The PHEX gene consists of 22 short exons with huge intronic regions. The coding region spans 2,250 bp and encodes for a 749-aminoacid transmembrane endopeptidase that belongs to the type II integral membrane zinc-dependent endopeptidase family [13].

Structurally, PHEX has a small aminoterminal intracellular tail, a short transmembrane domain and a large carboxyterminal extracellular domain with the catalytic and zinc-binding sites [14,15]. PHEX is mainly expressed in osteoblast, osteocytes and odontoblast [16,17].

More than 250 different PHEX mutations have been described to date, including nonsense (17-19%), missense (21-22%), deletions (24-30%), insertions (11-12%) and splice mutations (18-24%) unevenly distributed along the PHEX gene with three regions with high mutation density [18-20].

This study shows, to our knowledge, the highest mutation detection rate reported so far in HR patients. The percentage of detected mutations in other cohort studies ranged from 45% to 87% [13,20,23,26-29]. The high rate of mutation detection in the present study can be explained by several reasons. Firstly, it could be due to the careful clinical and biochemical selection of patients before referring their samples for genetic analysis. Also, the methodology used for the mutation detection could be an explanation. Our approach detects punctual mutations by sequencing but also large deletions or duplications by MLPA analysis. Conversely, researchers of other studies, lack a method for detection of single or contiguous exons deletions. The fact that such type of mutations were not observed in some studies despite a PHEX mutation detection rate of 87% in familial and 72% in sporadic cases, underlies the impact of the methodology used. That type of mutations represented 13.9% in our study, with 100% mutation detection rate. Furthermore, it has been pointed out that those single or contiguous exons deletions may account for up to 19% of PHEX mutated alleles [30]. Finally, differences in the genetic background between the series published cannot be ruled out. Moreover, in other studies, the mutations detected in familial cases were higher than in sporadic ones, whereas we observed the same mutation rate in both groups.

All types of mutations were found in this cohort and the majority of them were novel. This was in agreement with other cohort studies where most of the mutations observed were specific to each family and the minority of them was recurrent. Nonsense mutations represented a 30.5% of the causative mutations found in this study. Five were novel and four had been previously described, two of them were recurrent in our study. These mutations resulted in truncated proteins which lack either a large region with the highly conserved metallopeptidase domain or just the final three residues, like the p.Arg747X mutation. The PHEX arginine residue 747 has been shown to be involved in substrate interaction in the homologous protein neprilysin [31]. The nonsense mediated decay, (which selectively eliminates mRNAs containing premature stop codons), is an accepted mediator mechanism in the deleterious process of the nonsense mutations [32]. In fact, we were not able to detect any mRNA corresponding to the nonsense mutated alleles by RT-PCR. Deletions represent a 22.2% of the mutations in this series and all of them were novel: Two small frameshift deletions in exon 1 and 8, and four large deletions which involved several contiguous exons, mainly exons 13 to 22. These mutations were detected by MLPA analysis which was used when no punctual mutation was detected by gene sequencing. In male carriers the deletions were confirmed by the absence of PCR product of the corresponding exons. Two additional in frame deletions were found. A deletion that included the complete exon 3, which would result in a PHEX protein lacking 53 residues of the peptidase M13 and which was considered as deleterious for the genotype-phenotype correlation analysis. A triplet deletion that would result in a protein lacking the aspartic acid 646, which was considered as plausible causative mutation for genotype-phenotype correlation study. Duplications accounted for 13.8% of the mutations. Four were small frameshift duplications and one was a gross duplication that involved the complete exon 3. This last one was evident by RT-PCR and also by MLPA analysis. It is possible that during meiosis a chromatid misalignment occurred and resulted in either deletion or duplication of exon 3 because of more than 85% sequence homology between three stretches of intron 2 and 3. Splicing mutations accounted for 19.4% of the cases. Aberrant mRNA was detected in two of the patients and although no RNA of sufficient quality was available from the rest of them, the effect of the mutation was deduced based on splicing prediction algorithms and cosegregation of the mutation with the disease phenotype. The synonymous mutation c.591A > G created an abnormal splice donor site in the middle of exon 5. Only five missense mutations (13.8%) were found: p.Arg510Pro, p.Gly579Arg and p.Arg651Pro which had been already described associated with the disease [13,20,23,29]. Even more, the p.Gly579Arg mutant protein was shown to be sensitive to endoglycosidase H digestion and to remain trapped in the endoplasmic reticulum in contrast to the wild type PHEX, thus providing a mechanism for loss of PHEX function [33]. And two missense mutations not previously described: p.Asn71Ile and p.Ala262Pro that are very likely pathogenic. Both had appeared de novo in the affected individuals and they were not found in 130 normal chromosomes. These mutations resulted in the substitution of highly conserved PHEX aminoacids in different species (Figure ​(Figure2),2), which suggested that both residues are important for protein function and that the mutations are probably damaging.

The 36 different mutations found in our study were distributed along the gene and affected almost all exons or their adjacent introns. However, the main density of the mutations was located at the end of the gene, which is one of the three described regions with high mutation density [20]. When comparing the distribution of our mutations with the mutations from the PHEX mutation database [18], a lower mutation density was observed in the first and second part of the gene in our study (Figure ​(Figure33).

For the genotype-phenotype correlation analysis, the mutations were grouped based on their predicted effect rather than on their localization as it seemed a better approach in other investigations [26,34,35]. One group comprised all the mutations which were very likely deleterious because they would result in a truncated PHEX protein. This group included nonsense mutations, frameshift mutations due to small duplications and deletions, large duplications and deletions and splicing mutations. The second group included missense mutations and a triplet deletion which might result in a PHEX protein with reduced but not totally abolished activity. Indeed, significant correlation between TRP and 1,25(OH)₂D l serum levels with the mutation type was found. As could be expected, patients with deleterious mutations had lower TRP and lower 1,25(OH)₂D levels. Both might be the consequence of more active FGF23. Even so, we are conscious that the sample is small, that some parameters may have influenced the results and that larger series of patients should be analyzed to confirm our results. No additional statistical genotype-phenotype correlation was observed with other parameters. The mildest phenotype was shown by the missense p.Ala262Pro mutation carrier. Regarding the bone phenotype, the probands carrying deleterious mutations showed a more severe bone phenotype of no statistical significance, which was in agreement with other studies that reported a trend between truncating mutations and more severe skeletal phenotype [23,34,35]. However, no conclusion can be obtained because patients from the same family and with the same genotype showed bone phenotype differences (data not shown) and on the other hand, it is a subjective assessment made by different physicians and not a quantitative variable. We are aware that great caution should be exercised when assessing genotype-phenotype correlation. In fact, phenotype variation between individuals with the same genotype, particularly regarding the skeletal phenotype, is described [26,36].

To our knowledge, this is the largest study where a correlation between biochemical phenotype and genotype was analyzed. There is a previous study with 8 patients about the genotype correlation with P, AP and TRP which gave no positive results probably due to the small sample size [35].

Regarding the intra-genotype variability, patients 21 and 22 both carry the c.1645C > T (p.Arg549X) mutation and presented very similar phenotype except the development of nephrocalcinosis after treatment. Patients 32 and 33 carry the c.2104C > T (p.Arg702X) mutation and also have similar phenotype except for the development of hyper-PTH after treatment in one and the height SDS unexpectedly positive in the other. Regarding the intra-familial variability, the serum P values and TRP were similar among carriers of the same mutation whilst they showed greater skeletal variability (Data not shown). Nevertheless, the number and size of the families and the number of patients with recurrent mutations was too small to get any conclusive data.

Hyperparathyroidism is rather frequently detected in XLHR patients, which was traditionally explained by chronic phosphorus therapy [37]. However, it should be taken into account that disturbance of the PTH secretion circadian rhythm has been reported among XLHR patients, which increase in nocturnal pulses [38], In addition, animal models of XLHR showed hyperparathyroidism in the absence of treatment [39], and abundant levels of PHEX mRNA were observed in the parathyroid gland of two XLH patients [40]. In our cohort, three patients presented hyperparathyroidism before initiating medical therapy and eight more after treatment. Ten of these patients had deleterious mutations. All these data suggested that therapy with phosphorous is not the only cause of hyperparathyroidism in XLHR patients and that more severe PHEX mutations may have some influence.