Prevalence of Mendelian kidney disease among patients with high-risk APOL1 genotypes undergoing commercial genetic testing in the United States

Designated or self-reported race/ethnicity

The following race and ethnicity categories were present on the intake form are: African American, Ashkenazi Jewish, Caucasian, East Asian, French Canadian/Cajun, Hispanic, Mediterranean, Other, Sephardic Jewish, South Asian and South-East Asian. To achieve a sufficiently large sample size for further analysis, certain race/ethnic categories, namely French Canadian/Cajun, Sephardic Jewish, and Ashkenazi Jewish, were combined with the category labeled as ‘Other’ due to the limited number of individuals.

Sequencing

Genomic DNA was extracted from either the individual’s whole blood or saliva and processed for hybrid capture-based next generation sequencing. Massively parallel sequencing was performed at 150 base-pairs, paired-end reads on a clinical exome backbone. The percentage of coverage was determined at a minimum of 20X per gene on the panel and Sanger sequencing was used to fill in regions of low coverage. The sequencing data was aligned to the GRCh37/hg19 genome assembly, and the variants were called and annotated using a bioinformatics pipeline that applied the GATK framework (11). Single nucleotide variants (SNV), insertions and deletions (indels) as well as copy number variants (CNV) were detected by this assay. Orthogonal methods for confirmation of variants included Sanger sequencing for SNV and indels and quantitative Polymerase Chain Reaction or Multiplex-ligation dependent probe amplification for CNV.

Genetic Ancestry Determination

The population structure and admixture analyses were conducted using the continental ancestry groups from the combined, harmonized databases of the Human Genome Diversity Project (HGDP) and the 1000 Genomes Project (1KG) (12). We included SNVs with a minor allele frequency (MAF) >0.05 and missing in fewer than 5% of individuals within our dataset to determine ancestry. We retained samples for further analysis with an overall genotype rate greater than 90% (i.e., missing fewer than 10% of selected SNVs). We performed principal component analysis (PCA) using SNPWeights version 2 (10) and PLINK v1.9 (13), and estimated global genomic ancestry proportions using Admixture v1.3 (14) focusing on the following populations: African, admixed American, Central South Asia, East Asian, and European.

In additional analyses, we assigned a single ancestry group to each individual patient. Ancestry groups were determined using a clustering method derived from the PCA dimensions. We utilized the first 10 principal components (PCs) and applied the Uniform Manifold Approximation and Projection (UMAP) algorithm, via the umap package in R (15). The UMAP outputs were clustered using the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm, via the dbscan R package (16). Five distinct clusters were identified, corresponding to inferred continental ancestries. These clusters aligned well with the predominant admixture proportions for their respective ancestry groups (e.g., samples clustered in the African ancestry group had majority of global genomic proportions of African genetic population; Supplemental Figure 1). We categorized samples with unknown designated/self-reported race/ethnicity into one of the five distinct ancestry clusters.

Interpretation of Genetic Variants

Variants were analyzed in 343 genes associated with Mendelian forms of kidney disease (see Supplemental Table 1), including the risk alleles associated with APOL1. Detected germline variants were classified using a 5-tier classification system (B - benign, LB - likely benign, VUS - variant of uncertain significance, LP - likely pathogenic, and P - pathogenic) in accordance with the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) guidelines at the time of testing (17). As part of routine clinical care, reports of identified variants and their relevant ACMG/AMP classifications were returned to the referring clinicians and patients at the time of testing.

Identifying patients with molecularly diagnosed Mendelian kidney diseases

Accurate diagnosis of Mendelian forms of kidney disease requires genotype-phenotype correlation based on variant classification information combined with the mode of inheritance for the disorder. We classified individuals as having molecularly diagnosed Mendelian kidney disease based on the presence of P/LP variants, meeting criteria informed by empirical data (17). We considered a test “positive” or consistent with a molecular diagnosis of Mendelian kidney disease if: (1) the individual has a heterozygous or hemizygous P/LP variant in a gene associated with a dominant inheritance pattern; or (2) the patient has two P/LP variants (compound heterozygous or homozygous) in a gene with a recessive inheritance pattern. For genes with both dominant and recessive patterns of inheritance, we considered each variant’s specific phenotype and the associated pattern of inheritance was considered. For HBB, only results consistent with autosomal recessive sickle cell anemia were included as “positive.” We considered as carriers those individuals harboring heterozygous P/LP variants in genes associated with an autosomal recessive (AR) inheritance were considered carriers. The remaining samples were considered negative. In general, a variant classified as pathogenic (P) or likely pathogenic (LP) have met criteria informed by empirical data such that healthcare providers can use the results of molecular testing in clinical decision making (17).

APOL1 genotype status definition

We assessed two APOL1 high-risk alleles: (1) G1 (NM_003661.4:c.[1024A>G;1152T>G]) defined by the presence of two missense variants p.S342G (rs73885319) and p.I384M (rs60910145) that are nearly always in linkage disequilibrium; (2) G2 (NM_003661.4:c.1164_1169del) defined by the presence of an in-frame deletion of two amino acid residues at codons 388 and 389 respectively (rs71785313). We defined APOL1 high-risk genotypes by the presence of two high-risk alleles (G1/G1, G1/G2 or G2/G2). We defined single risk APOL1 allele carriers by the presence of only one risk allele (G1/G0 or G2/G0). The lack of APOL1 risk alleles was denoted as (G0/G0). We deliberately considered single risk APOL1 allele carriers because, by definition, these patients have recent African ancestry, in contrast to patients with G0/G0 APOL1 genotypes.

Statistical analysis

Continuous variables were summarized by means, medians and interquartile ranges, where appropriate. Comparisons of continuous variables across groups were made via Wilcoxon’s rank sum test. We described categorical variables using proportions and compared groups using Fisher’s exact test. P-values <0.05 were considered statistically significant. Analyses were conducted in R version 4.1.2.

The primary goal of our analysis was to explore the relationship between APOL1 genotypes (exposure) and Mendelian kidney disease (outcome). However, by definition, patients with APOL1 risk alleles have recent African ancestry, and African ancestry may confound the relationship between APOL1 genotype and Mendelian kidney disease. As such, we conducted a matched case-control study, matching patient with high-risk APOL1 genotypes to patients who were either single-risk APOL1 allele carriers or G0/G0 in a 1:1 ratio, matching on age, sex, genetic ancestry proportions from the five continental populations in our study, and the first four principal components from PCA. Matching was performed via the nearest neighbor method and generalized linear models for distance estimation using the MatchIt package in R (18). We compared the prevalence of molecularly diagnosed Mendelian kidney in the matched cohorts.

Prevalence of Mendelian kidney disease in the overall cohort

Of 15,181 patients, 3119 had a molecular diagnosis of Mendelian kidney disease (20.5%; Table 1), defined by the presence of pathogenic or likely pathogenic variants (P/LP) consistent with the inheritance pattern of the specific disease. 5049 (33.3%) patients were classified as carriers of Mendelian kidney disease, and 7013 (46.2%) were negative for P/LP variants in any of the 343 genes evaluated (Supplemental Table 1). Patients with a molecular diagnosis of Mendelian kidney disease were younger on average, at the time of testing, compared to carriers or patients with negative test results (40 v. 48 years, p<0.0005; Table 1).

We grouped Mendelian forms of kidney disease into five non-exclusive categories: (1) cystic and tubulointerstitial diseases, (2) glomerular diseases, (3) tubulopathies and other tubular diseases, (4) congenital anomalies of the kidney and urinary tract (CAKUT) and other structural disorders, and (5) complement-related kidney diseases (see Supplemental Table 1). Of patients with a molecular diagnosis of Mendelian kidney disease, 42.1% had cystic and tubulointerstitial disorders (dominated by variants in PKD1/2), 34.4% had glomerular disorders (dominated by variants in COL4A3/4/5), 13.9% tubulopathies and other tubular disorders, 8.0% had CAKUT and other structural disorders and 1.5% had complement-related diseases. In total (including molecular diagnosis positive and carriers), 4820 rare P or LP variants were identified across 286 of the 343 genes evaluated. Nearly half of these variants (46.8%, n=2,258) were not present in ClinVar or gnomAD databases at the time of the variants being assessed. The reported variants were categorized as frameshift (28%, n=1352), missense (24.3%, n=1170), nonsense (21.4%, n=1032), canonical splice site (14.5%, n=700), non-frameshift insertions/deletions (5%, n=240), copy number (CNV, 3.4%, n=166), or other (∼3.2%, n=160). The majority of patients with a molecular diagnosis of Mendelian kidney diseases (71.5%; n=2334) had variants in one (or more) of 10 genes: PKD1/2, COL4A3/4/5, HNF1B, SLC12A3, UMOD, GLA, and SLC7A9 (Supplemental Table 5).

Prevalence of APOL1 risk alleles

Of 15,181 individuals, 1035 (6.8%) had high-risk APOL1 genotypes (Table 2). Homozygous haplotypes G1/G1 and G2/G2 accounted for 43.7% (n = 452) and 12.9% (n = 134) of patients, respectively; 449 patients with high-risk APOL1 genotypes had one G1 allele and one G2 allele (43.4%; G1/G2). 1687 (11.1%) patients were single-risk APOL1 allele carriers, of which 1064 with G1 (63.1%), and 623 with G2 (36.9%); 12459 patients were G0/G0 (82.1%).

The large majority of patients with high-risk APOL1 genotypes had designated African American race (815/1035; 78.9%). Similarly, single-risk allele carriers had a high proportion of designated African American race (1158/1687; 68.6%), in contrast to individuals with G0/G0 genotypes, who had a low proportion of designated African American race (987/12459; 7.9%). We chose to deliberately study patients with single-risk APOL1 alleles, despite being low-risk, because, by definition, these have recent African ancestry, and as such make a relevant comparator group to patients with high-risk APOL1 genotypes. Though ∼23% of patients in our cohort had unknown designated race/ethnicity information, we were able to recategorize these patients according to their genomic ancestry clusters. Among patients with high-risk APOL1 genotypes and unknown designated race/ethnicity, the vast majority (94%) were classified as having recent African ancestry (Supplemental Figure 2 and Supplemental Table 6). Finally, 25.6% of patients in the genetic African ancestry cluster had high-risk APOL1 genotypes (n= 944/3678; Table 2).

Prevalence of Mendelian kidney disease stratified by APOL1 genotype and recent African ancestry

Of 1035 patients with high-risk APOL1 genotypes, 95 (9.2%) had a concomitant molecular diagnosis of Mendelian kidney disease (Table 2), a smaller fraction compared to single-risk APOL1 allele carriers (14.4%, p<0.0001), and patients with G0/G0 (22.3%, p<0.00001). The most common forms of Mendelian kidney disease among patients with high-risk APOL1 alleles were: Alport syndrome/thin basement membrane disease (3.7%), autosomal dominant polycystic kidney disease (ADPKD; 2.0%), and HNF1B-associated disease (0.5%). The full list of molecularly diagnosed Mendelian kidney disease are presented in Supplemental Table 5.

The overall prevalence of molecularly diagnosed Mendelian kidney disease was significantly lower among patients with recent genetic African ancestry compared to those with recent European ancestry (13.6% v. 23.8%, p<0.00001). Even among patients with G0/G0 genotypes, those with recent genetic African ancestry had a lower prevalence of molecular diagnosed Mendelian kidney disease compared to patients with European ancestry (16.6% v. 23.8%, p<0.00001). To evaluate the possibility that recent African ancestry might confound the association between APOL1 risk alleles and molecularly diagnosed Mendelian kidney disease, we matched patients with high-risk APOL1 genotypes to single-risk allele carriers and patients with G0/G0, in a 1:1 ratio, based on age, sex, the genomic proportions of continental populations, and the first four principal components (Supplemental Table 7). We found a significant lower prevalence of molecularly diagnosed Mendelian kidney disease among patients with high-risk APOL1 risk alleles (9.4%; n=92) compared to matched single-risk allele carriers (13.4%; n=128, p=0.006) and matched G0/G0 patients (16.2%; n=156, p=0.000009). We did not observe statistically significant difference in the prevalence of molecularly diagnosed Mendelian kidney disease between matched single-risk allele carriers and matched G0/G0 patients (p=0.09). These finding suggest that the lower prevalence of molecularly diagnosed Mendelian kidney disease among patients with high-risk APOL1 genotypes is not driven by differences in the proportions of patients with recent African ancestry across the APOL1 genotypes.