Genetic Variation in APOL1 Associates with Younger Age at Hemodialysis Initiation

Abstract

African Americans have a markedly higher incidence of ESRD compared with other racial groups. Two variants in the APOL1 gene, to date observed only among individuals of recent African ancestry, associate with increased risk for renal disease among African Americans. Here, we investigated whether these risk alleles also associate with age at initiation of chronic hemodialysis. We performed a cross-sectional study of 407 nondiabetic African Americans with ESRD who participated in the Accelerated Mortality on Renal Replacement (ArMORR) study, a prospective cohort of incident chronic hemodialysis patients. African Americans carrying two copies of the G1 risk allele initiated chronic hemodialysis at a mean age of 49.0 ± 14.9 years, which was significantly younger than both subjects with one copy of the G1 allele (55.9 ± 16.7 years; P = 0.014) and subjects without either risk allele (61.8 ± 17.1 years; P = 6.2 × 10⁻⁷). The association between the presence of the G1 allele and age at initiation of hemodialysis remained statistically significant after adjusting for sociodemographic and other potential confounders. We did not detect an association between the G2 risk allele and age at initiation of hemodialysis, but the sample size was limited. In conclusion, genetic variations in APOL1 identify African Americans that initiate chronic hemodialysis at a younger age. Early interventions to prevent progression of kidney disease may benefit this high-risk population.

African Americans have a fourfold greater risk of ESRD compared with white Americans.^1,2 In 2009, the mean age for African Americans at the start of renal replacement treatment was 59.2 years, compared with 66.8 years in Caucasians.² This may be due in part to an accelerated progression of renal disease in African Americans.^3–5 Several studies have found that the high prevalence of ESRD in African Americans cannot be fully explained by socioeconomic differences or differences in access to medical care.^1,6 Thus, it is thought that biologic factors, such as genetic differences, contribute to this disparity. Indeed, previous studies have demonstrated strong familial aggregation of kidney disease in African Americans.⁷ Two recent studies used genetic admixture mapping to identify a region of chromosome 22 that explained the increased kidney disease risk in African Americans.^8,9

Genovese et al. identified sequence variants in apolipoprotein L-1 (APOL1) as risk alleles for focal segmental glomerulosclerosis (FSGS) and hypertension-attributed ESRD (H-ESRD) in African Americans.^9,10 APOL1 is located adjacent to the MYH9 gene on chromosome 22, a locus that has previously been reported to explain the high risk of renal disease in African Americans.⁸ Interestingly, APOL1 risk proteins have lytic activity against a subspecies of trypanosomes known to cause African sleeping sickness. Carrier status may have provided a selective evolutionary advantage and thus maintained these risk alleles in the African population. The two risk alleles found to confer an elevated risk for FSGS and H-ESRD included “G1,” a two-locus allele found in a 10-kb region in the last exon of APOL1, and “G2,” a 6-bp deletion located in close proximity to the G1 risk allele.⁹ These risk alleles in APOL1 are only found in individuals of African descent, with allele frequencies of 38% for G1 and 8% for G2 in the African Yoruba population. These alleles appeared to act in a recessive manner, with a 7- to 10-fold increased risk of H-ESRD or FSGS conferred by the presence of a risk-associated allele in both copies of APOL1.

Given the association between APOL1 risk alleles and nondiabetic renal disease in African Americans, we hypothesized that African Americans with ESRD who were homozygous for APOL1 risk alleles would progress to ESRD at an earlier age than those who did not have these risk alleles. We tested this hypothesis in a cohort of nondiabetic African American subjects initiating chronic hemodialysis in the United States.

RESULTS

Subject characteristics, including demographic information, income, vascular access, cause of ESRD, and laboratory values are summarized in Table 1. The mean age of hemodialysis initiation among all subjects was 55.2 ± 17.1 years.

Table 1.

Subject characteristics (n = 407)

Characteristics	Mean ± SD or Percent (n)	Range (Min to Max)
Age at dialysis initiation	55.2 ± 17.1	18.9 to 94.7
Gender
male	52.3% (213)
female	47.7% (194)
Median income
tertile 3	31.9% (130)	$31,924 to $107,479
tertile 2	33.2% (135)	$21,076 to $31,611
tertile 1	32.4% (132)	$6,878 to $20,985
unknown	2.5% (10)
Census region
northeast	11.3% (46)
Midwest/west	17.0% (69)
south	69.8% (284)
unknown	2.0% (8)
Access
catheter	58.5% (238)
graft	11.1% (45)
fistula	23.3% (95)
unknown	7.1% (29)
Location of dialysis initiation
inpatient	83.5% (340)
outpatient	16.5% (67)
Cause of ESRD
hypertension	72.7% (296)
other	27.0% (110)
unknown	0.3% (1)
BMI	27.2 ± 7.8	13.8 to 67.5
Systolic BP, mmHg	145.0 ± 22.4	90.0 to 219.0
Diastolic BP, mmHg	79.4 ± 14.1	49.0 to 137.0
Albumin, g/dl	3.4 ± 0.6	1.3 to 4.7
Creatinine, mg/dl	8.1 ± 3.6	2.1 to 22.1
eGFR, ml/min per 1.73 m2	9.9 ± 5.6	3.0 to 45.4
PTH, pg/ml	387.9 ± 325.8	4.6 to 2353.4
Calcium, mg/dl	8.4 ± 1.0	4.3 to 12.7
Hemoglobin, g/dl	9.9 ± 1.3	5.9 to 14.9

When subjects were stratified into six unique groups according to the number of G1 or G2 risk alleles, only subjects with two G1 risk alleles had significantly lower mean age at hemodialysis initiation compared with subjects without these APOL1 risk alleles (wild type [Wt] + Wt = 61.8 ± 17.0 years versus Wt + G1 = 55.9 ± 16.7 years [P = 0.152]; versus G1 + G1 = 49.0 ± 14.9 years [P = 3.0 × 10⁻⁶]; versus G1 + G2 = 49.3 ± 17.0 years [P = 1.83 × 10⁻⁴], Figure 1). In contrast, subjects with one or two G2 risk alleles, but no G1 risk alleles, did not begin hemodialysis at an earlier age compared with subjects who lacked the G1 or G2 risk alleles (Wt + G2 = 58.1 ± 16.3 years [P = 0.96], G2 + G2 = 48.9 ± 15.3 years [P = 0.09]). However, the number of individuals with the G2 + G2 risk alleles was small. We therefore conducted subsequent analyses only with G1 risk alleles. Subjects carrying two copies of the G1 risk allele initiated chronic hemodialysis at a mean age of 49.0 ± 14.9 years, earlier than subjects with one copy of the G1 allele (55.9 ± 16.7 years; P = 0.014) or those without any risk allele (61.8 ± 17.1 years; P = 6.2 × 10⁻⁷). Decreased power from exclusion of subjects with G2 risk alleles did not dramatically change the results of the analysis. Including all subjects, regardless of G2 risk, mean age at hemodialysis was significantly lower among those with two G1 risk alleles compared with those without G1 risk alleles (P = 1.0 × 10⁻⁶).

Figure 1.

The mean age at dialysis initiation for subjects by APOL1 risk allele status. Because of the proximity of the alleles it is not expected for a diploid sample to have more then two risk alleles, therefore only six groups existed within this dataset: Wt + Wt, Wt + G1, G1 + G1, Wt + G2, G2 + G2, and G1 + G2. Bar height is the mean age. Error bars denote the standard error. *Significantly different from Wt (Wt + Wt).

Table 2 shows subject characteristics according to whether subjects had zero, one, or two copies of the G1 risk allele. The P values in Table 2 represent comparisons between the genotypes. G1 homozygotes, heterozygotes, and subjects without the G1 risk allele had a similar proportion of male subjects, similar income distribution, and similar locations of hemodialysis initiation. Subjects with and without G1 alleles had similar systolic and diastolic BP, parathyroid hormone (PTH) levels, calcium levels, hemoglobin concentration, and albumin concentrations. Subjects with G1 risk alleles tended to have higher serum creatinine levels (P = 1.0 × 10⁻⁶), perhaps because of their age. To investigate possible confounders between G1 risk allele and age at hemodialysis initiation, we also investigated factors independently correlated with mean age at hemodialysis initiation: body mass index (BMI) (r = −0.224, P = 1.2 × 10⁻⁴), serum creatinine (r = −0.439, P = 1.5 × 10⁻¹⁴); systolic BP (r = −0.188, P = 1.3 × 10⁻³), and diastolic BP (r = −0.490, P = 9.2 × 10⁻¹⁹). We estimated the estimated GFR (eGFR) levels using the Modification of Diet in Renal Disease formula and found that subjects with G1 risk alleles had lower eGFR levels at ESRD initiation (P = 8.1 × 10⁻⁵).

Table 2.

Subject characteristics by G1 risk allele status

Characteristics	Wt (n = 104)	Heterozygous (n = 101)	Homozygous (n = 85)	P
Age at dialysis initiation	61.8 ± 17.1a,b	55.9 ± 16.7a	49.0 ± 14.9b	1.0 × 10−6
Gender				0.9314
male	51.0% (53)	53.5% (54)	52.9% (45)
female	49.0% (51)	46.5% (47)	47.1% (40)
Median income				0.6869
tertile 3	26.9% (28)	34.7% (35)	35.3% (30)
tertile 2	29.8% (31)	33.7% (34)	32.9% (28)
tertile 1	37.5% (39)	30.7% (31)	31.8% (27)
unknown	5.8% (6)	1.0% (1)	0.0% (0)
Census region				0.7810
northeast	11.5% (12)	13.9% (14)	9.4% (8)
Midwest/west	20.2% (21)	15.8% (16)	17.7% (15)
south	64.4% (67)	69.3% (70)	72.9% (62)
unknown	3.9% (4)	1.0% (1)	0.0% (0)
Access				0.8344
catheter	57.7% (60)	63.4% (64)	58.8% (50)
graft	15.4% (16)	10.9% (11)	12.9% (11)
fistula	19.2% (20)	20.8% (21)	23.5% (20)
unknown	7.7% (8)	5.0% (5)	4.7% (4)
Location of dialysis initiation				0.4177
inpatient	78.9% (82)	85.2% (86)	84.7% (72)
outpatient	21.2% (22)	14.9% (15)	15.3% (13)
Cause of ESRD				0.9573
hypertension	73.1% (76)	71.3% (72)	71.8% (61)
other	26.9% (28)	28.7% (29)	27.1% (23)
unknown	0.0% (0)	0.0% (0)	1.2% (1)
BMI	25.9 ± 6.3	26.9 ± 7.5	27.6 ± 7.9	0.2838
Systolic BP, mmHg	146.6 ± 22.3	145.4 ± 23.5	141.3 ± 20.6	0.2514
Diastolic BP, mmHg	77.8 ± 12.9	79.3 ± 15.2	80.6 ± 14.7	0.3898
Albumin, g/dl	3.5 ± 0.6	3.4 ± 0.6	3.5 ± 0.6	0.6550
Creatinine, mg/dl	6.8 ± 2.8a,b	7.7 ± 3.4a	9.4 ± 3.8b	1.0 × 10−6
eGFR, ml/min per 1.73 m2	11.6 ± 6.6a,b	10.4 ± 5.5a	8.0 ± 3.5b	8.1 × 10−5
PTH, pg/ml	332.2 ± 318.7	337.2 ± 249.7	444.9 ± 382.1	0.0585
Calcium, mg/dl	8.5 ± 0.9	8.4 ± 1.0	8.3 ± 1.0	0.5680
Hemoglobin, g/dl	9.9 ± 1.3	9.8 ± 1.4	10.1 ± 1.2	0.3316

Values with the same letter differ significantly from each other on the basis of post hoc tests.

In consonance with other U.S. renal study populations, in nearly three fourths of our subjects ESRD was caused by hypertension. Other causes of ESRD in our population, including HIV, inflammation, toxins, etc., were grouped together as “other.” In stratified analyses by cause of ESRD, the mean age at initiation of hemodialysis was younger in H-ESRD subjects and in subjects with other reported causes of ESRD with G1 risk alleles. However this difference was only significant among H-ESRD subjects (P = 1.7 × 10⁻⁵; Figure 2).

Figure 2.

Left panel shows the mean age at dialysis initiation by G1 risk allele status in subjects with H-ESRD. Right panel shows age by G1 risk allele status in subjects with other ESRD causes, including HIV, inflammation, toxins, etc. Horizontal bars denote mean age whereas vertical bars denote standard error. *Significantly different from Wt (Wt + Wt).

Subjects with ESRD due to causes other than hypertension initiated chronic hemodialysis at an earlier mean age than subjects with H-ESRD (50.6 ± 18.0 years versus 58.1 ± 16.3 years, P = 7.8 × 10⁻⁴). In multivariate regression models, the presence of G1 risk alleles remained significantly associated with early hemodialysis initiation after adjustment for demographic and socioeconomic variables and cause of ESRD among nondiabetics (Table 3). The average values in Table 3 represent the predicted values estimated from multivariate regression equations controlling for sociodemographic and clinical characteristics. In a similar model, when we did not stratify by cause of ESRD (hypertension versus other), we found that the G1 allele remained associated with age of hemodialysis initiation.

Table 3.

Average predicted age at dialysis initiation from linear regression models by G1 risk allele

	Wt	Heterozygousc	Homozygousc
Age	61.8	55.9	49.0
	—	P = 0.011	P = 2.1 × 10−7
Age + H-ESRD	61.8	55.9	49.1
	—	P = 0.011	P = 1.7 × 10−7
Age + H-ESRD + male gender	61.8	55.9	49.1
	—	P = 0.012	P = 1.7 × 10−7
Age + H-ESRD + male gender + tertile 3 incomea + tertile 1 incomea	62.1	55.8	49.1
	—	P = 8.0 × 10−3	P = 1.2 × 10−7
Age + H-ESRD + male gender + tertile 3 incomea + tertile 1 incomea + inpatient dialysis	62.1	55.8	49.1
	—	P = 0.012	P = 2.1 × 10−7
Age + H-ESRD + male gender + tertile 3 incomea + tertile 1 incomea + inpatient dialysis + northeast regionb + south regionb	62.1	55.9	49.1
	—	P = 0.013	P = 1.7 × 10−7

P values represent significance model heterozygous and homozygous G1 risk allele coefficients compared with Wt controlling for all other variables in the model.

^aReference group is medium income.

^bReference group is Midwest/west.

^cReference group is Wt.

DISCUSSION

We aimed to determine if the G1 or G2 sequence risk alleles of the APOL1 gene (previously associated with an increased risk of renal disease in African Americans) were associated with initiating chronic hemodialysis at a younger age—a marker of the severity of progressive CKD. We found that African Americans with two copies of the G1 risk allele initiated chronic hemodialysis approximately 10 years earlier than those without this allele. Subjects with one copy of the G1 allele initiated hemodialysis an average of 6 years earlier. These estimates were unchanged after adjustment for various sociodemographic factors. Given the limited sample size, we were unable to determine whether G2 was associated with the age of chronic hemodialysis initiation.

In a previous case-control study by Genovese et al., APOL1 risk alleles were associated with FSGS and H-ESRD.¹⁰ However, the design of this previous study precluded conclusions about a possible association between the APOL1 risk allele and age of onset and progression of renal disease.⁹ Our study reinforces and extends the results of Genovese et al. by suggesting that these risk alleles are also associated with age at first start of chronic hemodialysis—a measure we used as a surrogate for age of developing ESRD. Our data support the conceptual model that APOL1 risk alleles trigger onset of renal disease at an earlier age or, once initiated, alter the rate of progression. Which mechanism predominates will be the subject of future studies because we did not have access to data on age of initial development of chronic kidney disease (CKD) or rate of progression of CKD in our subjects.

Previous studies have suggested that several factors influence increased risk for ESRD in African Americans. Interestingly, the overall prevalence of CKD in African Americans appears to be lower than or similar to the prevalence in white Americans. However, milder forms of kidney disease, CKD stages II to III, are less prevalent in African Americans whereas stage IV CKD and ESRD are much more common.^2,3,11 Consistent with this, in the Accelerated Mortality on Renal Replacement ArMORR study, the overall mean age at hemodialysis initiation for Caucasians is 65.5 ± 14.9 years, compared with 57.8 ± 15.4 years in African Americans (P < 1.0 × 10⁻⁴).¹² This is consistent with an accelerated progression of ESRD in these subjects. This could also be explained by an alternative pathogenesis for ESRD in some African Americans, leading to early clinical onset of ESRD, potentially mediated by mutations in the APOL1 gene. Several factors have been associated with the rate of progression to ESRD in African Americans. We were unable to test whether some of these other known risk factors such as proteinuria, 25-hydroxyvitamin D levels, and late referral to specialty care mediate the relationship between APOL1 risk alleles and ESRD in this population.¹³ In this study we found that BMI and income were unlikely to confound the association between APOL1 genetic risk alleles and earlier age at hemodialysis initiation. Future studies will explore these relationships and potential causal pathways in more detail.

Subjects with two copies of the G2 risk allele did appear to initiate hemodialysis at a younger age, but the low number of subjects with this genotype likely precluded reaching statistical significance. In contrast, subjects with one copy of the G2 risk allele initiated hemodialysis at approximately the same age as subjects without any risk alleles. It is possible that the G2 risk allele predisposes to renal disease, but it is not associated with earlier onset of ESRD.

Because there was less power to detect effects of the rarer G2 risk allele, it is less clear to what extent this characteristic affects age at hemodialysis initiation. Post hoc power analysis revealed that a sample size of 34 per G2 risk allele group is required to obtain a power of 0.8, and a sample size of 45 is required to obtain a power of 0.9. On the basis of our current sample sizes, it is likely that homozygous G2 risk allele groups were underpowered to find any significant differences in age of hemodialysis initiation. Also, in contrast to the prior study, just one copy of the G1 allele was associated with a younger age at hemodialysis initiation. This is not consistent with a completely recessive model of inheritance, in which the group with one G1 risk allele would have a mean initiation age of hemodialysis similar to the group with no G1 risk alleles, for the predisposition to early onset of ESRD associated with the G1 risk allele.

Our study has certain limitations. First, this was a cross-sectional analysis of baseline data from an observational cohort study. Therefore, we were unable to account for certain factors that antedated hemodialysis initiation, such as age of onset of CKD, predialysis CKD care, comorbidities, compliance with medical therapy, degree of control of hypertension, and other environmental variables. Laboratory measurements such as serum creatinine and eGFR that were significantly different between the genotypes could also be partially attributed to physician's biases based on culture, protocol, or access to health care. However, our findings remained robust to adjustment for sociodemographic factors, and the entire analysis was performed in a single racial group from several sites across the United States.

We were unable to control for other important factors. For example, we did not have information on level of BP control, BP medications, or the timing of BP measurements, all of which may have explained earlier onset of ESRD. However, we did include baseline BP in our analyses. We found that the G1 allele was associated with hypertensive ESRD and not other forms of ESRD. We could not discriminate between FSGS or other causes of ESRD, therefore misclassification between H-ESRD and FSGS was possible. Interestingly, prior studies suggest that the APOL1 gene is associated with H-ESRD and FSGS-related ESRD.^8,10,14

In conclusion, genetic variation in APOL1 is associated with earlier onset of ESRD in African Americans without diabetes mellitus as the etiology of end-stage renal failure. Data from prospective cohort studies of CKD patients are needed to confirm our findings and to more closely assess whether this genetic predisposition plays a role in the initiation or acceleration of CKD. If such studies support our results, preventative interventions based on APOL1 genetic screening could be developed with the goal of identifying and intervening much earlier than is currently practiced.

CONCISE METHODS

Subjects

Subjects were African Americans with nondiabetic ESRD enrolled in the ArMORR study, a prospective cohort study of 10,044 subjects who initiated chronic hemodialysis at any of 1056 U.S. hemodialysis centers operated by Fresenius Medical Care, North America between June 2004 and August 2005.¹⁴ Remnant whole blood samples from ArMORR subjects were available for DNA extraction and establishment of the ArMORR DNA Repository, which was approved by the Institutional Review Board of the Massachusetts General Hospital. The need for informed consent was waived because of the use of remnant samples and the irreversible removal of all personal identifiers (including all direct or indirect links to personal identifiers) and clinical information from samples before entry into the repository. Blood samples from 407 nondiabetic African Americans from 281 sites in 31 different states across the United States were available for DNA extraction.

Data Collection

Data were collected prospectively by caregivers and included demographics, BMI, comorbidities, hemodialysis access (catheter, graft, or fistula), and reported cause of ESRD. Laboratory tests on blood samples collected within 14 days of hemodialysis initiation were performed by a central laboratory (Spectra East, Rockland, NJ); included albumin, creatinine, calcium, phosphate, and hemoglobin; and were measured using standard multisample automated analyzers. Intact PTH was measured using the Nichols Bio-intact assay of full-length 1-84 PTH. Subject's eGFR levels were estimated using the Modification of Diet in Renal Disease formula.¹⁵

Each subject received chronic hemodialysis at an outpatient Fresenius Medical Center, North America facility. Because age of initiation of chronic hemodialysis may differ by region and income, median household income of African Americans living in the zip code for each facility was determined from U.S. Census data for the year 2000 and used as an estimate of socioeconomic status.¹⁶ We divided subjects into three equally sized groups (high, medium, and low socioeconomic status) on the basis of median household income of African Americans in the zip code in which they initiated dialysis. We also assigned subjects to one of three geographic regions on the basis of their hemodialysis facility's zip code and U.S. Census regions.¹⁷ These regions included northeast (Connecticut, Massachusetts, Maine, New Hampshire, New Jersey, Pennsylvania, Puerto Rico, Rhode Island), Midwest/west (Arizona, Colorado, Illinois, Kansas, New Mexico, Missouri, Minnesota, Montana, Ohio, Wisconsin), and south (Alabama, Arkansas, Washington DC, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, and West Virginia). Puerto Rico and the island areas are not part of any census region or census division. For this reason, Puerto Rico was assigned to the northeast region on the basis of standard federal regions where Puerto Rico is grouped in region III with New York and New Jersey.¹⁸

Genotyping

Genomic DNA was extracted from whole blood stored in PaxGene tubes using a protocol adapted from PreAnalytix using a QiagenAutoPure extraction robot (Harvard Partners Center for Genetics and Genomics, Cambridge, MA). In all samples, DNA quality was assessed with 260/280 optical density ratios. The patients' DNA samples were diluted in water to 10 ng/μl, and 300 ng of each DNA sample were sent to Polymorphic DNA Technologies in Alameda, California, which provided assay design, oligonucleotide primers, PCR amplification, and DNA sequencing. Polymorphic DNA Technologies uses Sanger dideoxy DNA sequencing instruments. G1 and G2 alleles were assayed in the same sequencing reaction. Sequencing is less prone to error than other genotyping methods, especially when two alleles, such as G1 and G2, are physically close to each other.

We considered G1 and G2 risk alleles and classified the subjects by APOL1 risk allele status into groups depending on the number of G1 or G2 alleles present. This created six unique groups (WT + WT, G1 + WT, G1 + G1, G2 + WT, G2 + G2, G1 + G2). Because of mutual exclusivity of G1 and G2, no subjects had more than two risk alleles in total (Figure 1). We then separately considered G1 and G2 risk alleles and compared the age of hemodialysis initiation in subjects with zero, one, or two copies of each allele.

Statistical Analysis

ANOVA with Sidak post hoc tests was used to compare mean ages at hemodialysis initiation and other relevant continuous variables across genotypic groups. Pearson correlation coefficients were used to examine the associations between continuous variables. χ² tests were used for categorical variables. Multivariate linear regression modeling was used to obtain the average predicted age of hemodialysis initiation for G1 risk alleles, excluding G2 risk alleles, after adjustment for socioeconomic, demographic, and clinical factors including age, H-ESRD, sex, income, inpatient versus outpatient hemodialysis, and residing region of the United States. All statistical analyses were performed using SAS version 9.2 software (Cary, NC) and STATA version 11 (College Station, TX). Two-tailed P values of <0.05 were considered significant.

DISCLOSURES

R.T. has a research grant from Abbott, and is a consultant to Fresenius Medical Care North America and Shire Human Genetics.