Introduction
Hypersensitivity to vitamin D (HVD) may be defined as an inability to regulate vitamin D metabolism which leads to calcium homeostasis deregulation, especially in case of vitamin D supplementation (1). This phenotype was classically associated with Idiopathic Infantile Hypercalcemia (IIH), a condition firstly described in the early fifties (2). Children present with non-specific signs of hypercalcemia (failure to thrive, vomiting, polyuria and polydipsia) with appropriately low parathormone (PTH) and inappropriately high 1,25-(OH)2D, suggesting deregulation of vitamin D metabolism.
In 2011, mutations in CYP24A1 encoding the vitamin D 24-hydroxylase were identified as the cause of Infantile Hypercalcemia type 1 (IH1) (MIM 143880) with HVD phenotype (3, 4). This enzyme is responsible for inactivating 1,25-(OH)2D and 25-OH-D (5–7). A high 25-OH-D3:24,25-(OH)2D3 ratio (substrate over product ratio, R ratio), which directly reflects the enzymatic defect, has been specifically associated with this condition and proposed as an effective screening tool (8–10). The clinical spectrum has been broadened to older children, teenagers and adults with nephrocalcinosis, renal stones and chronic fluctuating high serum and urine calcium with decreased serum PTH (9, 11–15). Patients present with a lifelong susceptibility to acute hypercalcemia and hypercalciuria, especially in case of vitamin D supplementation (16) and sunshine exposure (17).
In 2016, mutations in SLC34A1, encoding the renal phosphate transporter NPT2a, were reported as the cause of Infantile Hypercalcemia type 2 (IH2) (MIM 616963) with renal phosphate wasting and hypophosphatemia (18). A few cases have been published thereafter with clinical phenotype similar to type 1 (19–21). These observations suggest that phosphate wasting disorders may lead to HVD.
Other genes have been associated with renal phosphate wasting. Mutations in SLC34A3 encoding the renal phosphate transporter NPT2c have been associated with autosomal recessive hypophosphatemic rickets (ARHR) with hypercalciuria (MIM 241530) (22, 23). Its phenotype also includes biochemical features of HVD (low PTH, high serum 1,25-(OH)2D, high urine calcium with renal stones and nephrocalcinosis) and normal to low FGF23 serum concentrations. Furthermore, two patients with idiopathic hypercalciuria without bone signs and biallelic variations in SLC34A3 were described suggesting a potential role of SLC34A3 in HVD phenotype (24). Lastly, SLC9A3R1, encoding NHERF1 (sodium-proton exchanger regulatory factor 1) protein, was previously associated with autosomal dominant phosphate wasting (25, 26). This protein is implicated in transport regulation of NPT2a to the apical membrane of the proximal tubular cells and control of its retrieval by PTH (27).
We have previously published a cohort of 72 patients (1-72,
Supplementary Table S1
) with HVD tested for CYP24A1 mutations (9). We highlighted a wider phenotype depending on vitamin D intakes, from a balanced state with normal or high normal serum calcium and hypercalciuria to an unbalanced state with hypercalcemia and hypercalciuria, but consistent low PTH. In the present study, we aimed to evaluate the respective contribution of above mentioned genes involved in phosphate wasting in this cohort enriched with 113 new patients. Our objective was to search for a diagnostic algorithm to guide candidate gene testing in patients with features of HVD.
Patients and Methods
Patients
Over a 5 years period, we enrolled for molecular analysis 185 patients (index cases) presenting with a serum PTH concentration <20 pg/mL (2nd generation PTH assay were considered, normal values: 15-65 pg/mL) and a medical history of acute or chronic hypercalcemia (>2.6 mmol/L) and/or chronic hypercalciuria. Patients with medical history of calcium homeostasis deregulation with PTH > 20 pg/mL were not considered.
Recruitment was national through the network of the National Center for rare diseases of calcium and phosphate metabolism.
After genetic counseling, relatives of patients carrying biallelic variations were studied for genetic screening. Clinical and biochemical data were available in 17 relatives.
Written informed consent was obtained from the patients and/or their parents for retrospective collection of clinical and laboratory data, and for DNA collection to conduct molecular studies.
Biochemical and Clinical Parameters
Data on clinical symptoms were collected retrospectively using records from hospitals or primary care physicians.
Data on biochemical parameters (calcium and phosphate in serum and urine, renal function: creatinine and estimation of glomerular filtration rate with CKD-EPI for adults (28), PTH and vitamin D concentrations) were collected retrospectively using the same procedure, and thus correspond to a collection of different methods from many clinical chemistry departments. Approximate normal values for 1,25-(OH)2D were used as follows considering the use of different assays inherent in such retrospective study: 65-134 pmol/L for children and 60-108 pmol/L for adults (29). As most of routine assays for vitamin D cannot discriminate between vitamin D3 and D2, we used the term “vitamin D” to describe these measurements. As serum phosphate decreases from birth to adulthood, values were expressed as Z-scores to avoid the effects of the age-related variations of reference values (24, 30, 31).
Routine biochemical assays were mostly performed at the time of the diagnosis of acute hypercalcemia, or during follow up after normalization of the calcium level.
For urinary calcium: creatinine ratio, the different cutpoints of normal level were as follow: adults <0.33; children under 6 months <2.4; children 7-12 months <1.7; children 1-5 years <1.1 and children above 5 years <0.7.
Liquid Chromatography Tandem Mass Spectrometry (LC-MS/MS) analysis was performed as previously described using 100 µL of serum. Results were expressed as the molar ratio of 25-OH-D3: 24,25-(OH)2D3 (R ratio) (8). Values under 25 indicated no defect in 25-OH-D3-24-hydroxylase activity and were considered as normal.
Medullary nephrocalcinosis was assessed by renal ultrasonography (echogenic renal pyramids). Nephrolithiasis was defined as kidney stones on ultrasonography or a medical history of iterative renal colic.
Molecular Analysis
Genomic DNA was extracted from whole blood samples with QuickGene DNA Whole Blood kit (Kurabo Industries LTD, Osaka, Japan) and QuickGene-610L extraction system (Fujifilm LTD, Singapore, Republic of Singapore).
Coding exons and flanking regions of CYP24A1 (NM_000782.4), SLC34A1 (NM_003052.4), SLC34A3 (NM_001177317.1) and SLC9A3R1 (NM_004252.4) genes were studied by Massively Parallel Sequencing (MPS) using a custom Ampliseq library on the Ion Torrent Personal Genome Machine (PGM) (ThermoFisher Scientific, Waltham, Massachusetts, USA). Reads were aligned against the reference sequences and variants were annotated using IonSuit (ThermoFisher Scientific, Waltham, Massachusetts, USA) and Nextgene (SoftGenetics, State College, Pennsylvania, USA). A read depth (RD)-based approach was used to detect copy number variation (CNV) from targeted MPS data using the CovCop tool (32).
Sequences with coverage <30X and variations identified with MPS were studied with Sanger sequencing using Big Dye® Terminator v1.1 Cycle Sequencing kit (ThermoFisher Scientific, Waltham, Massachusetts, USA) on a ABI 3500 Sequencer (ThermoFisher Scientific, Waltham, Massachusetts, USA).
Variations of sequence were searched on Varsome database. Interpretation was based on standards and guidelines of the American College of Medical Genetics (33). Pathogenicity prediction programs PolyPhen-2, Align-GVGD, MutationTaster and SIFT were considered for in silico analyses. Allelic frequencies were evaluated through the GnomAD database. For research purposes, variations classified as of uncertain significance (III) were considered apart from mutations [i.e. likely pathogenic (IV) and pathogenic (V) variations] as some may represent pathogenic mutations and other benign variations.
Statistical Analysis
Data from patients with class IV and V variations were used to compare biochemical profiles.
Statistical analyses were made using Prism 7 (GraphPad Software Inc.). For each biochemical parameter, differences between groups were calculated using the Mann and Whitney nonparametric U test. A two-tailed p value <0.05 was considered statistically significant. Kruskal-Wallis test was used for multiple group comparison (one-way nonparametric ANOVA).
Sex was not considered a factor in the statistical analysis of the data.
Chi-squared test was used to compare the allelic frequency of recurrent variations in our cohort and in the general population (estimated from data of GnomAD database).
Discussion
This report presents extensive clinical, biochemical and molecular data on a large cohort of patients with HVD phenotype and suspicion of IH1. We found: i) Mutations in genes associated with renal phosphate wasting can be found with a lower frequency than loss-of-function mutations in CYP24A1 and digenism may represent a rare mechanism of HVD. ii) Our data indicate that hypophosphatemia is not specific to phosphate wasting defects, also observed in CYP24A1 mutation, explaining a wide phenotypic overlap.
In our population, we found biallelic variants in CYP24A1, SLC34A1 and SLC34A3, in accordance with an autosomal recessive inheritance, in 28-36 patients (15-19%), but no mutation in SLC9A3R1, suggesting that mutations in this latter gene do not cause HVD phenotype. We observed a majority of biallelic mutations in CYP24A1, in accordance with our inclusion criteria (low serum PTH and medical history of hypercalcemia and/or hypercalciuria). We found a high number of variants of uncertain significance in SLC34A1 and SLC34A3 (17/24) with difficulties interpreting their pathogenicity in a context of routine molecular diagnosis. Even the presence of the recurrent c.272_292del/p.Val91_Ala97del variant previously described in IH2 may still be difficult to interpret because of a high allelic frequency in general population and in our cohort, suggesting a likely benign variant rather than a pathogenic loss of function one. Although its biological activity was preserved, it had been reported as a pathogenic variant responsible for cellular localization defect (18, 37). Its association with another variant in SLC34A1 in several patients (18) cannot exclude pathogenicity, maybe as an hypomorphic allele. Interpretation was more complex considering the high number of rare variants reported in a heterozygous state in these genes in online population databases (i.e. more than 700 SLC34A3 variations in GnomAD). In vitro functional studies are not feasible in the context of routine diagnosis, justifying the need for specific clinical screening or diagnostic criteria and the collection of molecular data.
Most patients with biallelic variants presented with renal complications including chronic hypercalciuria, renal stones (6/28) or nephrocalcinosis (18/28); and were adults or children over 2 years old (24/28). The lower diagnosis rate in children (8-12% versus 45-51%) suggested that most of the children in our cohort may present with another disease or a transient HVD which may be due to the initial prematurity in renal function during early infancy (3). Indeed, relative renal immaturity may lead to a delay in the adaptation of mineral metabolism from a prenatal to a postnatal context, especially regarding the vitamin D 24-hydroxylase expression. Patients with biallelic variants in CYP24A1 and SLC34A1 presented a significant phenotypic overlap. However, these two disorders can be easily differentiated on the basis of the serum 25-OH-D3:24,25-(OH)2D3 ratio, which is distinctly elevated in patients with biallelic CYP24A1 variants reflecting the vitamin D 24-hydroxylase defect; whereas the ratio remains normal in patients with biallelic SLC34A1 variants (8–10). More comprehensive profiling of serum vitamin D metabolite profiles by LC-MS/MS including 23- and 24-hydroxylated metabolites (23,25,26-(OH)3D3, 25-OH-D3-26,23-lactone and 1,24,25-(OH)3D3), strengthen the differences in vitamin D metabolism in these two diseases, as well as other causes of vitamin D-related hypercalcemia (38). A few clinical criteria may also help distinguish CYP24A1 mutation from other diseases. The presence of prenatal hyperechogenic kidneys in 2 patients of this cohort with SLC34A1 class III variants and 2 patients in the literature (21) suggested a more precocious disease of the urinary tract. To the best of our knowledge, no patient with prenatal hyperechogenic kidneys and CYP24A1 mutation were mentioned in the literature or observed in our patients. Mutations in SLC34A3 are responsible for ARHR, with characteristic high urine calcium and serum 1,25-(OH)2D, but hypercalcemia in these patients was not documented to date. Among patients with biallelic SLC34A3 variations, we reported 2 patients with fortuitously discovered hypercalcemia: A patient (n°152) with initial clinical features of HVD whose clinical presentation evolved into classic ARHR (with short stature and legs bowing); and a patient (n°88) with HVD with nephrocalcinosis without bone disease nor hypophosphatemia and normal height. These observations widen the phenotypic spectrum of SLC34A3 mutations to HVD. Whether the HVD presentation corresponds to a sustainable mild form of phosphate tubular reabsorption defect or a precocious presentation of ARHR requires further investigation.
In our cohort, only 2 patients with biallelic variations in SLC34A3 (class V) or SLC34A1 (class III) presented with hypophosphatemia. Thus, most of them did not present with a biochemical profile suggestive of a phosphate wasting disorder prima facie, including 2 patients with biallelic SLC34A3 pathogenic variations. Similarly, approximately one third of patients (5/16) did not present with hypophosphatemia in the IH2 cohort published by Schlingmann et al. (18) and patients with biallelic mutations in SLC34A3 and normal serum phosphate have also been published (24). Case reports have thus suggested that supplementation with oral phosphate could help correct calcium metabolism in patients with SLC34A1 mutation (18). More surprisingly, 27-28% of our patients with CYP24A1 biallelic mutations also presented with hypophosphatemia.
It is important to consider these findings in relation to the role of FGF23 in HVD, also described by Brancatella et al. (18, 39). FGF23 is a key regulator of phosphate and vitamin D homeostasis, up-regulated by serum phosphate and 1,25-(OH)2D (40, 41). Binding of FGF23 to FGF receptors induces the retrieval of renal phosphate transporters (responsible for phosphate tubular reabsorption) and decreases the expression of vitamin D 1α-hydroxylase (CYP27B1) (42). In IH1, we suggest that reduced clearance of 1,25-(OH)2D up-regulates FGF23 production and consequently leads to phosphate wasting and hypophosphatemia. In contrast, in NaPiIIa/NPT2a/NaPiIIc/NPT2c defects, renal phosphate wasting is responsible for a decrease in serum phosphate which in turn down-regulates FGF23 and up-regulates 1α-hydroxylase. As a consequence, the increase in 1,25-(OH)2D favors phosphate and calcium intestinal absorption. The negative feedback exerted by the resulting increase in serum calcium is supported by the low serum PTH. Such mechanism had previously been proposed and corroborated by murine Slc34a1-/- studies, illustrating the crucial role of FGF23 as a key regulator of vitamin D metabolism (18). Measurement of FGF23 in serum of HVD patients may be useful in future studies to specify this mechanism.
In 15-25 patients, we identified heterozygous variants, which represented diagnostic challenges in the context of a classic autosomal recessive disease. We could not exclude a second undetected variant (e.g. localized in non-coding sequences or a complex rearrangement) or contributory effect of the heterozygous variant as a risk factor. Although an autosomal dominant trait (7) had been proposed for heterozygous CYP24A1 loss-of-function mutation (43), our previous work showed that heterozygotes (proband as well as relatives) had normal 24-hydroxylase activity (9, 13). Thus, a patient with a CYP24A1 mutation in a heterozygous state and a high 25-OH-D3:24,25-(OH)2D3 ratio may benefit from extensive genetic exploration. Considering heterozygous variations in other genes, we observed a typical HVD phenotype in SLC34A3 heterozygous relatives which had previously not been diagnosed as HVD patients. Similarly, it was suggested that SLC34A1 heterozygous mutation may be a risk factor for kidney stone disease (18, 44) and heterozygous relatives may have abnormal biochemical data. It is likely that heterozygous phenotype may depend on additional factors, including genetics and environmental factors. Indeed, families in which the probands harbored a variation in two different genes suggested that heterozygous variations in CYP24A1, SLC34A1 and SLC34A3 may have an impact on vitamin D metabolism with a deleterious synergistic effect as previously proposed (45), as illustrated by the lower serum PTH in these patients compared to their heterozygous parents. Such a digenic mechanism, with both heterozygous SLC34A1 and SLC34A3 mutations, has been postulated as the cause of hypophosphatemic rickets with renal stone disease in an American family (46).
No mutation was identified in 66-77% of this cohort, especially in children (n= 110-126/150, 73-84%; adults: n=12-16/35, 34-46%). In our hands, these differences between children and adults may rely on obvious differences in the aims of genetic analysis in children and adults. In most children (78% or 117/150 children under 2 years), the first aim was to quickly distinguish CYP24A1 mutations from other causes with transient HVD. We may wonder if these children had consistent phenotype over the years but this work does not include follow up data. Repeated evaluation of mineral metabolism including PTH after initial discontinuation of vitamin D supplementation, may be useful to detect a normalization of serum PTH (not in favor of a genetic defect) or a consistent low serum PTH (in favor of a genetic defect). Evaluation of 24-hydroxylated vitamin D metabolites may also represent an interesting tool in this purpose (8, 10, 38). On the other hand, genetic analyses in adult were conducted in a different context of longer medical history to explain a phenotype. We cannot assume that patients without mutation in CYP24A1, SLC34A1, SLC34A3 or SLC9A3R1 did not carry variations in other genes or regulating sequences which were not analyzed.
No mutation was identified in SLC3A3R1 encoding NHERF1 in our cohort, which may be explained by our inclusion criterion as serum PTH in patients with SLC3A3R1 mutation were above 20 pg/mL in previous publications (25, 26).
Unfortunately, unlike genetic data obtained from the same procedure, the data collected in this retrospective cohort may include bias due to the aggregation of biochemical results from various assays and routine medical laboratories, which may impact the interpretation of 1,25-(OH)2D particularly. Appropriate exploration of phosphate metabolism including serum FGF23 evaluation would be of great interest to further characterize this pathophysiologic mechanism in HVD. Lastly, several patients without mutation (n=24) presented with hypophosphatemia suggesting the existence of other different mechanisms of calcium, vitamin D and phosphate metabolism deregulation.