African ancestry-derived APOL1 risk genotypes show proximal epigenetic associations

Abstract

Apolipoprotein L1 (APOL1) coding variants, termed G1 and G2, are established genetic risk factors for a growing spectrum of diseases, including kidney disease, in individuals of African ancestry. Evidence suggests that the risk variants, which show a recessive mode of inheritance, lead to toxic gain-of-function changes of the APOL1 protein. Disease occurrence and presentation vary, likely due to modifiers or second hits. To understand the role of the epigenetic landscape in relation to APOL1 risk variants, we performed methylation quantitative trait locus (meQTL) analysis to identify differentially methylated CpGs influenced by APOL1 risk variants in 611 African American individuals. We identified five CpGs that were significantly associated with APOL1 risk alleles in discovery and replication studies, and one CpG-APOL1 association was independent of other genomic variants. Our study highlights proximal DNA methylation alterations that may help explain the variable disease risk and clinical manifestation of APOL1 variants.

Main text

Two APOL1 risk variants (G1 and G2) have been identified as genetic risk factors for a wide spectrum of diseases including chronic kidney disease (CKD), hypertension-attributed kidney failure (odds ratio [OR] 7), HIV-1 associated nephropathy (OR 29–89), focal segmental glomerulosclerosis (FSGS) (OR 17), and more recently, COVID-19-associated nephropathy and pregnancy-induced hypertension (preeclampsia) in individuals of African descent [1, 2]. Our studies in African American postmenopausal women additionally identified associations of APOL1 risk variants with heart failure [3]. These diseases contribute to substantial morbidity and mortality as well as poor maternal and fetal outcomes. APOL1 is a gene involved in innate immunity, and G1/G2 variants are believed to be under recent selective pressure driven by resistance to African trypanosomiasis [1]. The prevalence of APOL1 high risk genotypes comprising any two risk alleles is 13% in African American individuals and ~ 1% in Hispanics/Latinos of African descent [4, 5]. The mechanisms relating APOL1 to disease and possible treatment are under investigation [4, 6]. Some new therapies in this area are based on the hypothesis that APOL1 G1/G2 variants are related to gain-of-function protein toxicity [7] and exploit the observation that the APOL1 protein is not required for life since middle-aged humans carrying two null alleles are healthy, and most non-human primates and all non-primate mammals lack the APOL1 gene [8, 9].

The relationship between APOL1 risk variants and the epigenome has been understudied and could provide insights into targets to prevent or treat APOL1 associated diseases. Differences in DNA methylation (DNAm) can be driven by genomic sequence variants and can modulate local gene expression. Studies have shown that many functionally relevant changes in DNAm occur in regulatory elements such as enhancers [10]. Prior studies identified African ancestry-related differential gene expression of APOL1 in hepatocytes of African American individuals [11]. To understand the relationship between APOL1 risk variants and the epigenome, we performed DNA methylation quantitative trait locus (meQTL) analysis in African Americans. The number of G1 and G2 risk alleles were combined for analyses (see Methods).

Whole blood DNAm was measured using the Illumina 450 K array in two studies sampled from the Women’s Health Initiative (WHI), a cohort study of postmenopausal women aged 50–79 years. Analysis was restricted to African American participants (WHI-BAA23, n = 410 discovery; WHI-EMPC, n = 201 replication) who also had APOL1 risk genotypes. APOL1 G1 and G2 risk were computed as 0, 1 or 2 copies of the risk alleles. We used linear models to test the association of DNA methylation with APOL1 allele copies in models adjusted for age, recruitment center, smoking status (current, past, never) and smoking pack-years, blood cell composition, batch effects and the first 10 principal components derived from genome-wide genotypes. For each associated CpG we compared the consistency of findings between cohorts (WHI-BAA23 and WHI-EMPC studies).

The average age of participants was 61.8 and 61.0 years and the average estimated glomerular filtration rate (eGFR) was 92.3 and 93.6 ml/min/1.73m² for participants of WHI-BAA23 and WHI-EMPC, respectively. Chronic kidney disease (defined by an eGFR< 60 ml/min/1.73m²) was present in 6% of participants. Two copies of APOL1 risk alleles were present in 13% of participants (Table S1). We identified five CpGs that were significantly associated with the number of APOL1 risk alleles after Bonferroni correction, all of which replicated, showing consistent association and direction of effect (Table S2, Fig. S1). Figure 1 shows the location of identified meQTL CpGs and an example at cg10543947, which is in a candidate regulatory region (Fig. 1a) and, like other significant meQTL CpGs, shows consistent direction of effect in both cohorts (Fig. 1b, Table S2).

Fig. 1

The proximal epigenetic landscape of APOL1 risk alleles: A Shown is the chromosome region with genes (top), and histone marks (middle) from Encyclopedia of DNA Elements (ENCODE) data. B Shown are meQTL CpGs for APOL1 G1/G2 number of risk alleles (0, 1 or 2). An example of a DNAm position associated with APOL1 genotypes is highlighted at cg10543947 showing consistent direction of effect in two distinct studies (WHI-EMPC and WHI-BAA23). Replicated significant CpGs are shown in red

Table S3 shows additional single nucleotide polymorphisms (SNPs) that were significantly associated with four out of the five CpGs in models adjusting for the number of APOL1 risk alleles. Specifically, we identified additional significant meQTLs for 4 of the 5 CpGs (cg10543947, cg15716373, cg16121206, and cg21855316). The additional identified meQTLs for these 4 CpGs were also shown to attenuate APOL1 association with DNA methylation (Table S3). For cg21092464 no additional significant meQTLs were identified, even after rerunning analysis on SNPs within 5 kb of this CpG. In short, 4 out of the 5 CpGs show association with SNPs that were also found to attenuate APOL1 association with DNA methylation in these CpGs, while 1 out of the 5 CpGs did not. eFORGE analyses shows that 3 out of the 5 CpGs (cg10543947, cg15716373, and cg16121206) are in DNase I hotspots in kidney and other tissues, while CpG cg21855316 is in a blood enhancer, and cg21092464 is in a B cell and liver enhancer (Tables S4, S5 and S6). Taken together, these results indicate an meQTL effect of APOL1 risk alleles that is independent of additional variants in the region, and put APOL1 associations in context with additional genotype data.

There is little understanding of why only a subset of carriers of two APOL1 risk variants develop disease. Prior studies suggest that APOL1 G1 and G2 are gain-of-function variants [4]. In our study, we identify epigenetic differences at this locus, which may occur prior to chronic kidney disease, given that our participants had an average normal eGFR. Epigenetic regulation may contribute to the heterogeneity in disease risk and manifestation among individuals of African descent who carry APOL1 risk alleles. The relationship between these epigenetic differences and gene regulation will need to be further explored, as well as their potential contribution to kidney toxicity and hypertensive conditions.

Our findings highlight epigenetic differences associated with disease risk variants, including for variants derived from African ancestry that are common in African Americans. This is an important consideration for studies focusing on disease-associated ancestry-specific variants [12, 13], given that epigenetic marks such as DNAm may be targets to modify disease risk. Novel therapies tested for APOL1-associated kidney disease include small molecule compounds that bind to the APOL1 protein to inhibit APOL1 channel function [14]. Antisense oligonucleotides that block APOL1 transcription in Apol1-transgenic mice models have been shown in two independent studies to ameliorate proteinuria and reduce kidney dysfunction [15, 16]. However, at least 6 other mechanisms for APOL1 related disease have been proposed including mitochondrial and endolysosomal dysfunction, and inflammasome pathways, which may provide targets for therapeutic intervention [17]. Our study supports research to better understand differences in the epigenetic landscape associated with APOL1 risk variants in individuals of African descent.

Methods

Population

WHI is a study of postmenopausal women (aged 50–79 years), comprising 161,808 women recruited from 40 U.S. clinical centers to participate in an observational study or in clinical trials during 1993–1998 [18,19,20,21]. The study has comprehensive information on risk factors including lifestyle, medical history, medication, physical measures and biomarkers obtained at a baseline clinical examination and follow-up. All participants have provided informed consent for genetic research. Two DNA methylation studies of WHI African American participants were included.

DNA methylation profiling

Blood DNAm was assayed using the Illumina 450 K array in 3927 WHI participants from two studies: the Broad Agency Award 23 (WHI-BAA23), a case-control study of cardiovascular disease, and the Epigenetic Mechanisms of PM-Mediated Cardiovascular Disease Risk (WHI-EMPC), a stratified, random sample of participants examined between 1993 and 2001. To correct the beta value distributions of the two types of probes on the 450 K array, β-values were normalized using the beta-mixture quantile (BMIQ) normalization method, [22] extreme outliers were removed, and ComBat was used to adjust for technical artifacts across batches [23]. Cell proportions were estimated using the Houseman method [24].

APOL1 and genome-wide genotyping

APOL1 G1 (rs73885319 & rs60910145, two amino acid substitutions: S342G and I384M) and G2 (rs71785313, two-amino acid deletion: del388N389Y) (Table S2) were directly genotyped using the Taqman assay (Thermo Fisher Scientific). APOL1 G1 and G2 variants were coded as 0, 1 or 2 copies of the risk alleles. Genome-wide genotypes were available in approximately 8500 WHI African American women who were genotyped using the Affymetrix 6.0 array. Genetic data was imputed using TOPMed freeze 8 data. We included SNPs available in the 1000 Genomes Project AFR data, and trimmed SNPs based on linkage disequilibrium (r2 = 0.8) using SNPclip from LDlink [25]. Genome-wide genotypes were used to estimate principal components among unrelated individuals using standard methods [26].

Statistical analysis and functional annotation

We tested the association of methylation at CpGs within 2 MB of the APOL1 risk variants using linear models adjusted for age, recruitment center, smoking status (current, past, never), pack-years, cell composition and 10 principal components derived from genome-wide genotypes, and performed robust standard error calculations via the ‘sandwich’ package [27]. Statistical significance was considered after Bonferroni correction for the number of DNA methylation CpGs tested within the region (n = 972 CpGs). For each identified meQTL CpG, we compared the consistency of findings between WHI-BAA23 discovery and WHI-EMPC replication studies. We provided functional annotation using the UCSC genome browser and data from the Encyclopedia of DNA Elements (ENCODE) [28], and eFORGE (https://eforge.altiusinstitute.org/) [29, 30]. We also tested for association of SNPs with the 5 CpGs using the same covariates listed above, in models adjusted for the number of APOL1 risk alleles with a significance threshold based on the number of SNPs tested.