An Integrative Study of the Genetic, Social and Environmental Determinants of Chronic Kidney Disease Characterized by Tubulointerstitial Damages in the North Central Region of Sri Lanka

Abstract

Objectives: Previous investigations on chronic kidney disease of unknown etiology characterized by tubulointerstitial damages (CKDu) in the North Central Region (NCR) of Sri Lanka have supported the involvement of social, environmental and genetic factors in its pathogenesis. Methods: We conducted a social-environmental-and-genetic epidemiology study on a male population in NCR to investigate the genetic and environmental contributors. We recruited 311 case-series patients and 504 control candidates. Of the 504 control candidates, 218 (43%) were eliminated because of the presence of hypertension, proteinuria, high HbA1c, high serum creatinine or high alpha-1 microglobulin in urine. Results and Discussion: None of 18 metals measured (μg//) in urine, including Cd, As and Pb, showed significantly higher concentrations in cases compared with controls. As speciation results showed that 75–80% of total urinary As was in the form of arsenobetaine, which is non-toxic to humans. None of the metal concentrations in drinking water samples exceeded guideline values. A genome-wide association study (GWAS) was conducted to determine the genetic contributors. The GWAS yielded a genome-wide significant association with CKDu for a single nucleotide polymorphism (SNP; rs6066043; p=5.23 × 10⁻⁹ in quantitative trait locus analysis; p=3.73 × 10⁻⁹ in dichotomous analysis) in SLC13A3 (sodium-dependent dicarboxylate transporter member 3). The population attributable fraction and odds ratio for this SNP were 50% and 2.13. Genetic susceptibility was identified as the major risk factor for CKDu. However, 43% of the apparently healthy malepopulation suffers from non-communicable diseases, suggesting their possible influence on CKDu progression.

(J Occup Health 2014; 56: 28–38)

Introduction

Based on data from renal clinics at tertiary care hospitals, the prevalence of chronic kidney disease (CKD) in areas in and around the North Central Province of Sri Lanka (hereafter referred as the North Central Region, NCR) has been speculated to be relatively high¹⁾. Local health professionals claim that diabetes mellitus, hypertension and other identifiable causes of CKD are not responsible for CKD in this region. Since its pathogenesis is entirely unknown, it is designated as CKD of unknown etiology (CKDu).

In the previous study in NCR, we found that tubulointerstitial fibrosis was identified as the dominant histopathological lesion in CKDu²⁾, which supports opinions of local pathologists. In accordance with the pathological characteristics, we found that a tubular marker, alpha-1-microglobulin (A1M), was elevated in CKDu cases and was a sensitive marker for screening the population and diagnosing early-stage CKDu patients³⁾. In addition to these findings, we found that family history was a risk factor and observed familial clustering in some families, suggesting the involvement of genetic factors3). Over the last few years, research to explain tubulointerstitial damages focusing on the involvement of environmental factors, such as Cd, As, and fluoride, has been reported, but not confirmed^4–6). An epidemiological study revealed male paddy farmer preponderance and association with various life styles⁷⁾. The aim of this study was to comprehensively evaluate the possible involvement of environmental, social, and genetic factors in the pathogenesis of CKDu with tubulointerstitial damages.

Materials and Methods

Ethics statement

This study was approved by the ethics committees of Kyoto University, Japan, and the Faculty of Medicine, University of Peradeniya, Sri Lanka. All human samples, clinical reports, and questionnaire data were obtained after receiving written informed consent, and the study was performed in accordance with the guidelines of the Declaration of Helsinki.

Study populations for environmental and genetic studies

CKDu cases were defined as cases who developed CKDu in the clinical course of tubulointerstitial damages. Thus, we eliminated CKD that can be a secondary complication of diabetes (history of diabetes mellitus and HbAlc >6.5% at the time of diagnosis of CKD) or hypertension (history of chronic and/or severe hypertension at the time of diagnosis of CKD), or other known renal diseases such as autoimmune diseases, glomerular nephritis, Fanconi syndrome or IgA nephropathy (presence of histopathological and immunoflurocence evidence). Previous studies suggested that areas of residence and male sex are strong risk factors for CKDu^{2, 3, 5)}. Therefore, we selected two areas, Medawachchiya and Girandurukotte, where CKDu has been postulated to be more prevalent, and male sex for this study (Fig. 1A).

Figure 1B shows a schematic representation of the selection methodology for the cases and controls recruited for this study. We invited male CKDu caseseries patients (age range, 16–70 years) from a single ethnic group (Sinhalese), who were registered in Medawachchiya and Girandurukotte renal clinics from January 2005 to December 2010. The response rate was about 90%, and these patients were recruited as cases for this study (n=311). Cases were confirmed to have tubulointerstitial damage by either renal biopsy or clinical data; 93% (n=288) were biopsy-proven (interstitial fibrosis with or without interstitial inflammation and negative immunofluorescence for IgG, IgM, IgA, and C3) and 7% (n=23) had serum creatinine >1.2 mg/dl and/or AIM >15.5 mg/l. In parallel, we used the clinical records to confirm that these patients developed CKDu in the course of tubulointerstitial damage and the cases did not have uncontrolled hypertension or diabetes at the time of initial diagnosis.

With assistance from community administrative leaders, randomly selected apparently healthy Sinhalese males (age range, 16–70 years), who had no past history of hypertension, diabetes mellitus, or renal diseases, were not on treatment for any other disease condition, and had resided in the region for at least 10 years, were invited to participate. The response rate was about 60%. As illustrated in Figure 1B, blood pressure measurement, dipstick test for proteinuria and glycosuria in spot urine samples, HbA1c, serum creatinine, and A1M were used to recruit subjects without undiagnosed hypertension, diabetes, or renal function impairment.

All the selected cases and controls were interviewed using a structured questionnaire to collect personal, occupational, lifestyle-related, food habit-related, and clinical information. Blood samples (10 ml) were collected from peripheral veins into K-EDTA tubes. Serum was separated immediately by centrifugation at 3,000 rpm for 10 minutes. Spot urine samples were also collected from all the recruited cases and controls into empty polypropylene tubes.

Fig. 1

Study area and design. (A) North Central Province in Sri Lanka and sampling locations. The map shows the locations of the CKDu clinics (Medawachchiya and Girandurukotte) from where the human samples were collected. Water samples were collected from two areas in NCR (Medawachchiya and Girandurukotte). (B) Flow chart summarizing the selection process for the cases and controls. MW, Medawachchiya; GD, Girandurukotte; SCr, serum creatinine; A1M, alpha-1-microglobulin; GWAS, genome-wide association study. (C) Study design for the genetic analysis. GWAS, genome-wide association study; HWE, Hardy-Weinberg equation; QTL, quantitative trait locus; eGFR, estimated glomerular filtration rate, PCA, principal component analysis. (D) Manhattan plot. The Manhattan plot shows the significance of associations for all SNPs. The SNPs are plotted on the X-axis according to their position on each chromosome against the association with the estimated glomerular filtration rate on the Y-axis, shown as the −log₁₀(p-value). One SNP showing a genome-wide significant association and all SNPs achieving p<5 × 10⁻⁶ are shown with their p-values.

Water samples (1 l each) were collected into empty polypropylene bottles from drinking water sources in two areas of NCR (Medawachchiya, n=15; Girandurukotte, n=16) (Fig. 1A). The water samples were randomly selected from common drinking water sources in villages with a high incidence of CKDu.

All the drinking water, urine, serum, and whole-blood samples were immediately stored at 4°C and then transferred to −30°C within 6 hours of collection. The samples were shipped to Japan at −20°C and stored at −30°C until analysis.

The estimated glomerular filtration rate (eGFR) was obtained using the modified diet in renal disease (MDRD) formula, i.e. GFR (ml/min/1.73 m²)=186 x serum creatinine^−1.154 × (age)^−0.203 × (0.742 if female), based on the serum creatinine levels⁸⁾. In addition, we conducted a field survey to investigate food habits of the local residents.

Drinking water, serum and urine analyses

Drinking water and urine samples were analyzed for heavy metals using inductively coupled plasma mass spectrometry (ICP-MS). Urinary heavy metals were checked in all subjects recruited for the genome-wide association study (GWAS) (n=597). From the 286 controls, 154 subjects (54%; 77 per area) and from the 311 cases, 176 subjects (57%; 88 per area) were randomly selected for As speciation. Urinary As speciation was carried out using Agilent 1260 HPLC BIO inert, Agilent ICP-MS 7700x (HPLC-ICP-MS). Serum fluoride was analyzed using flow injection analysis with a fluoride ion-selective electrode. Urinary creatinine concentrations, urinary AIM and HbAlc were measured at Special Reference Laboratory (Tokyo, Japan). The detection limits for As species are 0.02μg/l and standard reference materials (Seronorm Trace Elements Urine L-l and L-2, SERO AS, Billingstad, Norway) were analyzed in every analytical batch. Coefficients of variance (n=26) ranged from 4.2% (Cs) to 42.3% (Ni) in the Urine L-l analysis and from 3.8% (Tl) to 22.6% (Fe) in the Urine L-2 analysis.

Genome-wide association study and direct sequencing

DNA was extracted from the whole-blood samples using a QIAamp DNA Blood Mini Kit (Qiagen, Chatsworth, CA) for the GWAS. The GWAS was performed on 597 subjects (n=311 for cases; n=286 for controls) using Illumina 500 K Chips (Human 6l0-Quad-Custom vl.0 DNA Analysis Bead Chip; Illumina, San Diego, CA) according to standard protocols. Prior to the analysis, the data sets were filtered on the basis of the single nucleotide polymorphism (SNP) genotyping call rates (>99% completeness), sample call rates (>95%), non-deviation from the Hardy-Weinberg equilibrium (p<0.00l), and evidence of cryptic family relationships (Fig. 1C).

Principal component analysis showed no outliers and all subjects were confirmed to be of a single ethnicity (Fig. S1). A quantile-quantile plot revealed that the distribution of observed p values followed the expected distribution (Fig. S2). The inflation factor was calculated to be l.0. The analyses were based on an additive genetic model. Analyses were performed using SVS version 7.6.10, GoldenHelix software (Golden Helix, Inc., Bozeman, MT, USA). Values of p<5.0 × l0⁻⁸ were considered to show genome-wide significance and values of 5 × l0⁻⁸<p<5 × l0⁻⁶ were considered to show suggestive evidence of an association.

Direct sequencing of the SLC13A3 gene and TP53RK gene including the 5'-UTR, 3'-UTR, exons and 100-bp covering splicing donor and acceptor sites was performed using the Sanger method in eight randomly selected cases and eight randomly selected controls. Information on the primer sequences and PCR conditions will be provided upon request to the corresponding author.

Determination of the numbers of CAA repeats

We determined the numbers of CAA repeats in the 5' promoter region of SLC13A3 in all cases and controls. The regions around the CAA repeats were amplified by PCR using fluorescently tagged primers (forward, 5'-FAM-AGC CTG GGT AAC AGA GTG AGA-3'; reverse, 5'-TCC CTT TAA GAC CTC ATC ACC-3'). The expected size of PCR products was 215 + 3 n (where n=CAA repeat number). The sizes of amplification products were analyzed by DNA sequencer.

Statistics

Univariate and multiple logistic analyses for demographic factors, past events and life-style related factors were conducted to identify the risk factors. To show the distribution of urinary creatinine levels and each metal parameter, we presented means with standard deviations, medians, minimums, maximums and percentiles (25%, 75%). Some urinary parameters were not normally distributed and we computed the statistical significance of differences between cases and controls using both parametric (Student's t-test) and non-parametric (Kolmogorov-Smirnov) methods. A comparison of serum fluoride levels among different CKD stages (stages l-5) was made using the Student-Newman-Keuls (SNK) multiple range test. A p value of <0.05 following adjustment by Bonferroni correction, using the number of urinary metal parameters in multiple comparisons, was considered significant. All statistical procedures were performed using STATISTICA 64 (supplied by StatSoft, OK, USA).

Results

Population characteristics and characterization of risk factors

As shown in Figure 1B, we recruited 504 candidate control subjects. Blood pressure measurements eliminated 154 subjects owing to hypertension. A further 24 subjects were eliminated because of proteinuria and/or glycosuria. Seventeen subjects were eliminated owing to high HbAlc levels and 23 subjects were eliminated because of elevated serum creatinine or urinary AIM levels. In total, 43% of the 504 participants were eliminated.

The demographic and clinical characteristics of the selected cases and controls are summarized in Table 1. The CKDu cases were older and had stayed longer in the region than the controls. On average, the subjects in both groups were neither overweight nor underweight. We identified several risk factors (Table S1). The percentage of farmers was significantly higher in the cases (97.1%) than in the controls (69.9%). Some lifestyle-related factors that are commonly associated with farming occupation, such as smoking, betel chewing, tobacco chewing, and history of snake bites were also significantly more common in the cases than in the controls. The presence of CKDu family history was significantly higher in the cases than in the controls. The significant factors (p<0.05) identified in univariate analyses were further analyzed using multiple logistic analysis. This analysis revealed that farming, family history of CKDu, tobacco chewing and history of snake bites were independent risk factors (Table 2).

Metal concentrations in drinking water

Metal concentrations in drinking water samples collected from Medawachchiya and Girandurukotte are summarized in Table S2 together with the maximum allowable limits in drinking water recommended by the WHO (http://www.who.int/water_sanitation_health/publications/2011/dwq_guidelines/en/index.html) and/or United States Environmental Protection Agency (http://water.epa.gov/drink/contaminants/index.cfm). We further compared the metal concentrations with the Japanese water quality standards (Ministry of Environment, Japan, 2004; http://www.env.go.jp/en/water/wq/wp.pdf). According to the available recommendations, none of the metals, which have guideline values including nephrotoxic metals such as Cd, As, and Pb, were present at toxic concentrations in the drinking water consumed daily by the CKDu cases in Medawachchiya and Girandurukotte.

Metal concentrations in urine

In Table 3, we presented metal concentrations (μg/l) in urine in cases and controls. We could not conduct urine analyses for 10 cases and 10 controls due to inadequate sample volume. To decrease false positive rates, a significant level (p=0.003=0.05/18) was adjusted by numbers of metals examined (n=18). None of the 18 metal concentrations was higher in urines in cases than in controls, including nephrotoxic metals such as Pb, Cd, As, Ni, V, and Al. Reversely, several metals were higher in controls than in cases (p<0.001); Pb, Tl, Cs, Cd, Mo, Sr, Rb, Se, As, Ni, Co, V, and Al. Cases had significantly lower creatinine concentrations in urine (n=301: 0.72 ± 0.53 g/l) than controls (n=276: 1.18 ± 0.70 g/l) (p<0.0001), suggesting impaired urine concentration ability due to CKDu. Thus, we also investigated the metal concentrations in urine adjusted by urinary creatinine levels (Table S3). Creatinine-adjusted As, Zn, Cu, Co, Mn, and Al concentrations in urine were significantly higher in cases than in controls (p<0.05/18=0.003) in both parametric and non-parametric analyses (Table S3). However, it should be addressed that irrespective of the adjustment for creatinine level, urinary Cd concentration was significantly higher in controls than in cases (Table 3 and S3), indicating that Cd is not a causative factor for CKDu.

Table 1. Demographic and clinical characteristics of the study subjects

Parameter	Cases		Controls		p
	Mean (SD)	Min-Max	Mean (SD)	Min-Max
n	311	—	286	—	—
Age (yr)	46.6 (9.0)	16–68	41.1 (8.1)	20–59	<0.0001
Duration of stay in NCR (yr)	39.0 (12.3)	7–68	33.1 (12.1)	10–59	<0.0001
Body mass index (kg/m2)	21.1 (3.4)	13.6–35.7	23.0 (4.3)	15.6–43.7	<0.0001
Blood pressure (systolic) (mmHg)	124.0 (18.9)	65–220	128.0 (10.6)	90–139	0.0013
Blood pressure (diastolic) (mmHg)	71.0 (12.6)	35–125	74.6 (9.0)	47–89	<0.0001
Serum creatinine (mg/dl)	2.12 (1.10)	0.46–11.1	0.88 (0.11)	0.56–1.18	<0.0001
eGFR (ml/min/1.73 m2)	45.7 (23.1)	5.3–204.3	104.0 (17.0)	67.4–180.9	<0.0001
Urinary alpha-1-microglobulin (mg/l)	65.9 (52.3)	0.1–410.7	1.6 (2.4)	0.1–13.8	<0.0001
Glycated hemoglobin (HbA1c %)	5.5 (0.6)	3.7–9.8*	5.4 (0.5)	2.9–6.5	0.0052
Serum fluoride (μg/l)	86.4 (51.6)	12.7–367	35.5 (16.2)	9.5–99.0	<0.0001

eGFR, estimated glomerular filtration rate; HbA1c, glycated hemoglobin.

*  Six cases were diagnosed with diabetes after the diagnosis of CKDu.

Table 2. Multiple logistic analysis for selected demographic factors, life-style related factors and past events

Parameter	Odds ratio (95% CI)	p
Farming vs. other occupations	9.17 (20.00–4.18)	<0.0001
Family history of CKDu	2.66 (4.15–1.70)	<0.0001
Tobacco chewing	2.59 (4.02–1.68)	<0.0001
History of snake bites	1.93 (3.16–1.18)	0.0093
History of malaria	1.31 (1.96–0.87)	0.1925
Smoking	1.25 (1.89–0.83)	0.2782
Betel chewing	1.38 (2.38–0.80)	0.2520

CI, confidence interval; CKDu, chronic kidney disease of unknown etiology.

Table 3. Metal concentrations in urine

Meta	Cases (μg/l) n=301				Controls (μg/l) n=276				pa value	pb value
	Mean (SD)	Geometric Mean	Min−Max	25% / 50% / 75%	Mean (SD)	Geometric Mean	Min−Max	25% / 50% / 75%
Pb	0.94 (0.89)	0.69	0.05–8.42	0.39 / 0.68 / 1.16	1.88 (2.55)	1.45	0.21–31.8	0.96 / 1.43 / 2.23	<0.001*	<0.001*
Tl	0.18 (0.19)	0.12	0–1.10	0.06 / 0.11 / 0.22	0.38 (0.26)	0.30	0.01–1.73	0.2 / 0.32 / 0.49	<0.001*	<0.001*
Cs	3.09 (2.77)	2.18	0.25–20.3	1.13 / 2.16 / 4.23	5.81 (4.08)	4.57	0.14–23.6	2.92 / 4.87 / 6.99	<0.001*	<0.001*
Cd	0.47 (0.78)	0.25	0.02–7.17	0.11 / 0.25 / 0.52	1.76 (1.89)	0.78	0.03–8.24	0.23 / 0.78 / 3.1	<0.001*	<0.001*
Mo	55.6 (48.9)	38.4	2.36–314	19.4 / 42.6 / 75.9	105 (97.5)	73.7	4.62–972	39.3 / 80.8 / 141	<0.001*	<0.001*
Sr	132 (115)	94.4	8.74–650	55.4 / 97.3 / 172	247 (163)	191	16.1–829	119 / 226 / 332	<0.001*	<0.001*
Rb	1,380 (1,250)	1,020	185–9,860	575 / 1,030 / 1,780	2,360 (1620)	1,900	130–9,900	1,270 / 1,980 / 2,810	<0.001*	<0.001*
Se	16.8 (15.0)	11.5	0.37–101	6.11 / 12.8 / 22.5	29.8 (20.0)	23.6	2.00–124	15.9 / 25.4 / 38.2	<0.001*	<0.001*
As	33.2 (37.5)	20.3	1.49–259	10.2 / 21.8 / 40.6	44.4 (43.5)	29.6	1.22–277	18.1 / 30.2 / 55.4	<0.001*	<0.001*
Zn	374 (322)	279	36.2–2,130	163 / 276 / 476	428 (453)	304	24.6–4,210	180 / 315 / 522	0.1	>0.10
Cu	16.2 (15.5)	12.0	0.99–153	7.70 / 11.9 / 19.3	15.7 (11.6)	12.7	1.41–116	8.39 / 13.1 / 19.9	0.628	>0.10
Ni	4.04 (3.06)	3.11	0.18–23.2	1.79 / 3.18 / 5.27	5.52 (4.19)	4.51	0.45–37.9	3.07 / 4.54 / 6.92	<0.001*	<0.001*
Co	0.42 (0.41)	0.30	0.04–3.30	0.17 / 0.32 / 0.48	0.55 (0.48)	0.42	0.03–3.28	0.26 / 0.42 / 0.65	<0.001*	<0.001*
Fe	10.8 (24.4)	5.83	0.04–280	3.37 / 5.89 / 10.4	11.5 (46.9)	6.82	1.36–773.18	4.3 / 6.59 / 9.86	0.817	<0.025
Mn	0.68 (1.14)	0.51	0.07–19.0	0.34 / 0.53 / 0.75	0.84 (1.97)	0.59	0.08–31.9	0.38 / 0.60 / 0.90	0.213	<0.05
Cr	0.19 (0.30)	0.12	0–3.62	0.08 / 0.13 / 0.19	0.24 (0.78)	0.15	0.01–12.63	0.10 / 0.15 / 0.23	0.291	<0.005
V	0.46 (0.31)	0.36	0.02–1.66	0.24 / 0.41 / 0.61	0.67 (0.31)	0.58	0.01–2.67	0.46 / 0.68 / 0.86	<0.001*	<0.001*
Al	4.60 (3.61)	3.70	0.31–30.0	2.54 / 3.67 / 5.52	6.60 (8.51)	4.95	0.03–114.98	3.37 / 4.78 / 7.39	<0.001*	<0.001*

^a  p value calculated by t-test;

^b  p value calculated by Kolmogorov-Smirnov (non-parametric) test;

*  Significant by Bonferroni correction (p=0.05/18=0.003).

In terms of As, our results were much lower than the reported concentrations in As-contaminated areas in countries such as Bangladesh and Taiwan^{10, 11)}. However, the total As concentrations were relatively higher than the concentrations observed in some uncontaminated areas in Europe, the United States, and Canada (http://www.hc-sc.gc.ca/ewh-semt/alt_formats/hecs-sesc/pdf/pubs/contaminants/chms-ecms/report-rapport-eng.pdf)^{12, 13)}. Therefore, we conducted speciation to investigate a possibility that modest contamination with As may be associated with CKDu. A total of 330 urine samples were duplicated and analyzed at two independent laboratories for total As and As speciation. The sum of arsenic speciation components (As-III, As-V, MMA, DMA, AB, Table 4) measured by HPLC ICP-MS nicely matched the total As levels measured by ICP-MS (Table 3 and Table 4, and Fig. S3: R²=0.97, p<0.0001). As speciation results revealed that 75–80% of total urinary As in both cases and controls was in the form of arsenobetaine (Table 4), which is non-toxic. Adjustment with creatinine showed that total arsenic concentration (μg/g-Cr) in urine was greater in cases than in controls. However, the difference of the total arsenic concentration was mostly attributable to the difference in arsenobetaine fraction (Table 4). Hata et al (2007) reported that the median total urinary arsenic level of a Japanese healthy male adult is 141.3 μg/l and the arsenobetaine level is 61.3 μg/l; these levels are about three times higher than those of CKDu cases⁹⁾. The modestly elevated levels of arsenobetaine in urine are in accord with the common habit of eating dry fish in these areas, as confirmed by the field study. Typical signs of chronic As exposure, such as skin keratosis and pigmentation, were observed in neither the cases nor the controls.

Table 4. As speciation results unadjusted and adjusted by urinary creatinine levels

Component	Controls ( n=154			Cases (n=176)			pa value	pb value
	Mean (SD)	Min−Max	25% / 50% / 75%	Mean (SD)	Min−Max	25% / 50% / 75%
Unadjusted for creatinine
As (III) (μg/l)	0.71 (0.94)	<0.02–9.71	0.14 / 0.59 / 1.05	0.39 (2.16)	<0.02–28.0	<0.02 / 0.10 / 0.27	0.095	<0.05
As (V) (μg/l)	0.22 (0.36)	<0.02–3.42	0 / 0.14 / 0.27	0.14 (0.24)	<0.02–2.42	<0.02 / 0.10 / 0.18	<0.05	>0.1
(μg/l)	0.57 (0.57)	<0.02–5.09	0.23 / 0.46 / 0.73	0.56 (0.58)	<0.02–4.49	0.21 / 0.37 / 0.62	0.859	>0.1
DMA (μg/l)	7.83 (7.75)	0.49–65.9	3.63 / 5.94 / 9.21	6.39 (7.40)	0.41–46.4	2.16 / 4.60 / 7.19	0.087	<0.05
AB (μg/l)	33.0 (35.7)	<0.02–215	10.0 / 20.5 / 41.0	25.5 (33.4)	0.17–227	6.96 / 15.2 / 28.8	<0.05	<0.05
(μg/l)	42.4 (38.6)	1.62–237	16.5 / 29.4 / 52.5	33.0 (37.8)	1.77–234	9.78 / 22.2 / 40.8	<0.05	<0.05
UCr (μg/l)	1.19 (0.7)	0.06–3.87	0.68 / 1.08 / 1.50	0.72 (0.53)	0.06–4.11	0.34 / 0.59 / 0.97	<0.05	<0.05
Adjusted for creatinine
As (III) (μg/g-Cr)	0.82 (1.36)	ND-14.5	0.18 / 0.45 / 1.23	0.77 (5.25)	ND-69.4	ND / 0.18 / 0.42	0.404	<0.05
As (V) (μg/g-Cr)	0.28 (0.52)	ND-5.12	ND / 0.12 / 0.33	0.35 (0.88)	ND-9.08	ND / 0.14 / 0.34	<0.05	<0.1
MMA (μg/g-Cr)	0.54 (0.40)	ND-2.60	0.3 / 0.48 / 0.68	0.92 (0.67)	ND-3.91	0.47 / 0.72 / 1.19	<0.05	<0.05
DMA (μg/g-Cr)	7.05 (6.02)	1.14–58.7	3.98 / 5.42 / 8.41	9.43 (6.06)	1.52–35.5	5.44 / 7.97 / 11.8	<0.05	<0.05
AB (μg/g-Cr)	29.0 (31.5)	ND-258	11.1 / 21.6 / 33.7	40.2 (44.9)	1.08–275	13.4 / 24.5 / 45.9	<0.05	<0.10
Total As (μg/g-Cr)	37.7 (32.7)	4.1–270	18.8 / 30.9 / 43.4	51.7 (47.6)	6.04–303	21.1 / 36.2 / 59.2	0.095	<0.05

* Significant by Bonferroni correction.

^a  p value calculated by t-test;

^b  p value calculated by Kolmogorov-Smirnov (non-parametric) test. MMA, Monomethylarsonic acid; DMA, Dimethylarsinic acid; AB, Arsenobetaine; UCr, urinary creatinine; ND, not detected.

Fluoride concentration in serum

An inverse relationship was observed between the eGFR and the serum fluoride concentration (Fig. S4). However, the eGFR stratified analysis demonstrated that serum fluoride levels were not different between cases and controls (Table S4) with eGFR >60 ml/ min/1.73 m². Moreover, the serum fluoride concentrations in the majority of the unaffected and early-stage CKDu patients were below the reported upper normal concentration of 50 μg/l¹⁴⁾. Thus, we concluded that higher serum fluoride is a result of lowered eGFR (impaired renal function) and not the primary cause of CKDu.

GWAS

A total of 543,848 GWAS SNPs in 577 samples were left after the quality control process for the associations test with the eGFR (quantitative trait locus (QTL) analysis) or the affected status (dichotomous analysis) (Fig. 1C). A genome-wide view of the single point association data is shown in Fig. 1D.

A summary of the association results (QTL analysis and dichotomous analysis, with and without age adjustment) for 10 candidate SNPs located in four genes is shown in Table 5 (see also Fig. 1D). The top hit that reached the genome-wide significance level (p<5 × 10⁻⁸) in both the QTL analysis and dichotomous analysis was rs6066043 (G>A), which is localized on the SLC13A3 gene encoding sodium-dependent dicarboxylate transporter 3 (p=5.23 × 10⁻⁹ for age-adjusted QTL analysis; p=3.73 × 10⁻⁸ for age-adjusted dichotomous analysis). The population attributable fraction and odds ratio for this SNP were 50% and 2.13. The allele frequencies for this SNP vary among populations. The G allele is more common among European populations (88% G allele frequency), while it is found in less than 50% of individuals in Asian populations (HapMap data). In the present study, the G allele frequencies in the cases and controls were 77% and 62%, respectively.

Table 5. Ten SNPs with significant and suggestive evidence of associations identified in the genome-wide association study

SNP	Chr:position	Alleles	Gene	Distance	MAF (number)	MAF (number)	Slope (OR)	p value for QTL analysis (dichotomous analysis)
					Cases	Controls		Without age adjustment	With age adjustment
1 rsl7126268	1:102498987	T>C	OLFM3	264	0.03 (Wild=281, HT=19, Homo=1)	0.10 (Wild=225, HT=48, Homo=3)	17.42 (2.78)	7.19 × 10−7 (5.29 × 10−6)	1.58 × 10−5 (9.64 × 10−5)
2 rs7539242	1:102499390	G>A	OLFM3	264	0.03 (Wild=281, HT=19, Homo=1)	0.10 (Wild=225, HT=48, Homo=3)	17.42 (2.78)	7.19 × 10−7 (5.29 × 10−6)	1.58 × 10−5 (9.64 × 10−5)
3 rs7539242	1:102501863	C>T	OLFM3	266	0.03 (Wild=281, HT=19, Homo=1)	0.10 (Wild=225, HT=48, Homo=3)	17.42 (2.78)	7.19 × 10−7 (5.29 × 10−6)	1.58 × 10−5 (9.64 × 10−5)
4 rs12120258	1:102507539	G>T	OLFM3	272	0.03 (Wild=281, HT=19, Homo=1)	0.10 (Wild=225, HT=48, Homo=3)	17.42 (2.78)	7.19 × 10−7 (5.29 × 10−6)	1.58 × 10−5 (9.64 × 10−5)
5 rs12140211	1:102512105	C>T	OLFM3	276	0.07 (Wild=259, HT=41, Homo=1)	0.14 (Wild=203, HT=68, Homo=5)	13.71 (2.00)	9.54 × 10−7 (5.21 × 10−5)	1.48 × 10−5 (7.59 × 10−4)
6 rs2980098	4:4303857	A>G	TMEM128	13	0.37 (Wild=120, HT=138, Homo=43)	0.49 (Wild=73, HT=138, Homo=65)	9.11 (1.61)	3.80 × 10−6 (1.00 × 10−4)	2.16 × 10−6 (1.22 × 10−4)
7 rs10099338	8:9295149	A>G	LOC157273	65	0.05 (Wild=273, HT=28, Homo=0)	0.10 (Wild=221, HT=53, Homo=2)	13.28 (2.04)	4.05 × 10−6 (1.00 × 10−4)	4.50 × 10−4 (4.33 × 10−3)
8 rs1004571	20:44711576	A>G	SLC13A3	Intragenic	0.32 (Wild=32, HT=129, Homo=140)	0.44 (Wild=51, HT=142, Homo=83)	7.98 (1.63)	4.33 × 10−6 (2.78 × 10−5)	6.23 × 10−5 (1.73 × 10−4)
9 rs6066043	20:44721860	G>A	SLC13A3	Intragenic	0.23 (Wild=175, HT=113, Homo=13)	0.38 (Wild=104, HT=132, Homo=40)	12.17 (2.13)	2.40 × 10−9 (1.13 × 10−8)	5.23 × 10−9 (3.73 × 10−8)
10 rs4810537	20:44740335	A>G	SLC13A3	Intragenic	0.33 (Wild=32, HT=137, Homo=132)	0.45 (Wild=54, HT=142, Homo=80)	8.02 (1.60)	3.47 × 10−6 (2.78 × 10−5)	6.08 × 10−5 (2.67 × 10−4)

Chr, chromosome; MAF, minor allele frequency; OR, odds ratio; HT, heterozygous; SNP, single nucleotide polymorphism; QTL, quantitative trait locus, p values for both the QTL analysis and dichotomous analysis are shown. Reference sequence: GRCh36 (2006).

Since the SNP rs6066043 is located in the intergene region of SLC13A3 and TP53RK (Fig. S5), it is hard to explain the functional impairment by this SNP. We thus screened for variants in these two genes. Direct sequencing was conducted for all coding exons and the 5'UTR and 3'UTR for these two genes. The results of direct sequencing are shown in Table S5. We identified variations in the CAA repeat number in the 5' promoter region, 14.6 Kb upstream of exon 1 of SLC13A3 (p<0.01). Minor allele frequencies of other variants did not seem to differ in cases and controls. Thus, the sizes of the CAA repeat expansion were determined using fluorescently tagged primers in all cases and controls and this revealed variation in the CAA repeat expansion from (CAA)₁₀ to (CAA)₁₈. We investigated the association between repeat numbers and CKDu. Considering (CAA)₁₅ as the wild-type, an odds ratio was calculated for the repeat polymorphism (Table S6) to be 1.6, being smaller than odds ratio of rs6066043 (2.13) (Table 5). No significant difference in mean eGFR was observed among various CAA repeat combinations (Table S6). Thus, we considered that (CAA) repeats were not a causative variant for CKDu.

The GWAS study revealed three other genes with suggestive evidence (5 × 10⁻⁸<p<5 × 10⁻⁶) of an association: OLFM3, TMEM128 and LOC157273 (Table 5). However, OLFM3 can be discarded because it is an olfactory gene.

Discussion

This is the first comprehensive study addressing both the genetic, social, and environmental contributors of CKDu with tubulointerstitial damages in NCR of Sri Lanka. Ecological investigations in the present study confirmed the absence of nephrotoxic heavy metal contamination in drinking water. We found that the SNP rs6066043 of SLC13A3 loads discernible risk of CKDu in the local population as indicated by a large population attributable risk of 50%. It should also be addressed that occupation, family history of CKD, tobacco chewing, and history of snakebites were also identified as independent risk factors.

None of 18 metals had urinary concentrations (μg/l) greater in cases than in controls. Only when adjusted to the urinary creatinine levels (μg/g-cre), it was found that the Zn, Cu, Co, Mn, and Al levels in urine were higher in cases than in controls. However, comparison of these levels with those in other populations is difficult due to the limited availability of published data. We considered that due to the impaired urine concentration ability in cases, those apparent elevations of urine concentrations for some metals might emerge by adjustment with creatinine levels. Further studies are needed to verify the individual and combined effects of these metal parameters on CKDu.

It is worthwhile mentioning that urinary concentrations of total As levels in cases and controls were suggestive of modest exposure to As. Thus, we conducted speciation of As compounds in urine. Speciation revealed that arsenobetaine is the major As compound, suggesting that seafood is the major sources of As¹³⁾. The high urinary As concentrations in countries like Japan and Korea were found to result from the high amounts of seafood intake, and As speciation studies showed high proportions of arsenobetaine in urine, which is non toxic^{9, 15)}. A field survey on the food habits of the population in NCR also supports regular intake of dried sea fish which is likely to be the source of high urinary arsenobetaine levels.

It is worth highlighting the fact that some studies have reported a possible relationship between As exposure and CKD^{16, 17)}. Hsueh et al. (2009) reported a positive association between urinary total As concentration and CKD for the first time in Taiwan, and the total As concentrations in their study population closely resemble the observations in the present study (total mean urinary As in the CKD group and healthy controls: 31.95 and 20.71 versus 50.3 and 39.4 ßg/gCr, respectively)¹⁷⁾. However, the dominant component of total As was arsenobetaine in this study, but dimethylarsinic acid in the Taiwan study. Urinary Cd levels in controls were higher than those in cases, suggesting a minimal involvement of Cd in CKDu. The Cd levels reported in this study are comparable to other published data^{18, 19)}.

The present study revealed that a common genetic variant close to SLC13A3 is associated with CKDu. SLC13A3 encodes the high-affinity sodium dicarboxylate transporter 3 located in the basolateral membranes of human renal proximal tubules, liver, brain, and placenta²⁰⁾. On the other hand, TP53RK, encodes a protein, TP53-regulated kinase, which is a serine/threonine protein kinase²¹⁾ and functions in a P53 signal pathway. SLC13A3 has been identified as one of the most sensitive marker genes for predicting the clinical course to end-stage renal disease in type 2 diabetes mellitus and has an association with blood pressure^{22, 23)}. Recently, SLC13A3 was identified as one of the 43 genes that can be used in protein expression signatures to predict progressive renal fibrosis in mice, and was suggested to be a potentially useful molecular predictor for CKD progression in humans²⁴⁾. SLC13A3 has also been shown to play a role in the accumulation of Hg-thiol conjugates in the basolateral membrane vesicles in proximal tubular cells in the rat²⁵⁾. Those pieces of evidence strongly support that SLC13A3 is more likely mechanistically associated with CKD rather than TP53RK. However, further studies are needed to explain a mechanism by which CKDu is associated with rs6066043.

It was interesting to note that 43% of the apparently healthy male population, most of whom were farmers, had, at least, one undiagnosed non communicable disease, mainly high blood pressure. Mendis et al. (1988) reported that the prevalence of high blood pressure among Sri Lankans is one of the highest reported in the world, and Wijewardene et al (2005) reported that the area in Sri Lanka with the highest prevalence of high blood pressure is the Uva province^{26, 27)}. One of our study areas (Girandurukotte) is located in this province. Therefore, it is likely that undiagnosed hypertension may accelerate CKDu progression. Even in one of our previous studies, we found hypertension as a risk factor for CKDu progression²⁶⁾.

Some limitations of this study warrant mention. We used the eGFR based on the MDRD formula rather than using the direct glomerular filtration rate, since it was not feasible to measure the direct glomerular filtration rate in this setting. However, the MDRD formula is not validated for the Sri Lankan population. Owing to the discrepancies present in the validated MDRD formulas for other Asian populations, we could not use these coefficients for this population^29–32). A smaller sample size without replication evidence may decrease the robustness of the results, and we have planned to incorporate it in the next stage of this study. Although we confirm the farmer preponderance among CKDu patients, this finding needs cautious interpretation. Given that we selected patients by a case-series method to eliminate possible biases, there may be selection bias in selecting controls. We eliminated 43% of controls due to undiagnosed non-communicable diseases, most of them were farmers. Such an elimination of a large portion of the population from the control candidates suggests that local farmers have less chance to access medical care or preventive care than controls with other occupations. Thus, our conclusion on being a farmer as a risk of CKDu is very likely confounded by the accessibility to medical care or public health services. On the other hand, the elimination of such subjects from control candidates strengthened our genetic analysis. Because we primarily aimed to determine genetic factors for tubulointerstitial damages, we selected cases who developed CKDu in the course of primary tubulointerstitial damages without hypertension or diabetes mellitus or other comorbidities. To unify the selection criteria, controls should be selected with the same backgrounds. Applying the same backgrounds made our genetic analysis for tubulointerstitial damages scientifically sound. If we included controls with hypertension or other morbidities, phenotypic heterogeneity between cases and controls may contaminate our study.

In conclusion, our results do not support the involvement of a founder mutation or single heavy metal in the pathogenesis of CKDu. However, the present study found a significant SNP in SLC13A3 with an odds ratio of 2.13 and a 50% population attributable fraction, indicating major genetic susceptibility to CKDu. Thus, further study is warranted to elucidate causal link of the SNP rs6066043 with CKDu.

Acknowledgments: We thank Dr Meiko Takahashi (Kyoto University) for support in performing the GWAS, and the staff of the Medawachchiya and Girandurukotte renal clinics for their assistance with the sampling program. The authors would like to acknowledge the Special Coordination Funds for Promoting Science and Technology from the Ministry of Education, Culture, Sports, Science and Technology in Japan. The funding agency had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

• 1)   Athuraliya  TN,  Abeysekera  DT,  Amerasinghe  PH,  Kumarasiri  PV,  Dissanayake  V. Prevalence of chronic kidney disease in two tertiary care hospitals: high proportion of cases with uncertain etiology Ceylon Med J 2009; 54: 23-5.
• 2)   Nanayakkara  S,  Komiya  T,  Ratnatunga  N, et al. Tubulointerstitial damage as the major pathological lesion in endemic chronic kidney disease among farmers in North Central Province of Sri Lanka Environ Health Prev Med 2012; 17: 213-21.
• 3)   Nanayakkara  S,  Senevirathna  ST,  Karunaratne  U, et al. Evidence of tubular damage in the very early stage of chronic kidney disease of uncertain etiology in the North Central Province of Sri Lanka: a crosssectional study Environ Health Prev Med 2012; 17: 109-17.
• 4)   Bandara  JM,  Senevirathna  DM,  Dasanayake  DM,  Herath  V,  Abeysekara  T,  Rajapaksha  KH. Chronic renal failure among farm families in cascade irrigation systems in Sri Lanka associated with elevated dietary cadmium levels in rice and freshwater fish (Tilapia) Environ Geochem Health 2008; 30: 465-78.
• 5)   Chandrajith  R,  Nanayakkara  S,  Itai  K, et al. Chronic kidney diseases of uncertain etiology (CKDue) in Sri Lanka: geographic distribution and environmental implications Environ Geochem Health 2011; 33: 267-78.
• 6)   Jayasumana  MA CS,  Pa ranagama  PA,  Amarasinghe  MD, et al. Possible link of Chronic arsenic toxicity with chronic kidney disease of unknown etiology J Nat Sci Res 2011; 13: 64-73.
• 7)   Athuraliya  NT C,  Ab eysekera  TD J,  Am erasinghe  PH, et al. Uncertain etiologies of proteinuric-chronic kidney disease in rural Sri Lanka Kidney Int 2011; 80: 1212-21.
• 8)   Kooman  JP. Estimation of renal function in patients with chronic kidney disease J Magn Reson Imaging 2009; 30: 1341-6.
• 9)   Hata  A,  Endo  Y,  Nakajima  Y, et al. HPLC-ICP-MS speciation analysis of arsenic in urine of Japanese subjects without occupational exposure J Occup Health 2007; 49: 217-23.
• 10)   Chen  Y,  van  Ge en  A,  Graziano  JH, et al. Reduction in urinary arsenic levels in response to arsenic mitigation efforts in Araihazar, Bangladesh Environ Health Perspect 2007; 115: 917-23.
• 11)   Chen  JW,  Chen  HY,  Li  WF, et al. The association between total urinary arsenic concentration and renal dysfunction in a community-based population from central Taiwan Chemosphere 2011; 84: 17-24.
• 12)   Caldwell  KL,  Jones  RL,  Verdon  CP,  Jarrett  JM,  Caudill  SP,  Osterloh  JD. Levels of urinary total and speciated arsenic in the US population: National Health and Nutrition Examination Survey 2003-2004 J Expo Sci Environ Epidemiol 2009; 19: 59-68.
• 13)   Saoudi  A,  Zeghnoun  A,  Bidondo  ML, et al. Urinary arsenic levels in the French adult population: the French National Nutrition and Health Study, 2006-2007 Sci Total Environ 2012; 433: 206-15.
• 14)   Joshi  S,  Hlaing  T,  Whitford  GM,  Compston  JE. Skeletal fluorosis due to excessive tea and toothpaste consumption Osteoporos Int 2011; 22: 2557-60.
• 15)   Lee  JW,  Lee  CK,  Moon  CS, et al. Korea National Survey for Environmental Pollutants in the Human Body 2008: heavy metals in the blood or urine of the Korean population Int J Hyg Environ Health 2012; 215: 449-57.
• 16)   Zheng  LY,  Umans  JG, Tellez- Plaza  M, et al. Urine arsenic and prevalent albuminuria: evidence from a population-based study Am J Kidney Dis 2013; 61: 385-94.
• 17)   Hsueh  YM,  Chung  CJ,  Shiue  HS, et al. Urinary arsenic species and CKD in a Taiwanese population: a case-control study Am J Kidney Dis 2009; 54: 859-70.
• 18)   Huang  M,  Choi  SJ,  Kim  DW, et al. Evaluation of factors associated with cadmium exposure and kidney function in the general population Environ Toxicol 2013; 28: 563-70.
• 19)   Batariova  A,  Spevackova  V,  Benes  B,  Cejchanova  M,  Smid  J,  Cerna  M. Blood and urine levels of Pb, Cd and Hg in the general population of the Czech Republic and proposed reference values Int J Hyg Environ Health 2006; 209: 359-66.
• 20)   Pajor  AM. Molecular properties of the SLC13 family of dicarboxylate and sulfate transporters Pflugers Arch 2006; 451: 597-605.
• 21)   Miyoshi  A,  Kito  K,  Aramoto  T,  Abe  Y,  Kobayashi  N,  Ueda  N. Identification of CGI-121, a novel PRPK (p53-related protein kinase)-binding protein Biochem Biophys Res Commun 2003; 303: 399-405.
• 22)   Bento  JL,  Palmer  ND,  Zhong  M, et al. Heterogeneity in gene loci associated with type 2 diabetes on human chromosome 20q13.1. Genomics 2008; 92: 226-34.
• 23)   Simino  J,  Shi  G,  Arnett  D,  Broeckel  U,  Hunt  SC,  Rao  DC. Variants on chromosome 6p22.3 associated with blood pressure in the HyperGEN study: follow-up of FBPP quantitative trait loci. Am J Hypertens 2011; 24: 1227-33.
• 24)   Ju  W,  Eichinger  F,  Bitzer  M, et al. Renal gene and protein expression signatures for prediction of kidney disease progression Am J Pathol 2009; 174: 2073-85.
• 25)   Lash  LH,  Hueni  SE,  Putt  DA,  Zalups  RK. Role of organic anion and amino acid carriers in transport of inorganic mercury in rat renal basolateral membrane vesicles: influence of compensatory renal growth Toxicol Sci 2005; 88: 630-44.
• 26)   Mendis  S,  Ranasinghe  P,  Dharmasena  BD. Prevalence of hypertension in Sri Lanka A large population study in the central province. Public Health 1988; 102: 455-62.
• 27)   Wijewardene  K,  Mohideen  MR,  Mendis  S, et al. Prevalence of hypertension, diabetes and obesity: baseline findings of a population based survey in four provinces in Sri Lanka Ceylon Med J 2005; 50: 62-70.
• 28)   Senevirathna  L,  Abeysekera  T,  Nanayakkara  S, et al. Risk factors associated with disease progression and mortality in chronic kidney disease of uncertain etiology: a cohort study in Medawachchiya, Sri Lanka Environ Health Prev Med 2012; 17: 191-8.
• 29)   Praditpornsilpa  K,  Townamchai  N,  Chaiwatanarat  T, et al. The need for robust validation for MDRD-based glomerular filtration rate estimation in various CKD populations Nephrol Dial Transplant 2011; 26: 2780-5.
• 30)   Imai  E,  Horio  M,  Nitta  K, et al. Estimation of glomerular filtration rate by the MDRD study equation modified for Japanese patients with chronic kidney disease Clin Exp Nephrol 2007; 11: 41-50.
• 31)   Ma  YC,  Zuo  L,  Chen  JH, et al. Modified glomerular filtration rate estimating equation for Chinese patients with chronic kidney disease J Am Soc Nephrol 2006; 17: 2937-44.
• 32)   Matsuo  S,  Imai  E,  Horio  M, et al. Revised equations for estimated GFR from serum creatinine in Japan Am J Kidney Dis 2009; 53: 982-92.