Genetic Association Analysis of Chronic Kidney Disease Progression in a Small Korean Cohort Study

CKD is a widespread and significant global health problem, affecting an estimated 10% of the adult population, including that in Korea.¹ The occurrence of CKD is known to have a significant genetic component, with an estimated high degree of heritability (25%–44%).²

Decline in eGFR is usually determined using linear mixed regression models that (unlike simpler models) account for correlations between visits and time dependence.³ It can be used to define clinically relevant end points that occur earlier than end stage kidney disease.⁴

Genome-wide association studies (GWASs) in nephrology tend to be biased toward European ancestry, making studies with diverse ancestries particularly important.⁵ Most GWASs on CKD progression have been conducted in population-based studies,⁶ and the genomic regions identified have substantial overlap with cross-sectional kidney function loci.⁷ However, data from CKD case cohorts are limited, mainly because of power constraints.

Oh et al.⁸ performed a GWAS to identify genetic loci associated with eGFR slope among 1738 patients with CKD in the Korean CKD cohort study (KNOW-CKD). Analyses were stratified by diabetes mellitus (DM, n=771; non-DM, n=967) and adjusted for age, sex, and the first five genetic principal components. eGFR slope as the primary outcome was defined as eGFR change over time and estimated by a mixed model using repeated measurements at multiple time points. Replication analyses were conducted in the Chronic Renal Insufficiency Cohort (CRIC) study of 2498 patients with CKD of European ancestry (DM, n=511 non-DM, n=828) and African ancestry (DM, n=576; non-DM, n=583). Two single nucleotide polymorphisms (SNPs), rs59402340 (Pmeta=7.23×10⁻⁸) within TPPP from the non-DM GWAS and rs28629773 (Pmeta=1.87×10⁻⁷) in FAT1-LINC02374 from the overall GWAS, met the replication criteria. Both SNPs also showed significant associations in KNOW-CKD with binary kidney outcomes, such as 30% reduction in eGFR at 3 years and renal replacement therapy. The discovery results were further validated by two GWASs in the general population from the CKDGen Consortium, one cross-sectional GWAS for eGFR or BUN (n=765,348) and another longitudinal GWAS with or without CKD at baseline (n=63,558). Of the initially significant (but not replicated) SNPs, rs35914637 within TPPP and five SNPs (rs13103811, rs11132426, rs4861741, rs6553037, and rs28629773) near FAT1 were validated (P value <0.05).

To identify potential causal variants near TPPP and FAT1-LINC02374, statistical fine mapping analysis identified five SNPs, including rs28629773 in FAT1-LINC02374, with the highest posterior inclusion probability of 1.0.

Risk SNPs near TPPP and FAT1-LINC02374 had significant (P<0.05) expression quantitative trait loci association P values derived from the Nephrotic Syndrome Study Network and the Genotype-Tissue Expression project. Histone activity markers (H3K27ac and H3K4me1) showed peaks near both sets of risk SNPs at the TPPP and FAT1-LINC02374 loci, and open chromatin accessibility in kidney tissue around TPPP showed a similar pattern.

A transcription factor binding site disruption analysis identified 18 SNPs with a significant effect on the affinity of transcription factors (P<10⁻⁴), including rs59402340 of TPPP and rs4861741 of FAT1-LINC02374, which were predicted to disrupt MZF1 and SOX17/POU1F1 binding, respectively.

Oh et al. investigated relevant phenotypes and performed interesting in silico follow-up analyses. Strengths of their study include the availability of multiple longitudinal eGFR measurements per patient, the East Asian study population, the replication of their findings, and the in silico follow-up analyses.

Both the TPPP and FAT1 genes have been previously implicated in renal function and disease in large cross-sectional and longitudinal European ancestry population-based GWASs.^6,7 A monogenic (rare deleterious) mutation in FAT1 causes a syndrome that includes kidney disease.⁹

Adjustment for baseline eGFR in analyses of eGFR decline is controversial. As eGFR decline is independently associated with the risk of kidney failure and mortality after adjustment for baseline eGFR,⁴ additional genetic variants may be found by adding baseline eGFR as a covariate. A higher baseline eGFR is associated with a steeper subsequent absolute decline in eGFR, and by removing this effect through adjustment, it may be possible to discover genes on a common pathway for the progression of kidney disease. However, collider bias may be an issue,¹⁰ so sensitivity analyses are important. In the present study by Oh et al., adjusting the models for baseline eGFR did not change the associations presented.

As in other cohorts of participants with CKD, the relatively small sample size compared to large population-based studies limits the statistical power. Assuming an allele frequency of 5% and a SD for eGFR decline of 0.5 ml/min per 1.72 m² per year,¹¹ in a study, such as KNOW-CKD, with 2500 individuals, one can expect to detect markers with an effect difference on annual eGFR decline of ±0.25 ml/min per 1.73 m² with a power of >80% (own calculations). Given the mean annual eGFR decline of approximately 1 ml/min per 1.73 m² per year found across age groups in healthy participants,¹¹ these are only large effects. As seen in other GWASs, there are likely to be many more markers relevant to kidney function decline, each with smaller effect sizes and/or lower allele frequencies, that can be found when sample size is substantially increased (Figure 1). Combining many genetic variants for eGFR decline with individually small effects results in a polygenic risk score associated with increased risk of renal end points, such as acute and chronic kidney failure.⁶

Figure 1.

Absolute change in annual eGFR decline detectable with 80% power by minor allele frequency. Colors represent three cohort sizes (n=2500/10,000/100,000). Power calculations assume a P value threshold of 5×10⁻⁸ and a SD for eGFR decline of 0.42¹¹ and were performed using the R package gwas-power (https://github.com/kaustubhad/gwas-power).¹²

In summary, Oh et al. present variants in two loci that are significantly associated with decline in kidney function in a Korean CKD cohort. Given the published literature and supporting in silico data for TPPP and FAT1-LINC02374, it will ultimately be interesting to identify the underlying causal variants and investigate them in experimental models.

Disclosures

M. Wuttke reports the following: Employer: Freiburg University Hospital and Mesalvo GmbH; Ownership Interest: Minority shareholder of Mesalvo GmbH; and Advisory or Leadership Role: Chief Technology Officer of Mesalvo GmbH. The remaining author has nothing to disclose.

Funding

The work of S. Ghasemi and M. Wuttke was supported by German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) Project ID 431984000, SFB 1453.

Author Contributions

Formal analysis: Matthias Wuttke.

Software: Matthias Wuttke.

Writing – original draft: Sahar Ghasemi, Matthias Wuttke.

Writing – review & editing: Sahar Ghasemi, Matthias Wuttke.