Glutathione S-transferase µ1 (GSTM1), belongs to the family of glutathione-S-transferases that metabolize a broad range of reactive oxygen species and aldehydes (Hayes et al., 2005; Yang et al., 2009). Loss of GSTM1 is very common in the population; ∼50% of whites and 27% of blacks have 0 copies of GSTM1 (Garte et al., 2001). Deletion of one or both copies of the gene results in reduced amount of GSTM1 (Board et al., 1990) and could lead to increased oxidative stress due to diminished ability to neutralize reactive chemical species. An association between loss of GSTM1 and chronic kidney disease (CKD) progression has been reported in the African American Study of Kidney and Hypertension (AASK) (Chang et al., 2013). The association between GSTM1 copy number and kidney failure was reported in the Atherosclerosis Risk in Communities (ARIC) study, which had a larger sample size, both in blacks and whites (Tin et al., 2017). However, data from smaller case–control studies have had mixed findings (Yang et al., 2004; Agrawal et al., 2007; Tiwari et al., 2009; Gutierrez-Amavizca et al., 2013; Nomani et al., 2016). The inconsistent results may be due to insufficient sample size and different populations and disease background.

The Geisinger MyCode^® Community Health Initiative is an EHR-linked biobank for precision medicine research (Carey et al., 2016). The population served by Geisinger have low rates of out migration; thus, the electronic health record (EHR) data are relatively complete longitudinally, with a median of 14 years of follow-up (Carey et al., 2016). In addition, a high proportion of participants were found to have first- and second-degree relatedness through cryptic relatedness analysis (Staples et al., 2018). Through the ongoing DiscovEHR collaboration with the Regeneron Genetic Center, whole exome sequence (WES) data are available from ∼92,000 MyCode^® participants to date (Abul-Husn et al., 2016; Carey et al., 2016; Dewey et al., 2016). These comprehensive clinical data are linked to matched genetic data and provide power to identify disease-genetic associations (Rader and Damrauer, 2016).

Confirming whether or not loss of GSTM1 increases the risk of kidney failure is important, given the high prevalence of GSTM1 loss in the population and the serious morbidity and mortality associated with kidney failure. In this study, we examine the relationship between GSTM1 copy number and incident kidney failure in the Geisinger MyCode cohort.

To our knowledge, this is the largest study to investigate the association of GSTM1 copy number variation with kidney failure. There were a total of 46,187 unrelated whites and 796 blacks in the study with an average of 9.3 years’ follow-up. The frequency of the GSTM1 copy numbers were very similar to those reported previously (Garte et al., 2001; Chang et al., 2013; Tin et al., 2017). Our results showed no significant association between GSTM1 copy number and risk of kidney failure in unadjusted or adjusted analyses whether using a genotypic, dominant, or recessive genetic model.

Data from the ARIC and AASK cohorts suggested that the loss of GSTM1 increased the risk of kidney failure or accelerated CKD progression (Chang et al., 2013; Bodonyi-Kovacs et al., 2016; Tin et al., 2017). In ARIC, a community-based cohort of middle-aged black and white participants, there were 3,461 white participants with WES reads. In fully adjusted models among whites in ARIC, loss of GSTM1 was associated with risk of kidney failure (0 or 1 copy vs. 2 copies: HR 2.54; 95% CI: 1.32–4.88). It is possible that differences in baseline characteristics of the study populations could explain differing findings. Our cohort had a lower prevalence of smoking (never smoking 48 vs. 38%), a higher prevalence of diabetes (14 vs. 8%), and was more contemporary (median baseline year 2004 vs. ARIC baseline year 1987–1989). However, in sensitivity analyses in 14,572 white participants with baseline year between 1996 and 2002 in our cohort, there was no association between GSTM1 loss and risk of kidney failure.

There is also the possibility that GSTM1 may only be deleterious when GFR falls below certain levels or in the setting of specific types of kidney disease. In the AASK study, a randomized trial comparing different antihypertensive medications and levels of blood pressure control in blacks with CKD attributed to hypertension, 692 participants had GSTM1 genotyping completed. Loss of GSTM1 in AASK was associated with increased risk of CKD progression (HR 0 copy: HR 1.88; 95% CI: 1.07–3.30; HR 1 copy: 1.68; 95% CI: 1.00–2.84). While there were few blacks in our study population with CKD, we found no association between loss of GSTM1 and risk of kidney failure in 3,129 white participants with baseline eGFR < 60 ml/min/1.73 m². Some smaller case–control studies also tested a recessive genetic model (0 copies vs. 1 or 2 copies) of GSTM1 and found that GSTM1 0 copy was associated with kidney failure in several case–control studies (Agrawal et al., 2007; Chang et al., 2013; Gutierrez-Amavizca et al., 2013; Suvakov et al., 2013). In our study, however, we examined both recessive and dominant genetic models and found no associations between GSTM1 copy number and risk of kidney failure.

Some limitations of this study are worth noting. First, the genotypes of GSTM1 in this study were derived from WES, and the thresholds were determined empirically from the histogram. Multiplex PCR validation was not performed for this study. However, this method was used in the ARIC study, which showed 99.3% agreement when comparing the results of 0 copy with ≥ 1 copy from PCR (Tin et al., 2017). Moreover, the frequency of GSTM1 copy numbers in this study is similar to those reported previously for both whites and blacks (Chang et al., 2013; Tin et al., 2017), and it follows the Hardy–Weinberg equilibrium. We also examined the possibility of mismapping due to the complexity of GSTM locus. The histogram of the sum of normalized coverage of GSTM5, the paralog gene of GSTM1, is unimodal distribution. Based on these evidences, we believe that the copy number results derived from WES had a low error rate and can be used with confidence for analysis. Second, the sample size of blacks was small with only 796 patients available for analysis, limiting the power to examine an association between GSTM1 and kidney failure in this population. The strength of this study is the large sample size of whites, which included 46,187 patients, allowing for large subgroup analyses in older patients, those with baseline eGFR < 60 ml/min/1.73 m², and those with longer duration of follow-up. Finally, other factors such as SNPs and haplotypes in the GSTM cluster, CNV of other GST genes, such as GSTT1, and environmental factors such as diet were not examined. A recent study identified substantial differences of SNP-based haplotype in GSTM cluster among groups of individuals with different GSTM1 CNV, suggesting a role for local genetic context in the deletion frequency thus the deletion-related effects on various disease (Khrunin et al., 2016). Studies that examined the combined effect of GSTM1 and GSTT1 deletion showed increased risk for ESRD in patients with null genotype of both GSTM1 and GSTT1 (Agrawal et al., 2007; Suvakov et al., 2013; Nomani et al., 2016). GSTM1 deletion is known to be protective for lung cancer among individuals with a high intake of cruciferous vegetable and deleterious among those with low intake of cruciferous vegetable (London et al., 2000; Zhao et al., 2001; Wang et al., 2004; Brennan et al., 2005; Carpenter et al., 2009). We were not able to analyze the effect modification between GSTM1 copy number and cruciferous vegetable intake.

In conclusion, our study does not support an association between loss of GSTM1 and increased risk of kidney failure in whites. Additional research is needed to confirm whether loss of GSTM1 increases risk of kidney failure in certain subgroups such as blacks. The interaction with diet, and genetic context will also be examined in future.

The individual EHR and genetics datasets (even de-identified) used and/or analyzed during the current study are not publicly available. The availability restriction is a condition of the ethical approval and Geisinger policy and terms with RGC. Collaboration requests and data use agreements with Geisinger are necessary to obtain access to the deidentified EHR data.

All participants provided their informed written consent, and the study was approved by the Geisinger Institutional Review Board.

AT, AC, and ML designed the study; DH extracted data; YZ and WZ performed the analysis and drafted the manuscript. MW, ML, and AC did critical review on the paper. RGC supported the study and reviewed the manuscript. All authors approved the final version of the manuscript.

The Regeneron Genetics Center funded the collection of genomic study samples, the generation of genotype data, and genotype imputation. Geisinger provided funding for clinical data extraction and genetic association analysis. AC is supported by the National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases grant K23 DK106515-01

Regeneron Genetics Center authors were employed by Regeneron Pharmaceuticals. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.