Association of ABCB1 genetic variants with renal function in Africans and in Caucasians.

BACKGROUND: The P-glycoprotein, encoded by the ABCB1 gene, is expressed in human endothelial and mesangial cells, which contribute to control renal plasma flow and glomerular filtration rate. We investigated the association of ABCB1 variants with renal function in African and Caucasian subjects.
METHODS: In Africans (290 subjects from 62 pedigrees), we genotyped the 2677G>T and 3435 C>T ABCB1 polymorphisms. Glomerular filtration rate (GFR) was measured using inulin clearance and effective renal plasma flow (ERPF) using para-aminohippurate clearance. In Caucasians (5382 unrelated subjects), we analyzed 30 SNPs located within and around ABCB1, using data from the Affymetrix 500 K chip. GFR was estimated using the simplified Modification of the Diet in Renal Disease (MDRD) and Cockcroft-Gault equations.
RESULTS: In Africans, compared to the reference genotype (GG or CC), each copy of the 2677T and 3435T allele was associated, respectively, with: GFR higher by 10.6 +/- 2.9 (P < 0.001) and 4.4 +/- 2.3 (P = 0.06) mL/min; ERPF higher by 47.5 +/- 11.6 (P < 0.001) and 28.1 +/- 10.5 (P = 0.007) mL/min; and renal resistances lower by 0.016 +/- 0.004 (P < 0.001) and 0.011 +/- 0.004 (P = 0.004) mm Hg/mL/min. In Caucasians, we identified 3 polymorphisms in the ABCB1 gene that were strongly associated with all estimates of GFR (smallest P value = 0.0006, overall P = 0.014 after multiple testing correction).
CONCLUSION: Variants of the ABCB1 gene were associated with renal function in both Africans and Caucasians and may therefore confer susceptibility to nephropathy in humans. If confirmed in other studies, these results point toward a new candidate gene for nephropathy in humans.

Background

A better knowledge of the determinants of renal function is of paramount importance considering the high and increasing burden of chronic kidney disease worldwide. [1] The familial aggregation of renal function [2-4] suggests that genetic factors determine, in part, renal function and that there are candidate genes for nephropathy in humans.

The transmembrane efflux-P-glycoprotein (PGP; also known as multidrug resistance-associated protein 1) is encoded by the ABCB1 gene, several genetic variants of which have been shown to influence PGP expression in humans. PGP has been extensively studied in a pharmacogenetic context, in particular for its role in multi-drug resistance in cancer treatment. PGP is widely expressed in the human kidney [5-7] and is involved in ciclosporin-induced post-transplantation nephrotoxicity [5,8,9], possibly because of its influence on ciclosporin absorption. [10]ABCB1 genetic variants have been associated with post-transplantation ciclosporin-induced nephrotoxicity. [10-12] In contrast to the extensive data about the physiological role of PGP on the transport of xenobiotics, little is known about the role of PGP in the transport of endogenous substrates and to our knowledge, ABCB1 gene variants have not been previously associated with renal function in the absence of xenobiotics in humans. [13]

Since PGP is expressed in human endothelial and mesangial cells, which contribute to the control of renal plasma flow and glomerular filtration, we hypothesized that variants in the ABCB1 gene could be associated with renal hemodynamics and glomerular filtration rate (GFR), even in the absence of treatment by PGP substrates and/or modulators, such as ciclosporin. In particular, we investigated whether the 3435 C>T and the 2677G>T ABCB1 variants are associated with GFR, effective renal plasma flow (ERPF) and renal vascular resistance (RVR) in families of African descent. We then confirmed an association of variants in the ABCB1 gene with renal function using the 30 genetic markers located within the ABCB1 gene in a large ongoing population-based genome-wide association study of Caucasians employing the Affymetrix 500 K chip.

Methods

Seychelles study

Study population

In the Seychelles islands (Indian Ocean, African region), we enrolled 494 subjects of East African descent from 76 families enriched in hypertensive individuals between August 1999 and January 2002. The detailed family selection process has been previously described. [2] Renal hemodynamics and ABCB1 genotype were determined in 297 individuals. We excluded seven individuals with extreme outlier values for any of the renal measures, i.e. observations lying beyond three interquartile ranges from the first and third quartiles, leaving 290 individuals to analyze the associations of interest. The study was approved by the Ethical Committees of the Ministry of Health in the Seychelles and of the University of Lausanne (Switzerland). All participants provided written informed consent.

Phenotype and covariate measurements

In the Seychelles study, antihypertensive therapy, if any, was stopped for 2 weeks before clearance protocols. After a two-hour equilibration period, two one-hour inulin and PAH clearances were obtained to measure GFR and ERPF, respectively [see Additional file 1]. The inulin and PAH clearances (Cx) were calculated with the formula Cx = Ux*V/Px, where Ux and Px are urinary and plasma concentrations of the x solute, and V is the urine flow rate in ml/min. Renal blood flow (RBF) was calculated as ERPF/(1-(haematocrit/100)) and RVR as (mean arterial blood pressure)/RBF. GFR was also estimated using the abbreviated Modification of the Diet in Renal Disease (MDRD) equation [14]. Creatinine concentration was measured by the picric acid method (Cobas-Mira, Roche, Basel, Switzerland). Mean arterial blood pressure (MAP) was calculated as diastolic blood pressure plus 1/3 of (systolic minus diastolic blood pressure) from the mean of 6 measurements taken with a mercury sphygmomanometer (3 on the day preceding clearances and 3 on the morning of the clearances). On the day preceding clearance, participants collected urine for 24 hours. Urinary and plasma sodium and potassium concentrations were measured by flame photometry (IL-943, Instrumentation Laboratory, Milan, Italy). Participants on antidiabetic treatment during the preceding month, or with fasting blood glucose ≥ 7.0 mmol/l (measured on venous whole blood in duplicate using a Glycotronic® C reflectometer, Macherey-Nagel, Düren, Germany), were considered as diabetics. Body surface area was calculated using the Dubois formula. [15] Body mass index (BMI) was calculated as weight (kg) divided by squared height (m2).

Genetic analyses

DNA was isolated using standard methods from blood drawn into K-EDTA tubes and stored at 4°C. The ABCB1 genotypes (exon 21, 2677 G>T [rs2032582] and exon 26, 3435 C>T [rs1045642]) were determined by real-time PCR with TaqMan®, as previously described. [16,17] The marker genotypes did not significantly deviate from Hardy-Weinberg proportions (P > 0.09) in founders, and 2677 G>T and 3435 C>T were in strong linkage disequilibrium (D' = 0.90) with each other.

Statistical analyses

We used the ASSOC program in S.A.G.E. (V5.3) [18], which accounts for familial correlations by implementing simultaneous maximum likelihood estimation of both familial components of variance and covariate coefficients, to conduct linear regressions with GFR, ERPF or RVR as dependent variables. Based on the findings from descriptive analyses in our sample (i.e. GT heterozygotes had phenotypic values lying between those of the GG and TT homozygotes), we assumed an additive mode of action for the ABCB1 3435T and 2677T alleles. To conduct analyses stratified by hypertension status, we defined hypertensive subjects as those on antihypertensive treatment at entry into the study or having, respectively, a systolic and/or diastolic blood pressure = 140/90 mm Hg. For multiple linear regression models, we used as predictors age, sex, BMI, 24-hour urinary sodium and potassium excretion, fasting blood glucose, diabetes and MAP. All multivariable models included ascertainment as previously described. [19] We conducted sensitivity analyses that included extreme outlier values. We formally tested whether normotensive and hypertensive subjects had similar results by including a test of hypertension status by genotype interaction (i.e. a likelihood ratio test with 2 degrees of freedom) in multivariable models. To ensure that any associations were not due to population stratification: (1) because some subjects were of mixed descent, we conducted a sensitivity analysis that included only those subjects (78%) with known grand-parent's ethnicity who reported having at least 3 grand-parents of African descent (n = 153), and (2) we extended our analysis of GFR by a separate analysis of 119 subjects from the same families with missing inulin clearance but available MDRD. We also conducted analyses based on haplotypes. For the 61 doubly heterozygous participants (CT/GT) with a potentially ambiguous phase, ABCB1 haplotypes were inferred using the HAPLORE program that uses an expectation-maximisation (EM) algorithm in pedigree data. [20]

CoLaus study

A simple non-stratified random sample (n = 56,694) representing 35% of the overall population of the city of Lausanne (Switzerland) aged 35–75 years was drawn. Inclusion criteria were: a) written informed consent; b) aged 35–75 years; c) available examination and blood sample and d) Caucasian descent. The study was approved by the ethical committee of the Faculty of Medicine of Lausanne. Recruitment began in June 2003 and ended in May 2006. Among those eligible, the participation rate was 41%. Of the 6188 participants, 781 were excluded because of missing or low quality genotyping data, genetic relatedness with other subjects, or missing phenotyping data, leaving 5407 subjects for the present analyses.

Venous blood samples (50 mL) were drawn after an overnight fast, and clinical chemistry assays were performed by the CHUV Clinical Laboratory on fresh blood samples on a Modular P apparatus (Roche Diagnostics, Switzerland) within 2 hours of blood collection. Serum creatinine was measured by the Jaffe kinetic compensated method (2.9% – 0.7% maximum inter and intra-batch CVs). GFR was estimated using the abbreviated MDRD equation [14] and the Cockcroft-Gault (CG) formula [21]. Body weight and height were measured with participants standing without shoes in light indoor clothes. Smoking was defined as present if a participant reported to be a current smoker at the time of examination, and alcohol consumption as present for participants reporting drinking alcohol at least once a day. Diabetes was defined as being on current antidiabetic treatment or having fasting plasma blood glucose ≥ 7 mmol/L.

Nuclear DNA was extracted from whole blood for whole genome scan analysis. Genotyping was performed on 6014 participant samples, using the Affimetrix 500 K SNP chip, as recommended by the manufacturer. Individuals with less than 95% genotyping efficiency overall (or <90% efficiency on either array, N = 399), and individuals with possible gender inconsistencies (N = 5), were removed. Duplicate samples, and first and second degree relatives were identified by estimating genomic identity-by-descent (i.b.d.) coefficients; the younger individual from each such pair was removed (N = 200). A further 3 individuals had missing phenotype data, leaving 5407 individuals with complete data available for analysis. Monomorphic SNPs (N = 4052), SNPs with less than 70% genotyping efficiency (N = 157), and SNPs not in Hardy-Weinberg proportions (N = 35417) were excluded from analysis.

A more detailed description of the statistical analyses can be viewed in the Additional file 1. We excluded individuals with outlier phenotypic values defined as values lying more than 4 SD away from the mean (N = 25), leaving 5382 individuals for association tests. Phenotype data were transformed and adjusted for covariates using R. [22] We used age, height, weight and BMI as continuous covariates, and sex, alcohol consumption, smoking, diabetes, antidiabetic or antihypertensive treatment, the presence (or absence) of each of the major classes of antihypertensive drugs (ACE-inhibitors, angiotensin receptor blockers, beta-blockers, channel calcium blockers and diuretics) and currently (i.e. during the last 6 months) taking any prescription drug as binary covariates in normal linear regression. To confirm the association, we used only directly genotyped SNPs first and imputed SNPs in further exploratory analyses. Association tests were performed using PLINK (v0.99s) [23], with an additive mode of action.

Results

Seychelles Study

The allele frequency was 34.1% overall and 30.8% in founders for 2677T, and 54.5% and 53.9%, respectively, for 3435T. Participants' characteristics by genotypes and allele carrier status are presented in Table 1. There was no significant trend across genotypes. Carriers of the 3435T allele had a higher BMI than non-carriers (28.8 ± 0.4 versus 27.3 ± 0.5 kg/m2; P = 0.02).

Table 1

Distribution of selected characteristics by genotypes for each ABCB1 variant in the Seychelles

	2677 G>T			3435 C>T
	GG	GT	TT	CC	CT	TT
N	191	84	15	132	118	40
Age (years)	46.8 ± 12.2	46.2 ± 11.8	44.6 ± 12.5	46.0 ± 12.2	47.3 ± 11.5	45.8 ± 13.3
Proportion of women	0.57 ± 0.50	0.51 ± 0.50	0.47 ± 0.52	0.59 ± 0.49	0.51 ± 0.50	0.53 ± 0.51
BMIa (kg/m2)	28.0 ± 5.6	28.5 ± 5.0	26.8 ± 4.2	27.3 ± 5.2	29.0 ± 5.7	28.0 ± 4.9
MAPb (mm Hg)	101 ± 13	103 ± 14	98 ± 15	100 ± 14	103 ± 13	99 ± 14
FBGc (mmol/L)	4.6 ± 1.8	4.6 ± 1.7	4.5 ± 1.6	4.6 ± 1.6	4.7 ± 1.9	4.6 ± 1.5
Diabetes (prevalence)	0.08 ± 0.28	0.11 ± 0.31	0.07 ± 0.26	0.08 ± 0.27	0.09 ± 0.29	0.13 ± 0.33
Plasma Na (mmol/L)	140 ± 4	140 ± 4	139 ± 3	140 ± 4	140 ± 4	139 ± 3
Plasma K (mmol/L)	3.8 ± 0.3	3.7 ± 0.3	3.7 ± 0.4	3.7 ± 0.3	3.8 ± 0.3	3.7 ± 0.3
Urine Na (mmol/24 h)	106 ± 54	107 ± 56	104 ± 67	101 ± 54	108 ± 56	118 ± 57
Urine K (mmol/24 h)	44 ± 18	46 ± 18	35 ± 16	44 ± 17	45 ± 19	44 ± 17

Results presented are the mean ± s.d. For each variant, there was no significant trend across genotypes. aBMI: body mass index. bMAP: mean arterial pressure. cFBG: fasting blood glucose.

The 2677T allele was associated, in an additive manner, with higher GFR (P = 0.003) and ERPF (P = 0.005) and lower RVR (P = 0.001) (Figure 1). Results were similar, although slightly less significant with the 3435T allele. These results suggest that the T allele is a protective allele with respect to renal function. We obtained similar results for renal haemodynamic phenotypes adjusted for body surface area (Table 2). It is therefore unlikely that the difference in BMI observed between 3435T carriers and non-carriers represents a major confounding factor for these associations.

Figure 1

Renal haemodynamics by genotypes for each . Bar heights are means and vertical lines standard errors. P: P value for linear trend across genotypes using the ASSOC program in S.A.G.E. in models without additional covariates. GFR: glomerular filtration rate measured using inulin clearance. ERPF: effective renal plasma flow measured using PAH clearance. RVR: renal vascular resistance.

Table 2

Renal haemodynamics by genotypes for each ABCB1 variant in the Seychelles

	2677 G>T			3435 C>T
	GG	GT	TT	CC	CT	TT
N	191	84	15	132	118	40
GFRa (mL/min)	110 ± 32	126 ± 35	128 ± 29b	110 ± 29	118 ± 37	128 ± 31c
GFRa (mL/min/1.73 m2)	103 ± 25	114 ± 26	118 ± 26b	104 ± 24	107 ± 27	117 ± 25c
ERPFd (mL/min)	440 ± 143	488 ± 126	542 ± 109b	434 ± 128	470 ± 150	513 ± 129b
ERPFd (mL/min/1.73 m2)	412 ± 119	445 ± 100	500 ± 102b	413 ± 112	426 ± 117	470 ± 108c
RVRe (mm Hg/min/mL)	0.16 ± 0.06	0.13 ± 0.04	0.12 ± 0.04b	0.16 ± 0.06	0.15 ± 0.06	0.13 ± 0.04c
RVRe (mm Hg/min/mL/1.73 m2)	0.15 ± 0.07	0.13 ± 0.04	0.11 ± 0.05b	0.15 ± 0.07	0.14 ± 0.06	0.12 ± 0.04c

Results are the unadjusted mean ± s.d. aGFR: glomerular filtration rate (inulin clearance). b P < 0.01, c 0.01

Results are the unadjusted mean ± s.d. aGFR: glomerular filtration rate (inulin clearance). b P < 0.01, c 0.01