Contribution of variants across the allelic frequency spectrum to cystic kidney disease

Introduction Cystic kidney disease (CyKD) is frequently a familial disease, with ~85% of probands receiving a monogenic diagnosis. However, gene discovery has been led by family-based and candidate gene studies, limiting the ascertainment of non-Mendelian genetic contributors to the disease. Using whole genome sequencing data provided by the 100,000 Genomes Project (100KGP), we used hypothesis-free approaches to systematically characterize and quantify the genetic contributors to CyKD across variant types and the allele frequency spectrum. Methods We performed a sequencing-based genome-wide association study in 1,209 unrelated patients recruited to the 100,000 Genomes Project with CyKD and 26,096 ancestry-matched unaffected controls. The analysis was inclusive of individuals with diverse genetic ancestries. Enrichment of common, low-frequency (minor allele frequency [MAF] > 0.1%) and rare (MAF < 0.1%) single-nucleotide variants (SNV), indels and rare structural variants (SV) on a genome-wide and per-gene basis was sought using a generalised linear mixed model approach to account for population structure. Meta-analysis of CyKD cohorts from Finngen, the UK Biobank and BioBank Japan was performed. Results In 995 of the 1209 (82.30%) CyKD cases a likely disease-causing monogenic variant was identified. Gene-based analysis of rare SNVs/indels predicted to be damaging revealed PKD1 (P=1.13x10-309), PKD2 (P=1.96x10-150), DNAJB11 (P=3.52x10-7), COL4A3 (P=1.26x10-6) and truncating monoallelic PKHD1 (P=2.98x10-8) variants to be significantly associated with disease. Depleting for solved cases led to the emergence of a significant association at IFT140 (P=3.46x10-17) and strengthening of the COL4A3 (P=9.27x10-7) association, driven exclusively by heterozygous variants for both genes. After depleting for those harbouring IFT140 and COL4A3 variants , no other genes were identified. Risk of disease attributable to monoallelic defects of multiple genes linked with CyKD was quantified, with lower risk seen in rarer and more recently described genetic diagnoses. Genome-wide structural variant associations highlighted deletions in PKD1 (P=2.17x10-22), PKD2 (P=7.48x10-12) and the 17q12 locus containing HNF1B (P=4.12x10-8) as statistically significant contributors to disease. Genome-wide analysis of over 18 million common and low-frequency variants in the Finnish population revealed evidence of association (P=1.4x10-149) of a heterozygous stop-gain variant in PKHD1 that is endemic (MAF=4.7x10-03) in this population. Meta-analysis of 2,923 cases and 900,824 controls across 6,641,351 common and low frequency variants including UK, Japanese and Finnish biobanks did not reveal any novel significant associations. SNVs with a MAF>0.1% accounted for between 3 and 9% of the heritability of CyKD across three different European ancestry cohorts. Conclusions These findings represent an unbiased examination of the genetic architecture of a national CyKD cohort using robust statistical methodology. Causative monoallelic mutations in IFT140 have recently been reported in other cohorts associated with a milder phenotype than PKD1/2-associated disease. The association with COL4A3 suggests that in some circumstances CyKD may be the presenting feature of collagen IV-related kidney disease and the significant association observed with monoallelic predicted loss-of-function PKHD1 variants extends the spectrum of phenotypic abnormalities associated with this gene. In addition to quantification of the contribution of non-coding and structural variants to CyKD, the per gene quantification of CyKD risk presented could be used to inform genetic testing and counselling strategies clinically and we also show that common variants make a small contribution to CyKD heritability. Keywords: genomics, cystic kidney disease, renal, ADPKD