Abstract
Background
Two APOL1 nephropathy variants confer substantial risk for non-diabetic end-stage kidney disease (ESKD) in African Americans (AAs). Since not all genetically high-risk individuals develop ESKD, modifying factors likely contribute. Forty-two potentially interactive single nucleotide polymorphisms (SNPs) from a genome-wide association study in non-diabetic ESKD were tested for interaction with APOL1 to identify genes modifying risk for non-diabetic nephropathy.
Methods
SNPs were examined in an expanded sample of 1367 AA non-diabetic ESKD cases and 1504 AA non-nephropathy controls, with validation in an independent family-based cohort containing 608 first-degree relatives of index cases with non-diabetic ESKD. Logistic regression and mixed models were fitted to test for interaction effects with APOL1 on ESKD, estimated kidney function and albuminuria.
Results
Among ESKD samples, 14 of 42 SNPs demonstrated suggestive APOL1 interaction with P-values <0.05. After Bonferroni correction, significant interactions with APOL1 were seen with SNPs in podocin (rs16854341; NPHS2, P = 8.0 × 10−4), in SDCCAG8 (rs2802723; P = 5.0 × 10−4) and near BMP4 (rs8014363; P = 1.0 × 10−3); with trends for ENOX1 (rs9533534; P = 2.2 × 10−3) and near TRIB1 (rs4457349; P = 5.7 × 10−3). The minor allele in NPHS2 markedly changed the APOL1-ESKD association odds ratio (OR) from 7.03 to 1.76 (∼50% reduction in effect per copy of the minor allele), rs2802723 changed the OR from 5.1 to 10.5, and rs8014363 increased the OR from 4.8 to 9.5. NPHS2 (P = 0.05) and SDCCAG8 (P = 0.03) SNPs demonstrated APOL1 interaction with albuminuria in independent family-based samples.
Conclusions
Variants in NPHS2, SDCCAG8 and near BMP4 appear to interact with APOL1 to modulate the risk for non-diabetic ESKD in AAs.
Keywords: African American, APOL1, bone morphogenetic protein 4 (BMP4), kidney disease, podocin (NPHS2), serologically defined colon cancer antigen 8 (SDCCAG8)
INTRODUCTION
Mounting evidence supports the presence of modifiable environmental and inherited factors underlying the risk for apolipoprotein L1 gene (APOL1)-associated nephropathy in African Americans (AAs) [1–3]. Despite possession of two APOL1 nephropathy risk variants with markedly increased risk for non-diabetic end-stage kidney disease (ESKD), all such individuals do not develop nephropathy [4–6]. Renal and urothelial viral infections, including human immunodeficiency (HIV) and JC polyoma virus, may modify inherited risk and serve as treatment targets for prevention of ESKD [1, 7]. In a similar manner, additional genes could interact with APOL1 to alter the risk for the development of progressive nephropathy [8].
In contrast to many complex diseases, the impressive APOL1 effect provides improved power for detection of interactive genetic variants; odds ratios (ORs) for nephropathy association of 29, 17 and 7.3 are reported in HIV-associated nephropathy (HIVAN), idiopathic focal segmental glomerulosclerosis (FSGS) and hypertension-attributed nephropathy [focal global glomerulosclerosis (FGGS), with interstitial fibrosis and arteriolar changes], respectively [4, 7]. These and additional APOL1 variants have recently been implicated in CKD among African cohorts, as well [9, 10].
We previously reported a genome-wide association study (GWAS) aimed at identifying genes associated with non-diabetic ESKD in AAs, as well as an additional analysis testing for genes that might interact with APOL1 to modify nephropathy risk [8]. The present report tested the top 42 APOL1-second gene interactive single nucleotide polymorphisms (SNPs) that demonstrated the initial interaction P-values <0.01 in the GWAS, in an expanded sample of unrelated AAs with and without non-diabetic ESKD. To confirm interaction, validation testing was performed in an independent sample of asymptomatic first-degree relatives of AAs with non-diabetic ESKD. These subjects lacked ESKD and had either normal kidney function or mild kidney disease manifested as reduced estimated glomerular filtration rate (eGFR) or albuminuria.
MATERIALS AND METHODS
Participants
The ESKD case–control sample included subjects from the original GWAS [8], with an additional 503 unrelated AAs with non-diabetic etiologies of nephropathy and 892 non-nephropathy controls. New participants were recruited in the same fashion as original cases and controls. Cases with ESKD were recruited from dialysis facilities in North Carolina, South Carolina, Georgia, Virginia and Tennessee; all were self-described AAs and lacked diabetes mellitus at the initiation of renal replacement therapy. Non-diabetic ESKD was diagnosed in subjects with hypertension or a primary or secondary chronic glomerular disease listed as cause of ESKD on the Centers for Medicare and Medicaid Services 2728 form. Subjects confirmed the onset of hypertension prior to renal replacement therapy and the absence of other kidney disease risk factors. Hypertension-attributed ESKD was diagnosed with proteinuria ≤1.5 g/day, urinalysis ≤100 mg/dL protein or spot urine protein:creatinine ratio ≤1.5 g/g (when data were available), with evidence of other hypertensive target organ damage. Chronic glomerulonephritis was diagnosed in those with kidney biopsy evidence or proteinuria ≥1.5 g/day. Cases with polycystic kidney disease, Alport's syndrome, IgA nephropathy, urologic disease or surgical nephrectomy were excluded. AA controls over age 18 years were recruited from community sources and medical clinics in North Carolina. Controls denied a personal history of kidney disease.
The validation cohort included 608 related and unrelated AAs from ‘Natural History of APOL1-associated Nephropathy’ study families [1, 11]. A maximum of three relatives (siblings or children) were recruited per family for the measurement of albuminuria and kidney function, each subject had a first-degree relative with non-diabetic ESKD. Probands in families had ESKD attributed to hypertension, FSGS, HIVAN or unknown cause in the absence of diabetes. Parents of probands were not recruited, nor were relatives with ESKD. Siblings and offspring could not be younger than 15 years below the age at development of ESKD in the family proband. In this analysis, relatives without kidney disease >5 years younger than the age at ESKD in their family proband were excluded, since future development of nephropathy could not be excluded. All the participants provided blood for DNA extraction after giving written informed consent. Study procedures were performed in compliance with the Wake Forest School of Medicine Institutional Review Board and Declaration of Helsinki.
Genotyping
DNA extraction from whole blood was performed using the PureGene system (Gentra Systems, Minneapolis, MN, USA). Two SNPs in the APOL1 G1 nephropathy risk variant (rs73885319; rs60910145) and an indel for the G2 risk variant (rs71785313) were genotyped using a custom assay designed and genotyped in the Wake Forest School of Medicine Center for Genomics and Personalized Medicine Research using the Sequenom platform (San Diego, CA, USA). In addition, 106 bi-allelic ancestry informative markers (AIMs) were genotyped to provide African-ancestry proportion estimates. The maximum likelihood approach of Tang et al. [12] as coded in the package FRAPPE was used to obtain the proportion of African and European ancestry for each individual. Genotype data at these markers were obtained from 44 HapMap Yoruba individuals (YRI) and 39 European American controls as ‘anchors’ and provided startup values for the Expectation–Maximization algorithm used in FRAPPE. The top interactive SNPs with interaction P-values <0.01 identified in the discovery ESKD GWAS [8] (n = 42) were genotyped in new unrelated ESKD cases and non-nephropathy controls, and in family-based samples on the Sequenom platform (San Diego, CA, USA).
Statistical analysis
Tests for Hardy–Weinberg equilibrium and genotypic association as well as associated summary statistics were computed using the analysis program SNPGWA (https://www.phs.wfubmc.edu/public/bios/gene/downloads.cfm). Tests of interaction in the ESKD data set were computed adjusting for age, gender and African-ancestry proportion using a logistic regression model where the log odds of ESKD served as the outcome in case–control analyses. The odds of carrying the APOL1 risk alleles in the analyses were modeled where we stratified by the ESKD status. The homogeneity of the OR test was computed for the stratified analyses.
We considered two outcomes in the family-based cohort: urine albumin:creatinine ratio (UACR) and the four-variable Modification of Diet in Renal Disease (MDRD) eGFR [13]. The Box-Cox method [14] identified the logarithm transformation for UACR in the family-based cohort. We added 1 to all UACR values before taking the logarithm to avoid generating missing values in case a participant had a value of 0. Linear mixed models were fitted for the continuous outcomes and non-linear mixed models for the mild kidney disease outcome, which took the value of 1 if an individual had UACR >30 mg/g and/or MDRD eGFR <60 mL/min/1.73 m2 (and 0 otherwise). The mixed model framework allowed us to account for the familial relationship using the expected kinship coefficient matrix. This framework has been shown to provide valid inference, when the observed data are a combination of unrelated and related individuals [15]. We could not compute the observed relationship matrix, since genotype data were available only on the APOL1 markers and the 106 AIMs.
Based on the known linkage disequilibrium pattern, where G1 and G2 risk variants are very rarely observed on a single chromosome, we constructed a binary variable representing the compound G1/G2 risk across these two markers, modeling APOL1 risk as the response for all individuals with recessive haplotypes at either G1 or G2 or heterozygosity at both G1 and G2. To test for an interaction between the APOL1 risk loci and each of the 42 potentially interactive SNPs in the ESKD case–control sample, a logistic regression model was fitted with age, gender, African-ancestry proportion, the binary variable representing the G1/G2 compound risk and the individual SNP as covariates and the interaction was tested as modeled using the standard centered cross-product of G1/G2 and the SNP. We assume an additive mode of inheritance (MOI) for the SNP. In addition, we computed the logistic regression model without the interaction term to test for association between each SNP and ESKD after accounting for the effects of G1/G2 and all the other covariates in the model. We also conducted analyses stratified by the APOL1 risk status. Controlling for ancestry and in accordance with Mendel's second law of inheritance, independence is guaranteed whenever the SNP being considered was not located on chromosome 22. In each stratified analysis, based on the APOL1 genotype, comparison of the effects observed between strata was performed by testing for equality between the parameter estimates obtained in each stratum for each model of inheritance. This test was based on the inverse variance weighted difference between these two parameters, and is expected to be more powerful than the Cochran–Mantel–Haenszel test when the effects are in opposite direction [16]. Statistical significance in the ESKD case–control analysis was determined based on a straight application of the Bonferroni correction rule which places the significance threshold at α = 0.05/42 = 0.0012; trends toward association had an overall P < 0.05 with consistent directions of effect when evaluated solely in the newly added cases versus controls as were present in the original analysis [8]. Statistical significance in the family-based validation cohort was considered with a P < 0.05, due to a priori association in the ESKD analysis.
RESULTS
Table 1 contains age at recruitment, African-ancestry proportion and age at renal replacement therapy in the 1367 unrelated AA cases with non-diabetic ESKD and the 1504 unrelated non-nephropathy controls. Table 2 contains demographic and kidney disease parameters in the 608 first-degree relatives of AA index cases with non-diabetic ESKD. Most of these family members were asymptomatic and had a normal MDRD eGFR with relatively low-level albuminuria.
Table 1.
Demographic characteristics of unrelated African American non-diabetic ESKD cases and non-nephropathy controls
| Group | # APOL1 risk variants | n | Age at recruitment, years |
African ancestry, % |
Age at ESKD, years |
||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean | SD | Median | Mean | SD | Median | Mean | SD | Median | |||
| Controls | 0/1 | 1283 | 49.4 | 12.0 | 48.5 | 0.77 | 0.11 | 0.79 | NA | NA | NA |
| 2 | 221 | 49.6 | 13.7 | 48.0 | 0.80 | 0.09 | 0.81 | NA | NA | NA | |
| All | 1504 | 49.4 | 12.3 | 48.0 | 0.77 | 0.11 | 0.79 | NA | NA | NA | |
| ESKD cases | 0/1 | 726 | 57.5 | 14.1 | 57.0 | 0.80 | 0.11 | 0.81 | 53.9 | 38.3 | 52.0 |
| 2 | 641 | 51.4 | 13.8 | 51.0 | 0.80 | 0.09 | 0.81 | 45.2 | 14.8 | 44.0 | |
| All | 1367 | 54.6 | 14.3 | 54.0 | 0.80 | 0.10 | 0.81 | 49.8 | 30.0 | 49.0 | |
| Combined | 0/1 | 2009 | 52.6 | 13.5 | 51.0 | 0.78 | 0.11 | 0.80 | 53.9 | 38.3 | 52.0 |
| 2 | 862 | 51.0 | 13.8 | 50.0 | 0.80 | 0.09 | 0.81 | 45.2 | 14.8 | 44.0 | |
| All | 2871 | 52.1 | 13.6 | 51.0 | 0.79 | 0.11 | 0.80 | 49.8 | 30.0 | 49.0 | |
Table 2.
Demographic characteristics of African American family-based study samples
| Kidney diseasea | # APOL1 risk variants | N | Age, years |
African ancestry, % |
BMI, kg/m2 |
Urine ACR, mg/g |
MDRD eGFR, mL/min/1.73 m2 |
|||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean (SD) | Median | Mean (SD) | Median | Mean (SD) | Median | Mean (SD) | Median | Mean (SD) | Median | |||
| No | 0/1 | 340 | 49.5 (13.0) | 50.0 | 0.79 (0.11) | 0.81 | 33.0 (9.1) | 32.0 | 6.7 (6.2) | 4.1 | 99.2 (22.0) | 98.3 |
| 2 | 89 | 48.1 (14.6) | 49.0 | 0.82 (0.10) | 0.83 | 31.5 (8.6) | 30.2 | 6.6 (5.1) | 4.8 | 97.0 (20.6) | 96.0 | |
| All | 429 | 49.2 (13.3) | 50.0 | 0.80 (0.11) | 0.82 | 32.7 (9.0) | 31.5 | 6.7 (6.0) | 4.3 | 98.8 (21.8) | 98.0 | |
| Yes | 0/1 | 136 | 53.6 (12.7) | 55.0 | 0.81 (0.10) | 0.83 | 35.1 (10.9) | 32.8 | 444.9 (1200.0) | 65.7 | 77.2 (30.3) | 76.1 |
| 2 | 43 | 47.1 (13.6) | 49.0 | 0.83 (0.08) | 0.83 | 34.3 (10.3) | 30.7 | 242.5 (287.0) | 89.1 | 81.3 (33.8) | 81.0 | |
| All | 179 | 52.0 (13.2) | 52.0 | 0.82 (0.10) | 0.83 | 34.9 (10.7) | 32.3 | 396.3 (1058.0) | 68.3 | 78.2 (31.1) | 76.8 | |
| Total | 0/1 | 476 | 50.7 (13.0) | 52.0 | 0.80 (0.11) | 0.82 | 33.6 (9.7) | 32.1 | 131.9 (669.9) | 6.5 | 92.9 (26.6) | 93.4 |
| 2 | 132 | 47.8 (14.2) | 49.0 | 0.82 (0.09) | 0.83 | 32.4 (9.2) | 30.5 | 83.5 (196.8) | 7.8 | 91.9 (26.6) | 91.5 | |
| All | 608 | 50.0 (13.3) | 51.0 | 0.80 (0.10) | 0.82 | 33.3 (9.6) | 31.9 | 121.4 (599.9) | 6.9 | 92.7 (26.6) | 93.1 | |
aKidney disease defined as eGFR < 60 mL/min/1.73 m2 and/or urine ACR > 30 mg/g.
BMI, body mass index; ACR, albumin:creatinine ratio; eGFR, estimated glomerular filtration rate.
Table 3 displays APOL1-interactive genetic association results for the ESKD samples for the 14 (of 42) SNPs that were identified in the initial report and displayed the same direction of interaction as the initial analysis [8] with association P-values <0.05 in the full ESKD case and non-nephropathy control sample. We present the additive-by-recessive interaction models, where all SNPs are tested assuming an additive MOI and APOL1 is modeled under a recessive MOI. The other 28 SNPs failed to meet these criteria and were felt likely to be false positives (data not shown); they were not evaluated further. Supplementary Table S1 contains minor allele frequencies and Hardy–Weinberg equilibrium results for all 42 SNPs. Podocin gene (NPHS2) SNP rs16854341 significantly interacted with APOL1 G1/ G2 risk (interaction OR = 0.6; P = 8.0 × 10−4) in the non-diabetic ESKD cases versus controls, as did rs2802723 in the serologically defined colon cancer antigen 8 gene (SDCCAG8; interaction OR = 1.8; P = 5 × 10−4), and rs8014363 located near BMP4 (interaction OR = 1.6; P = 1 × 10−3). All met the Bonferroni-adjusted level of significance. Several other SNPs trended toward interaction with APOL1 in the ESKD analysis, including rs9533534 located on chromosome 13 in the ecto-NOX disulfide-thiol exchanger 1 gene (ENOX1; interaction OR = 0.7; P = 2.2 × 10−3) and rs4457349 near TRIB1 (interaction OR = 0.7; P = 5.7 × 10−3). A SNP located at ∼174 kb from the stanniocalcin-1 gene (STC1; rs1586171) was weakly interactive in the combined analysis; another STC1 variant was associated with GFR in a prior GWAS [17].
Table 3.
Top 14 SNPs showing an interaction effect with APOL1 in non-diabetic ESKD
| Chromosome/nearest gene | Position | SNP | OR | P-valuea |
|---|---|---|---|---|
| 1/NPHS2 | 177795549 | rs16854341 | 0.6 | 0.0008 |
| 1/SDCCAG8 | 241564934 | rs2802723 | 1.8 | 0.0005 |
| 7/∼381 kb to POM12IL12 | 52689925 | rs7810220 | 1.4 | 0.03 |
| 7/∼238 kb to PTPRZ1 | 121062464 | rs10253361 | 0.7 | 0.0099 |
| 8/∼174 kb to STC1 | 23942506 | rs1586171 | 1.6 | 0.0108 |
| 8/∼44 kb to TRIB1 | 126563498 | rs4457349 | 0.7 | 0.0057 |
| 9/∼447 kb to TLE4 | 80929505 | rs11138082 | 0.7 | 0.0192 |
| 10/∼175 kb to GPR26 | 125250912 | rs7897598 | 1.3 | 0.0816 |
| 13/ENOX1 | 43050726 | rs9533534 | 0.7 | 0.0022 |
| 13/∼54 kb to SUCLA2 | 47361123 | rs7986369 | 0.7 | 0.0209 |
| 14/LOC728755 | 27053341 | rs12587505 | 0.7 | 0.0274 |
| 14/∼8 kb to BMP4 | 53501324 | rs8014363 | 1.6 | 0.0010 |
| 15/∼137 kb to FOXB1 | 58222638 | rs6494167 | 1.3 | 0.065 |
| 19/NPHS1 | 41030902 | rs392702 | 0.7 | 0.0135 |
aReflects the additive effect of each SNP with APOL1 risk (recessive).
All models adjusted for age, gender and proportion of African ancestry.
Interaction effects were quantified for the top SNPs in or near the interactive genes demonstrating the lowest P-values in Table 3; NPHS2, SDCCAG8 and rs8014363 located near BMP4 were significantly interactive; ENOX1 and near TRIB1 trended toward significance. In addition to inheriting two APOL1 risk variants, each copy of the rs16854341 (NPHS2) minor allele reduced the ESKD OR by ∼50%; from 7.03 for individuals who are homozygotes at the major allele to 3.52 for heterozygote individuals and 1.76 for minor allele homozygotes. Inheriting two APOL1 risk variants with zero, one and two rs2802723 (SDCCAG8) minor alleles changed the OR for ESKD from 5.1 to 7.3 to 10.5, respectively. Inheriting two APOL1 risk variants with zero, one and two rs8014363 (near BMP4) minor alleles changed the OR for association from 4.8 to 6.8 to 9.6, respectively and inheriting two APOL1 risk variants with zero, one and two rs9533534 (ENOX1) minor alleles changed the OR for association from 7.45 to 6.32 and to 4.4, respectively.
To confirm these findings, it remained critical to replicate genetic association results in independent samples. To validate ESKD results, potentially APOL1-interactive SNPs were tested for effects on milder kidney disease phenotypes in participants from the ‘Natural History of APOL1 Nephropathy’ study [11]. Assessed phenotypes included MDRD eGFR <60 mL/min/1.73 m2 and/or UACR >30 mg/g (defined as the presence of kidney disease); and the natural logarithm of the (UACR + 1). Although milder renal phenotypes display weaker association with APOL1 relative to ESKD, several SNPs replicated interaction for albuminuria with the same directions of effect as ESKD (Table 4). Importantly, rs16854341 in NPHS2 demonstrated an APOL1 interaction with albuminuria (P = 0.05, dominant model), and a trend for interaction with the binary trait of kidney disease (P = 0.08, recessive model). Similar results were seen for rs2802723 in SDCCAG8 with albuminuria (P = 0.03, dominant model) and a trend for interaction with kidney disease (P = 0.11; dominant model); as well as for rs10253361 near PTPRZ1 (albuminuria; P = 0.02, recessive model and kidney disease; P = 0.01, recessive model). Although rs7897598 appeared to interact with APOL1 for kidney disease and albuminuria in family-based samples, the direction of effect opposed that seen in the ESKD analyses and failed to validate the original finding.
Table 4.
Interaction effect between the 14 SNPs and APOL1 in the family-based cohort (age, gender and African ancestry adjusted)
| Chromosome (nearest gene) | Position | SNP | Outcome | Additive |
Dominant |
Recessive |
|||
|---|---|---|---|---|---|---|---|---|---|
| Parameter | P-value | Parameter | P-value | Parameter | P-value | ||||
| 1 NPHS2 | 17779554 | rs16854341 | Kidney disease | −0.4 | 0.44 | −0.4 | 0.44 | −1.9 | 0.08 |
| 17779554 | rs16854341 | Log(UACR + 1) | −0.5 | 0.06 | −0.63 | 0.05 | −0.28 | 0.73 | |
| 1 SDCCAG8 | 24156493 | rs2802723 | Kidney disease | 0.3 | 0.84 | 0.73 | 0.11 | −0.47 | 0.38 |
| 24156493 | rs2802723 | Log(UACR + 1) | 0.46 | 0.07 | 0.63 | 0.03 | −0.16 | 0.85 | |
| 7 ∼381 kb to POM12IL12 | 52689925 | rs7810220 | Kidney disease | −1.5 | 0.24 | −0.46 | 0.33 | 0.48 | 0.17 |
| 52689925 | rs7810220 | Log(UACR + 1) | 0 | 1.00 | 0.14 | 0.65 | −0.62 | 0.34 | |
| 7 ∼238 kb to PTPRZ1 | 12106246 | rs10253361 | Kidney disease | −2.18 | 0.04 | −0.44 | 0.34 | 1.04 | 0.01 |
| 12106246 | rs10253361 | Log(UACR + 1) | −0.02 | 0.93 | 0.26 | 0.39 | −1.49 | 0.02 | |
| 8 ∼174 kb to STC1 | 23942506 | rs1586171 | Kidney disease | 0.29 | 0.80 | −0.59 | 0.25 | 0.23 | 0.60 |
| 23942506 | rs1586171 | Log(UACR + 1) | −0.2 | 0.46 | −0.38 | 0.23 | 0.56 | 0.47 | |
| 8 ∼44 kb to TRIB1 | 12656349 | rs4457349 | Kidney disease | −0.04 | 0.94 | −0.38 | 0.40 | 0.15 | 0.52 |
| 12656349 | rs4457349 | Log(UACR + 1) | −0.04 | 0.85 | −0.02 | 0.94 | −0.09 | 0.80 | |
| 9 ∼447 kb to TLE4 | 80929505 | rs11138082 | Kidney disease | −0.25 | 0.87 | −0.57 | 0.22 | −1 | 0.13 |
| 80929505 | rs11138082 | Log(UACR + 1) | −0.27 | 0.28 | −0.46 | 0.13 | 0.37 | 0.59 | |
| 10 ∼175 kb to GPR26 | 12525091 | rs7897598 | Kidney disease | −1.18 | 0.33 | −0.59 | 0.19 | −0.83 | 0.03 |
| 12525091 | rs7897598 | Log(UACR + 1) | −0.38 | 0.09 | −0.58 | 0.04 | −0.14 | 0.78 | |
| 13 ENOX1 | 43050726 | rs9533534 | Kidney disease | 0.29 | 0.65 | 0.17 | 0.71 | −0.26 | 0.32 |
| 43050726 | rs9533534 | Log(UACR + 1) | 0.06 | 0.76 | −0.13 | 0.65 | 0.44 | 0.26 | |
| 13 ∼54 kb to SUCLA2 | 47361123 | rs7986369 | Kidney disease | 0.19 | 0.71 | 0 | 1.00 | 0.22 | 0.28 |
| 47361123 | rs7986369 | Log(UACR + 1) | 0.19 | 0.36 | 0.22 | 0.50 | 0.3 | 0.39 | |
| 14 LOC728755 | 27053341 | rs12587505 | Kidney disease | −0.15 | 0.80 | 0.56 | 0.23 | 0.41 | 0.07 |
| 27053341 | rs12587505 | Log(UACR + 1) | 0 | 1.00 | 0.12 | 0.69 | −0.18 | 0.64 | |
| 14 ∼8 kb to BMP4 | 53501324 | rs8014363 | Kidney disease | 0.5 | 0.46 | 0.38 | 0.41 | 0.43 | 0.09 |
| 53501324 | rs8014363 | Log(UACR + 1) | 0.15 | 0.50 | 0.07 | 0.81 | 0.48 | 0.31 | |
| 15 ∼137 kb to FOXB1 | 58222638 | rs6494167 | Kidney disease | −1.27 | 0.09 | 0.11 | 0.82 | −0.41 | 0.11 |
| 58222638 | rs6494167 | Log(UACR + 1) | −0.12 | 0.56 | 0 | 0.99 | −0.38 | 0.29 | |
| 19 NPHS1 | 41030902 | rs392702 | Kidney disease | −0.3 | 0.76 | 0.05 | 0.90 | 0.38 | 0.32 |
| 41030902 | rs392702 | Log(UACR + 1) | −0.11 | 0.62 | −0.14 | 0.63 | −0.13 | 0.82 | |
Kidney disease - MDRD eGFR < 60 mL/min/1.73 m2 and/or urine albumin:creatinine ratio (UACR) > 30 mg/g.
Table 5 displays results of the top 14 interactive SNPs from the ESKD analysis, stratifying the data set composed of non-diabetic ESKD cases and non-nephropathy controls into those with 2 APOL1 risk variants (carriers) and 0 or 1 risk variants (non-carriers). NPHS2 SNP rs16854341 demonstrated significant interactive effects solely in those with 2 APOL1 risk variants (OR = 0.6, P = 2.0 × 10−4; dominant model), with a statistically significant, different interactive effect in the 0/1 APOL1 risk variant group (P = 0.02 between strata). Other potentially interactive SNPs often displayed differential effects in APOL1 2 risk variant carriers versus 0/1 risk variant non-carriers. For example, SDCCAG8 SNP rs2802723 displayed an OR of 1.4 for APOL1 interaction in carriers, with an OR opposite in direction (0.8) in non-carriers (P = 0.002 between strata). Significantly, different effects for the APOL1-second gene interaction based on the presence of 2 risk variants (versus <2) were seen for 13 of the 14 interactive genes.
Table 5.
Association between the top 14 SNPs and ESKD stratified by APOL1 risk status
| Chromosome (nearest gene) | Position | SNP | APOL1 | Additive |
||
|---|---|---|---|---|---|---|
| OR | P-value* | P-value** | ||||
| 1 NPHS2 | 177795549 | rs16854341 | 0/1 | 1 | 0.73 | 0.02 |
| 177795549 | rs16854341 | 2 | 0.6 | 0.0002 | ||
| 1 SDCCAG8 | 241564934 | rs2802723 | 0/1 | 0.8 | 0.02 | 0.002 |
| 241564934 | rs2802723 | 2 | 1.4 | 0.04 | ||
| 7 ∼381 kb to POM12IL12 | 52689925 | rs7810220 | 0/1 | 1.1 | 0.26 | 0.02 |
| 52689925 | rs7810220 | 2 | 1.6 | 0.002 | ||
| 7 ∼238 kb to PTPRZ1 | 121062464 | rs10253361 | 0/1 | 1.0 | 0.64 | 0.006 |
| 121062464 | rs10253361 | 2 | 0.6 | 0.0006 | ||
| 8 ∼174 kb to STC1 | 23942506 | rs1586171 | 0/1 | 0.9 | 0.54 | 0.02 |
| 23942506 | rs1586171 | 2 | 1.5 | 0.01 | ||
| 8 ∼44 kb to TRIB1 | 126563498 | rs4457349 | 0/1 | 1.2 | 0.02 | 0.0005 |
| 126563498 | rs4457349 | 2 | 0.7 | 0.008 | ||
| 9 ∼447 kb to TLE4 | 80929505 | rs11138082 | 0/1 | 1.0 | 0.63 | 0.11 |
| 80929505 | rs11138082 | 2 | 0.8 | 0.07 | ||
| 10 ∼175 kb to GPR26 | 125250912 | rs7897598 | 0/1 | 0.9 | 0.12 | 0.01 |
| 125250912 | rs7897598 | 2 | 1.3 | 0.0499 | ||
| 13 ENOX1 | 43050726 | rs9533534 | 0/1 | 1.2 | 0.003 | 0.002 |
| 43050726 | rs9533534 | 2 | 0.8 | 0.16 | ||
| 13 ∼54 kb to SUCLA2 | 47361123 | rs7986369 | 0/1 | 1.2 | 0.005 | 0.0008 |
| 47361123 | rs7986369 | 2 | 0.8 | 0.05 | ||
| 14 LOC728755 | 27053341 | rs12587505 | 0/1 | 1.1 | 0.49 | 0.04 |
| 27053341 | rs12587505 | 2 | 0.8 | 0.02 | ||
| 14 ∼8 kb to BMP4 | 53501324 | rs8014363 | 0/1 | 0.8 | 0.02 | 0.0004 |
| 53501324 | rs8014363 | 2 | 1.4 | 0.01 | ||
| 15 ∼137 kb to FOXB1 | 58222638 | rs6494167 | 0/1 | 0.8 | 0.02 | 0.03 |
| 58222638 | rs6494167 | 2 | 1.1 | 0.44 | ||
| 19 NPHS1 | 41030902 | rs392702 | 0/1 | 1.2 | 0.02 | 0.003 |
| 41030902 | rs392702 | 2 | 0.8 | 0.06 | ||
*P-value computed in each stratum.
**P-value of the equality of the effects between strata.
DISCUSSION
This series of APOL1 by second gene interaction analyses in AAs with non-diabetic nephropathy confirms a relationship between APOL1 with the NPHS2, SDCCAG8 and near BMP4 genes in susceptibility to non-diabetic ESKD, as well as with albuminuria for NPHS2 and SDCCAG8.
With detection of the impressive APOL1 association with a range of non-diabetic kidney diseases in AAs, including HIVAN, FSGS, FGGS, severe lupus nephritis and sickle cell nephropathy came the realization that modifying genetic and environmental factors were likely present [3–5, 18, 19]. Two APOL1 risk variants are necessary, but not sufficient, for development (or progression) of kidney disease. Environmental modifiers have been identified in the form of viral infections; both HIV and JC polyoma virus maintain renal/uroepithelial reservoirs of infection and appear to interact with APOL1-mediated genetic risk [1, 7]. The incidence rate of HIVAN has been steadily declining since widespread use of highly active anti-retroviral therapy (HAART) [20]. Environmentally acquired infections interact with APOL1 genotypes to increase nephropathy risk; ∼50% of AAs with untreated HIV infection possessing 2 APOL1 risk variants develop HIVAN. The present analyses include the first replication efforts for genes potentially interacting with APOL1 in non-diabetic nephropathy.
An APOL1-interactive SNP was detected in the podocin gene (NPHS2) on chromosome 1q25.2. Podocin is exclusively expressed in the glomerular podocyte [21] and mutations are associated with autosomal recessive FSGS [22], as well as proteinuria, microalbuminuria and clinical nephropathy in diverse populations [23–27]. Podocin is a membrane anchored protein that interacts with other components of the glomerular slit diaphragm, including nephrin and CD2-associated protein [21, 22, 28–30]. Our group previously examined NPHS2 for its role in AA non-diabetic ESKD and uncommon variants were independently associated [31]. Although the presence of two functional NPHS2 mutations is a requirement for developing steroid-resistant nephritic syndrome (homozygosity or compound heterozygosity), we speculated that one NPHS2 risk variant might interact with additional mutations in a regulatory region of the gene or with another gene to produce kidney disease. In this series of replication analyses, it appears that APOL1 and NPHS2 interact to impact the risk for non-diabetic ESKD, as well as albuminuria. Until the mechanisms whereby APOL1 variants leading to severe nephropathy and disease progression are clarified, it may prove difficult to determine the mechanisms whereby these genes interact to cause kidney disease.
In addition to NPHS2, a known glomerulosclerosis gene, SNPs in novel genes and genomic regions appear to interact with APOL1 and impact nephropathy risk. The significant association with rs2802723 (SDCCAG8) was interesting since mutations in this gene play roles in retinal and nephronophthisis-related ciliopathies, and severe nephropathy in Bardet–Biedl syndrome [32, 33]. A trend toward APOL1 interaction with rs8014363 near BMP4 is also of potential interest due to the apparent role of BMP pathway genes in susceptibility to nephropathy [34, 35]. Finally, the NPHS2 and SDCCAG8 validation in family-based samples with mild kidney disease was striking, since APOL1 associates more strongly with ESKD than mild nephropathy [11, 36]. Several of these genes (BMP4, NPHS2 and APOL1) are expressed by podocytes [37, 38]. The renal diseases in which they are involved often result in FSGS and FSGS, collapsing variant. It is therefore possible that podocyte injury explains their combined involvement, with potential for podocyturia and subsequent loss of glomerular structure.
This report has strengths and some limitations. We evaluated a relatively large number of DNA samples from AAs with non-diabetic ESKD and confirmed statistical significance and directions of interactive association with APOL1 based on an initial analysis after including additional ESKD cases and non-nephropathy controls. Initial and replication study samples were combined due to the paucity of available DNA samples in AAs with non-diabetic ESKD. Evaluation of independent family-based samples, first-degree relatives of AAs with non-diabetic ESKD, was a unique resource that allowed for extension of findings in asymptomatic individuals enriched for APOL1 nephropathy risk variants and with milder kidney disease. Confirmation of APOL1 interaction with NPHS2 and SDCCAG8 in both ESKD samples and close relatives of non-diabetic ESKD index cases strongly supports true effect. It is likely that the high ORs for APOL1 association in ESKD made it possible to identify interactive gene variants; a situation different from that in many complex diseases. Potential limitations include the fact that causative variants that interact with APOL1 have not been identified. We note that the present analysis results were performed slightly differently than those previously published [8], since 106 AIMs were employed for African-ancestry adjustment (relative to 70 AIMs previously). The number of markers used to estimate this proportion plays an important role in the ability to control for the type I error rate [39].
In conclusion, this series of gene–gene interaction analyses in AAs with non-diabetic forms of nephropathy identified evidence of significant interaction between NPHS2, SDCCAG8 and near BMP4 with APOL1-associated ESKD and with albuminuria for the first 2 genes; ENOX1 and a variant near TRIB1 trended toward interaction. Identification of interactive genes for APOL1-associated nephropathy will improve our discriminatory ability when screening high-risk populations of African ancestry for risk of progressive nephropathy, possibly for screening potential deceased and living kidney donors prior to transplantation. These results may improve understanding of the pathogenetic mechanisms underlying the progression of this diverse spectrum of non-diabetic kidney diseases and assist in explaining why not all individuals possessing two APOL1 risk variants ultimately develop nephropathy.
SUPPLEMENTARY DATA
Supplementary data are available online at http://ndt.oxfordjournals.org.
CONFLICTS OF INTEREST STATEMENT
The authors report no conflicts of interest. The results presented in this paper have not been published previously in whole or part, except in abstract format.
Supplementary Material
ACKNOWLEDGEMENTS
This work was supported by NIH grants R01 HL56266, RO1 DK070941 and RO1 DK084149 (BIF).
REFERENCES
- 1.Divers J, Nunez M, High KP, et al. JC Polyoma virus interactions with APOL1 in African Americans with non-diabetic nephropathy. Kidney Int. 2013 doi: 10.1038/ki.2013.173. doi:10.1038/ki.2013.173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Skorecki KL, Wasser WG. Hypertension-misattributed kidney disease in African Americans. Kidney Int. 2013;83:6–9. doi: 10.1038/ki.2012.369. [DOI] [PubMed] [Google Scholar]
- 3.Freedman BI, Bowden DW, Rich SS. Genetic basis of kidney disease. In: Taal MW, editor. Brenner & Rector’s The Kidney. Philadelphia: Elsevier Saunders; 2012. pp. 1554–1569. [Google Scholar]
- 4.Genovese G, Friedman DJ, Ross MD, et al. Association of trypanolytic ApoL1 variants with kidney disease in African Americans. Science. 2010;329:841–845. doi: 10.1126/science.1193032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Tzur S, Rosset S, Shemer R, et al. Missense mutations in the APOL1 gene are highly associated with end stage kidney disease risk previously attributed to the MYH9 gene. Hum Genet. 2010;128:345–350. doi: 10.1007/s00439-010-0861-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Freedman BI, Kopp JB, Langefeld CD, et al. The apolipoprotein L1 (APOL1) Gene and nondiabetic nephropathy in African Americans. J Am Soc Nephrol. 2010;21:1422–1426. doi: 10.1681/ASN.2010070730. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kopp JB, Nelson GW, Sampath K, et al. APOL1 genetic variants in focal segmental glomerulosclerosis and HIV-associated nephropathy. J Am Soc Nephrol. 2011;22:2129–2137. doi: 10.1681/ASN.2011040388. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bostrom MA, Kao WH, Li M, et al. Genetic association and gene-gene interaction analyses in African American dialysis patients with nondiabetic nephropathy. Am J Kidney Dis. 2012;59:210–221. doi: 10.1053/j.ajkd.2011.09.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ko WY, Rajan P, Gomez F, et al. Identifying Darwinian selection acting on different human apol1 variants among diverse African populations. Am J Hum Genet. 2013;93:54–66. doi: 10.1016/j.ajhg.2013.05.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Tayo BO, Kramer H, Salako BL, et al. Genetic variation in APOL1 and MYH9 genes is associated with chronic kidney disease among Nigerians. Int Urol Nephrol. 2013;45:485–494. doi: 10.1007/s11255-012-0263-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Freedman BI, Langefeld CD, Turner J, et al. Association of APOL1 variants with mild kidney disease in the first-degree relatives of African American patients with non-diabetic end-stage renal disease. Kidney Int. 2012;82:805–811. doi: 10.1038/ki.2012.217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Tang H, Peng J, Wang P, et al. Estimation of individual admixture: analytical and study design considerations. Genet Epidemiol. 2005;28:289–301. doi: 10.1002/gepi.20064. [DOI] [PubMed] [Google Scholar]
- 13.Levey AS, Bosch JP, Lewis JB, et al. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med. 1999;130:461–470. doi: 10.7326/0003-4819-130-6-199903160-00002. [DOI] [PubMed] [Google Scholar]
- 14.Box GEP, Cox DR. An analysis of tranformations. J R Statis Soc B. 1964;26:211–246. [Google Scholar]
- 15.Manichaikul A, Chen WM, Williams K, et al. Analysis of family- and population-based samples in cohort genome-wide association studies. Hum Genet. 2012;131:275–287. doi: 10.1007/s00439-011-1071-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Agresti A. An Introduction to Categorical Data Analysis. Hoboken, NJ: John Wiley & Sons, Inc.; 2007. [Google Scholar]
- 17.Kottgen A, Glazer NL, Dehghan A, et al. Multiple loci associated with indices of renal function and chronic kidney disease. Nat Genet. 2009;41:712–717. doi: 10.1038/ng.377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ashley-Koch AE, Okocha EC, Garrett ME, et al. MYH9 And APOL1 are both associated with sickle cell disease nephropathy. Br J Haematol. 2011;155:386–394. doi: 10.1111/j.1365-2141.2011.08832.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Larsen CP, Beggs ML, Saeed M, et al. Apolipoprotein L1 risk variants associate with systemic lupus erythematosus-associated collapsing glomerulopathy. J Am Soc Nephrol. 2013;24:722–725. doi: 10.1681/ASN.2012121180. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.U.S. Renal Data System. National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 2012. USRDS 2012 Annual Data Report, Vol 1: Atlas of Chronic Kidney Disease and End-Stage Renal Disease in the United States. Ref Type: Generic. [Google Scholar]
- 21.Roselli S, Gribouval O, Boute N, et al. Podocin localizes in the kidney to the slit diaphragm area. Am J Pathol. 2002;160:131–139. doi: 10.1016/S0002-9440(10)64357-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Boute N, Gribouval O, Roselli S, et al. NPHS2, encoding the glomerular protein podocin, is mutated in autosomal recessive steroid-resistant nephrotic syndrome. Nat Genet. 2000;24:349–354. doi: 10.1038/74166. [DOI] [PubMed] [Google Scholar]
- 23.Karle SM, Uetz B, Ronner V, et al. Novel mutations in NPHS2 detected in both familial and sporadic steroid-resistant nephrotic syndrome. J Am Soc Nephrol. 2002;13:388–393. doi: 10.1681/ASN.V132388. [DOI] [PubMed] [Google Scholar]
- 24.Frishberg Y, Rinat C, Megged O, et al. Mutations in NPHS2 encoding podocin are a prevalent cause of steroid-resistant nephrotic syndrome among Israeli-Arab children. J Am Soc Nephrol. 2002;13:400–405. doi: 10.1681/ASN.V132400. [DOI] [PubMed] [Google Scholar]
- 25.Caridi G, Bertelli R, Scolari F, et al. Podocin mutations in sporadic focal-segmental glomerulosclerosis occurring in adulthood. Kidney Int. 2003;64:365. doi: 10.1046/j.1523-1755.2003.00084.x. [DOI] [PubMed] [Google Scholar]
- 26.Caridi G, Berdeli A, Dagnino M, et al. Infantile steroid-resistant nephrotic syndrome associated with double homozygous mutations of podocin. Am J Kidney Dis. 2004;43:727–732. doi: 10.1053/j.ajkd.2003.12.034. [DOI] [PubMed] [Google Scholar]
- 27.Pereira AC, Pereira AB, Mota GF, et al. NPHS2 R229q functional variant is associated with microalbuminuria in the general population. Kidney Int. 2004;65:1026–1030. doi: 10.1111/j.1523-1755.2004.00479.x. [DOI] [PubMed] [Google Scholar]
- 28.Schwarz K, Simons M, Reiser J, et al. Podocin, a raft-associated component of the glomerular slit diaphragm, interacts with CD2AP and nephrin. J Clin Invest. 2001;108:1621–1629. doi: 10.1172/JCI12849. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Sellin L, Huber TB, Gerke P, et al. NEPH1 defines a novel family of podocin interacting proteins. FASEB J. 2003;17:115–117. doi: 10.1096/fj.02-0242fje. [DOI] [PubMed] [Google Scholar]
- 30.Huber TB, Kottgen M, Schilling B, et al. Interaction with podocin facilitates nephrin signaling. J Biol Chem. 2001;276:41543–41546. doi: 10.1074/jbc.C100452200. [DOI] [PubMed] [Google Scholar]
- 31.Dusel JA, Burdon KP, Hicks PJ, et al. Identification of podocin (NPHS2) gene mutations in African Americans with nondiabetic end-stage renal disease. Kidney Int. 2005;68:256–262. doi: 10.1111/j.1523-1755.2005.00400.x. [DOI] [PubMed] [Google Scholar]
- 32.Otto EA, Hurd TW, Airik R, et al. Candidate exome capture identifies mutation of SDCCAG8 as the cause of a retinal-renal ciliopathy. Nat Genet. 2010;42:840–850. doi: 10.1038/ng.662. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Billingsley G, Vincent A, Deveault C, et al. Mutational analysis of SDCCAG8 in Bardet-Biedl syndrome patients with renal involvement and absent polydactyly. Ophthalmic Genet. 2012;33:150–154. doi: 10.3109/13816810.2012.689411. [DOI] [PubMed] [Google Scholar]
- 34.Martini S, Nair V, Patel SR, et al. From SNP to transcriptional mechanism: a model for FRMD3 in diabetic nephropathy. Diabetes. 2013;62:2605–2612. doi: 10.2337/db12-1416. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Palmer ND, Freedman BI. Diabetic nephropathy: FRMD3 in diabetic nephropathy-guilt by association. Nat Rev Nephrol. 2013;9:313–314. doi: 10.1038/nrneph.2013.81. [DOI] [PubMed] [Google Scholar]
- 36.Friedman DJ, Kozlitina J, Genovese G, et al. Population-based risk assessment of APOL1 on renal disease. J Am Soc Nephrol. 2011;22:2098–2105. doi: 10.1681/ASN.2011050519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Ueda H, Miyazaki Y, Matsusaka T, et al. Bmp in podocytes is essential for normal glomerular capillary formation. J Am Soc Nephrol. 2008;19:685–694. doi: 10.1681/ASN.2006090983. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Madhavan SM, O'Toole JF, Konieczkowski M, et al. APOL1 Localization in normal kidney and nondiabetic kidney disease. J Am Soc Nephrol. 2011;22:2119–2128. doi: 10.1681/ASN.2011010069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Divers J, Vaughan LK, Padilla MA, et al. Correcting for measurement error in individual ancestry estimates in structured association tests. Genetics. 2007;176:1823–1833. doi: 10.1534/genetics.107.075408. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.