Abstract
Albuminuria demonstrates significant heritability in multiply affected hypertensive and diabetic families. The role of endothelial nitric oxide synthase (eNOS) gene variants as risk factors for albuminuria was investigated in 590 European American siblings from 230 families in the Diabetes Heart Study. Two polymorphisms in the eNOS gene (T-786C in the promoter region and Glu298Asp in exon 7) were genotyped. Albuminuria was defined as an albumin:creatinine ratio (ACR) ≥17 mg/g in men and ≥25 mg/g in women. Tests of association were based on generalized estimating equations, and tests of linkage disequilibrium were based on the quantitative pedigree disequilibrium test. A total of 83% of participants had type 2 diabetes. The median ACR was 10.7 mg/g (interquartile range, 5.1 to 32.8), and 34% (202 of 590) of participants had an elevated ACR. The eNOS −786C allele but not the Glu298Asp was associated with increased ACR (31% increase in absolute level of ACR for each additional copy of the −786C allele; P < 0.0001) and a higher risk for albuminuria (odds ratio, 1.55 for each additional copy of the −786C allele; P = 0.0005). Adjustment for the nongenetic determinants of ACR had no significant effect on the results; neither did stratification by gender, presence of diabetes, and the Glu298Asp genotype. Results were confirmed by quantitative pedigree disequilibrium test analysis and were consistent with haplotype analysis. The −786C eNOS variant was positively correlated with a higher prevalence and a greater degree of albuminuria in European American families in both diabetic and nondiabetic family members.
Vascular endothelial nitric oxide (NO) regulates endothelial function and maintains endothelial-dependent vasodilation in multiple organs, including the kidney. NO is synthesized from l-arginine catalyzed by NO synthase (NOS). In rodent models, chronic NO blockade using substituted l-arginine compounds to compete for NOS produces elevations in glomerular capillary pressure, proteinuria, and glomerulosclerosis (1). Three distinct isoforms of NOS exist in humans, endothelial constitutive NOS (eNOS), neuronal NOS, and inducible NOS. The NOS3 gene encoding eNOS is located on chromosome 7q35–36 and expressed by endothelial cells.
Three polymorphisms in NOS3 are reportedly associated with coronary artery disease (CAD) and/or nephropathy: The Glu298Asp missense mutation, a 27-bp repeat in intron 4, and the T-786C single nucleotide polymorphism (SNP) in the promoter. The Glu298Asp polymorphism is associated with CAD, ESRD, and diabetic nephropathy in some (2–7) but not all studies (8–13). Plasma concentrations of NO metabolites are reduced in carriers of the “a” allele in intron 4 (four repeats in intron 4) (14). However, variable results have been reported for “a” allele associations with diabetic nephropathy (5–17). The T-786C SNP is strongly linked to the intron 4 polymorphism (18,19), and functional studies reveal that the T-786C mutation reduces NOS3 gene promoter activity (20). Two reports have identified a T-786C association with the presence of CAD (10,21), and two other studies, one family-based, reported that the progression of renal disease was influenced by the T-786C polymorphism (12,18).
Microalbuminuria portends an increased risk for the development of progressive renal failure in individuals with diabetes, as well as increased risk for cardiovascular morbidity and mortality. Previous studies by our group have identified strong familiality in urine albumin:creatinine ratio (ACR) among hypertensive (22) and diabetic families (23), suggesting that genetic factors exert a significant influence on albuminuria. In this report, we explored the roles of the NOS3 polymorphisms T-786C and Glu298Asp in susceptibility to albuminuria among European American siblings from multiplex families with type 2 diabetes in the Diabetes Heart Study (DHS). Given the functional studies and high level of linkage disequilibrium with the T-786C polymorphism, (18–20) we did not genotype the intron 4 variant.
Materials and Methods
Study Design and Population
The DHS was conducted in Forsyth County, NC, to study the genetic epidemiology of cardiovascular disease–related phenotypes (e.g., carotid intima-media thickness, calcification of the carotid and coronary arteries, lipids, urine ACR) in families with multiple individuals who have a diagnosis of type 2 diabetes. Type 2 diabetes is clinically defined as diabetes developing after age 34 and treated with insulin and/or oral agents, in the absence of historic evidence of ketoacidosis. At least two siblings with type 2 diabetes must be available in each family and willing to participate. Siblings who are concordant for type 2 diabetes are recruited, along with one nondiabetic sibling in each family when possible. Nondiabetic siblings had fasting blood glucose (<126 mg/dl) to confirm nondiabetic status. Diabetic index cases with renal insufficiency, defined by a serum creatinine ≥1.5 mg/dl or blood urea nitrogen ≥35 mg/dl, were excluded because of the elevation of serum advanced glycosylation end product levels.
Genotype information on NOS3 was available on 590 European American siblings from 230 DHS families. The study was approved by the Institutional Review Board at the Wake Forest University School of Medicine, and all participants gave informed consent.
Laboratory Measurements
Subjects presented fasting to the General Clinical Research Center at the Wake Forest University School of Medicine for measurement of urine ACR from a spot morning urine sample, BP, serum chemistries (including serum creatinine), hemoglobin A1c, fasting glucose, and lipids. To maximize agreement between serial specimens of the study population, gender-specific cut points were chosen to delineate normal from elevated albuminuria: ACR ≥17 mg/g for men and ACR ≥25 mg/g for women. These values are nearly equivalent to albumin excretion rates of 30 μg/min for men and 31 μg/min for women (24).
Morning urine samples from study participants were collected in a resting state and run in duplicate for albumin, total protein, and creatinine concentration. Urine albumin was measured on a Model 1650 Advia (Bayer Diagnostics, Tarrytown, New York) using an automated immunoturbidity analysis. Urine creatinine was measured using the picric acid reaction on the Advia. In an alkaline medium, creatinine reacts with picric acid to form a yellow-orange complex. Rate of color formation is proportional to the concentration of creatinine present and is measured photometrically at 505 nm.
Genomic DNA was extracted from peripheral lymphocytes. The NOS3 T-786C and Glu298Asp variants were genotyped using a MassARRAY SNP genotyping system (Sequenom, San Diego, CA).
Statistical Analyses
Each polymorphism was tested for departures from Hardy-Weinberg proportions using a χ2 goodness-of-fit test. The pairwise linkage disequilibrium (LD) coefficients, D′ and r2, were calculated (25). Albuminuria (ACR) was analyzed as both a continuous and a dichotomous variable. The distributions of urine ACR were highly skewed to the right; thus, natural logarithmic transformations were applied to ACR to approximate better the distributional assumptions of conditional normality and homogeneity of variance. To account for familial correlation, generalized estimating equations (GEE1) assuming exchangeable correlations and using a robust estimator of the variance were calculated for all regression analyses using SAS version 8.0 (SAS Institute, Cary, NC) statistical software. The multivariate GEE1 models adjusted for potential nongenetic risk factors for albuminuria, including age; gender; smoking; diabetes duration; hemoglobin A1c; systolic and diastolic BP; cholesterol concentrations; prevalent cardiovascular disease (CVD); and use of angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and lipid-lowering medications. Models were conducted in the overall sample and by strata of gender, diabetes, and the Glu298Asp genotype. P ≤ 0.05 was considered significant in the analyses.
To minimize the type 1 error rate, we used an approach similar to Fisher protected least significant difference multiple comparisons procedure (26). First, the genotype-based two degree-of-freedom general association test was computed. When this test of general association was significant, the three a priori genetic models (dominant, additive, and recessive) were tested without further adjustment for multiple comparisons. When the general test of association was not significant, the three a priori genetic models were tested after making a sequential Bonferroni adjustment for the three comparisons.
The quantitative pedigree disequilibrium test (QPDT) (27) was performed to assess association of a specified allele or haplotype with urine ACR. A likelihood ratio test was used, using the expectation-maximization algorithm to account for haplotype ambiguities. The QPDT analysis was further conducted with residuals adjusted for potential nongenetic risk factors for albuminuria, including age; gender; smoking; diabetes duration; hemoglobin A1c; systolic and diastolic BP; cholesterol concentrations; prevalent CVD; and use of angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and lipid-lowering medications. Unlike the population-based GEE1 approach, the family-based QPDT analysis tends to be modestly conservative in the presence of population stratification and is valid even under selection bias.
Results
Cross-Sectional Study
Table 1 contains the characteristics of the 590 European Americans who were genotyped for NOS3. A total of 83% of participants had type 2 diabetes. The median of ACR was 10.7 (interquartile range [IQR], 5.1 to 32.8), and 34% (202 of 590) of participants presented with an elevated ACR. In the 230 unrelated probands, the genotype frequencies of the T-786C (C allele frequency, 0.37) and the Glu298Asp (Asp allele frequency, 0.33) polymorphisms were consistent with Hardy-Weinberg expectations, and the observed LD coefficients (D′ = 0.49, r2 = 0.21) did not suggest strong LD. Figure 1 shows the distribution of urine ACR by the genotype of the two polymorphisms. The −786C allele, not the Glu298Asp, was associated with higher ACR levels across the entire range of urine ACR.
Distribution of urine albumin-to-creatinine ratio (ACR) according to the genotype of T-786C (A) and Glu298Asp (B) polymorphisms in NOS3 gene.
Characteristics of 590 European American participants in Diabetes Heart Studya
The genotype-based two degrees of freedom general association test was significant for the T-786C polymorphism (P = 0.0002). Genotype-specific means of ln ACR (95% confidence intervals [CI]) are presented in Figure 2A. Univariate GEE1 analysis revealed that the −786C allele was significantly associated with the level of ACR (P < 0.0001 for trend) in an additive association model. The Glu298Asp was not associated with ACR (P = 0.89; data not shown).
Mean levels of ln ACR and odds ratio of albuminuria according to the T-786C genotype. *P values were controlled for familial correlation.
Compared with −786T homozygotes, the odds ratio (OR) for albuminuria was 1.71 (95% CI, 1.15 to 2.56) for −786T/C heterozygotes and 2.31 (95% CI, 1.40 to 3.81) for −786C homozygotes (OR, 1.55 for each additional copy of the −786C allele; P = 0.0005; Figure 2B). Those associations remained highly significant after adjustment for age; gender; smoking; diabetes duration; hemoglobin A1c; systolic and diastolic BP; serum cholesterol concentrations; prevalent CVD; and the use of angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and lipid-lowering medications (Table 2). The observed differences were consistent in the various subgroups, with stratification of the overall group by gender, presence of diabetes, and the Glu298Asp genotype (Table 2). Some observed null associations in the subgroup analyses might be due to the small sample size.
Association of −786C allele with risk for albuminuria and urine ACRa
Family-Based Study
The findings of the population-based GEE1 approach were examined further in a family-based approach using the QPDT to test for LD between the polymorphism and ACR. Transmission of the C allele of the T-786C polymorphism (P = 6.25e-05) but not the Glu298Asp allele (P = 0.31) contributed to the higher levels of ACR (Table 3). Transmission of the four NOS3 haplotypes defined by the two polymorphisms was also examined (Table 3). The three degrees of freedom overall test for transmission difference was statistically significant (P = 0.01). The haplotype that contained the −786C allele (C-Asp or C-Glu) was transmitted significantly more often to offspring with higher ACR. These transmission distortions remained significant after adjustment for age; gender; smoking; diabetes duration; hemoglobin A1c; systolic and diastolic BP; serum cholesterol concentrations; prevalent CVD; and use of angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and lipid-lowering medications (Table 3).
QPDT analysis of the NOS3 gene and urine ACRa
Discussion
In this population of European American families enriched for type 2 diabetes, we found an association between genetic variation in the eNOS gene and albuminuria. The excess risk for albuminuria was specifically associated with the −786C allele in the promoter. The association was present in both diabetic and nondiabetic individuals, suggesting an important role of the NO pathway on ACR, independent of diabetes status.
Few studies have examined the association of these NOS3 polymorphisms with renal disease, particularly using the phenotype of albuminuria. Our results are consistent with those in two small Japanese and white studies, including a family-based study, which revealed that the −786C allele influenced the progression of renal disease in nondiabetic, type 1 diabetic, and type 2 diabetic causes of nephropathy (12,18). A positive association between the −786C variant and CAD was also demonstrated in two other reports (10,21). Furthermore, an additive association model that we observed is consistent with the previous report (21). In agreement with the family-based study (12), we did not detect association of the Glu298Asp missense mutation with albuminuria, although mixed results have been reported for its relationship with ESRD and CAD (2–13). The majority of the positive associations for the missense mutation were in small studies, as the three largest studies (9–11) failed to detect association.
The mechanisms that are potentially responsible for the association of T-786C with albuminuria can be postulated on the basis of recent advances in molecular biology. Luciferase reporter gene functional analyses demonstrate that the T-786 to C mutation substantially reduces NOS3 gene promoter activity (20). The presence of the −786C allele may reduce NO production, increase renal vascular tone, and accelerate the progression of nephropathy, as has been demonstrated in the rat (1,28,29). Furthermore, reduction of NO expression may accelerate the accumulation of extracellular mesangial matrix proteins and secondarily influence glomerular function and structure, as NO inhibits the synthesis of collagen and fibronectin in cultured rat mesangial cells (30). These pathways likely account for the hypothesized effects of NOS3 gene variation, in both individuals with and without diabetes.
Our study has several limitations. We measured urine ACR in duplicate on only one occasion, whereas the classical definition of microalbuminuria requires two concordant results of three independent samples. Variation in ACR is known to occur when frequent measurements are performed (31). However, unmeasured within-individual variation would be expected to reduce the magnitude of the association. Although these analyses are based exclusively on self-reported European Americans, it is possible that there exists within-ethnicity population structure. The consistency of the family-based QPDT results with the GEE1-based association results suggests that any potential population substructure within our sample is not influencing our results. The observed association could be due to a highly linked disease variant in or near NOS3. The intron 4 polymorphism is almost completely linked to the T-786C variant; however, it has no clear functionality. Furthermore, this study was limited to European Americans; therefore, the generalizability of our findings to other races is uncertain.
In summary, the −786C NOS3 gene variant independently predicted urine ACR and a higher risk for albuminuria in European American families. Our results underscore the role of inherited factors and the NO pathway in the development of albuminuria. The results of family-based studies are most reliable for drawing conclusions about putative genetic associations (32). Confirmation of the independent effect of the −786C NOS3 gene variant on susceptibility to albuminuria will strengthen our understanding of the link between albuminuria and both cardiovascular disease and nephropathy. It could also prompt the development of novel prevention and treatment strategies in renal disease. It will be necessary to provide functional data directly supporting the T-786C variant’s contribution to albuminuria. Future studies are needed to address this consideration.
Footnotes
Published online ahead of print. Publication date available at www.jasn.org.
- © 2005 American Society of Nephrology