Novel associations of Claudin gene variants with kidney stone disease

ABSTRACT

Background

The tight junction family of claudin proteins regulates paracellular ion transport in the nephron and across epithelial and endothelial barriers throughout the body. Rare recessive variants in CLDN16 and CLDN19 cause familial hypomagnesemia with hypercalciuria and nephrocalcinosis (FHHNC), while common variants in CLDN2, CLDN10, and CLDN14 have been associated with kidney stones. Despite evidence that most claudin proteins are expressed within the nephron, their genetic contribution to kidney stone susceptibility has not been systematically explored.

Methods

We analyzed 415 237 UK Biobank participants (9009 kidney stone cases and 406 228 controls) using exome sequencing and imputed genotype data spanning all 24 human CLDN genes. Single-variant association testing, gene burden analysis, and phenome-wide association studies were employed to identify novel associations.

Results

Forty-three novel coding variant associations were identified through exome sequencing, including two CLDN16 variants: p.Gly57Arg (rs747654138; odds ratio [OR] = 42.2, P = 9.5 × 10⁻⁸) and p.Arg146His (rs772241737; OR = 60.5, P = 8.1 × 10⁻⁴). We replicated a known association with a CLDN19 variant, p.Arg200Gln (rs116804195; OR = 1.23, P = 2.6 × 10⁻⁴), which was also associated with decreased serum phosphate and increased alkaline phosphatase. We also identified 13 novel susceptibility loci within the imputed genotype dataset, including single nucleotide polymorphisms near CLDN2, CLDN10, CLDN11, CLDN18, and CLDN22–CLDN24 (all P < 0.00025), and confirmed a known CLDN14 association.

Conclusions

Beyond established FHHNC genes, multiple CLDN family members harbor variants potentially influencing kidney stone risk, possibly through altered paracellular transport and mineral homeostasis. These findings expand kidney stone genetic architecture and suggest that claudin proteins merit further investigation in its pathogenesis.

KEY LEARNING POINTS.

What was known:

• Most claudins are expressed within the nephron where they regulate paracellular transport of calcium, magnesium, sodium, and other ions depending on the combination of claudins present in different segments of the nephron.

• Rare recessive loss-of-function mutations in CLDN16 and CLDN19 cause familial hypomagnesemia with hypercalciuria and nephrocalcinosis (FHHNC), a severe monogenic kidney stone disorder leading to renal failure in early adulthood.

• Common variants in CLDN2, CLDN10, and CLDN14 have been associated with kidney stone risk in candidate gene and/or genome-wide association studies.

This study adds:

• Heterozygous CLDN16 variants as risk factors: Heterozygous coding variants in CLDN16 (p.Gly57Arg, rs747654138; and p.Arg146His, rs772241737) are significantly associated with increased kidney stone risk in this population-based dataset, demonstrating that CLDN16 may contribute to kidney stones beyond its role in recessive FHHNC.

• Expansion of CLDN genetic architecture in kidney stone disease: Multiple newly identified CLDN coding and non-coding variants associate with kidney stones, including single nucleotide polymorphisms in previously implicated genes (CLDN10, CLDN14, CLDN16, CLDN19) and a novel one near CLDN11.

• CLDN variants have effects across mineral homeostasis pathways: Phenome-wide associations implicate CLDN variants in mineral homeostasis and recapitulate features of claudin-related disorders.

Potential impact:

• Reinterpreting heterozygous CLDN variants in clinical practice: Variants in CLDN genes previously thought to cause pathology only under a recessive model may also influence kidney stone pathophysiology in heterozygous individuals, highlighting the need to consider rare variants when evaluating the etiology of common kidney stone disease.

• Implications for families with pathogenic CLDN variants: First-degree relatives who carry pathogenic CLDN16 and CLDN19 variants may have increased kidney stone susceptibility. Testing of such individuals could improve our understanding of variant penetrance and the phenotypic spectrum of heterozygous carriers.

• Broader claudin relevance for kidney stone pathophysiology: A larger group of claudin proteins is relevant to kidney stone pathophysiology and implicates claudin-mediated mineral homeostasis in common kidney stone disease.

INTRODUCTION

Kidney stone disease is characterized by the formation of large, typically calcium-based, stones that obstruct urinary drainage [1]. Obstruction causes acute pain and inflammation, and recurrent stones can lead to progressive loss of renal function [1]. The claudin tight junction genes (CLDNs) encode transmembrane CLDN proteins that are expressed throughout the nephron epithelium [2]. CLDN proteins play a critical role in regulating bidirectional paracellular flux of calcium, sodium, and other ions implicated in kidney stone pathogenesis [2, 3]. Tight junction proteins form paracellular barriers or pores with selectivity determined by CLDN composition [3–5].

Rare loss-of-function and missense mutations in CLDN16 and CLDN19, inherited in an autosomal recessive manner, cause familial hypomagnesemia with hypercalciuria and nephrocalcinosis (FHHNC; Online Mendelian Inheritance in Man [OMIM] #248250 and #248190), a severe form of kidney stones that leads to kidney failure in early adulthood [6, 7]. Additionally, common single nucleotide polymorphisms (SNPs) in CLDN10 and CLDN14 have been associated with kidney stones in genome-wide association studies (GWAS) [8, 9]. Importantly, most of these CLDNs (10, 14, 16, and 19) are expressed in the thick ascending limb of the loop of Henle (TAL), a nephron segment critical for calcium, magnesium, and sodium paracellular transport [10–14]. Targeted analyses have also identified non-coding CLDN2 variants associated with kidney stones, expanding the scope of risk variants beyond the TAL to the proximal tubule [15]. Disruptions in CLDN function may affect sodium and/or calcium flux, promoting supersaturation within the urinary lumen, resulting in crystal aggregation and ultimately, stone formation [2]. Given this potential for broad physiological impact, we sought to identify the role of genetic variants within the entire CLDN gene family for kidney stone susceptibility by leveraging exome sequencing and imputed genotype data to analyze SNPs and small insertions/deletions.

MATERIALS AND METHODS

Ethics approvals

UK Biobank (UKB) data were analyzed under Application #124751 with Research Ethics Board approval from the Research Institute of the McGill University Health Centre (RI-MUHC; MP-37-2024-9259). The UKB has ethics approval from the North West Centre for Research Ethics Committee (11/NW/0382) and written informed consent from all participants.

Population and phenotype classification

The UKB is a population-based cohort of ~500 000 individuals with extensive phenotypic data, including kidney stones [16]. Unrelated participants of European genetic ancestry with complete data for age, sex, body mass index (BMI), and hypertension were included (n = 415 237) in the primary analysis. Kidney stone cases (n = 9009) were identified using standard International Classification of Diseases (ICD-9 and ICD-10) and OPCS Classification of Interventions and Procedures (OPCS-4) codes (Table S1A) previously used in this cohort [17]. Remaining participants served as controls (n = 406 228). Significant results were assessed in participants of South Asian (n = 7140; 194 cases) and African genetic ancestry (n = 7278; 128 cases), using identical inclusion and phenotyping criteria. Significance was defined as a nominal two-sided P ≤ 0.05 and a directionally consistent effect. Hypertension was defined by hospital diagnosis, medication use, or blood pressure criteria (Table S1B).

Genetic analyses

We conducted parallel analyses using exome sequencing and imputed genotype data (Fig. 1), for all 24 human CLDN genes: CLDN1–12, CLDN14–20, CLDN22–25, and CLDN34.

Figure 1:

Methods pipeline used to identify CLDN variant associations with kidney stones. The UKB exome sequencing release includes ~470 000 participants; analyses were restricted to those with both exome sequencing and imputed genotype data available, yielding a European genetic-ancestry sample of n = 415 237. Classification of kidney stone cases (n = 9 009) was based on ICD and OPCS codes, and the remaining served as controls (n = 406 228). Two complementary approaches were used in parallel: exome sequencing, examining 464 CLDN loss-of-function and missense variants annotated with ANNOVAR; and imputed genotyping, covering CLDN genes with 100 kb flanking regions, with clumping (r² < 0.1) to identify lead SNPs. Single variant association analyses were performed for both datasets, and significant variants were further assessed via PheWAS.

Exome sequencing data

Exome sequencing data (UKB Data-Field 23158) were aligned to the GRCh38 reference genome and underwent standard quality control (Table S2). Variants within CLDN gene regions (Table S3A) were annotated for functional consequence, deleteriousness scores (e.g. Combined Annotation Dependent Depletion [CADD]), and other functional predictions using ANNOVAR [18]. Variants were retained if they were classified as exonic and protein-altering (frameshift/non-frameshift substitution, non-synonymous single nucleotide variant, stop-gain/loss, start-loss) with a minor allele count ≥1 in cases.

Imputed genotype data

Imputed genotype data (UKB Data-Field 22828) were constrained to CLDN gene regions (GRCh37 coordinates ±100 kb). Since some CLDN family genes are located close to one another [19], we included intergenic regions uninterrupted by other protein-coding genes (maximum length 60 726 bp, coordinates in Table S3B). Variants with imputation quality scores <0.8 were excluded.

Single-variant analysis

Single-variant associations were tested using PLINK2 [20], employing age- and sex-adjusted logistic regression, and Firth regression when standard models failed to converge. Exome sequencing data were also evaluated using Fisher’s exact test with PLINK1.9. Effect sizes correspond to the effect allele (EA), regardless of frequency. Benjamini–Hochberg false discovery rate (FDR) correction was applied separately to exome and imputed analyses (α = 0.05).

Exome sequencing analysis results underwent permutation testing (1000 permutations) to calculate a family-wise empirical P-value. We permuted disease status while preserving case–control ratios and covariates to identify the significance of the lead variant for each permutation.

Identification of independent variants

Pairwise linkage disequilibrium (LD) was computed in PLINK2 within the primary cohort for FDR-significant biallelic variants. Independent associations were identified by grouping variants (r² ≥ 0.1) with the most significant lead variant. From lead variants, novel variants were defined if r² < 0.5 with known kidney stone associations.

Gene burden analysis

Gene-based burden tests and the Optimal Sequence Kernel Association Test (SKAT-O) were performed on exome variants [21], using REGENIE v3.3 [22]. Variants were grouped by functional consequence (Table S4) and analyzed using age- and sex-adjusted logistic regression with Firth correction. FDR correction was applied across all tests, accounting for the total number of genes and masks used. The most significant association across variant masks was then retained for each gene.

Phenome-wide association analysis

Phenome-wide association studies (PheWAS) were conducted on binary phenotypes and continuous traits to assess pleiotropic and systemic effects of significant variants.

For binary phenotypes, ICD-10 codes (UKB Data-Field 41234) were converted to phecodes and analyzed using the R PheWAS package [23] (Bonferroni correction: α = 0.05/N_phecodes, where N_phecodes ranged from 1267 to 1361). Phecodes were omitted if the minimum case or control threshold of 20 was not met for each given variant. For continuous traits, 33 quantitative variables (Table S5) were analyzed in PLINK2 using age- and sex-adjusted linear regression on their raw scales. A Bonferroni correction (α = 0.05/33) was applied to the results.

RESULTS

Participants

Of 415 237 participants, 9009 had kidney stones. Cases were more likely to be male, older, hypertensive, and to have higher BMI (Table 1), consistent with established risk factors [24].

Table 1:

Characteristics of participants in the UK Biobank with and without kidney stones.

	Overall	Controls	Cases	P
n	415 237	406 228	9009
Age	58.0 [51.0, 64.0]	58.0 [51.0, 63.0]	60.0 [53.0, 64.0]	<0.001
Male	190 995 (46.0)	184 696 (45.5)	6299 (69.9)	<0.001
BMI	26.7 [24.1, 29.9]	26.7 [24.1, 29.8]	28.2 [25.4, 31.5]	<0.001
Hypertension	223 344 (53.8)	217 441 (53.5)	5903 (65.5)	<0.001

Cohort characteristics for UKB genetic study of kidney stone disease (cases = 9009; controls = 406 228). Continuous variables are reported as median (interquartile range); categorical variables as n (%).

Associations of missense variants in CLDN16 and CLDN19 with kidney stones

We identified 43 novel significant associations with CLDN coding variants (Table S6), including multiple variants in CLDN16 and CLDN19.

Exome sequencing analysis identified a novel association between CLDN16 p.Gly57Arg (c.169G>A; rs747654138) and kidney stones (P = 9.5 × 10⁻⁸, odds ratio [OR] = 42.2; 4/9009 cases vs 5/406 208 controls) that persisted after permutation testing (P = 0.009). This variant, located in the first extracellular loop (Fig. 2), has a high CADD score (29.0) and was also associated with salivary gland hypertrophy (Phecode 527.1, OR = 484, 95% confidence interval [CI] 60–3909, P = 6.6 × 10⁻⁹). We also identified a significant association with the known pathogenic FHHNC variant CLDN16 p.Arg146His [25] (c.437G>A; rs772241737; P = 8.12 × 10⁻⁴, OR = 60.5; 2/9009 cases vs 1/406 213 controls), located in the second extracellular loop (Fig. 2B).

Figure 2:

Significant coding CLDN variants and their position on a canonical CLDN protein. (A) CLDN16 and CLDN19 coding variants associated with kidney stone disease in the UKB. Statistics for CLDN19 p.Arg200Gln (rs116804195) are derived from the imputed genotype analysis. AA change, amino acid change; NEA, non-effect allele; 0EA/1EA/2EA, 0, 1, or 2 copies of the effect allele. (B) Schematic of canonical claudin topology featuring four transmembrane domains and two extracellular loops (EL1, EL2). EL1 mediates ion selectivity, while EL2 facilitates cell–cell interactions. Colored circles indicate approximate variant positions for CLDN16 and CLDN19, distinguishing between UKB kidney stone associations (hatched) and variants previously reported in FHHNC literature (OMIM #248250 and #248190). The CLDN16 R146H variant is shown with both solid and hatched patterning, as it is both a kidney stone association and reported in FHHNC literature [25]. CLDN19 has multiple protein isoforms; the variant is labelled p.Arg200Gln (R200Q) here according to the canonical isoform (UniProt Q8N6F1-1, RefSeq NP_683763.2). EL1, extracellular loop 1; EL2, extracellular loop 2.

Imputed genotype analysis confirmed the previously reported CLDN19 coding variant association (p.Arg200Gln; c.599G>A; rs116804195; annotated as p.Gly172Ser for an alternative isoform; P = 2.6 × 10⁻⁴, 95% CI 1.10–1.37, OR = 1.23) (Fig. 2; Table 2) [26]. This finding was supported by concordant effect estimates within our exome sequencing dataset (Table S6). Phenotypic characterization of this variant revealed associations with increased serum alkaline phosphatase (U/L; P = 3.4 × 10⁻⁶, β = 1.09) and decreased serum phosphate levels (mmol/L; P = 9.4 × 10⁻⁶, β = −0.0065), consistent with imputed exome-wide association results reported by Barton et al. [27]. SKAT-O analysis of medium-impact variants revealed a significant association for CLDN19 (P = 0.0013). However, this association was lost when p.Arg200Gln was removed, suggesting that this variant is driving the association.

Table 2:

Significant lead variants in genomic regions (±100 kb) of CLDN genes.

CHR	POS	Lead SNP	CLDN region	Classification	EA	NEA	EAF	OR	95% CI	P
1	43201576	rs116804195	CLDN19  a	Missense	T	C	0.01560	1.23	(1.10, 1.37)	2.57E−04
1	43262514	rs1011416844	CLDN19	Intronic (TMEM269)	G	A	0.00026	3.16	(1.71, 5.81)	2.22E−04
3	137690920	rs547158477	CLDN18	Intergenic	C	T	0.00001	28.98	(5.00, 167.82)	1.72E−04
3	170097606	rs80304993	CLDN11	Intronic (SKIL)	G	A	0.76600	1.07	(1.03, 1.11)	2.27E−04
4	184150182	rs570418127	CLDN22–CLDN24	Intronic (WWC2)	A	T	0.00001	47.33	(7.67, 291.95)	3.25E−05
4	184202715	rs554535587	CLDN22–CLDN24	Intronic (WWC2)	T	TTA	0.00003	18.91	(4.64, 77.10)	4.13E−05
8	8584994	rs982244980	CLDN23	Intergenic	G	T	0.00018	4.13	(2.25, 7.59)	4.74E−06
13	96019637	rs1027426856	CLDN10	Intergenic	A	G	0.00004	8.92	(2.99, 26.59)	8.61E−05
13	96038436	rs147264740	CLDN10	Intergenic	T	G	0.00075	2.22	(1.46, 3.36)	1.69E−04
13	96072416	rs143641306	CLDN10	Intergenic	GTAA	G	0.00030	3.21	(1.76, 5.87)	1.42E−04
13	96194882	rs9516600	CLDN10  b	Intronic	G	A	0.61825	1.08	(1.05, 1.11)	1.32E−06
16	3027276	rs145053019	CLDN9–CLDN6	Intronic (PKMYT1)	T	C	0.00444	1.56	(1.29, 1.89)	4.98E−06
21	31570468	rs146511225	CLDN17–CLDN8	Intergenic	C	A	0.00004	18.25	(5.45, 61.20)	2.52E−06
21	37818871	rs2776288	CLDN14  a	Intergenic	A	G	0.62623	1.14	(1.10, 1.18)	3.05E−16
21	37884313	rs534259169	CLDN14	Intronic	G	A	0.00029	3.13	(1.75, 5.61)	1.21E−04
X	106177754	rs144531837	CLDN2	Intergenic	T	C	0.00001	64.95	(9.06, 465.48)	3.27E−05

^aPreviously reported association.

^b r ² > 0.6 with previously reported association.

Using imputed genotype data from the UKB, we identified 16 significant (FDR α = 0.05) and independent variants located ±100 kb of CLDN genes associated with kidney stone disease. Of these, 13 are novel, while the remaining three replicate known associations: one in LD (r² = 0.69) with a previously reported variant [9], and two previously reported in CLDN14 and CLDN19 [8, 26].

CHR, chromosome; POS, genomic position; NEA, non-effect allele; EAF, effect allele frequency.

CLDN14 association and cross-ancestry evaluation

Although synonymous variants were excluded from our primary exome analysis, we specifically evaluated a previously identified synonymous CLDN14 variant [8]. It reached genome-wide significance in the European genetic-ancestry cohort with a consistent effect (rs219780[C]: OR = 1.12, 95% CI 1.08–1.16, P = 2.7 × 10⁻⁹). Notably, we observed a directionally consistent, nominally significant association in African genetic-ancestry participants (OR = 1.39, P = 0.042), suggesting that the CLDN14 variant is associated with kidney stones in an additional ancestry group.

Non-coding CLDN variants associated with kidney stones

We examined non-coding variants within and flanking CLDN loci. We identified 16 independent variants significantly associated with kidney stone disease (Table 2, Table S7), 13 of which are novel; in combination with coding-variant findings, these comprised 56 distinct novel CLDN-associated signals.

In addition to the CLDN19 missense variant, we identified a novel independent variant, rs1011416844 (OR = 3.16, 95% CI 1.71–5.81, P = 2.2 × 10⁻⁴), located 56.6 kb upstream of CLDN19. It was also associated with disorders of mineral metabolism (Phecode 275, OR = 98.1, 95% CI 12.9–746.0, P = 9.4 × 10⁻⁶).

We identified three novel variants flanking CLDN10: rs143641306 (13.6 kb upstream; OR = 3.21, 95% CI 1.76–5.87, P = 1.4 × 10⁻⁴), associated with tooth-position anomalies and/or malocclusion (Phecode 524.3, OR = 17.5, 95% CI 5.1–60.2, P = 5.2 × 10⁻⁶); rs1027426856 (66.4 kb upstream; OR = 8.92, 95% CI 2.99–26.59, P = 8.6 × 10⁻⁵), linked with keratoconjunctivitis (Phecode 370.3, OR = 328, 95% CI 43–2513, P = 2.4 × 10⁻⁸); and rs147264740 (47.6 kb upstream; OR = 2.22, 95% CI 1.46–3.36, P = 1.7 × 10⁻⁴).

Additionally, rs9516600 (OR = 1.08, 95% CI 1.05–1.11, P = 1.3 × 10⁻⁶) in LD (r² = 0.69) with a known variant rs57719175 [9], was associated with increased cystatin C (mg/L; β = 0.0020, P = 1.5 × 10⁻⁷) and urea (mmol/L; β = 0.015, P = 2.1 × 10⁻⁶). This risk allele is an expression quantitative trait locus (eQTL), and reduces CLDN10 expression in visceral adipose (normalized effect size [NES] = −0.59, P = 7.5 × 10⁻²⁶), tibial nerve (NES = −0.58, P = 1.9 × 10⁻²⁴), and colonic tissues (NES = −0.53, P = 5.3 × 10⁻¹⁴), according to Genotype-Tissue Expression (GTEx) V10 [28].

We identified a novel kidney stone susceptibility locus at CLDN11 tagged by rs80304993 (OR = 1.07, 95% CI 1.03–1.11, P = 2.3 × 10⁻⁴). This risk allele has a strong cis-eQTL effect, associated with decreased CLDN11 (tibial nerve; NES = −0.22, P = 9.6 × 10⁻⁹) [28]. The variant is also associated with reduced levels of insulin-like growth factor 1 (IGF-1; nmol/L; β = −0.053, P = 2.9 × 10⁻⁴), apolipoprotein A-I (ApoA-I; g/L; β = −0.0024, P = 6.9 × 10⁻⁴), and high-density lipoprotein cholesterol (HDL-C; mmol/L; β = −0.0032, P = 8.6 × 10⁻⁴).

Consistent with prior evidence linking CLDN2 to kidney stones [15], we identified a novel association with rs144531837 (OR = 65.0, 95% CI 9.1–465.5, P = 3.3 × 10⁻⁵). We additionally identified CLDN18 as a novel risk locus; the lead variant rs547158477 (26.7 kb upstream; OR = 29.0, 95% CI 5.0–167.8, P = 1.7 × 10⁻⁴) was also associated with abnormal kidney function (Phecode 589, OR = 333, 95% CI 33–3359, P = 8.2 × 10⁻⁷). Finally, two novel associations were observed at the CLDN22–CLDN24 locus: rs570418127 (89.0 kb downstream; OR = 47.3, 95% CI 7.7–291.9, P = 3.2 × 10⁻⁵) and rs554535587 (36.5 kb downstream; OR = 18.9, 95% CI 4.6–77.1, P = 4.1 × 10⁻⁵).

Across all reported loci, adjustment for BMI and hypertension did not meaningfully alter our results.

DISCUSSION

We examined variation across the CLDN gene family, identifying multiple putative contributors to kidney stone susceptibility. Exome sequencing enabled detection of rare variants that contribute to kidney stone risk in the general population. While all members were analyzed, associations were concentrated in CLDN16, CLDN19, CLDN10, CLDN14, and CLDN11. These genes are expressed in the TAL [13, 29–31], a critical site for calcium and magnesium reabsorption [32].

Recent large-scale GWAS, such as those by Hao et al. and Lovegrove et al. [9, 26], have mapped the broader polygenic landscape of kidney stone disease, implicating pathways including calcium and phosphate metabolism. To our knowledge, this is the first study to evaluate both rare and common, coding and non-coding variation across all CLDNs for kidney stone disease within a single family-wide targeted association analysis. This present study deepens the genetic characterization of the CLDN tight junction gene family, which has established and emerging roles in renal mineral handling [2, 33]. It also implicates heterozygous variants in CLDN16 and CLDN19, previously identified only under a recessive model [6, 7]. Together, these findings situate CLDN-mediated mineral homeostasis within the broader genetic architecture of kidney stone disease.

CLDN16 and CLDN19 coding variants define an intermediate phenotype

Rare, missense and loss-of-function variants in CLDN16 and CLDN19 underlie FHHNC, a recessive monogenic form of kidney stone disease. Our findings are consistent with the idea that heterozygous variants in these genes are also associated with increased kidney stone risk in the general population. Supporting this, an FHHNC family had two heterozygous CLDN16 p.Leu151Phe carriers with hypercalciuria (the mother and maternal grandfather of the index case); the grandfather also had a history of kidney stones [34].

CLDN16 and CLDN19 heterodimerize to form a putative TAL cation channel, mediating magnesium, and calcium reabsorption [11, 12, 35]. Disruption of this channel can reduce paracellular Mg²⁺ and Ca²⁺ reabsorption, shifting the divalent cation balance and increasing urinary Ca²⁺ concentration, which can promote calcium–phosphate stone formation.

We identified a novel association with CLDN16 p.Gly57Arg, which substitutes a small, uncharged glycine with a large, positively charged arginine in the first extracellular loop. This domain is enriched in negatively charged residues critical for shaping paracellular cation selectivity [2, 13]. The substantial change in charge and side-chain size in a critical domain is consistent with its high CADD score (29).

In contrast, CLDN16 p.Arg146His is a known pathogenic variant recently reported in a male with FHHNC and recurring nephrolithiasis [25]. This variant introduces an aromatic ring into the second extracellular loop, potentially impairing CLDN–CLDN interactions. Its identification in a population-based cohort confirms that established recessive disease variants can confer kidney stone risk in carriers in the general population.

The association of CLDN19 p.Arg200Gln with decreased serum phosphate and increased alkaline phosphatase may reflect systemic disturbances in mineral metabolism involving the calcium–phosphate axis. During our analysis, Lovegrove et al. also identified this variant as a kidney stone locus in their genome-wide meta-analysis [26].

Experimental studies provide a mechanistic basis for these associations and have demonstrated that pathogenic CLDN16 and CLDN19 variants may not be effectively transported to the tight junction like their wildtype counterparts [36–38]. Heterozygous variants, potentially causing partial reductions in function, may therefore also affect TAL divalent cation handling and increase kidney stone risk.

Pleiotropic effects of CLDN10 variants

We identified three novel CLDN10 variants associated with kidney stones (rs143641306, rs1027426856, and rs147264740). These signals are independent (all r² < 0.001) of the previously reported rs57719175 [9], which we also replicated (OR = 1.07, 95% CI 1.0–1.1, P = 1.2 × 10⁻⁵) (see below). The additional associations we identified possibly reflect the pleiotropic effects of CLDN10 across renal, oral, and ocular systems. CLDN10 encodes a tight junction protein expressed in the TAL alongside CLDN14, CLDN16, CLDN19, and CLDN11 [11, 29, 32, 39]. In mice, Cldn10 loss causes hypermagnesemia and nephrocalcinosis [40]. Thus, these CLDN10 variants may contribute to altered calcium reabsorption within the TAL and stone formation [40].

CLDN10 is responsible for autosomal recessive HELIX syndrome (Hypohidrosis, Electrolyte imbalance, Lacrimal dysfunction, Ichthyosis, and Xerostomia; OMIM #617671) [41], is expressed in dental epithelium, and is implicated in enamel formation [42]. Thus, the association of rs143641306 with malocclusion may reflect either secondary xerostomia-related enamel wear, or a primary developmental effect. Similarly, the association of rs1027426856 with keratoconjunctivitis is consistent with the lacrimal dysfunction reported in HELIX syndrome. However, these secondary associations are marked by wide confidence intervals reflecting sparse case counts, and should be considered hypothesis-generating.

We replicated a previously reported association at the CLDN10 locus [9]; our lead variant, rs9516600, is in moderate LD with the originally reported rs57719175 (r² = 0.69) [9]. Our results additionally show novel associations of rs9516600 with increased serum cystatin C and urea. These elevations reflect reduced filtration efficiency: cystatin C is a sensitive, muscle–mass-independent marker of early glomerular dysfunction [43], while elevated urea is common in moderate-to-advanced chronic kidney disease [44].

These observations suggest that common CLDN10 variants may partially recapitulate the renal, oral, and ocular features of HELIX syndrome.

CLDN11 and kidney stone pathogenesis

We identified a novel association with rs80304993 near CLDN11. CLDN11 is also expressed in cortical kidney tissue within the proximal and distal tubules and the TAL, suggesting a potential role in renal physiology [10, 29, 45]. This variant is also a significant cis-eQTL, associated with decreased CLDN11 expression in tibial nerve tissue in GTEx. While this tissue is not directly relevant to kidney stone formation, this eQTL likely serves as a proxy for disease–relevant tissues, as kidney tissue is among the least sampled tissues in the GTEx database, limiting statistical power to detect kidney-specific eQTLs [46]. This interpretation is supported by evidence that regulatory variants frequently share effects across tissues [47]. Thus, reduced CLDN11 expression is consistent with a role in kidney stone susceptibility.

The rs80304993 variant demonstrated significant associations with several circulating metabolites, including decreased IGF-1 and HDL-C levels. IGF-1 plays a role in calcium homeostasis through direct interactions with calcium channels [48]; the observed association with decreased IGF-1 levels may therefore represent an additional pathway by which this variant influences kidney stone susceptibility. Similarly, decreased HDL-C is associated with increased kidney stone risk [49]. Together, these findings suggest that the CLDN11 variant has pleiotropic effects on kidney stone pathophysiology warranting further exploration.

Strengths and limitations

Our study has several strengths. We leveraged the large-scale, well-phenotyped UK Biobank to analyze both exome sequencing and imputed genotype data, ensuring substantial statistical power and standardized case definitions. We applied rigorous multiple-testing corrections and validated our primary European genetic-ancestry analysis findings within South Asian and African genetic-ancestry subsets.

Despite these strengths, several limitations exist. First, reliance on hospital and procedure codes likely excludes kidney stones managed outside the hospital setting, reducing power and biasing associations toward the null. Second, the UKB reflects relatively high socioeconomic status, which may limit generalizability [50, 51]. Third, reliance on array and exome data precludes assessment of large structural variations. Fourth, due to the UKB’s composition, the primary analysis was limited to European genetic ancestry, which may limit the generalizability across diverse populations. Fifth, some PheWAS associations, particularly those with wide confidence intervals, reflect sparse case counts rather than robust pleiotropic signals. Sixth, odds ratios for rare coding-variant associations are based on small carrier counts and should be interpreted with caution, as effect size estimates are unstable under these conditions. Seventh, permutation testing was limited to 1000 permutations, which restricts the minimum attainable empirical P-value to 0.001 and provides an upper bound for significance rather than a precise estimate. Additionally, the UKB does not have 24 h urinary calcium and phosphate measures. Future research in cohorts with these key urinary measures will be needed to fully assess the functional impact of associated variants. Finally, the lack of functional validation limits causal interpretation of our results.

CONCLUSIONS

Both common and rare variants, within and near CLDN genes, contribute to kidney stone susceptibility in the general population. We identified novel and replicated known associations in CLDN10, CLDN14, CLDN16, and CLDN19, and expanded the range of kidney stone-associated CLDN family members through newly identified associations in CLDN11. Several of these loci exhibit pleiotropic effects consistent with known CLDN physiology [7, 13, 41].

Importantly, we highlight the existence of a risk continuum: genes underlying monogenic kidney stone syndromes also harbor additional impactful variants, such as the strong-effect CLDN16 variants (p.Gly57Arg, p.Arg146His) and the common CLDN19 risk allele (p.Arg200Gln), that are associated with increased susceptibility in this population-based dataset. This has important implications for future research: while the penetrance and absolute risk in heterozygous carriers have yet to be established, our results suggest that first-degree relatives of FHHNC patients who carry CLDN16 or CLDN19 variants may have increased susceptibility to kidney stones. Future studies in large biobanks and well-phenotyped families will be essential to clarify the penetrance and phenotypic spectrum of heterozygous CLDN16 and CLDN19 variants to help guide clinical practice.

AUTHORS' CONTRIBUTIONS

I.Y.L.: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, writing—original draft, writing—review and editing. J.H.: conceptualization, project administration, resources, supervision, writing—original draft, writing—review and editing. E.M.: data curation, formal analysis, investigation, methodology, software, writing—review and editing. L.D.: conceptualization, data curation, formal analysis, methodology, writing—review and editing. A.K.R.: conceptualization, funding acquisition, resources, writing—review and editing. J.C.E.: conceptualization, methodology, resources, supervision, writing—review and editing. I.R.G.: conceptualization, funding acquisition, supervision, writing—original draft, writing—review and editing. All authors have read and agreed to the submitted version of the manuscript.

CONFLICT OF INTEREST STATEMENT

None declared.

FUNDING

This study was supported by the Kidney Foundation of Canada (A.K.R., I.R.G.), the RI-MUHC–Montreal Children’s Hospital (MCH) Foundation Studentship and Fellowship in Child Health Research (I.Y.L.), and DERIVE, supported by an anonymous donor through the MCH Foundation (A.K.R.).

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available within the article and its supplementary materials. Individual-level data were derived from the UK Biobank and are available to approved researchers upon application. Further details can be found at https://www.ukbiobank.ac.uk/use-our-data/apply-for-access/.