have discovered that the proximal cause of the impaired renal retention of phosphate is increased serum levels of fibroblast growth factor 23 (FGF23), a phosphate regulating hormone expressed in bone by osteocytes and osteoblasts (Beck-Nielsen et al., 2019). X-linked hypophosphatemia (XLH; MIM# 307800), the most common form of hereditary HR, is caused by variants in the PHEX gene located at Xp22.1 (Francis et al., 1995). XLH shows X-linked inheritance and affects~1/20,000 males and females of all ages globally (Beck-Nielsen et al., 2010).

| PHEX and the XLH disease pathway
PHEX (Phosphate regulating gene with Homology to Endopeptidases that maps to the X chromosome) belongs to a well-defined family of zinc metalloendopeptidases involved in cancer, bone-renal diseases, cardiovascular disease, Alzheimer's disease, arthritis, and inflammatory disorders (Rowe, 2004;Turner & Tanzawa, 1997). The human PHEX gene encodes a 749 amino acid type II single integral transmembrane protein with most of the protein forming a large extracellular domain that contains the enzymatic active site, three zinc coordination sites, and multiple glycosylation sites and disulfide bonds (Figure 1a,b).
Multiple mouse models that harbor pathogenic variants in PHEX have been used to study XLH and its potential treatments. These include Hyp mice, which contain a 58 kb deletion of the 3′-end of the gene extending into the untranslated region (UTR) (Sabbagh et al., 2002), Gy mice, which contain a large deletion that encompasses PHEX exons 1-3 and also extends into the upstream gene (Lorenz et al., 1998), Ska1 mice, which contain a single nucleotide variant in a splice donor site near the 5′-end of the gene (Carpinelli et al., 2002), Jrt mice, which contain a stop codon at amino acid 496 (Owen et al., 2012), and Hyp-2J and Hyp-Duk mice (Han et al., 2012), which contain frameshift deletions affecting exons 13-14 and exon 15, respectively. The phenotypes of these mice vary somewhat depending on the background strain, but they all recapitulate the clinical features of XLH (hypophosphatemia, elevated serum FGF23, rickets and osteomalacia) and have been instrumental in elucidating the molecular basis of the disease.
The pathogenesis of phosphate wasting and impaired mineralization in individuals with XLH is generally attributed to inappropriately increased FGF23 levels (Figure 1c). PHEX is predominantly expressed in osteoblasts and osteocytes and is hypothesized to play a role in the phosphate sensing mechanism (Beck-Nielsen et al., 2019). Although the molecular mechanism by which PHEX influences FGF23 levels is not known, it is clear that inactivating PHEX variants lead to increased FGF23 expression and increased circulating FGF23, which causes renal phosphate wasting, decreased synthesis of the active metabolite of vitamin D (1,25dihydroxyvitamin D), and enhanced metabolism of 1, 25dihydroxyvitamin D to 1,24,25-trihydroxyvitamin D (Beck-Nielsen et al., 2019). Although the physiological substrate for PHEX is not yet know, variants that disrupt the folding, trafficking, or enzymatic function of the protein have all been linked to the development of XLH in patients.
PHEX may also have important functions independent of phosphate regulation through FGF23. In vitro assays have demonstrated that PHEX can directly cleave osteopontin (OPN), an extracellular matrix protein found in bone and teeth and a potent inhibitor of mineralization (Barros et al., 2013;Qin et al., 2004). In mouse models with PHEX inactivating variants, full-length OPN accumulates in the matrix of bone and is thought to contribute to impaired mineralization (Barros et al., 2013).

| XLH clinical presentation and diagnosis
XLH patients typically begin to show signs of the disorder by 1-2 years of age and present clinically with growth retardation, rickets/ bone deformity, and gait abnormalities Pavone et al., 2014). Pediatric patients have also demonstrated osteomalacia, craniosynostosis, and dental abscesses in otherwise healthy teeth Pavone et al., 2014;Reid et al., 1989). In addition to consequences of the progressive accumulation of skeletal defects (e.g., short stature, lower limb deformities, and osteoarthritis), adult patients experience ongoing disease manifested by impaired muscle function, pseudofractures, dental complications, enthesopathy (mineralization of ligaments and tendons), osteoarthritis and hearing loss (Beck-Nielsen et al., 2009). In both children and adults, XLH symptoms can include chronic pain and functional disability, leading to a diminished quality of life (Dahir et al., 2020).
A definitive diagnosis of XLH is established with the combination of clinical and family history, physical examination, radiographic and laboratory findings, and often, genetic testing. Clinical and radiographic findings of rickets combined with laboratory findings of hypophosphatemia, and increased renal phosphate clearance are highly indicative of HR. XLH, and most other etiologies of familial HR, also exhibit low-normal levels of circulating 1,25-dihydroxyvitamin D. The ratio of renal tubular maximum reabsorption rate of phosphate to glomerular filtration rate (TmP/GFR) is reduced and circulating intact FGF23 levels are usually elevated or at least higher than the normal population mean. Information regarding potentially affected family members can help to determine the underlying cause of HR, but sporadic occurrences are common and other genetic causes are possible. A definitive genetic diagnosis of XLH requires confirmation of a pathogenic variant in the PHEX gene (Carpenter et al., 2011; Haffner et al., 2019).

| PHEX variants genotype and phenotype
Pathogenic variants in human PHEX have been identified throughout the entire length of the gene and include frameshift, splicing, copynumber, nonsense, and missense variants (Beck-Nielsen et al., 2019;Sabbagh et al., 2000). These variants are predicted to cause loss of function protein changes, with the majority (>70%) producing a truncated PHEX protein. Studies of recombinant PHEX proteins in human cell culture assays have shown that missense variants can impair protein function by disrupting cellular processing, endopeptidase activity, or protein conformation (Li et al., 2020;Sabbagh et al., 2003Sabbagh et al., , 2001Zheng et al., 2020).
Controversy exists regarding potential correlation between PHEX genotypes and the severity of XLH phenotypes. It has been reported that female patients with certain variations in the PHEX gene may have less severe hypophosphatemia and milder skeletal deformity compared to males with the same PHEX variants, and some reports have suggested that patients with truncating variants or variants in  (Kelley et al., 2015); labels call out the enzymatic active site and amino acids altered in the 10 most common missense variations. (c) In the XLH disease state, decreased PHEX activity leads to an increase in serum FGF23, which decreases blood phosphate levels due to increased renal phosphate wasting. FGF23 also decreases the synthesis and increases the metabolism of the active vitamin D metabolite, ultimately leading to reduced bone mineralization. PHEX, Phosphate regulating gene with Homology to Endopeptidases that maps to the X chromosome; XLH, X-linked hypophosphatemia; UTR, untranslated region the C-terminal half of the PHEX gene have more severe biochemical or skeletal phenotypes (Holm et al., 2001;Morey et al., 2011;Song et al., 2007). Other studies were unable to establish a genotypephenotype correlation when comparing patients with truncating variants to those with missense variants (Cho et al., 2005;Rafaelsen et al., 2016;Reid et al., 1989;Zhang et al., 2019). There are also multiple studies describing broad and clinically significant variations in XLH phenotype among patients with the same genotype, including among members of the same family (Holm et al., 2001;Rafaelsen et al., 2016;Rodríguez-Rubio et al., 2021).

| A new locus-specific database for PHEX variants
Early and accurate diagnosis is beneficial for XLH patients as treatment leads to dramatic reduction in morbidities and improvement in patient quality of life. However, this is complicated by the rarity of the disease, phenotypic variability, similarities to other forms of congenital and sporadic hypophosphatemias, and the absence of a comprehensive source of PHEX disease-associated variants needed for accurate and timely interpretation of genetic testing results.
A new locus-specific database for PHEX gene variants, PHEX LSDB, was established to collect and disseminate information to the scientific community and affected families about disease-causing PHEX variants (https://www.rarediseasegenes.com/). The database compiles variants from four sources: an older, archived PHEX locusspecific variants database (Sabbagh et al., 2000), variants identified in a recent hypophosphatemia genetic testing program (Rush et al., 2021), unpublished variants identified in previous XLH clinical studies, and previously published variants identified from a comprehensive literature review. The PHEX LSDB will be updated regularly with new information on PHEX variants as reported in the literature or submitted directly to the database website. The purpose of this report is to describe the new PHEX LSDB and to present findings from an analysis of the initial set of variants in the database.

| Assembling the database
The database was developed by integrating PHEX variants from four different sources: (1) a now inactive McGill University PHEX locusspecific database (Sabbagh et al., 2000), last updated April, 2017, (2) genetic testing results from burosumab clinical trials, (3) a comprehensive literature review, and (4) a sponsored hypophosphatemia genetic testing program initiated by Ultragenyx Pharmaceutical Inc. and Invitae Corporation, which provided no-charge next-generation sequencing with a multi-gene panel to confirm a clinical XLH diagnosis or to aid diagnosis of suspected XLH or other genetic hypophosphatemia (Rush et al., 2021). The hypophosphatemia genetic testing program was designed to detect single nucleotide variants (SNV), small and large insertions/deletions (indels), sub-genic structural variants, and exon-level copy number variants (CNV).
A comprehensive literature review was performed using Mastermind, a database of variants with evidence cited in the medical literature (Genomenon Inc.) (Chunn et al., 2020) and considered all publications indexed in PubMed as of April 7, 2020. We followed stated guidelines for performing meta-analyses from biomedical literature. Before completing variant curation, the data were cross-

| Patients and privacy
All individuals who provided samples for genetic testing, through the gene panel program or busosumab clinical trials, consented to have their deidentified genetic information published. To avoid accidental reidentification in this rare disease population, phenotypic data are reported in aggregate for each variant, and lab values are not reported for variants that occur fewer than three times in the database.

| Analysis
We used serum phosphorus values and reported clinical phenotypes to examine correlations between PHEX genotype with XLH phenotype. To control for age-related variance, serum phosphorus values were analyzed as a percent of the lower limit of the provided normal ranges; values reported without an associated normal range were excluded from the analysis.
A 3-D visualization of human PHEX protein was modeled based on homologous endopeptidases and using the PyMOL Molecular-Graphics System, Version 2.4 (Schrödinger, LLC) with atomic coordinates from the Phyre2 web portal for protein modeling, prediction, and analysis (Kelley et al., 2015).

| Spectrum of PHEX variants
As of April 30, 2021, the PHEX LSDB reports 2578 total XLH-associated PHEX variants representing 870 unique variants (Table S1). The literature review data set contributed 223 unique variants that were not found in any other source, the genetic se-  (Table S2).
The most prevalent variant in the database is a c.*231A>G substitution in the 3′-UTR, which was frequently reported cooccurring with an exon 13-15 duplication, suggesting that these two variants co-segregate and constitute a single allele. In an analysis reported from the hypophosphatemia gene panel program, these variants were found together in both males and females in 65 of 66 probands and were shown to be in cis in all 51 individuals for whom phasing information was available (Rush et al., 2021).
Individuals represented in the database have a broad geographical distribution encompassing North America, Europe, Middle East, and Asia. Ten of the 12 most common variants were found across multiple geographies while the most common bi-variant allele, c.*231A>G 3'-UTR + Ex13-15 duplication, was found only in North America to date (Table 1).  (Li et al., 2020;Sabbagh et al., 2001Sabbagh et al., , 2003Zheng et al., 2020). In the same studies, missense variants p.Asp237Gly, p.Tyr317Phe, p.Gly553Glu, and p.Phe731Tyr were efficiently transported to the cell membrane but showed substantially reduced enzymatic activity or altered protein conformation (Sabbagh et al., 2003;Zheng et al., 2020).

| Clinical significance and biological relevance of PHEX variants
Even in the absence of data from animal models or in vitro assays, the position of missense variants can sometimes provide clues about their potential impact on protein function. The database includes 28 unique missense variants at nine different residues forming disulfide bonds (Cys59, Cys77, Cys85, Cys142, Cys406, Cys617, Cys693, Cys733, and Cys746), which could lead to protein conformation/trafficking defects, and two missense variants that occur at a glycosylation site (Asn71), which could lead to defects in protein trafficking. There are also four missense variants at zinc coordination sites (His580 and His584) and one missense variant at each of the active site residues (Glu581, Asp646), which are likely to disrupt enzymatic activity.  F I G U R E 3 Distribution of 870 unique PHEX variants by variant type. CNV, copy number variant (size greater than 100 nucleotides [nt]); small deletion (size smaller than 100 nt); small duplication (size smaller than 100 nt); small insertion (size smaller than 100 nt); SNV, single nucleotide variant. One allele that contains two variants (Ex13-15 dup and c.*231A>G) was counted in both the CNV and SNV categories. "Other" included 1 start codon change and 1 deep intronic variant. PHEX, Phosphate regulating gene with Homology to Endopeptidases that maps to the X chromosome T A B L E 1 Geographic distribution of the 11 most common alleles in PHEX LSDB   (Table 2).
Females had higher mean serum phosphorus levels compared with males in each of these groups, but this elevation was not found to be statistically significant.
We also examined serum phosphorus levels for the 11 most common PHEX alleles, which occurred at least 15 times in the database. Only nine of these alleles had one or more phosphorus levels reported with respect to a normal range, and we found no statistically significant differences in the relative phosphorus values between these groups (  (Figure 4a). Table 4 and the position of a selected variant relative to structural features of the gene and protein is displayed above the table. The interactive data pages allow users to explore genotype-phenotype correlations using biochemical data (serum phosphorus, TmP/GFR, and bone-specific alkaline phosphatase) or reported clinical phenotypes (Figure 4b).

The table can be searched by any of the fields listed in
The  (Rush et al., 2021). This underscores the utility of genetic testing methods that can detect sub-genic CNV, deep intronic variants, and large structural variations to establish a molecular diagnosis for XLH.
Such methods include RNA sequencing and whole-genome sequencing with computational analysis to detect intronic and structural variants.
The most common variant in the database is the bi-variant allele, c.*231A>G, Ex13-15 dup, which has been described previously and reported to be specific to patients from the Midwest region of the United States (Mumm et al., 2019). The expanded data from the hypophosphatemia genetic testing program revealed that these two variants occur together across the United States in individuals of White/Caucasian, Hispanic, and Black/African American ancestries (Rush et al., 2021). This is also consistent with earlier findings of XLH families from outside of the US Midwest geography carrying the *231A>G variant (Ichikawa et al., 2008).
Earlier, small cohort studies identified c.*231A>G as a diseasecausing variant in multiple probands leading to XLH (Ichikawa et al., 2008;Smith et al., 2020), however, these studies did not detect CNV, so it is not known whether these patients also carried the Ex-13-15 duplication. The recent readout from the hypophosphatemia genetic testing program found that all XLH patients who had the 3′-UTR c.*231A>G variant also carried the Ex13-15 duplication, but one affected proband carried the duplication without the c.*231A>G variant, suggesting that the duplication can contribute to disease on its own (Rush et al., 2021 have a similar pathophysiological effect. As described here, the pathogenicity of certain PHEX missense variants has been demonstrated with functional assays (Li et al., 2020;Sabbagh et al. 2001Sabbagh et al. , 2003Zheng et al., 2020).

| XLH management and treatment
Historically, the standard of care for XLH patients has been a combination of calcitriol and phosphate supplementation, which may improve bone mineralization by providing transient incremental elevations in phosphate administered multiple times daily (Carpenter et al., 2011;Dahir et al., 2020). Recently, the novel therapeutic burosumab, a fully human monoclonal antibody against FGF23, was approved for the management of pediatric and adult patients with XLH (Imel et al., 2019;Insogna et al., 2018). In randomized clinical studies, burosumab, which binds FGF23, resulted in increased reabsorption of phosphate in the kidney and increased serum phosphorus levels. Treatment, especially when initiated early in the course of disease, can provide substantial improvements in outcomes.
Therefore, accurate early diagnosis of XLH helps to provide prognostic information, inform patients about the need for lifelong disease management, and guide medical therapy. This comprehensive searchable database of PHEX variants aggregates and organizes the evidence around variants for medical, scientific, and lay audiences, enabling earlier diagnosis and better patient outcomes.

| Limitations
We were not able to determine reliably whether clinical and biochemistry data provided in case reports or genetic testing forms is treatment naïve or not. There was limited evaluable biochemical data because the age of patients was not always known, and lab values were frequently provided without reported age-specific reference ranges. Clinical data were also extremely limited in this data set, and we were not able to quantitatively evaluate genotype/phenotype correlations for skeletal or other non-biochemical outcomes. In addition, it was not always possible to know when patients were related or whether data for the same individual was reported by more than one molecular diagnostics lab. For this reason, allele frequency determinations from this data set cannot reliably be made.

| PHEX LSDB curation and updates
The PHEX LSDB is overseen and curated by a qualified scientific committee. New PHEX variants originating from publications, online submissions, or genetic testing labs will be added regularly. Following review by the scientific committee, all data displays and analysis pages will be updated accordingly, and new variants will be shared twice-yearly with ClinVar and the Leiden Open Variation Database.