Mouse and human studies support DSTYK loss of function as a low-penetrance and variable expressivity risk factor for congenital urinary tract anomalies

Abstract

Purpose:

Previous work identified rare variants in DSTYK associated with human congenital anomalies of the kidney and urinary tract (CAKUT). Here, we present a series of mouse and human studies to clarify the association, penetrance, and expressivity of DSTYK variants.

Methods:

We phenotypically characterized Dstyk knockout mice of 3 separate inbred backgrounds and re-analyzed the original family segregating the DSTYK c.654+1G>A splice-site variant (referred to as “SSV” below). DSTYK loss of function (LOF) and SSVs were annotated in individuals with CAKUT, epilepsy, or amyotrophic lateral sclerosis vs controls. A phenome-wide association study analysis was also performed using United Kingdom Biobank (UKBB) data.

Results:

Results demonstrate ~20% to 25% penetrance of obstructive uropathy, at least, in C57BL/6J and FVB/NJ Dstyk^−/− mice. Phenotypic penetrance increased to ~40% in C3H/HeJ mutants, with mild-to-moderate severity. Re-analysis of the original family segregating the rare SSV showed low penetrance (43.8%) and no alternative genetic causes for CAKUT. LOF DSTYK variants burden showed significant excess for CAKUT and epilepsy vs controls and an exploratory phenome-wide association study supported association with neurological disorders.

Conclusion:

These data support causality for DSTYK LOF variants and highlights the need for large-scale sequencing studies (here >200,000 cases) to accurately assess causality for genes and variants to lowly penetrant traits with common population prevalence.

Introduction

Congenital Anomalies of the Kidney and Urinary Tract (CAKUT) comprise the most common cause(s) of pediatric kidney failure, with up to 50% of afflicted children worldwide eventually requiring kidney replacement therapy.^1–3 This is accompanied by a significant impact on the development and progression of chronic kidney disease well into adulthood in addition to the development of extra-renal pathologies, such as hypertension and cardiovascular disease.⁴ CAKUT encompasses a highly understudied spectrum of conditions with a wide range of developmental defects, including kidney agenesis, kidney hypodysplasia, congenital obstructive uropathy (COU), duplex kidney and duplex ureter, horseshoe kidney, vesicoureteral reflux, and posterior urethral valve.^4–6 Despite an overall high impact on outcomes, CAKUT subcategories display a widespread population prevalence and disease severity, ranging from ultra-rare and lethal conditions, such as bilateral kidney agenesis, to more common and benign conditions that can remain undetected throughout life, such as ureteric dilation, mild hydronephrosis, and pyelectasia (as part of COU), as well as duplicated ureters, kidneys, or both (duplicated collecting system).⁵ In particular, COU, identified as hydronephrosis by imaging studies, is the most frequent anomaly of the urinary tract detected by prenatal ultrasound in up to 2% of otherwise normal pregnancies^7,8 and diagnosed in 2 to 29 cases per 10,000 live births.⁹ Given that hydronephrosis is usually clinically silent, resolves or improves over time, and that both neonates and children do not undergo routine screening by imaging studies, its incidence is vastly underestimated.¹⁰ Thus, within the context of a relatively benign course of disease and high population frequency, negative purifying selection (ie, the inability to segregate pathogenic alleles among generations) is likely to play a smaller role compared with more rare and severe CAKUT phenotypes that otherwise affect survival and likelihood to procreate. As a result, the genetic architecture of COU is likely to be highly complex, and rare variants with moderate effect size plus relatively low penetrance are expected to play a role.

In a previous report, we identified rare variants in DSTYK in patients with dominantly inherited CAKUT, mostly consistent with COU.¹¹ This gene encodes a dual serine/threonine and tyrosine kinase, which localizes to cell membranes and was shown to be highly expressed in the maturing epithelia of all major organs, including the kidney. In zebrafish, the inactivation of dstyk resulted in phenotypes resembling a loss of fibroblast growth factor (FGF) signaling. Consistently, DSTYK colocalized with FGF receptors 1 and 2 (FGFR1 and FGFR2) in the ureteric bud and metanephric mesenchyme, and silencing of DSTYK in human HEK293T cells resulted in a reduction of FGF-dependent extracellular signal-regulated kinase (ERK) phosphorylation, suggesting that DSTYK is a positive regulator of FGF-mediated signaling during nephrogenesis.¹¹

The identification of DSTYK as a candidate susceptibility gene for CAKUT was predicated on a large family with 7 affected individuals in which a combination of genome-wide linkage analysis with exome sequencing (ES) prioritized a single splice-site variant (SSV), c.654+1G>A, demonstrated to result in a heterozygous 27-bp deletion resulting from the use of an alternative splice donor within exon 2, in turn, resulting in an in-frame deletion of 9 highly evolutionarily conserved amino acids. Corroborating causality was the identification of an early protein-truncating variant (c.24G>A, p.W8*) in a child with COU, along with the identification of additional rare missense variants at other highly conserved amino acid residues. This study was conducted before 2013, at the advent of exome capture coupled with massively parallel sequencing as a novel and powerful approach to simultaneously extract the full spectrum of coding genetic variation in a single individual. Early studies immediately proved successful, especially where ES was conducted on very rare conditions with recessive or dominant de novo modes of inheritance.^12–14 Using the widely accepted criteria at the time, the DSTYK splice donor variant was identified via robust linkage analysis, ES, and absence in 864 control chromosomes, including 96 that were matched for the same Sardinian village where the original family was recruited. However, in 2014, a large-scale publicly available aggregation of ES projects, then called exome aggregation consortium or ExAC¹⁵ was released, and the DSTYK c.654+1G>A was found at a population frequency of about 1 in 3000 individuals of European ancestry, thus complicating the inference of genetic causality for this variant.

In the current series of studies, we describe the phenotypic analysis of a genetrap mutant for Dstyk across 3 different mouse genetic backgrounds, coupled with human genetic studies, in order to resolve the link between DSTYK and human urinary tract malformations.

Materials and Methods

Animals and husbandry

All animal studies were approved by the Columbia University Institutional Animal Care and Use Committee of the Institute of Comparative Medicine. All breedings and embryo harvests were performed at Columbia University, in accord with IACUC-approved procedures and protocols.

To study the consequences of Dstyk inactivation on the development of the mouse kidney, we procured live mice carrying the IST10368F3 (Omnibank Vector 76) Dstyk gene trap mutation from Texas A&M Institute for Genomic Medicine (IGM) (Mouse Accession: NM_0099041; shorthand: Dstyk^−/−; strain name: C57BL/6N-^{Ripk5Gt(IST10368F3)Tigm}). As shown in Supplemental Figure 1A, the gene trap insertion site falls in the second intron, thereby disrupting downstream transcription. For genotyping, a 3-primer setup was used with the following primer names and sequences: IST10368F3-F: GCAGAGACGGAATTGAGCAG, IST103 68F3-R: GAACAGCCCTTACTGCCCT, Downstream rev/LTR: CCAATAAACCCTCTTGCAGTTGC. Wild-type reaction oligos (IST10368F3-F + IST10368F3-R) produced a 319 bp product, whereas mutant oligos (IST10368F3-F + Downstream rev/LTR) produced a 238 bp product (Supplemental Figure 1B). Congenic lines were produced by moving the genetrap mutation onto the FVB/NJ or C3H/HeJ inbred lines, in which original C57BL/6J Dstyk^−/− mice were first backcrossed onto each of the strains to generate a F1 intercross. Progeny carrying the mutation, as identified by genotyping, were then selected for backcrossing onto each respective inbred strain for at least 10 generations (N10). Kidneys and urinary tracts were then evaluated from P0/newborn pups derived from crosses between parents at or greater than N10.

Validation of Dstyk suppression via rt-qPCR

Suppression of Dstyk expression was validated in mutant mice via rt-qPCR.¹⁶ Total RNA was extracted with TRIzol from mutant and wild-type brains and lungs, which are tissues known to normally express Dstyk.¹⁷ cDNA was subsequently synthesized using SuperScrip III One-Step RT-PCR System, and data were acquired using QuantStudio3 qPCR System (Applied Biosystems), which were then analyzed using the ΔΔCt relative quantification method. Dstyk primers were designed toward the 5′ end of Ex11 (mDSTYK_Ex11-F1: AAAC-CAGAAGCTATGATGTCAGG) and 3′ end of Ex12 (mDSTYK_Ex12-R1: GTCCCTCTGCGCACATTATT), yielding a 209 bp product (NCBI Reference Sequence NM_172516.4). β-actin was used as the reference with the following primer sequences: β-actin_Ex3-F: TGTTACCAACTGGGACGACA, β-actin_Ex4-R: GGGGTGTTGAAGGTCTCAAA.

Tissue processing and histochemistry

Tissues processed for basic histochemistry were dissected at respective ages, harvested, and processed for hematoxylin and eosin and periodic acid-schiff staining according to standard laboratory practices (as shown in Figures 1 and 2). Briefly, tissues from mice at each age were removed from the embryo and fixed in 4% paraformaldehyde overnight at 4 °C. Samples were then transferred to 70% ethanol, embedded in paraffin, and sectioned at 5 μm. Slides stained using standard procedures were then imaged on an Olympus IX-73 equipped with a DP80 camera. Where necessary, color levels were adjusted in Photoshop (Adobe), and calibrated scalebars added.

Figure 1. Phenotypes observed in Dstyk mutants on a C57BL/6J background.

Dstyk^−/− mutants exhibit low penetrance of perinatal mortality (A) from birth (P0) up until 2 days post-partum (P2) and low penetrance of urinary tract phenotypes (hydronephrosis and hydroureter) during development (E17.5-E19.5) and early post-natal life (P0-P3) (B). These phenotypes were never observed in wild-type littermates. Microscopic analysis of urinary tract tissue (C-F) shows proximal ureteric dilation and significant hydronephrosis with compression of the renal papilla and cortex (E, arrowhead) with overall preserved nephrogenic zone (F). WT, wild type; SAC, sacrificed/euthanized.

Figure 2. Congenic Dstyk−/− mutants in FVB/NJ and C3H/HeJ genetic background display additional urinary tract phenotypes and increased penetrance of proximal ureteric dilation, respectively.

Transfer of the Dstyk gene trap construct onto the FVB/NJ background resulted in replication of the proximal ureteric dilation phenotype (B, H, and H’) as observed in the B6 background, and elicited new phenotypes, including partial kidney and ureter duplication with intrarenal septation (C, I, and I’; yellow arrowheads). Transfer of the Dstyk gene trap construct onto the C3H/HeJ background resulted in the occurrence of mild kidney hypoplasia and mild hydronephrosis and hydroureter (E, F, J, and J’). The penetrance of urinary tract phenotypes was 20% in congenic FVBN Dstyk^−/− mice (K), similar to Dstyk^−/− in the B6 background. Congenic C3H Dstyk^−/− penetrance of urinary tract phenotypes was higher compared with the other genetic backgrounds (approximately 40%, L), but they were mild to moderate in severity.

Study participants

The study involving human subjects was conducted in accordance with the Declaration of Helsinki. All participants and/or guardians provided written informed consent, and the study was approved by the Institutional Review Board of Columbia University Irving Medical Center (CUIMC) and the local ethics committee at the IRCCS Giannina Gaslini Institute of Genoa, Italy.

Exome and genome sequencing

Genomic DNA for the family study subjects was isolated from whole blood according to standard protocols. ES was performed using the Illumina Hiseq2500 sequencing platform, using capture kit IDTERPv1, on 7 affected individuals from the DSTYK family (IDs 3,10,7,8,13,18, and 19; Figure 3A). The DRAGEN v3 platform was utilized to map sequenced reads to the reference genome (hs37d5.fa, Ensembl -GRCh37.73), and GATK 3.6 was subsequently used for base quality recalibration, indel realignment, and variant calling. ClinEff was used for variant annotation with Ensembl (version GRCh38), EVS-v.0.0.30, ExAC 0.3,¹⁵ gnomAD Exome and gnomAD Genome version 2.1,¹⁸ dbNSFP 4.1a, HGMD 2021.4, Clinvar 2022-01-10,¹⁹ American College of Medical Genetics (ACMG) v3, and REVEL 2016-06-03. Resulting variant calls, sample-level site coverages data, and annotations were stored in the Analysis Tool for Annotated Variants (ATAV) centralized database and queried.²⁰ Individual ID 10 was also subjected to 30X GS. Sequencing libraries were prepared from 100 ng of genomic DNA using the KAPA Hyper Library preparation kit (Kapa Biosystems). Libraries were subjected to 2 × 150 bp paired-end sequencing on HiSeq X instrument (Illumina), to achieve a mean sequencing depth of 30x. Variant analysis was performed using the GATK best practices guidelines.²¹ Raw reads were aligned to NCBI genome build GRCh37 using Burrows-Wheeler Aligner (BWA).²² PicardTools was used to mark duplicate reads. Base Quality Score Recalibration was performed using BaseRecalibrator tool included in Genome Analysis Toolkit v3.6 (GATK).²³ Local realignment around indels was performed using GATK IndelRealigner. Variant calls were generated using GATK HaplotypeCaller and GenotypeGVCFs. Variant Quality Score Recalibration was performed using GATK VariantRecalibrator. Truth sensitivity threshold of 99.6 was used for both single-nucleotide variants and insertions and deletions. Variant effects were predicted using Ensembl Variant Effect Predictor v91.²⁴ Genomic copy number alterations were identified using Canvas—an algorithm for calling copy number variants from a diploid sample.²⁵ Structural variations were discovered using Manta—an algorithm that discovers candidate structural variations from discordant pair and split-read alignments, followed by local assembly and realignment to refine candidates.²⁶

Figure 3. Human genetic studies on the extended family segregating the DSTYK c.654+1G>A splice-site variant and in large exome sequencing cohorts.

A. Updated pedigree structure of the study family segregating the DSTYK c.654+1G>A splice-site variant. Squares represent males, circles females. Deceased individuals are marked with a slash. Dark-filled symbols indicate cases with CAKUT, empty ones represent unaffected individuals, and gray-filled ones indicate individuals whose phenotype remains unknown after clinical and/or imaging analysis. Individuals with green outer circle had history of seizures and were treated with anti-seizure medications. Members carrying the DSTYK variant are indicated by solid red circles, whereas empty red circles indicate non-heterozygotes. The previously reported family is shown on the top right in the turquoise box and shaded in turquoise within the extended pedigree. IDs in red denote individuals with newly obtained clinical data and/or imaging studies. B. Burden of loss-of-function (LOF) DSTYK variants in 32,973 individuals across different phenotypes: CAKUT, N = 3682; epilepsy, N = 7591; amyotrophic lateral sclerosis (ALS), N = 8264; and controls, N = 13,436. A nominally significant excess burden of LOF variants was identified for CAKUT and epilepsy compared with controls. C. Enrichment analysis for the DSTYK c.654+1G>A splice-site variant in the same 32,973 individuals. Nominally significant enrichment was present for epilepsy; the variant was over 2-fold enriched in CAKUT compared with controls, but the result was not statistically significant, suggesting the need for much larger cohorts of cases and controls in order to clarify the presence or absence of association.

Variant-level quality control and prioritization

We used a manually curated list of 382 genes that, when mutated, are known to cause Mendelian forms of isolated or syndromic CAKUT.²⁷ Variant-level data from the 7 study ID individuals within the 382 prioritized genes were queried using the variant annotation function implemented in the ATAV,²⁰ the analysis variant engine that powers our exome-genome sequencing warehouse (Supplemental Table 1). Variant filtering was performed to require a quality score >50, quality by depth score ≥2, genotyping quality score ≥20, mapping quality score ≥40, and coverage ≥10. Alternate read percentages were within the range of 0.3 and 0.7 for heterozygous genotypes. To further ensure the removal of sequencing artifacts, variants that occurred ≥20 within the CAKUT cohort and variants that appeared ≥500 within the internal ATAV controls cohort were removed. For variant prioritization, we used Varsome (https://varsome.com/) web-based platforms that implement the ACMG guidelines²⁸ as a first-pass screen to predict an ACMG verdict for each uploaded variant for further clinical variant interpretation and genotype-phenotype correlation for all 382 genes. We used the following criteria to define a positive genetic finding for our clinical research variant adjudication. First-tier positive findings were considered if the genotype was already reported as pathogenic or likely pathogenic in ClinVar¹⁹ or classified as pathogenic or likely pathogenic by strict ACMG criteria via individual variant curation in Varsome. Because missense variants and variants never observed in public databases, such as gnomAD, rarely meet ACMG “pathogenic” or “likely pathogenic” criteria and are often classified as VUS (uncertain significance), in order to define our second-tier positive genetic finding, we used the following criteria: absent of exceedingly rare in public databases, as well as in our in-house 11,818 multiethnic population controls from the IGM,²⁰ a Revel score ≥0.5,²⁹ and plausibility of the genetic variant to be associated to the observed COU phenotype. Additionally, variants with the PVS1 classifier (ie, null variant in a gene where loss of function [LOF] is a known mechanism of disease) were classified as positive findings even if other criteria were not fulfilled. We next confirmed prioritized variants through Sanger sequencing in the patient DNA and, when available, in family members for segregation analysis in order to add support to our pathogenicity adjudication.

DNA array genotyping and copy number variation analysis

Individual ID 10 was also subjected to chromosomal microarray genotyping using the Illumina Infinium 660 chip. Copy number variation analysis was conducted as previously described^30–35 for the detection of pathogenic or likely pathogenic genomic disorders, as well as gene-disrupting structural variants at any of the 382 CAKUT Mendelian genes as above.

Large-scale ES investigations for DSTYK variants

To determine if DSTYK variants, specifically LOF in aggregate or the c.654+1G>A splice-site single-nucleotide variant are enriched in CAKUT and/or epilepsy at population level, we aggregated and analyzed ES or genome sequencing (GS) from a cohort of 32,973 individuals across different phenotypes: CAKUT, N = 3682; epilepsy, N = 7591; amyotrophic lateral sclerosis (ALS), N = 8264; and controls, N = 13,436. The control subjects were children or adults without overt clinical conditions, including neurologic or kidney disease. All genetic data were aggregated, harmonized using an identical base and variant calling pipeline as above, and analyzed at the IGM at CUIMC.²⁰ Pairwise comparisons were conducted using Fisher’s exact test, R version 4.3.0.

Nonsense-mediate decay (NMD) analysis of LOF variants

To assess if any of the DSTYK LOF variants identified here are predicted to escape NMD, we used the novel method aenmd R package (v0.3.11). (Klonowski J, Liang Q, Coban-Akdemir ZC, Lo C, Kostka D. aenmd: annotating escape from nonsense-mediated decay for transcripts with protein-truncating variants. bioRxiv. 2023. http://doi.org/10.1101/2023.03.17.533185). Variants were converted to VCF format and queried using GRCh37 assembly. The “process_variants” and “annotate_nmd” functions were applied for NMD escape prediction and annotation.

Phenome-wide association study (PheWAS)

We performed a phenome-wide association analysis for DSYTK predicted LOF, including stop-gain, frameshift, stop-lost, start-lost, and essential splice variant heterozygotes, defined previously in the UK Biobank (UKBB) data set.³⁶ There were 10,221 ICD-9 codes for UKBB participants (N = 460,363) with imputed genotype data mapped to 1817 distinct encodes. Phenome-wide associations were performed using the PheWAS R package.³⁷ The package uses 2 ICD-9 code occurrences within a given code grouping to define a case and pre-defined “control” groups for each encode. All 1817 encodes were tested using logistic regression with case-control status as the outcome and genotype, sex, age, batch, and 5 principal components of ancestry as predictors. We set the Bonferroni corrected statistical significance threshold for phenome-wide significance at 2.75 × 10⁻⁵ (0.05/1817 phecodes tested).

Sample-size calculations

The sample size required to achieve 80% power in detecting association with binary outcomes was calculated with the genpwr.calc function in genpwr version 1.0.4.³⁸ The calculations were performed under a dominant model, considering a range of minor allele frequencies (minor allele frequencies [MAF]: 0.1% to 0.001%), odds ratios (OR: 1.5, 2, and 2.5), case rates (0.01%, 0.1%, and 1%), and at the significance level of P = 5 × 10⁻⁸. Sample size with an 80% power to detect rare variant associations in a genomic region (per-gene burden test) was determined with Sequence Kernel Association Test (SKAT) version 2.2.5.³⁹ The default SKAT.haplotypes data set was used for simulation. For power analysis, a subregion length equivalent to the size of the DSTYK transcript, which included UTR (69,097 in GRCh37) was chosen. MAF was set to 0.001 (0.1%), case proportion 0.1, prevalence 0.01. The Get_RequiredSampleSize function was used to calculate sample sizes at the significance levels of 2.5 × 10⁻⁶ and 5 × 10⁻⁸.

Results

Dstyk knockout results in low penetrance of isolated ureteric obstruction in C57BL/6J mice

To understand the underlying etiology of COU associated with LOF variants in Dstyk, a genetrap mutant was procured from the Texas A&M Institute of Genomic Medicine (see Materials and Methods). As shown (Supplemental Figure 1), the gene trap insertion site is located between the 2nd and 3rd exon and is upstream of the sequence coding for the kinase domain, which spans from amino acid position P650 to L904 (UniProt ID: Q6XUX1). Decreased expression of Dstyk was confirmed via rt-qPCR, where homozygous mutants showed significantly decreased expression values that were negligible (<5%) compared with wild-type littermates in various tissues, including kidney, brain, and lung (Supplemental Figure 2). Dstyk mutant mice, generated on a pure inbred C57BL/6J genetic background, showed incomplete penetrance of perinatal lethality in the absence of overt lethal congenital malformations (Figure 1A) with 2 of 25 heterozygous mice (8%) and 7 of 24 (29.2%) homozygous mutants not surviving beyond 2 days post-partum. For animals surviving past postnatal day P2, survival rates were unchanged compared with wild-type mice.

Phenotypic analysis of the kidneys and urinary tract was carried out at 3 time points: during development but after the onset of glomerular filtration (E17-E19), at birth (P0), and between P1 and P3. Dstyk^−/− mutants exhibited incomplete penetrance of urinary tract defects recapitulating human COU. At E17-E19, 7/34 (20.6%) of embryos exhibited varying degrees of hydronephrosis, from moderate to severe (Supplemental Figure 3). At P0, 4 of 13 (30.8%) homozygous mutants exhibited a similar obstructive uropathy phenotype (Figure 1B and E), and the penetrance and expressivity returned to 22% for P1 to P3 homozygous mutants (5/23, 21.8%) (Figure 1C). Hence, in total, we observed obstructive uropathy phenotypes in 16 of 70 (22.8%) mutants compared with 0 of 134 wild-type littermates (Fisher’s Exact P = 1.05 × 10⁻⁸) across the same developmental time points (Figure 1B and C). These data implicate LOF of Dstyk in low penetrance of COU, even in a pure mouse genetic background.

Generation of congenic Dstyk knockouts in 2 additional inbred strains increases breadth and penetrance of CAKUT phenotypes

Since C57BL/6 mice are generally more resistant to kidney diseases,⁴⁰ we sought to determine whether the low penetrance of CAKUT phenotypes (~20%) observed in Dstyk^−/− mutants could be influenced by the different polygenic background of inbred mice. We first generated a new congenic line for the Dstyk genetrap mutation by consecutively backcrossing mutants for at least 10 generations into the FVB/NJ strain, which is a strain considered to be more susceptible to kidney and urinary tract disease, particularly glomerulosclerosis,^41–43 cisplatin nephrotoxicity,⁴⁴ and diabetic nephropathy.^45,46 As with mutants on the C57BL/6 background, Dstyk^−/− FVB/NJ congenic mice were analyzed at birth/P0. Although suppression/knockout of Dstyk in FVB/NJ mice did not increase the penetrance of CAKUT by histological analysis (8/40, 20%; Figure 2B, 2C, 2H, 2H’, 2I, 2I’, and 2K), we observed additional urinary tract defects, including partial kidney, and duplication of the pelvis and proximal ureter in at least 6 of 40 homozygous mutants (Figure 2I and I’, Table 1). The partial duplication of the kidney was characterized by elongation and misshaping of the kidney at macroscopic observation (Figure 2C) and by a clear parenchyma bridge, ie, a demarcation and separation of the 2 hemi-kidneys by a cellular boundary of the cortical nephrogenic zone (Figure 2I and I’, arrowheads).

Table 1.

Kidney and urinary tract developmental phenotypes observed in Dstyk mutants in 3 different genetic backgrounds

Genetic Background	Observed CAKUT Phenotypes
B6 Dstyk−/−	UPJO, hydronephrosis, megaureter (16/70 = 22.8%)
FVB Dstyk−/−	Kidney: Misshapen + bifid kidneys (6/40 = 15%), kidney hypoplasia (1 = 2.5%) Ureteric: hydroureter (2/40 = 5%), dilated pelvis (2/40 = 5%)
C3H Dstyk−/−	Kidney: kidney hypoplasia (1/23 = 4.3%) Ureteric: mild-to-moderate obstructive phenotype, including hydroureter or UPJO (9/23 = 39.1%)

CAKUT, congenital anomalies of the kidney and urinary tract; UPJO, ureteropelvic junction obstruction.

Because the penetrance of CAKUT remained low in Dstyk^−/− mice that were bred onto a more permissive genetic background, such as FVB/NJ, we generated an additional congenic line by similarly backcrossing Dstyk C57BL/6 mutant mice for at least 10 generations into the C3H/HeJ background, which was selected because of the presence of a natural predisposition to vesicoureteral reflux,^47–49 which therefore represents another permissive model for CAKUT. Accordingly, suppression of Dstyk in this genetic background led to an increase in penetrance of CAKUT phenotypes. However, this was still incomplete as hydronephrosis and proximal ureter dilation were observed in 9 of 23 (39.1%) of homozygous mutants at birth (Figure 2E, 2F, 2J, J’, and 2L). The urinary tract phenotypes mostly consisted of mild-to-moderate hydronephrosis/proximal hydroureter and, in 1 instance, kidney hypoplasia (Table 1).

Expansion and re-analysis of the pedigree segregating the DSTYK c.654+1G>A variant shows low penetrance of CAKUT and seizure disorder and no alternative plausible candidates

In light of the incomplete penetrance of CAKUT in knockout Dstyk mice across 3 different inbred genetic backgrounds, we decided to re-characterize and expand studies for the original family that led to the mapping of the DSTYK c.654+1G>A splice-site variant.¹¹ To do so, a team composed of a pediatric nephrologist certified in ultrasonography, a nephrology fellow, and a nurse was sent to the hometown of the family in Sardinia to conduct clinical and chart review, as well as kidney and urinary tract ultrasound. The original family spanned 4 generations and included 7 affected individuals, 7 unaffected individuals, and 6 individuals for whom clinical or imaging studies to determine their status were not available (Figure 3A, turquoise box). The new genealogy spanned 6 generations and included a total of 69 individuals for whom history, clinical data, and/or imaging studies were available (Figure 3A). Blood samples for DNA analysis were re-collected for all available individuals that were included in our previous analysis (IDs 4, 5, 9, 10, 8, 11, 12, 13, 14, 15, 16, 17, 18, 19, and 20) and obtained for 11 additional individuals (IDs 21 to 31). Targeted Sanger sequencing was conducted on all individuals with DNA available to test for the presence of the c.654+1G>A variant.

The analysis resulted in 5 individuals, previously with an unknown phenotype, being reclassified as unaffected (IDs 1, 4, 15, 17, and 20), and multiple novel family members, including 3 new variant heterozygotes who were also classified as unaffected by clinical and ultrasonographic analysis (IDs 21, 22, and 23). We also confirmed the clinical history of epilepsy in 3 individuals (IDs 7, 8, and 13). Based on this analysis, we identified 16 DSTYK variant heterozygotes, of which 7 were characterized as affected and 9 as unaffected, leading to an estimated penetrance of 43.8% for the candidate variant. This is likely an overestimation given the absence of family history for CAKUT in many members from the individual ID 1 family lineage.

To further investigate whether potential pathogenic variants were missed in the original analysis, we performed ES in all affected individuals (IDs 3, 10, 7, 8, 13, 18, 19) and Illumina DNA microarray genotyping and 30X GS in individual ID 10. Using the ACMG guidelines for clinical variant interpretation,²⁸ all ES cases were analyzed independently for pathogenic and likely pathogenic variants in a panel of 382 known CAKUT genes.²⁷ None of the 7 affected individuals carrying the DSTYK c.654+1G>A SSV carried additional pathogenic or likely pathogenic variant in DSTYK or in another CAKUT-associated gene, thus excluding obvious recessive or digenic inheritance. Moreover, GS in individual ID 10 did not reveal any additional causal variant, either single nucleotide or structural, not captured by ES, in any of the known genes nor in novel genes underlying the linkage regions from the original study.¹¹ Finally, using published criteria,^31–35 no pathogenic copy number variant was identified via chromosomal microarray analysis.

Altogether, these studies do not support alternative plausible genetic scenarios explaining the CAKUT phenotype in this family. Hence, the most parsimonious hypothesis supports, together with the mouse studies above, a low penetrance effect of the c.654+1G>A splice-site variant on the susceptibility to CAKUT in this family.

Interestingly, we also confirmed the clinical diagnosis of seizure disorder/epilepsy in 3 variant heterozygotes (IDs 7, 8, and 13) who also had CAKUT. After our original report, a founder homozygous intragenic deletion affecting the last 2 exons of DSTYK was identified in a consanguineous family segregating spastic paraparesis and a cutaneous pigmentation disorder.⁵⁰ The index case of this family had gait abnormalities at age of 2 to 3 preceding the onset of paraparesis, and abdominal imaging studies showed horse-shoe kidney. His younger brother had febrile seizures at the age of 6 years. More recently, another group reported a family with a novel heterozygous DSTYK missense variant segregating in a family affected by mild lower urinary tract dysfunction and mild spastic paraparesis.⁵¹ Interestingly, both dizygotic twins from this family presented with absence seizures at 4 years of age and were started on antiepileptic medications.

Large-scale ES and GS studies show an enrichment of DSTYK LOF variants in CAKUT and epilepsy

Considering that the mouse data support causality for DSTYK LOF, and literature reports, as well as a review of cases from our original family, suggest that DSTYK is involved in variable urinary tract defects and movement disorders (spastic paraplegia, seizures, and epilepsy), we next resorted to human genetics studies in large cohorts with available exome or GS data. Here, we tested the hypothesis that, at the population level, there would be an enrichment of protein-truncating variants in individuals with CAKUT and, possibly, related movement disorder phenotypes, especially epilepsy, compared with controls. To do so, we annotated DSTYK LOF variants in aggregate and the c.654+1G>A SSV separately, in a set of 32,973 individuals across different phenotypes: CAKUT, N = 3682; epilepsy, N = 7591; ALS, N = 8264; and controls, N = 13,436. We detected burden of LOF (including stop-gain, frameshift, and splice-site variants) DSTYK variants, with excess for CAKUT (5/3,682 cases; P = .0065; OR = 9.13, 95% CI 1.49-95.96) and epilepsy (7/7591 cases; P = .0135; OR = 6.20, 95% CI 1.18-61.21), compared with 2/13,436 controls (Figure 3B). Enrichment analysis for the DSTYK c.654+1G>A SSV in the same 32,973 individuals resulted in significant enrichment for epilepsy (P = .0172; OR = 6.04, 95% CI 1.10-9.12), and a >2-fold enrichment in CAKUT cases compared with controls (although not statistically significant) (Figure 3C). On the contrary, there was no significant enrichment, either at the burden or single-variant analysis, for ALS, a late-onset neurologic disease, used here as a disease control data set. When combining all LOF with the c.654+1G>A variant, 9/3682 CAKUT cases (P = .0070; OR = 3.66, 95% CI 1.28-10.40) and 19/7591 epilepsy cases (P = .0011; OR = 3.75, 95% CI 1.62-9.41) carried DSTYK variants compared with 9/13,436 controls, respectively.

Finally, when examining the 5 CAKUT individuals carrying LOF variants (Table 2), 4 of them had relatively mild developmental urinary tract disease characterized by hydronephrosis and/or kidney pelvis dilation, and only 1 had a more overt kidney trait (hypodysplasia) combined with a ureteric defect (ureterocele, case D4). In 2 cases (D1 and D5), the urinary tract defect was present in a variant heterozygote family member (for D1 in the mother, for D5 in the monozygotic twin fetus). Extra-urinary defects included hemangiomas (2/5 cases) and neurodevelopmental disease: D1 with childhood ataxia (and affected mother with childhood febrile seizures) and D5 monozygotic twin fetus with anencephaly. Given the variability of phenotypes and severity in these cases, we sought to ascertain if any of these variants result in escape from NMD, potentially leading to gain-of-function or dominant-negative effects. We utilized the aenmd R package, a newly developed tool for prediction of effect on NMD from genetic variants. This led to the identification of the frame-shift variant 1-205138725-TC-T as an NMD escape candidate. Altogether, of the 6 LOF variants (the c.654+1G>A SSV from the original report, and the 5 reported here) identified in CAKUT cases, 4 are predicted to result in LOF/haploinsufficiency and 2 in NMD escape, possibly indicating an alternative mode of disease causation, such as gain-of-function or dominant-negative effect.

Table 2.

Clinical characteristics of CAKUT cases with DSTYK protein truncating variants

ID	Variant ID	Variant	Consequence	gnomAD Global AF (N = 125,748 ES; 15,708 GS)	Internal Controls AF (N = 13,436)	Segregation	Sex	Urinary Phenotype	Non-urinary Phenotype
D1a	1-205180640-C-T	c.24G>A	p.Trp8*	0	0	Segregating from mother with unilateral left caliectasis and history of childhood febrile seizures	M	Left UPJO	Cutaneous hemangioma, childhood ataxia, hearing loss
D2	1-205138810-G-A	c.805C>T	p.Arg269*	3.98 × 10−6	0	Unknown	F	Persistent left caliectasis/hydronephrosis; small non-obstructing stone	Liver hemangioma, thrombotic thrombocytopenic purpura, cerebrovascular accident
D3	1-205138725-TC-T	c.889delG	p.Glu297Argfs*11	0	0	Unknown	M	Left Kidney hypoplasia and ectopy, left ureterocele	-
D4	1-205138462-G-A	c.1153C>T	p.Gln385*	0	0	Unknown	F	Bilateral proximal ureter dilation/extrarenal pelvises	Right subclavian artery stenosis, chronic headaches
D5a	1-205133085-T-C	c.1325-2A>G	Splice acceptor	0	0	Monozygotic twin fetus with hydronephrosis/pyelectasis and anencephaly	F	Bilateral hydronephrosis/pyelectasis, improved post-natally	Anencephaly in monozygotic twin

AF, allele frequency; ES, exome sequencing; F, female; M, male; UPJO, ureteropelvic junction obstruction

^aPreviously reported^11,27,41.

A pheWAS suggests DSTYK LOF variants as susceptibility factors for movement disorders

In order to gain insight into the role of DSTYK genetic variation on susceptibility to human disease at population level, we next performed a PheWAS using ES data from the UKBB data set to assess phenome-wide associations for LOF variants in aggregate (Figure 4A, Supplemental Table 1) and for the c.654+1G>A splice-site variant alone (Figure 4B, Supplemental Table 1), in the DSTYK gene. The top associated phecodes for the LOF model were “Myopathy” (P = 8.76 × 10⁻⁸; OR = 15.02), “Muscular dystrophies and other myopathies” (P = 5.42 × 10⁻⁶; OR = 9.94), and “Malignant neoplasm of renal pelvis” (P = 2.50 × 10⁻⁵; OR 20.66). Similarly, the top associations for the c.654+1G>A splice-site variant were “Myopathy” (P = 8.76 × 10⁻⁸; OR = 16.35), “Muscular dystrophies and other myopathies” (P = 2.51 × 10⁻⁶; OR = 10.79). Additional suggestive associations were found for related neurologic phecodes such as “Abnormal involuntary movements” (P = 2.38 × 10⁻⁴, OR = 4.55 for LOF; P = 1.16 × 10⁻⁴, OR = 4.90 for the single variant) and “Hereditary and idiopathic peripheral neuropathy” (P = 1.29 × 10⁻³, OR = 9.90 for LOF; P = 8.91 × 10⁻⁴, OR = 10.68 for the single variant). These data, although preliminary, suggest involvement of DSTYK LOF in neuromuscular conditions and, possibly, in susceptibility to urinary tract cancer. Moreover, these neurologic phecodes potentially correlate with the neurologic findings described in our cases and in the literature.^11,50,51 The UKBB, being composed of only adult and relatively healthy individuals, is depleted from cases with hydronephrosis (N = 1778; this phecode likely includes not only congenital forms but also secondary and non-developmental causes of kidney pelvis dilation) and epilepsy (N = 1467; similarly, only a fraction of these cases are represented by pediatric-onset epilepsy). Despite these limitations, we observed enrichment of DSTYK variants for these phecodes with direction-consistent odds ratios of similar magnitude to our estimates from the ES studies above (Supplemental Table 1).

Figure 4. Exploratory phenome-wide association study for DSTYK protein-truncating variants and sample size simulations.

PheWAS for (A) DSTYK pLOF variant heterozygotes and (B) splice variant (c.654+1G>A) heterozygotes data from UKBB with total N = 460,360 with both genotype and phenotype data available. This analysis was performed under the dominant inheritance model and adjusted for age, sex, diabetes, batch, and ancestry. The red horizontal lines indicate a phenome-wide significance level after accounting for the number of phecodes tested (P = 2.8E-05). y axis: −log10 (P value) from fixed effects meta-analysis. P values are 2 sided and provided without accounting for multiple testing. x-axis: system-based phecode groupings. An upward-pointing triangle indicates increased odds for a given phecode, downward-pointing triangle indicates reduced risk. C-E. Sample size estimates for an 80% power in detecting association of a rare, single variant across different effect sizes under the dominant model at genome-wide significance level (α = 5 × 10⁻⁸). Sample sizes are log scaled.

Simulation studies indicate the need for extremely large sample size for refuting genetic causality for the DSTYK c.654+1G>A splice-site variant

Based on the above mouse and human genetic findings supporting the causal role of DSTYK LOF and the c.654+1G>A variants but with incomplete penetrance and variable expressivity, we conducted sample size simulation under different, but realistic, scenarios. To do so, we estimated the number of cases and controls needed to unequivocally implicate or refute genetic association with variants of similar MAF and effect size as the DSTYK c.654+1G>A splice-site variant, ie, MAF between 1 × 10⁻³ and 1 × 10⁻⁴, OR between 1.5 and 2.5, and for a realistic window of population frequency for congenital hydronephrosis between 1% and 0.01% (Figure 4C–E, Supplemental Table 2). Under the single-variant model, the power analyses conducted using genpwr³⁸ revealed very large sample sizes of cases and controls needed to detect association, even in the most optimistic (yet less realistic) scenarios. In fact, for a hypothetical scenario (“best-case scenario”) of a disease prevalence of 1%, OR=2.5,andMAF=1 × 10⁻³ (Figure 4E), it would still take nearly 13,000 COU cases and 1.3 million controls to detect association for the c.654+1G>A splice-site variant at a genome-wide significance threshold of 5 × 10⁻⁸. For yet another hypothetical scenario (“worse-case scenario”) of MAF < 1 × 10⁻⁴, disease prevalence of 1 in 10,000, and OR 1.5, it would require, approximately, the entire world population (9.1 billion individuals; Figure 4C). Using point estimates derived from our family-based and large-scale ES above as the most likely and realistic scenarios (Figure 4D), we would need approximately 25,000 to 250,000 cases. Even by analyzing DSTYK in a gene-based, burden level using SKAT,^52,53 we estimated a requirement for 148,554 individuals to detect gene-level association at an exome-wide significance level of 2.5 × 10⁻⁶.

Discussion

The recent advancement of sequencing technologies accompanied by worldwide efforts in data sharing has resulted in the appreciation of rare genomic variation at a new scale. Before the release of ES/GS databases of genomic variants, such as the ExAC in 2014,¹⁵ and then gnomAD,¹⁸ as well as others, the traditional framework used to infer causality for a novel variant relied on the targeted sequencing of a few hundred ethnically matched controls. With deeper appreciation of human genetic variation, it became apparent that many genetic variants predicted to be deleterious and disruptive of protein coding genes are present at low frequency in the general population. This observation prompted reassessment of genetic causality inference for many variants discovered in the pre-genetic databases era, posing significant interpretative challenges. In fact, rare, but not ultra-rare, variants previously implicated in human disease but that exist with a background frequency in the general population can indicate several possible alternative scenarios. First, the variant was misclassified as causal while, indeed, representing a rare variant of no causal or clinical significance. Second, the variant has incomplete penetrance, and although conferring a true and measurable risk for disease, it does not conform to Mendelian genetics rules and provides limited actionability. Third, the variant, although present in the population, has large effect size but perhaps variable expressivity in its clinical manifestations. Finally, the variant is causal in a Mendelian-like fashion and suggests that the condition of interest is more common (and commonly caused by the variant) than previously anticipated. Discriminating between these possible scenarios requires significant effort in devising the appropriate functional and human genetic studies.

In this study, we addressed the inference of genetic causality for rare genetic variations in DSTYK and in particular for LOF variants and for the low-frequency c.654+1G>A splice-site variant. To do so, we conducted a series of studies in the mouse and in humans. First, we characterized the kidney and urinary tract phenotypes resulting from the loss of Dstyk in the mouse, in 3 different genetic backgrounds. These analyses showed that Dstyk is involved in normal urinary tract and kidney development and that its loss causes phenotypic correlates to human CAKUT and, especially, urinary tract defects, such as COU, which was the most commonly observed phenotype in individuals with DSTYK genetic variants.^11,51 These data strongly support genetic causality for the DSTYK LOF variants identified in CAKUT subjects. Nevertheless, CAKUT phenotypes were observed at incomplete penetrance (20% to 39%) even in Dstyk^−/− mice, suggesting interpretative challenges in clinical genetics settings. In light of the incomplete penetrance even in inbred knockout mice, we revisited the original pedigree segregating the low-frequency c.654+1G>A splice-site variant and conducted extensive clinical and genetic studies that confirmed a causal effect of this variant as the most parsimonious explanation for the observed CAKUT phenotypes but also demonstrated low penetrance in this family. Subsequent large-scale sequencing analyses showed enrichment for both DSTYK LOF variants in aggregate and for the c.654+1G>A alone, for CAKUT (especially hydro-nephrosis/ureteric dilation), and epilepsy compared with healthy or disease controls. Finally, an exploratory PheWAS using ES from the UKBB suggests myopathy and movement disorders as possible components of the phenotypic spectrum of DSTYK-associated conditions and shows direction- and effect-size-consistent enrichment of rare variant for epilepsy and hydronephrosis, although underpowered. Interestingly, among the significant phecodes associations, there was “Malignant neoplasm of the renal pelvis,” in line with recent evidence of shared risk between CAKUT and urinary tract cancers.⁵⁴

The results from our study are also consistent with other reports of DSTYK variants in which subjects showed paraparesis, horseshoe kidney, and febrile seizures⁵⁰; mild paraparesis, seizures, and lower urinary tract dysfunction⁵¹; and Vertebral, Anal atresia, Cardiac, Tracheal, Esophageal, Renal, Limbs association,⁵⁵ a syndrome often involving the urinary tract. In our main pedigree, 3 variant heterozygotes had a diagnosis of seizure disorder and 2 out of 5 families with DSTYK LOF (Table 2) displayed neurological phenotypes. These recurrent associations of mild or incompletely penetrant urinary tract defects with neurodevelopmental conditions suggest DSTYK dose reduction as a small-to-moderate risk factor for urinary tract and neurologic developmental conditions. The estimated effect size of DSTYK risk variants also suggest that current studies that apply the rules of Mendelian genetics are inadequate to directly establish causality, nor to refute it. In fact, our simulation studies indicate, in the most realistic scenario, a need for approximately 250,000 CAKUT cases to detect or reject associations at single-variant level and nearly 150,000 cases at the gene level. In practical terms, considering the low penetrance of DSTYK variants and that COU is often clinically silent and improves or resolves during childhood,¹⁰ when conducting population-based genetic screenings in adults, a variant heterozygote is more likely to be asymptomatic than symptomatic.

Based on the aggregated evidence from the literature and this study, we propose, from a clinical genetics perspective, the following recommendations: (1) DSTYK missense variants and the low-frequency c.654+1G>A splice donor variants should not be used in routine clinical genetic settings to ascertain diagnosis or risk, (2) DSTYK LOF variants might be used in clinical settings when coupled with inheritance information (de novo or segregating variants) and with associated clinical plausibility,⁵⁶ and (3) no DSTYK variants, in general, should currently be used for risk stratification in pre-natal settings or in asymptomatic individuals.

In conclusion, mouse and human data support causality for DSTYK LOF variants but with low penetrance and variable expressivity, and very large sequencing studies are required to implicate or disprove rare variants with small-to-moderate effect size underlying traits with relatively common population prevalence and mild or subclinical presentation, such as COU.

Supplementary Material

The online version of this article (https://doi.org/10.1016/j.gim.2023.100983) contains supplemental material, which is available to authorized users.

Data Availability

All mouse data are readily available from the corresponding author upon reasonable request. A reasonable request is defined as a request to review data for purposes of further study, including data re-analysis. Human data describing the original variant were previously published in their entirety.¹¹ Simulations were carried out using data from the UK Biobank and are publicly available.