Apolipoprotein L1 gene variants associate with hypertension-attributed nephropathy and the rate of kidney function decline in African Americans

Abstract

Despite intensive anti-hypertensive therapy there was a high incidence of renal end-points in participants of the African American Study of Kidney Disease and Hypertension (AASK) cohort. To better understand this, coding variants in the apolipoprotein L1 (APOL1) and the non-muscle myosin heavy chain 9 (MYH9) genes were evaluated for an association with hypertension-attributed nephropathy and clinical outcomes in a case-control study. Clinical data and DNA were available for 675 AASK participant cases and 618 African American non-nephropathy control individuals. APOL1 G1 and G2, and MYH9 E1 variants along with 44 ancestry informative markers were genotyped with allele frequency differences between cases and controls analyzed by logistic regression multivariable models adjusting for ancestry, age, and gender. In recessive models, APOL1 risk variants were significantly associated with kidney disease in all cases compared to controls with an odds ratio of 2.57. In AASK cases with more advanced disease, such as a baseline urine protein to creatinine ratio over 0.6 g/g or a serum creatinine over 3 mg/dL during follow-up, the association was strengthened with odds ratios of 6.29 and 4.61, respectively. APOL1 risk variants were consistently associated with renal disease progression across medication classes and blood pressure targets. Thus, kidney disease in AASK participants was strongly associated with APOL1 renal risk variants.

Introduction

African Americans are more likely to develop progressive kidney disease compared to European Americans (1) and this risk extends across the leading causes of end-stage kidney disease (ESKD): diabetes, hypertension-attributed kidney disease, and glomerulonephritis (2). In particular, for hypertension-attributed nephropathy, there is a three-fold increase in incidence of ESKD for African Americans overall that is magnified at younger ages.

The genetic contribution to the increased prevalence of hypertension-attributed ESKD is now better understood. In the last several years, a region on chromosome 22 associated with renal disease in African Americans was identified using mapping by admixture disequilibrium. This region contains the non-muscle myosin heavy chain type 2 (MYH9) and apolipoprotein-L1 (APOL1) genes. Initial studies that detected the association in this region showed a very strong association of MYH9 with HIV-associated nephropathy and focal segmental glomerulosclerosis (FSGS) (3, 4) as well as an association with non-diabetic forms of ESKD (3, 5).

More recent studies have shown that there is an even stronger association of the APOL1 gene with FSGS and hypertension-attributed ESKD than with MYH9 among African Americans. Controlling for APOL1 risk alleles showed no residual effect of the MYH9 gene, while controlling for MYH9 genotype did not significantly weaken the effect of APOL1 (6, 7). While these data suggested little role for the MYH9 risk alleles, it has subsequently been shown that MYH9 risk alleles are associated with increased risk of non-diabetic chronic kidney disease (CKD) and diabetic ESKD in individuals of European ancestry (8, 9); in these studies there was no risk attributable to APOL1 genotypes, however the frequency of the APOL1 risk alleles was extremely low among European Americans.

To further understand the role of APOL1 and MYH9 genetic variants in hypertensionattributed CKD, we have performed a case control study and evaluated the rate of decline of measured iothalamate glomerular filtration rate (GFR). African American cases were patients with CKD from the African American Study of Kidney Disease and Hypertension (AASK) who at baseline had subnephrotic proteinuria or no proteinuria, measured iothalamate GFR <65 ml/min/1.783m² and clinically diagnosed hypertensive nephropathy (10, 11). Cases were compared to African American controls with normal kidney function as determined by history and serum creatinine measurement, some with mild or moderate hypertension.

Results

The characteristics of the African American cases and African American control subjects are presented Table 1. All of the cases had hypertension and CKD. Of the controls, 410 were questioned about presence of hypertension, and of these 41.7% (171) reportedly were hypertensive based on physician report or use of anti-hypertensive medications. Those with CKD were predominantly male and older compared to controls and had similar body mass index (BMI) and population admixture. Cases had a low baseline GFR and elevated serum creatinine; while controls had a mean serum creatinine concentration of 1.0 mg/dl.

Table 1.

Demographic and clinical information (% or mean (median) ± SD)

	Cases	Controls	P-value
	N=675	N=618
Sex (% female)	40%	57%	<0.0001
Age (years)	54.1(55) ± 10.6	49.1(48) ± 11.7	<0.0001
BMI (kg/m2)	31.0(30) ± 6.6	30.1(29) ± 7.1	0.0103
Serum creatinine (mg/dl)	1.99(1.8) ± 0.7	1.00(0.9) ± 0.26	<0.0001
Mean baseline GFR (ml/min/1.73m2)	47.2(49) ± 13.4	N/A	-
Hypertension (% yes)	100%	41.7% of patients who were surveyed	<0.0001
YRI admixture (%)	0.89(0.93) ± 0.11	0.89(0.92) ± 0.10	0.34

African American cases were from the AASK study; 618 African American controls were recruited at Wake Forest School of Medicine, and 171 of 410 evaluated reported hypertension.

BMI – body mass index; GFR – glomerular filtration rate; YRI – Yoruban; N/A - not available

All reported APOL1 and MYH9 SNPs were in Hardy-Weinberg equilibrium in the AASK participants and in the controls. Although analyses were performed for the additive, dominant, and recessive models, the largest effect sizes (odds ratios (OR)) were observed for the recessive model. This is consistent with the results of other studies (6, 7). The tables show the results for only the recessive model; associations for the additive and dominant models are presented in Supplementary Table 1 and APOL1 genotype frequencies are presented in Supplementary Table 2. The results of the primary single SNP analysis are shown in Table 2. The associations of APOL1 G1 SNPs with CKD were highly significant and associated with approximately a 3.5 fold risk for CKD for G1. The G2 SNP was not significant, but, as in other studies (6, 7) was present at lower allele frequencies than the other SNPs. It should be noted that the more frequent G1 allele tends to mask the effect of G2 as the two variants occur on opposing chromosomes.

Table 2.

Chromosome 22 genotype results

Gene	SNP	C22 position	Risk allele	Risk allele fraction, cases	Risk allele, fraction controls	P, recessive model	OR, recessive model	95% Confidence Interval
APOL1	rs16996616	36661891	A	0.09	0.07	3.18E-01	2.1	(0.49, 9.04)
APOL1G1	rs73885319	36661906	G	0.28	0.21	2.97E-06	3.47	(2.06, 5.85)
APOL1G1	rs60910145	36662034	G	0.28	0.20	2.77E-06	3.54	(2.09, 6.00)
APOL1G2	rs71785313	36662051	C	0.16	0.13	1.46E-01	1.72	(0.83, 3.57)
APOL1G1G2				0.44	0.34	1.39E-08	2.57	(1.85, 3.55)
MYH9	rs11912763	36684722	A	0.24	0.19	1.36E-03	2.56	(1.44, 4.54)
MYH9 E1	rs2032487	36695428	C	0.69	0.61	2.11E-04	1.56	(1.23, 1.97)
MYH9 E1	rs4821481	36695942	C	0.70	0.60	3.41E-05	1.65	(1.30, 2.09)
MYH9	rs5750250	36708483	A	0.42	0.50	1.07E-03	0.62	(0.47, 0.83)
MYH9 EI	rs3752462	36710183	T	0.76	0.73	1.81E-02	1.33	(1.05, 1.68)

Genotype results are shown for chromosome 22 alleles, including those in APOL1, including those comprising the G1 and G2 alleles, and in MYH9, including those SNPs comprising the E1 haplotype). Data were adjusted for age, sex, and population admixture. Admixture outliers were excluded from analysis, leaving 675 cases and 618 controls.

MYH9 SNPs showed significant associations with CKD, but weaker than the APOL1 G1 and combined G1–G2 association. Analyses were repeated controlling for either the combined APOL1 G1–G2 risk alleles or the MYH9 E1 SNP rs4821481 using the recessive model and the data are shown in Table 3. Controlling for the APOL1 risk alleles attenuated the effect of MYH9 SNPs: the E1-tagging SNP rs4821481 remained of borderline statistical significance. Controlling for rs4821481 did not markedly decrease the effect of the APOL1 G1 alleles.

Table 3.

Case versus control comparisons, adjusted for MYH9 SNP rs4821481 or APOL1 combined risk alleles

Adjusted for MYH9 SNP rs48212481	APOL1 SNP	P value, cases/controls	Odds ratio	95% Confidence Interval
	rs16996616	2.44E-01	2.49	(0.54, 11.55)
	rs73885319 (G1)	1.87E-04	2.78	(1.62, 4.76)
	rs60910145 (G1)	1.69E-04	2.83	(1.65, 4.88)
	rs71785313 (G2)	2.85E-01	1.50	(0.72, 3.14)
Adjusted for APOL1 combined risk alleles	MYH9 SNP
	rs11912763	4.08E-01	1.32	(0.69, 2.52)
	rs2032487 (E1)	1.20E-01	1.23	(0.95, 1.59)
	rs4821481 (E1)	4.67E-02	1.30	(1.00, 1.69)
	rs5750250	6.92E-02	0.76	(0.56, 1.02)
	rs3752462 (E1)	6.73E-01	1.06	(0.82, 1.36)

The top half of the table shows the effect of APOL1 SNPs, adjusted for the MYH9 SNP rs48212481 using a recessive model. The bottom half of the table shows the effect of MYH9 SNPs, adjusted for APOL1 combined risk alleles. Data were adjusted for age, sex, BMI and population admixture.

Association results for the combined risk of APOL1 G1–G2 alleles on CKD are presented in Table 4 and the MYH9 E1 haplotype in Table 5. The statistical significance and odds ratio increased when comparing patients with more advanced renal disease as measured by serum creatinine concentration and proteinuria. This was particularly dramatic for the APOL1 risk alleles, but was also apparent for MYH9 E1 risk haplotype. The sensitivity analysis comparing subsets of controls phenotyped for hypertension and normal blood pressure (BP) shows similar odds ratios to the controls as a whole; the differences in p-values and odds ratios when using these as control subsets, compared to using combined controls, may be due to the differences in sample size.

Table 4.

Logistic regression model of the effect of APOL 1 risk alleles on clinical phenotype

Outcome	N (cases)	N (controls)	P-value	OR, recessive model	95% CI
All AASK cases vs controls	663	579	1.39E-08	2.57	(1.85, 3.55)
Cases with serum creatinine > 2 mg/dL or ESKD vs controls	330	579	1.99E-12	3.64	(2.54, 5.21)
Cases with serum creatinine > 3 mg/dL or ESKD vs controls	216	579	5.60E-15	4.61	(3.14, 6.76)
Cases with urine PCR < 0.22 g/g vs controls	457	579	0.0219	1.55	(1.07, 2.26)
Cases with urine PCR > 0.22 g/g vs controls	204	579	2.70E-15	4.85	(3.28, 7.18)
Cases with urine PCR > 0.60 g/g vs controls	105	579	2.62E-14	6.29	(3.92, 10.11)
Hypertensive kidney disease vs hypertensive controls	663	158	0.0013	2.40	(1.41, 4.08)
Hypertensive kidney disease vs non-hypertensive controls	663	220	8.52E-07	3.62	(2.17, 6.05)

Analysis using APOL1 risk alleles (G1 and G2) was adjusted for age, sex and admixture. Urine PCR denotes protein/creatinine ratio.

Table 5.

Logistic regression analysis for the effect of the MYH9 E1 haplotype on clinical outcomes

Outcome	N (cases)	N (controls)	P-value	OR, recessive model	95% CI
Case-control	622	592	0.0002	1.60	(1.25, 2.05)
Serum creatinine > 2 mg/dL or ESKD	311	592	5.99E-06	1.99	(1.48, 2.68)
Serum creatinine > 3 mg/dL or ESKD	201	592	1.05E-08	2.68	(1.91, 3.76)
Urine PCR < 0.22 g/g	431	592	0.0461	1.33	(1.01, 1.76)
Urine PCR > 0.22 g/g	189	592	1.26E-05	2.38	(1.68, 3.36)
Hypertensive controls	622	167	0.0619	1.42	(0.98, 2.06)
Non-hypertensive controls	622	227	0.0019	1.74	(1.23, 2.47)

Shown are the results of logistic regression analysis, using a recessive model, for the outcomes among cases compared to controls, adjusted for age, sex, and population admixture. PCR denotes urine protein/creatinine ratio.

Results of analyses testing APOL1 genotype effects on renal disease progression, accounting for BP medication and goal BP, are reported in Table 6. This analysis demonstrated that APOL1 risk (recessive) predicted the rate of progression of renal disease; the non-significant heterogeneity p-value reveals that the association between APOL1 groups and progression did not differ significantly across medication class and BP treatment arm.

Table 6.

Effects of APOL1, blood pressure target and medication class on rate of decline of GFR in the AASK Trial

Slope of iothalamate GFR
Medication Class	BP Arm	APOL1 non-risk	APOL1 risk	P-value	Heterogeneity P-value
ACE inhibitor	Low	−1.47±0.22	−2.68±0.40	0.0202
	Usual	−1.50±0.23	−2.84±0.41	0.0023
Beta blocker	Low	−1.59±0.24	−2.22±0.41	0.3776
	Usual	−2.01±0.24	−2.70±0.40	0.1736
Calcium Channel Blocker	Low	−2.05±0.34	−2.72±0.60	0.2542
	Usual	−2.18±0.33	−3.17±0.62	0.2050
Meta-analysis				4.29E-05	0.6257

BP – blood pressure; GFR glomerular filtration rate; APOL1 non-risk = less than two (G1 + G2) risk variants; APOL1 risk = two (G1 + G2) risk variants

Based on least squares projected slope of iothalamate GFR in AASK Trial phase, beginning six months after enrollment. APOL1 genotypes were significantly associated with rate of kidney function decline, in contrast to no effect of blood pressure treatment goal or medication class. The meta-analysis compares APOL1 risk and non-risk genotypes combining all 6 treatment cells.

Analysis of the achieved BP at one year post-randomization in these 675 AASK participants did not reveal significantly different systolic BP, diastolic BP, or mean BP in cases randomized to either low or usual BP treatment arm based on presence of two vs. less than two APOL1 risk variants. In the case of diastolic BP, the usual treatment arm had a trend toward a lower mean±SD diastolic BP in the APOL1 two risk variant group (84.2±1.29 mmHg) versus the APOL1 less than two risk variant group (86.2±0.76 mmHg; p=0.0565). Heterogeneity p-values comparing usual and low BP treatment arms based on APOL1 risk and non-risk were nonsignificant, revealing that the effect of the APOL1 risk variants on BP did not differ based on treatment arm (heterogeneity p=0.6240 systolic BP; p=0.3721 diastolic BP; p=0.4447 mean BP).

To test whether the AASK Trial inference changed after adjusting for the strong effect of the APOL1 risk variants, we computed a general linear model with age, gender, and APOL1 G1/G2 risk variants under a recessive model as covariates. We tested the effects of the BP arm adjusting for the medication class and the above covariates. We also tested for the effects of the medication class adjusting for BP arm and the above covariates. Specifically, there was no effect of the BP arm adjusting for age, gender, APOL1 G1/G2 risk variants under a recessive model and medication class group (P=0.19). There was no effect of the medication class group adjusting for age, gender, APOL1 G1/G2 risk variants under a recessive model and BP arm (P=0.13). Finally, there was no effect of the combined medication treatment group and BP arm, adjusting for age, gender, APOL1 G1/G2 risk variants under a recessive model (P=0.23). There was no evidence that the magnitude of the effect of the APOL1 variants differed as a function of the BP arm (P= 0.56) or medication class (P=0.18).

Discussion

The present results demonstrate that nephropathy risk variants in the APOL1 gene, and to a lesser extent MYH9, are significantly associated with CKD attributed to essential hypertension in non-diabetic AASK participants compared to controls. Evidence of genetic association was most robust in individuals with progressive renal functional decline and higher baseline levels of proteinuria. It is unlikely that these variants are associated with essential hypertension per se, as the results were consistent when comparing hypertensive AASK cases to controls with or without high BP. These results strongly suggest that progressive kidney disease attributed to hypertensive nephrosclerosis in AASK participants, particularly in those with higher baseline levels of proteinuria, lie in the spectrum of FSGS-related kidney disease, as in idiopathic FSGS, as well as HIV-related, C1q-related and idiopathic collapsing forms of FSGS. Focal global glomerulosclerosis, arteriolar nephrosclerosis and interstitial scarring are commonly present in the renal biopsies of AASK participants (12) and appear to reside in this disease-spectrum based on APOL1 association.

Subjects in the AASK trial were randomized to angiotensin converting enzyme (ACE) inhibitor, dihydropyridine calcium channel blocker or beta blockade with low and usual BP targets (e.g., a 3 × 2 factorial design). Subsequently, all participants in the continuation study, the AASK Cohort Study, received ACE inhibitors with a low BP goal. Despite this aggressive therapy, 54% of AASK patients experienced the primary outcome (doubling of creatinine, ESKD, or death) (13). These analyses demonstrate that APOL1 risk (recessive model) predicted the least squares projected slope of iothalamate GFR during the AASK Trial phase. The non-significant heterogeneity p-value reveals that medication class and BP treatment arm did not significantly differ in effect on progression of kidney disease across groups, accounting for APOL1-associated genetic risk. This is consistent with AASK Trial results demonstrating that a lower BP goal did not affect renal disease progression. However, ACE inhibitors did slow the progression of renal disease in AASK relative to beta blockers and calcium channel blockers; the inability to detect this in the current study may be due to the smaller sample size. These results suggest that the failure of intensive treatment of BP to halt the progression of renal disease is associated with the role of APOL1 gene variants. New clinical targets are urgently required to combat this severe genetic form of kidney disease.

It is unclear why ACE inhibition, which often benefits patients with heavy proteinuria, had less of a protective effect in AASK. In this subset of AASK participants, ACE inhibition (regardless of usual or low blood pressure treatment arm) did not significantly impact renal disease progression after accounting for APOL1. Not all AASK Trial or AASK Cohort participants were included in these genetic analyses, hence low power may contribute. In addition, AASK excluded participants with >2.5 grams of proteinuria per day at baseline, most had a far lower level of proteinuria. Therefore, the protective effect of ACE inhibition might have been reduced due to the generally lower levels of proteinuria.

Important clinical and histologic differences exist between African Americans and European Americans in the kidney disease that is labeled hypertensive nephrosclerosis. African Americans develop ESKD attributed to high BP earlier in life than European Americans (2). In addition, successful treatment of hypertension more effectively slows nephropathy progression in European Americans, relative to African Americans (14, 15). Despite being labeled with the same clinical diagnosis (hypertensive nephrosclerosis), African Americans exhibit greater degrees of solidified glomerulosclerosis and arteriolonephrosclerosis, whereas European Americans have greater degrees of obsolescent, collapsed glomeruli (12, 16). These racial differences have long hinted at different causative factors for renal disease progression. APOL1 nephropathy risk variants are present at high frequency in African-derived populations and are virtually absent in European-derived and Asian populations. Therefore, the clinical differences in non-diabetic renal disease may primarily relate to variation in APOL1, as it has been associated with shorter survivals of African American-donated kidneys in the setting of deceased donor transplantation (17). Importantly, kidneys from African American donors without APOL1 risk variants demonstrated excellent allograft survival, similar to European American donated kidneys. Thus, APOL1 genotypes, not race, convey risk for nephropathy.

Aggressive treatment of high BP in AASK failed to significantly slow nephropathy progression in non-diabetic African Americans with hypertension and CKD, particularly among those who lacked proteinuria at baseline (11); use of ACE inhibitors slowed progression compared to beta-blocker or calcium channel blocker (18), but long term progression rates on ACE inhibitors remain high (13). Genetic influences that may not be sensitive to ACE inhibitors or BP control have previously been shown to contribute to this propensity for progressive loss of kidney function among AASK participants, including variants in the adrenergic pathway (19) and the homocysteine pathway (20). We now show that in this region of chromosome 22, genetic variation in APOL1, and to a lesser extent MYH9, is strongly associated with this renal progression and may offer a new perspective on hypertension-attributed renal disease. MYH9 has effects independent of APOL1 on nephropathy risk in those with sickle cell disease (21) and in individuals of European ancestry (8, 9). There is now abundant data that current therapies have limited ability to slow progression of kidney disease. It is therefore critical to understand the function of APOL1 in order to develop improved therapeutic options to slow progression of non-diabetic kidney disease.

Methods

Subjects

The AASK clinical trial compared the effects of 3 different drugs and 2 BP targets on progression of hypertension-attributed nephropathy in 1094 African Americans recruited at 21 clinical sites in the United States. Study subjects were self-identified African Americans between 18 and 70 years old, had a diastolic BP >95 mmHg, and a measured iothalamate GFR between 20 and 65 ml/min/1.73 m². Selected exclusion criteria included evidence of a clinical cause other than hypertension for kidney disease, diabetes mellitus or fasting blood glucose level > 140 mg/dl, urine PCR >2.5 g/g, secondary hypertension, or accelerated or malignant hypertension in the preceding 6 months.

Control subjects were recruited at Wake Forest School of Medicine and consisted of 948 African American subjects with serum creatinine concentrations < 1.5 mg/dl in men or < 1.3 mg/dl in women, of whom 410 were questioned about the presence of hypertension and 171 (41.7%) reported having high BP.

Ethical approvals

The AASK clinical protocol was approved by the Institutional Review Board (IRB) of each participating institution and each patient provided informed consent. After the trial was underway, subjects were approached to contribute DNA as part of an Ancillary Study that was approved by the Institutional Review Board at each participating study site, 850 subjects consented and contributed blood for DNA extraction. Control participants provided DNA and clinical information through a protocol approved by the Wake Forest School of Medicine IRB. All controls provided written informed consent.

Genetic analysis

DNA was genotyped for 10 SNPs on chromosome 22 using TaqMan assays (Applied Biosystems, Foster City, CA), including the 4 MYH9 E1 haplotype SNPs (rs4821480, rs2032487, rs4821481, rs3752462) and 2 other SNPs in the MYH9 region (rs5750250 and rs11912763), the APOL1 G1 nonsynonymous SNPs (rs73885319, rs60910145), the APOL1 G2 six-base pair insertion/deletion (rs71785313) plus nonsynonymous SNP rs16996616 in the APOL1 gene. An additional 44 ancestry informative marker (AIM) SNPs were genotyped in cases and controls. AIMs included chromosome (chr) 1: rs1334336; rs2365669; rs9725312; chr 2: rs11890727; rs1868929; chr 3: rs7627605; chr 4: rs2725267; rs12503758; rs2545258; chr 5: rs7727623; rs6881896; rs7712675; chr 6: rs4946888; rs9388989; rs686406; chr 7: rs3823831; rs10488004; rs520556; chr 8: rs4732942; rs13439780; chr 9: rs449090; rs7873820; chr 10: rs1733742; rs570677; chr 11: rs647756; rs7952397; chr 12: rs7963493; rs4767461; chr 13: 1572018; rs11148886; chr 14: rs1956424; rs12897140; chr 15: rs4451923; chr 16: rs10775349; rs7198976; chr 17: rs8074370; chr 19: rs901792; rs2231738; chr 20: rs1418032; rs6046024; rs1028546; chr 21: rs220245; chr 22: rs12159761; and chr 23: rs6520141. Admixture analysis was performed with ADMIXMAP, using allele frequencies from CEU and YRI HapMap samples. Genotype data for CEU and YRI were included in the analysis as anchoring populations. Samples with YRI proportion < 0.4 (3 cases) were excluded from the analysis. After quality control for missing phenotypic data, duplicates, genotyping failures and admixture outliers, 675 cases and 618 controls were available for this study. Results for SNP rs4821480 were not included in the analysis because of technical difficulties with the TaqMan assay; its proxy, rs4821481 was retained in the analysis. All reported SNPs conformed to Hardy-Weinberg expectations for genotype frequencies. Slightly different cell counts relate to co-variate availability across analyses.

Analyses were performed to examine the combined risk of APOL G1–G2 alleles and the MYH9 E1 haplotype for different clinical phenotypes. These analyses were performed in subgroups of AASK cases that had or developed severe CKD during the study, defined as: 1) developing ESKD or serum creatinine >2 mg/mL; or 2) developing ESKD or serum creatinine > 3.0 mg/dL. Stratification by proteinuria was evaluated as in prior AASK analyses, which demonstrated patients with urine protein/creatinine ratio (PCR) > 0.22 g/g have a greater risk of progression to ESKD than those with less proteinuria (11). Sensitivity analyses were performed comparing cases to controls who had reported hypertension or normal BP.

Analyses testing APOL1 genotype effects on renal disease progression, accounting for BP medication and goal BP were performed on AASK participants. Renal disease progression was based on the least squares projected slope of iothalamate GFR during the AASK Trial phase (beginning at 6 months to avoid acute medication effects). A test for whether the association varied by BP target and/or medication class (heterogeneity p-value), as well as the 6 individual tests of the association within the BP target-medication class combinations was performed.

Quality Control

The accuracy of the E1 MYH9 and APOL1 TaqMan assays previously were validated by re-sequencing in a group DNAs from 60 African American donors and by genotyping several hundred African and European controls, of all which conformed to Hardy Weinberg equilibrium (HWE) expectations and had allele frequencies similar to those reported in dbSNP (http://www.ncbi.nlm.nih.gov/snp). In this study, we included 10% duplicate samples within and between plates to control for genotyping errors; there was 100% concordance for all SNPs. All but one SNP, rs4824180, were in HWE and showed expected patterns of linkage disequilibrium. We therefore elected to re-genotype all MYH9 E1 SNPs and the APOL1 SNPs to confirm genotype assignment; there was a 100% match for all SNP genotypes. SNP rs4824180 was removed from the analysis both because it was out of HWE (p=0.0001) and because the SNP showed an implausible pattern of linkage disequilibrium with it neighboring SNPs.

Data analysis

Single SNP analysis was performed using logistic regression via SNPGWA (http://www.phs.wfubmc.edu/public/bios/gene/downloads.cfm), adjusting for age, gender and admixture. The primary analysis compared all the cases to all the controls. Subgroup analyses were done stratifying the cases by degree of renal failure: serum creatinine >2 mg/dL or ESKD vs controls or serum creatinine >3 mg/dL or ESKD vs controls) and degree of proteinuria (urine PCR > 0.22 g/g vs controls and urine PCR < 0.22 g/g vs controls). Since not all the controls were phenotyped for self-reported hypertension, sensitivity analyses were performed comparing cases to controls with documented hypertension or with documented normotension.

To determine whether both MYH9 and APOL1 risk alleles were associated with CKD, we performed additional single SNP analyses adjusting for APOL1 G1–G2 combined allele risk (defined as being homozygous for the G1 allele (rs73885319 allele G) or homozygous for the G2 allele (rs71785313 deletion), or heterozygous for both G1 and G2 alleles (one copy of the risk ‘allele’ for each marker) to assess for residual risk of the MYH9 SNPs.

The role of the MYH9 risk haplotype (recessive model for E1 risk haplotype, consisting of SNPs rs2032487, rs4821481, rs3752462) was analyzed using logistic regression controlling for age, gender, admixture, and G1–G2 alleles combined risk.