Introduction

Familial hypomagnesemia with hypercalciuria and nephrocalcinosis (FHHNC) is an ultra-rare autosomal recessive renal tubular disease with an incidence of <1/1.000.000 individuals, characterized by severe urinary Ca²⁺ and Mg²⁺ wasting and progression to chronic kidney disease (CKD) and renal failure [1,2]. Disease loss-of-function causing mutations have been identified in the tight junction (TJ) proteins claudin 16 (CLDN16) [3] and 19 (CLDN19) [4]. Both proteins are co-expressed in the TJ of the thick ascending limb of Henle’s loop and form a paracellular cation-selective pore that induces a NaCl gradient and a lumen-positive transepithelial voltage driving Ca²⁺ and Mg²⁺ reabsorption [5,6]. Additionally, CLDN19 is also expressed in the retinal epithelium and a subset of FHHNC patients with CLDN19 mutations develop severe ocular impairment [4]. No specific therapy for FHHNC exists and patients rely only on supportive treatment (high fluid intake, dietary restrictions, magnesium salts and thiazide diuretics) attempting to delay the progression to renal failure. Kidney replacement therapy remains the only curative option in end-stage renal disease patients and renal transplant is usually required during the second to third decades of life [1,2], causing a severe impairment on quality of life of patients from a very young age.

Major research efforts in the last decade have been directed towards identifying novel mutations in FHHNC patients [7] and, as a result, 73 and 24 different mutations in CLDN16 and CLDN19, respectively, have been described and annotated in the Human Gene Mutation Database (HGMD) [8]. Most FHHNC patients have been found to carry CLDN16 mutations, although in the South of Europe (mainly Spain and France) CLDN19 mutations are more prevalent and a specific CLDN19 so-called Hispanic founder mutation (c.59G>A; p.G20D) has been described [9,10]. Remarkably, carriers of the p.G20D mutation exhibit a prominent, but yet unexplained, phenotypic variability which is observed even between homozygotic siblings [2,9,11,12]. This suggests the existence of other not-so-far-identified molecular events, such as phenotype modifier genes, that might contribute to the clinical variability beyond the causal genes. Here we present the first study focusing on the identification of phenotype modifier genes in a unique cohort of 45 FHHNC Spanish patients through whole exome sequencing (WES) analysis comparing extreme phenotypes, as it has been successfully applied to other complex Mendelian diseases [13–15]. Candidate modifier genes were selected by a thorough prioritization strategy combining publicly available information regarding described variants associated to disease-specific phenotypic traits from the GWAS catalog together with kidney-specific expression data from the Human Protein Atlas and Genotype Tissue Expression (GTEx) databases. Our results led to a panel of genetic variants in potential candidate modifier genes which may contribute to our understanding of the factors influencing the rate of renal failure progression in FHHNC patients (see graphical abstract in Fig 1). While these findings provide a foundation for future research and could potentially inform more personalized therapeutic approaches, they should be interpreted with caution given the inherent challenges of studying ultra-rare diseases. Further validation studies are necessary to confirm the role of these potential modifier genes in FHHNC progression.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Graphical abstract.

https://doi.org/10.1371/journal.pgen.1011568.g001

RESULTS

Clinical characterization of the FHHNC cohort

The discovery cohort was composed of 30 patients from 22 unrelated families. As previously described [12], pathogenic CLDN16 mutations were confirmed in 3 patients, while the remaining 27 patients carried CLDN19 mutations (74% were homozygous for the p.G20D mutation). Patients were classified based on the annual renal function decline, following the clinical criteria previously reported by our group [12].

Patients were then stratified in two extreme phenotypes for WES analysis, FAST vs. noFAST. In the discovery cohort (n=30), the FAST group included patients (n=6) with the fastest renal function decline, while the noFAST group (n=24) comprised patients with a moderate decline in eGFR rate and stable renal function (Fig 2 and S1 Table). Remarkably, two pairs of siblings belonging to families carrying the p.G20D in homozygosis showed different renal phenotypes (fast vs. moderate) and were classified in different groups for WES analysis.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Annual renal function decline.

Patterns of eGFR change over time were plotted for each patient and stratified as fast, moderate or slow according to the curve slopes determined by linear regression. Each coloured line indicates the annual eGFR loss. Dashed lines indicate patients for whom we only had one value of eGFR corresponding to the last follow-up.

https://doi.org/10.1371/journal.pgen.1011568.g002

During the course of this study, 15 new FHHNC patients from 14 families were recruited for validation. One of the patients carried a CLDN16 mutation, while pathogenic mutations in CLDN19 were confirmed in the remaining 14 patients (64% harboured the p.G20D in homozygosis). All of them were classified in the noFAST group. Additionally, we recruited 4 healthy controls among the non-affected siblings of patients, and 20 healthy controls, recruited among healthy siblings of patients with glomerulopathies from a different study, were also available.

Candidate gene prioritization identifies variants associated to a faster progression to end-stage renal disease in FHHNC patients

To provide a clear overview of our gene prioritization approach, we have developed a flow diagram (Fig 3) that illustrates the step-by-step process used to identify potential modifier genes associated with FHHNC progression.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Prioritization strategy.

Flow diagram illustrating the step-by-step process used to identify potential modifier genes associated with FHHNC progression. CKD, chronic kidney disease; CLDN, claudin; eGFR, estimated glomerular filtration rate; eQTL, expression quantitative trait loci; FHHNC, familial hypomagnesemia with hypercalciuria and nephrocalcinosis; GATK, Genome Analysis Toolkit; GWAS, genome wide association study; PTH, parathyroid hormone; SKAT-O, optimal sequence kernel association test.

https://doi.org/10.1371/journal.pgen.1011568.g003

First, the SKAT-O genes were filtered by selecting those with a P-value <0.01 and described to be expressed in the human kidney in the Human Protein Atlas database, at both RNA and protein evidence level. These filters reduced the candidate gene list to 165 genes. As a second step to further prioritize potential modifier genes, we searched the GWAS catalog database to identify gene variants associated to FHHNC specific phenotypic traits. Those specific traits were selected based on the renal phenotype described in our patient cohort (annual glomerular filtration rate decline) plus the wide range of common FHHNC symptoms which include polyuria-polydipsia, failure to thrive, recurrent urinary tract infections, hypercalciuria, nephrocalcinosis, hypomagnesemia, and hyperparathyroidism [2,16,17]. In addition, we also included traits related to the ocular impairment described for some carriers of CLDN19 mutations. S2 Table summarizes the selected phenotypic traits, the number of associated variants found in the GWAS catalog and the number of matched genes from our SKAT-O analysis after filtering by P<0.01 and kidney expression.

As a result, the list of potential candidates for further analysis was reduced to 44 candidate modifier genes. We performed a second round of data matching between SKAT-O genes and GWAS catalog data including genes not expressed in the kidney in order to not exclude the possibility that genes from other tissue and/or cell types may have an impact on FHHNC progression. This second data-matching analysis identified only another 5 candidates. Additionally, as the claudin family of proteins are highly involved in the physiopathology of FHHNC, we included in the list of selected candidates the gene CLDN17 that was significantly associated (P<0.01) to the FAST phenotype. Applying the described strategy for candidate prioritization, we finally selected 175 variants mapping to 50 genes for further analysis (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. List of 50 prioritized candidate modifier genes.

https://doi.org/10.1371/journal.pgen.1011568.t001

To uncover the biological significance of the prioritized genes, an overrepresentation analysis (ORA) was performed using the GO-BP followed by clustering analysis to summarize the results obtained (summORA). Among the top 15 terms identified by ORA, those related to apical junction regulation and assembly, cell adhesion and immune-related terms were the most significant biological processes associated to our 50 phenotype modifier candidates (Fig 4A). After clustering by similarity the enriched GO-BP terms obtained with an adjusted P-value < 0.25, 12 clusters were identified (Fig 4B). Both the most representative term found in each cluster and the 10 most frequent words within them, highlighted apical junction regulation and assembly, immune-related terms, ionic transport, protein modifications (mainly phosphorylation), cellular responses to different stimulus, and cell cycle regulation as the most significant biological processes enriched in the FHHNC prioritized candidate modifier genes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Overrepresentation analysis of the 50 prioritized candidate modifier genes in FHHNC patients.

A. Dot plot showing the top 15 most significant biological processes overrepresented. The size of the dots relates to the number of genes/proteins in the data that belong to that pathway, the colour of the plot refers to significance level (P-value). The terms are ordered by P-value and Gene ratio, which is the ratio between the genes/proteins in the data that belong to that term and the total number of genes/proteins in the term. B. Heatmap of clustering analysis summarizing overrepresentation results. Clusters are annotated with the most representative term within each cluster (the term with the lowest P-value and the higher number of candidate genes) and with a word cloud representing the top 10 most frequent words found within term names in each cluster. The heatmap shows the terms found within each list, coloured according to their adjusted P-value (P.Adjust).

https://doi.org/10.1371/journal.pgen.1011568.g004

Since SKAT-O does not provide any information regarding whether the candidate genes confer a risk or a protective effect, we took a complementing approach by performing Fisher’s exact test and calculating odds ratios for each individual variant identified in the SKAT-O prioritized candidate genes. Ten single nucleotide variants (SNVs) in homozygosis and nine in heterozygosis were significantly associated to renal progression phenotypes (Fisher P-value <0.05 and FDR adjusted P-value <0.15; Odds ratio ≠1) considering the alternative allele both in homozygosis and in heterozygosis (Fig 5 and S3 Table). From those, the alternative allele in homozygosis for SNVs mapping to genes CD86, CEP89, GALP, IL12RB1, NFU1 and SLC17A1 was associated to the FAST phenotype; while rs228406 (DMD) and rs8065903 (SPATA20) were associated to the noFAST phenotype (Fig 5A). Additionally, the alternative allele in heterozygosis for SNVs mapping to genes AGBL2, ANKLE2, CLDN17, CLDN8, DMD, HSP5, IL12RB1, ITGA1 and ZNF607 was associated to the FAST phenotype (Fig 5B). Variants in these genes are annotated in the GWAS catalog associated to Ca²⁺ levels (CD86, NFU1, AGBL2), eGFR (CEP89, IL12RB1, SPATA20, ZNF607), blood urea nitrogen levels (DMD, HPS5, ZNF607), microalbuminuria (GALP), bone mineral density (ITGA1) and high myopia (ANKLE2) (Table 1). In addition, most of these genes were found in clusters related to immune system activity and regulation of apical junctions in the summORA analysis (S4 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Forest plot showing the significant variants in candidate modifier genes associated to the FAST phenotype of FHHNC patients.

A. Odds ratios and 95% confidence intervals were calculated by comparing the distribution of the alternative allele in homozygosis versus the reference allele in homozygosis. B. Odds ratios and 95% confidence intervals were calculated by comparing the distribution of the alternative allele in heterozygosis versus the reference allele in homozygosis.

https://doi.org/10.1371/journal.pgen.1011568.g005

Identified phenotype modifier variants are overrepresented in FHHNC patients with a faster renal function decline compared to patients with moderate/mild phenotype and healthy individuals

In order to understand the functional implications of SNVs associated to renal progression in FHHNC, we retrieved expression quantitative trait loci (eQTL) analysis data for each significant variant in 659 microdissected (309 glomeruli and 350 tubules) human kidney samples from the Human Kidney eQTL Atlas [30] and in human kidney cortex tissues (n=73) from the GTEx database. Variant rs4453725 for NFU1, which we have identified as a risk factor for eGFR decline, was found to be a significant eQTL in microdissected renal tubules from the Human Kidney eQTL Atlas (P-value = 2.14^-12). Conversely, analysis of whole renal cortex tissue from GTEx revealed no significant impact of the identified gene variants on gene expression, suggesting that the impact of NFU1 variant on disease progression may be mediated through gene expression changes restricted to renal tubules, rather than through broad alterations in overall kidney gene expression.

Notably, the risk alleles for SNVs associated with rapid renal progression were significantly underrepresented in GTEx kidney tissues (Fig 6 and S5 Table). Furthermore, the FAST renal progression phenotype showed a markedly higher proportion of patients carrying these risk alleles compared to both the noFAST group and the expected frequency in the general healthy population, as represented by control tissues from GTEx (Fig 7). Interestingly, not all initially prioritized variants were found to be overrepresented in the FAST cohort. This discrepancy may indicate that some of the initially prioritized variants could have been false positives. Nonetheless, this analysis has allowed us to further refine the candidate variants, narrowing down those consistently associated with faster progression of renal failure across different analytical approaches.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Violin plots of allele-specific cis-eQTLs according to genotypes for each significant variant in human kidney cortex tissues from the GTEx database.

The teal region indicates the density distribution of the samples in each genotype. The allelic effect of variants on normalized gene expression levels are shown by boxplots within violin plots. The white line in the box plot (black) shows the median value of the expression of each genotype. The alleles and number of subjects analysed is shown under each genotype. A. Variants in homozygosis for the alternative allele associated to fast renal progression. B. Variants in heterozygosis associated to fast renal progression. The odds ratio and 95% confidence intervals calculated from Table 2 are shown next to each plot for reference.

https://doi.org/10.1371/journal.pgen.1011568.g006

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Distribution of the three possible genotypes for each candidate modifier gene variants across two FHHNC phenotypes and the control group from GTEx.

A. Variants in homozygosis for the alternative allele associated to fast renal progression. B. Variants in heterozygosis associated to fast renal progression. Comparisons were performed by the Fisher’s exact test.

https://doi.org/10.1371/journal.pgen.1011568.g007

Download:

• PNG
  larger image
• TIFF
  original image

Table 2.

Comparison of the genotype distribution obtained by Open Array genotyping data in the validation cohort and by whole exome sequencing data from the discovery cohort. Data from the three possible genotypes is shown for each candidate modifier gene variant across the two FHHNC renal phenotypes (FAST and noFAST), the healthy control cohort and the control group from GTEx.

https://doi.org/10.1371/journal.pgen.1011568.t002

Next, to validate our results, the 19 variants associated to a faster renal progression were genotyped in a second cohort. To do this, 15 additional FHHNC patients were recruited together with a healthy control cohort. However, none of the new patients met the criteria to be classified in the FAST group. Therefore, and to maximize the study power, we genotyped the full cohort of patients using a customized OpenArray panel (n=45). Our data showed that, in the FAST group the distribution of the three possible genotypes for each variant were identical to the data obtained by WES, while adding 15 patients to the noFAST group rendered highly similar results to those in the WES analysis (Table 2). Moreover, the healthy control cohort showed also a genotype distribution similar to those from the general healthy population (GTEx).

Next, we aimed to identify whether the prioritized SNVs showed a combined effect on the renal phenotype. Unfortunately, polygenic risk scores could not be determined as these SNVs are not annotated in any genome-wide study for FHHNC-related traits. However, most of the patients in the FAST group (67%) carried between 9 and 15 SNVs in the candidate modifier genes, while the maximum number of SNVs found in the noFAST group was 4 (S6 Table). This distribution highlights a potential cumulative impact of these SNVs on the progression of renal failure in the FAST cohort.

Discussion

The FHHNC clinical course is highly variable, whereas some individuals require kidney replacement therapy early in childhood, others maintain a more stable renal function into adulthood suffering with only mild to moderate renal disease [12]. Major research efforts in the field are focused on identifying novel mutations in the known causing genes, CLDN16 and CLDN19, while other potential genetic factors underlying phenotypic variability within patients have been ignored, likely due to the difficulties inherent to performing exome-wide association approaches in an ultra-rare disease. In the exploratory study presented here we describe the pipeline for WES analysis and candidate prioritization that has allowed us to identify potential phenotype modifier gene variants associated to a faster renal progression in FHHNC patients.

Our study was specifically designed to maximise the chances of identifying genes with significant associations in a small patient cohort, which benefits from the following key features. First, the unique cohort encompassing nearly all FHHNC patients diagnosed in Spain prior to January 2021, with detailed clinical and genetic characterization. Second, the high genetic homogeneity of our cohort with 90% of patients harbouring CLDN19 mutations, including 70% with the homozygous p.G20D founder mutation. Third, we employed an extreme phenotype study design based on the assumption that risk variants would segregate with a more severe phenotype. This approach has been successfully applied to other diseases and has been proven to increase power, particularly when dealing with limited sample sizes as in rare diseases [14,15]. Therefore, we stratified our patients at the extremes of a quantitative trait as the annual decline on the estimated glomerular filtration rate as previously reported by our group [12]. And fourth, we applied a WES data analysis pipeline assessing mixed effects of variants using SKAT-O. SKAT-O is a unified approach that optimally combines the burden and non-burden sequence kernel association test (SKAT) and increases statistical power to detect associations, particularly when used in combination with extreme phenotypes analysis strategy [18].

By using SKAT-O and comparing patients with a fast renal progression to patients with a moderate/mild renal progression, we obtained a high number of genes associated to the most severe renal phenotype. As expected, due to the limited number of samples analysed, none of those candidate genes resisted multiple testing correction. To mitigate the risk of false positives, we implemented a prioritization strategy focused on selecting genes relevant to FHHNC based on current state-of-the-art knowledge of the disease. This strategy consisted on i) reducing our threshold for significance to P<0.01, ii) filtering the list for those genes expressed in renal tissue, and iii) further filtering the list selecting genes that had been previously associated to FHHNC key phenotypic traits by GWAS. Remarkably, overrepresentation analysis on prioritized candidates identified apical junction regulation and assembly, cell adhesion and immune system-related processes as the most significant terms associated to the modifier variants in the fast renal progression group which indicates that our analysis pipeline is selecting the candidates that are better aligned to the physiopathological hallmarks of the disease and, thus, most likely to have a true involvement in its clinical course. We acknowledge, however, that the extra filtering steps by tissue expression and GWAS annotation may be suspicious of introducing some bias in the enrichment analysis (terms related to kidney or to the traits used for GWAS-annotation would probably be favoured due to the pre-filtering). Nonetheless, it should be noted that the overrepresentation analysis has been applied only as means to functionally annotate the selected genes and it did not have any influence on the candidate prioritization strategy.

Taking all the above considerations together, we took a further step to minimize the chances of selecting false positives and to prioritize gene variants associated to a higher risk of a faster renal progression by calculating the Fisher’s exact test and odds ratios. This additional analysis reduced the prioritized genes to 19 variants in 15 genes associated to a faster renal progression. Variants in these genes are annotated in the GWAS catalog associated to Ca²⁺ levels, eGFR, blood urea nitrogen levels, microalbuminuria, bone mineral density and high myopia, although the precise mechanisms underlying these associations have not been explored. Moreover, the variants found in our FHHNC cohort are different from those annotated in the GWAS catalog and their involvement in renal failure progression remains to be elucidated. Interestingly, among the genes conferring increased risk to faster renal progression, the biggest effect size corresponded to the homozygotic variant p.M25K in gene NFU1. NFU1 is involved in iron–sulphur [Fe‐S] cluster assembly, a highly complex system involved in several cellular processes including mitochondrial respiratory chain activity and various other enzymatic and regulatory functions [19–21]. Defects in this process have a severe impact on human health [19,22] and, indeed, loss-of-function mutations in NFU1 and other proteins involved in [Fe‐S] cluster assembly have been described in multiple mitochondrial dysfunction syndromes (MMDS), a rare disease characterized by neurological regression, reduced motor control (dystonia) and pulmonary hypertension. Although almost all NFU1 pathogenic mutations described so far are located in the C-terminal NifU [Fe‐S] cluster binding domain [23], the NFU1 p.R21P mutation located in the N-domain, which is only 4 aa apart from p.M25K and affects the mitochondrial target signal [24], has been described in the only MMDS patient showing renal tubular impairment [23,25]. Renal involvement has been also described for MMDS patients harbouring mutations in BOLA3 [23], a protein that cooperates with NFU1 in the last steps of maturation and transfer of [4Fe‐4S] clusters into target proteins, further supporting the role of [Fe‐S] cluster assembly defects on kidney disease. Based on our data, we suggest that [Fe‐S] cluster homeostasis may be imbalanced in renal tubular cells from FHHNC patients with a more severe renal phenotype contributing, therefore, to the aggravation of the high cellular stress already produced by the CLDN19 mutation. The potential involvement of the p.M25K NFU1 variant in FHHNC renal failure progression is further supported by the fact that it is considered a common variant with a MAF of 0.40 according to the gnomAD browser. Therefore, its potential impact on cellular homeostasis is likely to be easily accommodated in healthy individuals, particularly when the variant is present in heterozygosis, as it happens in the majority of cases genotyped. Indeed, the GTEx data showed that the p.M25K NFU1 variant was present in homozygosis in 11% of the genotyped renal tissues and in 8% of the healthy subjects in our study. Remarkably, our multi-step analytical approach revealed that only 5% of FHHNC patients (2 out of 39) with a moderate/mild progression had this variant in homozygosis, while it was present in 5 out of 6 patients with a fast renal progression.

This overrepresentation in patients with a more severe phenotype was not unique to the NFU1 variant but was a common observation for almost all the other variants prioritized in our study. Of particular interest are the variants found in the genes encoding dystrophin (DMD) and Hermansky-Pudlak Syndrome protein 5 (HPS5), as mutations in both genes have been associated to rare diseases accompanied with compromised renal function [26]. Indeed, progressive muscle degeneration in Duchene’s muscular dystrophy, caused by mutations in DMD, seems to be associated to subclinical kidney injury and the subsequent development of chronic kidney disease in paediatric patients [27]. Furthermore, and although insufficiently explored, neuromuscular impairment has been suggested to be part of the clinical spectrum associated with CLDN19 mutations as intolerance to muscular exercise that persist after kidney transplant and normalization of magnesium levels has been described in a few patients [28]. On the other hand, the HPS5 protein is highly expressed in human renal tubules according to the Human Protein Atlas and HPS5 knock-down resulted in a severe renal phenotype in a zebrafish model through a mechanism that involves intracellular accumulation of cell debris due to lysosomal impairment [26].

We also identified variants in genes not typically associated with the thick ascending limb of Henle where CLDN16 and CLDN19 function. Of particular interest are variants in SLC17A1 and SLC9A3, which play important roles in other segments of the nephron. SLC17A1 is a sodium-dependent phosphate transporter primarily expressed in the proximal tubule and involved in urate excretion [29]. SLC9A3 is a sodium/hydrogen exchanger found in the proximal tubule and thick ascending limb of Henle, crucial for sodium reabsorption and pH regulation [30]. Although not directly linked to magnesium and calcium handling in FHHNC, variants in these transporters could potentially influence overall renal function and electrolyte balance. Their association with enhanced progression to renal failure in FHHNC patients suggests possible indirect effects on disease progression, perhaps through alterations in tubular function or acid-base homeostasis. In addition, variants in other members of the claudin family of proteins, CLDN8 and CLDN17, are also associated to a more aggressive phenotype in our FHHNC cohort. Although information on the role of CLDN17 in the human kidney is limited, it has been identified in the proximal tubules of both mice and humans [6]. Recently, it was reported to play a role in renal function by regulating serum electrolyte levels and tissue reactive oxygen species [31]. On the other hand, CLDN8 has been implicated in the regulation of paracellular permeability to sodium and water in the distal nephron [32]. The identification of these variants highlights the complex interplay of various renal transporters and channels in maintaining kidney function. While the primary defect in FHHNC lies in the thick ascending limb, our findings suggest that genetic variations in other nephron segments may contribute to the variability in disease progression. This underscores the importance of considering the kidney as an integrated system, where dysfunction in one area can have far-reaching effects on overall renal health.

We acknowledge several important limitations in our study. First, the ultra-rare nature of FHHNC presents significant challenges in achieving robust statistical power. Our cohort, while comprehensive for Spain, remains small, which limits the generalizability of our findings and increases the risk of false positives. Replication studies in independent cohorts are mandatory. Our replication cohort, though valuable, has limitations. While we were able to recruit 15 additional patients, a significant achievement for an ultra-rare disease, and all prioritized candidate variants showed concordant results by a different technique in a larger patient cohort, we acknowledge this cannot be considered as a full replication study as this cohort lacked fast progressors, limiting our ability to fully validate the association of identified variants with rapid disease progression. We recognize that a more comprehensive replication study including patients with the most severe renal phenotype, would strengthen our findings. In addition, we are aware that the multi-step methodology taken may appear as somehow circular. Our approach involved an initial identification of variants associated with renal progression through WES analysis and statistical criteria, followed by prioritization based on biological relevance to FHHNC traits. The subsequent overrepresentation analysis in the FAST group served as an independent validation step and, indeed, not all initially prioritized variants were overrepresented in this analysis, which further supports the robustness of our approach. This discrepancy allowed us to focus on variants consistently associated with faster renal failure progression across different analytical steps. Despite all these measures, the potential for unidentified confounding factors and the challenge of distinguishing true modifier effects from statistical noise in a small cohort remain significant concerns. While we have identified promising potential modifier variants in NFU1, DMD, HPS5, CLDN8 and CLDN17, their functional impact on FHHNC progression towards kidney failure remains to be characterized. However, we would like to remark that even gathering information regarding the impact of the genetic variants on their respective protein function in cell and/or animal models, there would be still a huge gap to understand their contribution to such a complex phenotype as in FHHNC. Furthermore, since these variants have not been described to be pathogenic per se, in light of our results we can speculate that they interact together as part of the genetic context of patients with a more severe renal phenotype. How these variants are combined to interact with the FHHNC causative CLDN19 mutation resulting in a particular renal phenotype is far more difficult to approach. Appropriate models for this disease, which arises from the thick ascending limb of Henle’s loop of the nephron, simply do not exist and the knock-in mouse model carrying the CLDN19 p.G20D causative mutation has not been generated either. Moreover, ethical constraints prevent renal biopsies in FHHNC patients, limiting our access to patient-specific tissue for direct molecular studies. Our reliance on public databases like GTEx and Human Kidney eQTL Atlas, while informative, may not fully capture the unique genetic context of FHHNC patients. In summary, while our findings are promising, they should be interpreted with caution. Further validation in additional cohorts or through functional studies will indeed be crucial to confirm the utility of these results before they can be considered for clinical decision-making. Despite these limitations, we believe our study provides valuable insights into potential genetic modifiers in FHHNC and lays the groundwork for future, larger-scale investigations in this challenging field.

In summary, we have applied a pipeline specifically tailored for an ultra-rare renal disease leading to the identification of several novel gene variants potentially associated with a more aggressive renal phenotype. Notably, the most promising candidates include variants that may affect the functional stability of critical cell organelles such as mitochondria and lysosomes. The orchestrated action of these organelles is key for cellular homeostasis and survival under pathological stress conditions [33] and may explain the variability of the clinical phenotype in FHHNC in which the endoplasmic reticulum (ER) is likely to be under high stress due to the reported retention of CLDN19 p.G20D mutant protein [5]. Furthermore, the association of our candidate modifier genes with other diseases involving signs of kidney disease raises the interesting possibility that their involvement in renal failure progression may not be FHHNC-specific. Consequently, potential prognostic and therapeutic tools developed for these targets could offer broader applicability across various renal and extra-renal diseases. Finally, our approach can easily be customized to identify and study modifier genes for other renal rare diseases and can contribute to strengthen the interpretation of personalized exome sequencing approaches, providing, therefore, a valuable framework for the identification and understanding of phenotypic variability and disease risk.

Methods

Supporting information

Acknowledgments

We would like thank the FHHNC patient advocacy group HIPOFAM (http://hipofam.org), patients and families for their valuable support and contribution to our research activity. We gratefully acknowledge the contributions of the medical doctors responsible for the recruitment and clinical management of the FHHNC national patient cohort, whose collaboration was essential to this study. We also thank Dr. Xavier de la Cruz for helpful discussions and advise. The Genotype-Tissue Expression (GTEx) Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health, and by NCI, NHGRI, NHLBI, NIDA, NIMH, and NINDS.