Lack of Association of the APOL1 G3 Haplotype in African Americans with ESRD

Abstract

Apolipoprotein L1 gene (APOL1) G1 and G2 variants are strongly associated with progressive nondiabetic nephropathy in populations with recent African ancestry. Selection for these variants occurred as a result of protection from human African trypanosomiasis (HAT). Resequencing of this region in 10 genetically and geographically distinct African populations residing in HAT endemic regions identified eight single nucleotide polymorphisms (SNPs) in strong linkage disequilibrium and comprising a novel G3 haplotype. To determine whether the APOL1 G3 haplotype was associated with nephropathy, G1, G2, and G3 SNPs and 70 ancestry informative markers spanning the genome were genotyped in 937 African Americans with nondiabetic ESRD, 965 African Americans with type 2 diabetes–associated ESRD, and 1029 non-nephropathy controls. In analyses adjusting for age, sex, APOL1 G1/G2 risk (recessive), and global African ancestry, the G3 haplotype was not significantly associated with ESRD (P=0.05 for nondiabetic ESRD, P=0.57 for diabetes-associated ESRD, and P=0.27 for all-cause ESRD). We conclude that variation in APOL1 G3 makes a nominal, if any, contribution to ESRD in African Americans; G1 and G2 variants explain the vast majority of nondiabetic nephropathy susceptibility.

Two coding variants in exon 6 of the apolipoprotein L1 gene (APOL1), termed G1 and G2, are strongly associated with a spectrum of nondiabetic forms of ESRD in populations with recent African ancestry. Kidney disorders in the APOL1 spectrum include FSGS, focal global glomerulosclerosis with interstitial and vascular changes (FGGS or hypertension-attributed nephropathy), HIV-associated nephropathy, progressive lupus nephritis, and sickle cell nephropathy.^1–8 These nephropathy variants are more commonly found in African Americans than in other populations likely due to natural selection. Possession of one nephropathy variant in APOL1 protects from Trypanosoma brucei rhodesiense, a cause of African sleeping sickness in sub-Saharan Africa. Inheriting two APOL1 variants is strongly associated with susceptibility to ESRD.

Ko et al.⁹ resequenced a 1.4-kb region encompassing exon 6 of APOL1 from 10 geographically and genetically isolated African populations (n=187 individuals). These populations reside in regions exposed to T. b. rhodesiense and Trypanosoma brucei gambiense, causes of human African trypanosomiasis. This segment encodes three APOL1 functional domains identified as pore-forming, membrane-addressing, and serum-resistance–associated interacting domains, necessary for trypanosomal killing. Frequencies of the G2 variant were modest in all groups (3%–8%), whereas G1 was only common in Yoruba (38%). A novel G3 haplotype composed of eight single nucleotide polymorphisms (SNPs) in strong linkage disequilibrium (LD) was identified. G3 includes a nonsynonymous change (Lys255Arg, rs136176) at the membrane-addressing domain. The G3 haplotype was not detected in Yoruba. Evaluation of long-range LD revealed evidence for recent selection of G3 in Cameroon Fulani. The authors postulated that, in addition to G1/G2, other functional variants in APOL1 might exist and may be associated with resistance to human African trypanosomiasis and/or risk for kidney disease.

The APOL1 G3 haplotype has not been evaluated for association with nephropathy. As such, genetic association analyses were performed for G3 SNPs with common forms of ESRD in African Americans residing in the southeastern United States.

Table 1 contains demographic characteristics of patients with ESRD and non-nephropathy controls. Patients with type 2 diabetes (T2D)–associated ESRD (T2D-ESRD) and nondiabetic ESRD initiated RRT at mean ages of 58.0±10.9 and 48.7±15.5 years, respectively. Controls were recruited at 50.0±11.9 years. Overall mean African ancestry was 79.9%±11.4%, 80.0%±11.6%, and 78.2%±10.9% in patients with T2D-ESRD, patients with nondiabetic ESRD, and controls, respectively.

Table 1.

Clinical characteristics of African-American study samples

Characteristic	Patients with T2D-ESRD	Nondiabetic Patients with ESRD	Non-Nephropathy Controls
Participants (n)	965	937	1029
Age (yr)	59.2±16.4	54.3±14.6	50.0±11.9
Women (%)	61.20%	45.0%	57.30%
Body mass index (kg/m2)	29.7±7.0	26.9±6.8	30.0±7.0
Age at T2D (yr)	41.6±12.4	—	—
Age at ESRD (yr)	58.0±10.9	48.7±15.5	—
African ancestry (%)a	79.9±11.4	80.0±11.6	78.2±10.9

Results are expressed as the mean±SD unless otherwise indicated.

^aAfrican ancestry calculated from 70 ancestry informative markers.

Table 2 summarizes the association results for the G3 haplotype. In the unadjusted analysis, G3 SNPs demonstrated evidence of association with nondiabetic ESRD (P=1.2×10⁻⁴) and all-cause ESRD (combined nondiabetic patients with ESRD and patients with T2D-ESRD, P=0.001) with a trend toward association with T2D-ESRD (P=0.09). In the model adjusted for age, sex, African ancestry, and APOL1 G1/G2 compound risk, evidence of association was abolished (P=0.05, P=0.27, and P=0.57, respectively) when applying a Bonferroni correction for the number of tests performed (n=2 models with significance at P<0.03). Notably, best-guess genotypes, which form the basis of the haplotype analysis, can introduce a modest bias potential. In addition, the low frequency of two observed G3 haplotypes could affect the power to detect an association, which would be further diminished in omnibus tests of association. To complement this approach, single SNP association analysis using dosage genotypes (Table 3) revealed impressive evidence of association that was consistently markedly attenuated in the fully adjusted model. Only one G3 SNP (rs136170) remained significantly associated with nondiabetic ESRD and all-cause ESRD (P=3.9×10⁻⁴ and P=0.02, respectively).

Table 2.

APOL1 G3 haplotype association results for ESRD in African Americans.

Patient Group and Haplotypea	Frequency				Haplotypeb		Omnibusc
	Overall	Cases	Controls	Odds Ratio	STATd	P Value	STATd	P Value
T2D-ESRD versus control
Unadjustede
TGGAAAAA	0.94	0.96	0.94	1.31	4.17	0.041	4.89	0.087
CAAGCGGG	0.039	0.034	0.045	0.74	3.41	0.065
TGGAAAAG	0.011	0.0094	0.013	0.7	1.4	0.24
Adjustedf
TGGAAAAA	0.94	0.96	0.94	1.08	0.25	0.62	1.12	0.57
CAAGCGGG	0.039	0.032	0.049	0.81	1.09	0.3
TGGAAAAG	0.011	0.0099	0.013	1.07	0.034	0.85
Nondiabetic ESRD versus control
Unadjustede
TGGAAAAA	0.94	0.97	0.94	1.61	10.3	0.0014	18.1	1.2E-04
CAAGCGGG	0.033	0.019	0.045	0.4	18	2.2E-05
TGGAAAAG	0.014	0.015	0.013	1.12	0.16	0.69
Adjustedf
TGGAAAAA	0.94	0.97	0.94	1.05	0.1	0.75	6.14	0.046
CAAGCGGG	0.033	0.019	0.049	0.65	3.5	0.061
TGGAAAAG	0.014	0.014	0.013	1.74	2.82	0.093	0.093
All-cause ESRDg versus controls
Unadjustede
TGGAAAAA	0.94	0.96	0.94	1.43	9.56	0.002	13.2	0.001
CAAGCGGG	0.033	0.027	0.045	0.59	12.9	0.00032
TGGAAAAG	0.012	0.012	0.013	0.89	0.24	0.63
Adjustedf
TGGAAAAA	0.94	0.96	0.94	1.04	0.07	0.8	2.59	0.27
CAAGCGGG	0.033	0.026	0.049	0.8	1.8	0.18
TGGAAAAG	0.012	0.012	0.013	1.31	0.84	0.36

Patient group n_max values are as follows: 965 patients with T2D-ESRD versus 1029 non-nephropathy controls, 937 patients with nondiabetic ESRD versus 1029 non-nephropathy controls, and 1902 patients with all-cause ESRD versus 1029 non-nephropathy controls.

^ars136170, rs28391521, rs136169, rs28480494, rs136174, rs136175, rs136176, and rs136177.

^bLogistic regression analysis using each haplotype versus all others.

^cLogistic regression analysis to jointly estimate and test all haplotype effects in aggregate.

^dTest statistic (t from Wald test).

^eUnadjusted with no covariates.

^fAdjusted with covariates including age, sex, APOL1 G1/G2 risk, admixture.

^gAll-cause ESRD includes the combined sample of nondiabetic patients with ESRD and patients with T2D-ESRD.

Table 3.

APOL1 G3 single SNP association results for ESRD in African Americans

Patient Group and SNP	Positiona	Minor Allele Frequency			Unadjustedc			Adjustede
		Allelesb	Cases	Controls	β	SE	P Valued	β	SE	P Valued
T2D-ESRD versus controlsf
rs136170	22:36660921	C/T	0.043	0.057	0.33	0.16	0.03	0.24	0.20	0.22
rs28391521	22:36661061	A/G	0.034	0.046	0.32	0.16	0.05	0.24	0.21	0.24
rs136169	22:36661149	A/G	0.034	0.046	0.33	0.16	0.05	0.25	0.21	0.23
rs28480494	22:36661152	G/A	0.034	0.046	0.33	0.16	0.05	0.25	0.21	0.23
rs136174	22:36661536	C/A	0.041	0.052	0.29	0.16	0.07	0.17	0.20	0.40
rs136175	22:36661566	G/A	0.041	0.052	0.29	0.16	0.07	0.17	0.20	0.40
rs136176	22:36661646	G/A	0.034	0.046	0.34	0.17	0.04	0.26	0.21	0.21
rs136177	22:36661842	G/A	0.052	0.066	0.30	0.15	0.04	0.12	0.18	0.52
Nondiabetic ESRD versus controls
rs136170	22:36660921	C/T	0.018	0.057	1.09	0.18	4.8E-10	0.74	0.21	3.9E-04
rs28391521	22:36661061	A/G	0.021	0.046	0.76	0.18	2.8E-05	0.36	0.21	0.09
rs136169	22:36661149	A/G	0.023	0.046	0.70	0.18	1.1E-04	0.27	0.21	0.20
rs28480494	22:36661152	G/A	0.025	0.046	0.63	0.18	4.1E-04	0.20	0.21	0.32
rs136174	22:36661536	C/A	0.028	0.052	0.65	0.17	1.3E-04	0.18	0.20	0.37
rs136175	22:36661566	G/A	0.029	0.052	0.64	0.17	1.6E-04	0.18	0.20	0.36
rs136176	22:36661646	G/A	0.024	0.046	0.64	0.18	3.0E-04	0.20	0.20	0.32
rs136177	22:36661842	G/A	0.041	0.066	0.54	0.15	3.2E-04	0.08	0.18	0.63
All-cause ESRD versus controls
rs136170	22:36660921	C/T	0.032	0.057	0.74	0.15	8.2E-07	0.41	0.17	0.02
rs28391521	22:36661061	A/G	0.028	0.046	0.56	0.15	2.7E-04	0.24	0.17	0.16
rs136169	22:36661149	A/G	0.029	0.046	0.54	0.15	4.9E-04	0.19	0.17	0.27
rs28480494	22:36661152	G/A	0.030	0.046	0.50	0.15	9.7E-04	0.16	0.17	0.35
rs136174	22:36661536	C/A	0.035	0.052	0.50	0.15	7.9E-04	0.12	0.17	0.48
rs136175	22:36661566	G/A	0.035	0.052	0.49	0.15	8.5E-04	0.13	0.17	0.45
rs136176	22:36661646	G/A	0.029	0.046	0.52	0.15	7.1E-04	0.19	0.17	0.27
rs136177	22:36661842	G/A	0.047	0.066	0.45	0.13	6.6E-04	0.08	0.15	0.61

Patient group n_max values are as follows: 965 patients with T2D-ESRD versus 1029 non-nephropathy controls, 937 patients with nondiabetic ESRD versus 1029 non-nephropathy controls, and 1902 patients with all-cause ESRD versus 1029 non-nephropathy controls.

^aPosition in hg19.

^bMajor/minor alleles.

^cNo covariates.

^dCovariates include age, sex, APOL1 G1/G2 risk, and admixture.

^eAdditive P value.

^fImputation quality, as assessed by the information content from IMPUTE2, was rs136170 (0.85), rs28391521 (0.94), rs136169 (0.94), rs28480494 (0.94), rs136174 (0.85), rs136175 (0.85), rs136176 (0.93), and rs136177 (0.82).

This is the first report to assess genetic association between SNPs comprising the APOL1 G3 haplotype and ESRD in a large sample of African Americans. Participants with ESRD were recruited from dialysis facilities in the southeastern United States and had nephropathy clinically ascribed to nondiabetic forms of glomerulosclerosis (e.g., FGGS or FSGS) or diabetic kidney disease. After adjusting for age, sex, African ancestry, and APOL1 G1/G2 risk, a lack of evidence of association was detected between the APOL1 G3 haplotype and ESRD. A single SNP rs136170 was associated with nondiabetic ESRD, with nominal significance maintained in a combined all-cause ESRD analysis. None of the other SNPs in G3 were associated with ESRD in the adjusted model. In support of these findings, single SNP association of APOL1 G3 SNPs in patients with SLE with and without ESRD² revealed a lack of association (P>0.54). We therefore conclude that the G3 haplotype, potentially protective from trypanosomal infection, does not meaningfully contribute to the risk of ESRD in African Americans residing in the southeastern United States.

There are many reasons to suspect that variants on 22q13, beyond APOL1 G1/G2, contribute to nephropathy. The initial nephropathy association between the E1 haplotype in the nonmuscle myosin heavy chain 9 (MYH9) gene in African Americans, adjacent to APOL1, proved to relate to LD between MYH9 E1 and the coding APOL1 G1/G2 variants. Among African Americans with the MYH9 E1 haplotype, 89% had APOL1 G1 and 76% APOL1 G2. However, MYH9 SNPs were strongly associated with nephropathy susceptibility in European- and Asian-ancestry populations; these groups lack APOL1 G1/G2 variants.^10–13 Additional coding or regulatory variants to account for MYH9 association with ESRD have not yet been detected. As such, it remains likely that other variants in the APOL1-APOL6 region, likely in LD, are associated with nephropathy susceptibility. In addition, it is apparent that the majority of participants inheriting two copies of APOL1 G1 and/or G2 variants will not develop nephropathy. Additional modulating factors or second hits are likely present.^14–16 However, results in this report cast doubt on a major role for the APOL1 G3 haplotype on risk for ESRD in African Americans. Genovese et al.¹⁷ hypothesized that the overwhelming majority of APOL1-mediated risk for nondiabetic nephropathy in African Americans is attributable to G1/G2 alone.

We analyzed SNPs comprising the APOL1 G3 haplotype in a large sample of African-American patients with common complex forms of nondiabetic and diabetic ESRD, as well as in non-nephropathy controls. Lack of association between the APOL1 G3 haplotype and ESRD was observed although nominal evidence of association was detected between a single intronic SNP (rs136170) and nondiabetic ESRD. It was recently shown that the APOL1 G3 haplotype is but one of several “out-of-Africa” APOL1 nonrisk haplotypes.¹⁸ These haplotypes selectively “returned to Africa” and are found only in certain regions of the continent into which such return migration occurred. These important findings have recast the meaning and significance of the G3 haplotype, especially because it is known to affect the membrane-addressing domain of APOL1 protein and is trypanolytic. Therefore, it was important that we formally evaluate, and in this case exclude, association between the G3 haplotype and common forms of kidney disease in African Americans residing in the southeastern United States. Our results support and extend these findings.

Concise Methods

Self-identified African-American patients with ESRD were recruited from dialysis facilities. Nondiabetic causes of ESRD included hypertension-attributed nephropathy, FSGS, FGGS, HIV-associated nephropathy, unspecified glomerulosclerosis, or unknown cause in the absence of a kidney biopsy, without other nephropathy risk factors. T2D-ESRD was based upon the diagnosis of diabetes after age 25 years without diabetic ketoacidosis or treatment with insulin-alone because initial diagnosis, in the presence of T2D duration >5 years before ESRD onset, diabetic retinopathy, or proteinuria >100 mg/dl, absent other risk factors for nephropathy. Patients with ESRD as a result of urologic disease, surgical nephrectomy, polycystic kidney disease, Alport’s syndrome, IgA nephropathy, or membranous or membranoproliferative GN were excluded.

Self-identified African-American controls lacked T2D or kidney disease (serum creatinine concentration <1.5 mg/dl for men or <1.3 mg/dl for women) and were recruited from the community and Wake Forest School of Medicine internal medicine clinics. Patients and controls were unrelated and were born in North Carolina, South Carolina, Georgia, Tennessee, or Virginia. The Wake Forest School of Medicine Institutional Review Board approved recruitment and sample collection procedures and all participants provided written informed consent.

Genotyping for 965 patients with T2D and ESRD and 1029 non-nephropathy controls was performed on the Affymetrix 6.0 array as previously described.¹⁹ For quality control, samples reflecting duplicates, a call rate <0.95, sex mismatch, or population outliers were excluded. Genotyped SNPs with a minor allele frequency <0.01, a call rate <0.95 or a Hardy–Weinberg P value <0.001 were excluded. By comparing with the expected allele frequency derived from the 1000 Genomes Project, observing an average of 0.79±0.11 African ancestry proportion in our samples, markers with allele frequency differences >0.3 were excluded. A total of 826,788 genotyped SNPs passing quality control and present in the 1000 Genomes Project were used for imputation. Imputation was performed using all populations from the integrated reference panel (March 2012). SHAPEIT2 was used to prephase the haplotypes of our samples and IMPUTE2²⁰ was utilized to impute missing genotypes. Imputed markers with a minor allele frequency <0.01 or with a SNPTEST score <0.5 or a confidence score <0.90 were excluded from association analyses.

We resequenced 1.3 kb of APOL1 in 937 nondiabetic patients with ESRD, which captured the eight SNPs defining G3: rs1360170, rs28391521, rs136169, rs28480494, rs136174, rs136175, rs136176, and rs136177. Sequencing was performed in 937 African-American nondiabetic patients with ESRD. PCR primers were designed de novo to specifically amplify a region of APOL1 that shares high sequence similarity with APOL2 (Supplemental Table 1). DNA sequencing was performed using Big Dye Ready Reaction Mix (Applied Biosystems, Foster City, CA) on an ABI3730xl sequencer (Applied Biosystems) and visualized using Sequencher v4.10 (GeneCodes Corporation, Ann Arbor, MI).

For all samples, 70 ancestry informative markers were directly genotyped on the Sequenom platform (San Diego, CA) to control for population stratification. In addition, two SNPs in the APOL1 G1 variant (rs73885319 and rs60910145) and the insertion/deletion for the G2 variant (rs71785313) were directly genotyped.

Haplotypes were inferred using the standard expectation–maximization algorithm and haplotype-specific and omnibus association analyses were evaluated using logistic regression as implemented in PLINK.²¹ Genotypic associations were evaluated using SNPTEST v2.5β.²² Frequentist association tests were calculated under the additive model using the score method. The percentage of African ancestry, as determined from direct genotyping of 70 ancestry informative markers, was calculated using FRAPPE.²³ APOL1 G1/G2 risk allele status was coded assuming a recessive model of disease risk. The full model adjusted for age, sex, APOL1 risk (recessive), and overall African ancestry. Significance was considered for P<0.05.

Disclosures

None.