ORIGINAL RESEARCH published: 26 February 2019 doi: 10.3389/fgene.2019.00118 Association of Candidate Gene Polymorphisms With Chronic Kidney Disease: Results of a Case-Control Analysis in the Nefrona Cohort Edited by: Joan Valls 1† , Serafí Cambray 2† , Carles Pérez-Guallar 1† , Milica Bozic 2 , Kelli K. Ryckman, The University of Iowa, United States Marcelino Bermúdez-López 2 , Elvira Fernández 2 , Àngels Betriu 2 , Isabel Rodríguez 3* and José M. Valdivielso2* on behalf of the NEFRONA Investigators Reviewed by: Bethany Wolf, 1 Biostatistics Unit, Institut de Recerca Biomèdica de Lleida and Redes – Instituto de Salud Carlos III, Lleida, Spain, Medical University of South Carolina, 2 Vascular and Renal Translational Research Group, Biomedical Research Institute, Institut de Recerca Biomèdica de Lleida United States and RedinRen-ISCIII, Lleida, Spain, 3 Bone and Mineral Research Unit, RedinRen-ISCIII, Hospital Universitario Central Martha Guevara-Cruz, de Asturias, Instituto de Investigación Sanitaria del Principado de Asturias, Universidad de Oviedo, Oviedo, Spain Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (INCMNSZ), Mexico Chronic kidney disease (CKD) is a major risk factor for end-stage renal disease, *Correspondence: cardiovascular disease and premature death. Despite classical clinical risk factors Isabel Rodríguez for CKD and some genetic risk factors have been identified, the residual risk isabelrodriguez2710@gmail.com José M. Valdivielso observed in prediction models is still high. Therefore, new risk factors need to be valdivielso@irblleida.cat; identified in order to better predict the risk of CKD in the population. Here, we valdivielso@medicina.udl.es analyzed the genetic association of 79 SNPs of proteins associated with mineral † These authors have contributed equally to this work metabolism disturbances with CKD in a cohort that includes 2,445 CKD cases and 559 controls. Genotyping was performed with matrix assisted laser desorption ionization– Specialty section: time of flight mass spectrometry. We used logistic regression models considering This article was submitted to Applied Genetic Epidemiology, different genetic inheritance models to assess the association of the SNPs with a section of the journal the prevalence of CKD, adjusting for known risk factors. Eight SNPs (rs1126616, Frontiers in Genetics rs35068180, rs2238135, rs1800247, rs385564, rs4236, rs2248359, and rs1564858) Received: 10 July 2018 were associated with CKD even after adjusting by sex, age and race. A model containing Accepted: 04 February 2019 Published: 26 February 2019 five of these SNPs (rs1126616, rs35068180, rs1800247, rs4236, and rs2248359), Citation: diabetes and hypertension showed better performance than models considering only Valls J, Cambray S, clinical risk factors, significantly increasing the area under the curve of the model Pérez-Guallar C, Bozic M, Bermúdez-López M, Fernández E, without polymorphisms. Furthermore, one of the SNPs (the rs2248359) showed an Betriu À, Rodríguez I and interaction with hypertension, being the risk genotype affecting only hypertensive Valdivielso JM (2019) Association patients. We conclude that 5 SNPs related to proteins implicated in mineral metabolism of Candidate Gene Polymorphisms With Chronic Kidney Disease: Results disturbances (Osteopontin, osteocalcin, matrix gla protein, matrix metalloprotease 3 and of a Case-Control Analysis 24 hydroxylase) are associated to an increased risk of suffering CKD. in the Nefrona Cohort. Front. Genet. 10:118. Keywords: chronic kidney disease, risk factors, genetic association study, single nucleotide polymorphism, doi: 10.3389/fgene.2019.00118 linkage disequilibrium, haplotype Frontiers in Genetics | www.frontiersin.org 1 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease INTRODUCTION and, furthermore, are associated with CKD progression (Schwarz et al., 2006). Thus, CKD-MBD alterations are not only a Chronic kidney disease (CKD) is a major risk factor for end-stage consequence of CKD, but can be also playing a role in the renal disease, cardiovascular disease and premature death (Go causative pathway. et al., 2004; van der Velde et al., 2011). Nowadays, its worldwide Therefore, in the present study, we assessed the association of prevalence is estimated in 7.2%, increasing dramatically among 79 SNPs of genes reported in the literature as being associated elderly people (Zhang and Rothenbacher, 2008), so it is predicted to CKD-MBD and vascular calcification, (Rong et al., 2014; that CKD will become a highly prevalent disease due to Tuñón-Le Poultel et al., 2014; Ammirati et al., 2015; Higgins population aging. Indeed, the Global Burden of Diseases, Injuries, et al., 2015; Xu and Sun, 2015) with CKD prevalence in the and Risk Factors Study 2015 calculated that, in absolute numbers, NEFRONA cohort. its worldwide prevalence increased a 26.9% over the last 10 years (2005–1015), and a similar increase has been observed in the average years living with disability of CKD patients, which MATERIALS AND METHODS increased a 23.8% (GBD 2015 Disease and Injury Incidence and Prevalence Collaborators, 2016). Thus, considered as a major Study Design and Participants public health problem, international efforts are needed for CKD The NEFRONA study is a prospective, multicenter, observational prevention, early detection and treatment. Different risk factors study, in which 2,445 CKD subjects were recruited from 81 for CKD have been successfully identified, being the leading ones nephrology services and dialysis units throughout Spain, from diabetes, hypertension and obesity, but also including gender, age October 2010 to June 2012 (Junyent et al., 2010). Patients and smoking (Kazancioğlu, 2013). between 18 and 74 years of age were eligible if they had CKD Currently, it is accepted that CKD and its associated risk stage 3 or higher as defined by current guidelines [glomerular factors present a considerable genetic component. Nowadays, filtration rate below 60 mL/min/1.73 m2 estimated using the Genome Wide Association Studies (GWAS) have identified Modification of Diet in Renal Disease (MDRD) 4 equation (Levey novel genetic variants associated with CKD, its progression, et al., 1999; Stevens et al., 2006)]. During the same period, 559 and its associated pathologies (Wuttke and Köttgen, 2016; controls (MDRD4 above 60 mL/min/1.73 m2 ) were recruited Limou et al., 2018). Some of these SNPs are located in genes from Primary Care centers. Information about classical risk coding for proteins whose alteration could be causative of factors for CKD (diabetes, hypertension, and tobacco) as well as CKD. For example, Köttgen et al. (2009) found association age and sex were recorded. of the UMOD gene with CKD; mutations on this gene were Exclusion criteria for both groups included: pregnancy, active associated to rare autosomal dominant tubulointerstitial disease infections, life expectancy lower than 12 months, history of a that leads to CKD (Devuyst and Pattaro, 2018). Later, in previous cardiovascular event, carotid artery surgery or any organ 2013, it was demonstrated that these polymorphisms lead transplantation. All patients signed an informed consent and the to increased uromodulin expression, and mice overexpressing local Ethics Committee of each hospital approved the protocol. UMOD showed salt-sensitive hypertension and age-dependent renal lesions (Trudu et al., 2013). SNP Selection and Genotyping Due to this huge bulk of knowledge and the clear association We used blood samples from the NEFRONA study stored in of some polymorphisms to CKD (Köttgen et al., 2009, 2010; the Biobank of the REDinREN (Calleros-Basilio et al., 2016) Pattaro et al., 2016; Parsa et al., 2017), efforts have been focused in Alcala de Henares, Madrid, for DNA extraction from all on designing test for CKD diagnosis or prognosis based on these 2,445 CKD cases and 559 controls. Seventy-nine SNPs located findings but, to this day, none has reached the clinical practice in twenty-nine genes (Supplementary Table S1) related with (Ma et al., 2017; Thio et al., 2018). Indeed, with vast number of CKD-MBD and vascular calcification that have been reported markers in GWAS analyses, true ‘hits’ may become lost in a sea in the scientific literature (Rong et al., 2014; Tuñón-Le Poultel of false positives, due to the false discovery rate used to adjust et al., 2014; Ammirati et al., 2015; Higgins et al., 2015; Xu and p-values. The candidate gene approach is another widely used Sun, 2015) were genotyped. We only considered those SNPs with option to assess genetic predisposition to diseases. Candidate potential functional implications, so only those present in coding, gene studies, although being incapable of discovering new genes promoter or 30 untranslated region were considered. or gene combinations, tend to have rather high statistical power, Genotyping was performed with matrix assisted laser and a scientific hypothesis behind its rationale. desorption ionization–time of flight mass spectrometry in the Mineral metabolism disturbances, defined as CKD mineral Sequenom MassARRAY platform , in CEGEN (Santiago de R bone disorder (CKD-MBD), are a very common complication Compostela). As a quality control, the clusters for each SNP were of CKD patients. The CKD-MBD is a disruption in the normal reviewed and the SNPs with low genotyping percentage or not interplay between the kidney, skeleton, and cardiovascular meeting Hardy Weinberg Equilibrium (HWE) were eliminated. system. Apart from biochemical and bone abnormalities, CKD- Assessment of HWE was performed by means of a chi-squared MBD is defined by an increase in vascular calcification. Those test and an exact test for each SNP, considering a SNP in disturbances occur very early in CKD [increases in Fibroblast HWE when this was not rejected by any test. Samples with Growth Factor 23 have been described from CKD stage 2 (Wolf, low genotyping percentage and with insufficient quality were 2012)], have a high impact in outcomes (Valdivielso et al., 2017) also discarded. We also checked consistency within samples Frontiers in Genetics | www.frontiersin.org 2 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease duplicating the same sample in the same plate and in different m the number of selected SNPs in the single association plates. In addition, a trio of samples from the Coriell Institute analysis. Once an initial set of associated SNPs was determined, biorepository, for which published data were available, were interaction effects were considered, both between SNPs and included in each chip to verify the effectiveness of the genotyping. between SNPs and CKD risk factors (diabetes and hypertension). A step-wise algorithm was again used to determine the final Statistical Analysis model including both SNPs and clinical risk factors, calculating The clinical heterogeneity between cases and controls was the corresponding p-values, OR and 95% confidence intervals. assessed considering demographic and clinical variables. Mean Potentially confounding variables were also included to calculate and standard deviation or absolute frequency and percentage adjusted p-values. were computed for quantitative and qualitative variables, Finally, sensitivity and specificity analyses were implemented respectively, assessing differences with Mann–Whitney’s U-test to compare the model with the SNPs, the model with and Fisher test. diabetes and hypertension and the model with all the Hardy Weinberg Equilibrium was checked for all SNPs variables, calculating the receiver operating characteristic using data from both controls and cases. Since HWE was (ROC) curves and the area under the curve (AUC). accepted for all SNPS, no specific evaluation was conducted to A bootstrap procedure, using 1,000 simulations, was used assess HWE only in controls. Genetic association was evaluated to obtain a robust estimation of the AUC and a confidence using logistic regression models with five different inheritance interval was computed using 2.5 and 97.5% percentiles models: dominant, recessive, codominant, overdominant and of the bootstrap distribution. DeLong tests were used to additive models. Akaike’s Information Criterion (AIC) was compare the curves among models and an optimal threshold used to select the model with the best fit and therefore for the predicted probability obtained from the logistic compute the corresponding p-value, odds-ratio (OR) and 95% multivariate model was determined and, thus, sensitivity and confidence intervals. A permutation p-value was also computed specificity with the corresponding 95% confidence intervals for those SNPs found significantly associated to CKD. For were calculated. this purpose, the likelihood ratio (LR) was assessed in 20,000 All statistical analyses were performed using R software (R random permutations of case and control labels, computing the Core Team, 2014), particularly using the packages SNPassoc, permutation p-value as the sample probability of obtaining a Genetics and haplo.stats. Threshold for statistical significance was LR greater or equal to the one initially obtained. Moreover, set at α = 0.05. potentially confounding variables were also included in each SNP model, performing a likelihood ratio test (LRT) to obtain an adjusted p-value. RESULTS To represent missing genotype data, SNPs were sorted by increasing percent of missing genotypes (Supplementary Heterogeneity Analysis in Case-Control Table S2). The percentage of missing genotype was lower than 7% Groups for 10 of the 12 SNPs that were significant in the single association Main clinical and demographic data are summarized in analysis (6.16% for rs35068180 and rs35068180; 6.19% for rs4236, Table 1. The CKD group of patients showed higher percentage rs731236 and rs9138, 6.26% for rs1126616 and rs2238135, 6.32% of hypertensives (91% in CKD; 35.4% in controls), and for rs679620, 6.39% for rs1800247 and rs385564), however, it was diabetics (25.4% in CKD; 10.7% in controls). Moreover, the quite high for two SNPs (46.3% for rs2248359 and 46.34% for CKD group also included a higher percentage of males, rs1564858). No imputation was considered for the analysis, so all smokers and of non-caucasians. Diabetes and hypertension of them were made on a complete-cases basis. Thus, for the single are well known causes of CKD and were included in association analysis, all the data cases available for each SNP were the model as potentially confounding variables. Sex, race used but, for the multivariate models a reduced data base with (Caucasian/non-Caucasian) and age were also included as 1603 out of 3004 cases (53.4%) containing complete data for all adjusting variables. 12 SNPS was considered. Linkage disequilibrium (LD) between all pairs of SNPs was Association of Single SNPs With CKD quantified assessing the two-loci r2 coefficient. The statistical All the evaluated SNPs met the HWE (Supplementary significance of the obtained r2 values was evaluated using the Table S3). Thirteen SNPs showed a 100% missing chi-square statistic for LD. genotypes in controls and were removed from the analysis Multiple logistic regression models were used to assess the (Supplementary Figure S1 and Supplementary Table S2). combined effect of single SNPs on CKD risk. All possible Thus, only 66 out of 79 initially selected SNPs were models containing any combination of the single SNPs found analyzed for its association to CKD. Then, we tested to be statistically significant, were fitted considering the different inheritance models for all the SNPs, and univariate inheritance setting determined in the single association analysis. analyses to assess its association with CKD under the Using a step-wise algorithm and the model with the lowest selected inheritance model (Supplementary Table S3). AIC was chosen.   Thetotal number of fitted models was After adjusting for confounding variables, only twelve m m m SNPs showed significant association to CKD (Table 2). + + ... + = (1 + 1)m − 1 = 2m − 1 being 1 2 m Complete genetic information about the SNPs is provided in Frontiers in Genetics | www.frontiersin.org 3 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease Supplementary Table S4. Minor allele frequencies (MAF) in the European populations analyzed in 1,000 Genomes of those SNPs in the Control group were not different Project (1000 Genomes Project Consortium et al., 2015) to those in the CKD group, neither to those included (Supplementary Table S5). TABLE 1 | Baseline demographic and clinical characteristics of the population. Total 3004 (100%) Controls 559 (18.6%) CKD 2445 (81.4%) p-value OR (95% CI) Sex 0.0003 Male 1806(60.1%) 298(53.3%) 1508(61.6%) 1.0 Female 1198(39.8%) 261(46.6%) 937(38.3%) 0.71 (0.59–0.85) Race 0.11 Caucasian 2910(96.9%) 549(98.2%) 2361(96.5%) 1.0 Black 17(0.57%) 0(0%) 17(0.7%) Inestimable Asian 9(0.3%) 0(0%) 9(0.37%) Inestimable Arabic 14(0.47%) 1(0.18%) 13(0.53%) 3.02 (0.6–54.98) Hispanic 54(1.8%) 9(1.61%) 45(1.84%) 1.16 (0.59–2.55) Caucasian Y/N 0.04 Caucasian 2910(96.87%) 549(98.2%) 2361(96.5%) 1.0 Non-Caucasian 94(3.1%) 10(1.8%) 84(3.4%) 1.95 (1.06–4.03) Smoker Y/N 0.07 No 1296(43.1%) 222(39.7%) 1074(43.9%) 1.0 Yes 1708(56.8%) 337(60.2%) 1371(56.0%) 0.84 (0.7–1.01) Smoker Status 0.09 Non smoker 1296(43.1%) 222(39.7%) 1074(43.9%) 1.25 (1.02–1.54) Former smoker 1109(36.9%) 228(40.8%) 881(36%) 1.0 Smoker 599(19.9%) 109(19.5%) 490(20%) 1.16 (0.9–1.5) Diabetes <0.00001 No 2323(77.3%) 499(89.2%) 1824(74.6%) Yes 681(22.6%) 60(10.7%) 621(25.4%) 2.83 (2.15–3.79) Hypertension <0.00001 No 579(19.2%) 361(64.6%) 218(8.9%) Yes 2425(80.7%) 198(35.4%) 2227(91%) 18.63 (14.94–23.31) Weight 76.49(15.17) 76.51(14.54) 76.49(15.3) 0.96 Height 1.64(0.09) 1.65(0.09) 1.64(0.09) 0.55 BMI 28.26(5.1) 28.12(4.5) 28.29(5.2) 0.81 Age 57.34(12.6) 54.61(11.6) 57.97(12.7) <0.00001 Qualitative data are presented as frequency (%). Fisher test is used to assess differences in the proportions between controls and CKD groups. Quantitative data are presented as mean value (standard deviation). Mann–Whitney’s U-test is used to assess differences in the means between groups. TABLE 2 | Univariate analysis of CKD associated SNPs including the chosen inheritance model. SNP Gene Model Adjusted p-value OR (95%CI) AIC Permutation p-value (1) rs1126616 SPP1 Dominant 0.005 1.31 (1.08–1.58) 2777 0.005 (2) rs35068180 MMP3 Overdominant 0.01 1.29 (1.07–1.55) 2780 0.01 (3) rs2238135 VDR Recessive 0.007 1.74 (1.14–2.76) 2776 0.008 (4) rs3102735 OPG Overdominant 0.009 1.32 (1.07–1.64) 2780 0.01 (5) rs1800247 BGLAP Overdominant 0.01 1.28 (1.06–1.55) 2774 0.01 (6) rs385564 KL Dominant 0.02 1.27 (1.05–1.54) 2758 0.02 (7) rs679620 MMP3 Overdominant 0.02 1.25 (1.04–1.51) 2779 0.02 (8) rs2248359 CYP24A1 Recessive 0.01 1.4 (1.06–1.86) 2063 0.01 (9) rs1564858 TNFRSF11B Dominant 0.03 1.32 (1.02–1.69) 2063 0.03 (10) rs4236 MGP Dominant 0.04 1.23 (1.01–1.49) 2782 0.03 (11) rs9138 SPP1 Dominant 0.04 1.22 (1.01–1.47) 2782 0.041 (12) rs731236 VDR Dominant 0.046 1.22 (1–1.48) 2782 0.043 SNPs, associated gene, chosen inheritance model and its p-value in the LRT adjusted by confounding variables are shown. Odds-ratio and its 95% confidence interval and the AIC computed from the unadjusted model are also displayed. Finally, permutation p-value is also included. Frontiers in Genetics | www.frontiersin.org 4 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease Out of the twelve selected SNPs, three pairs were in proportion of non-caucasian patients was low (3.61% in the LD: rs679620 and rs35068180 (located in the MMP3 gene, complete model), we assessed the potential interactions of the chromosome 11), rs9138 and rs1126616 (located in SPP1 gene, significant SNPs with race. We did not find any significant chromosome 4), and rs3102735 and rs1564858 (located in interaction (results not shown). In addition, we also fitted the TNFRSF11B gene, chromosome 8). These SNPs, together with model with only Caucasian patients and we did not find any rs2238135 and rs4236 (located in the chromosome 12 in the VDR notable differences in the estimated parameters of the model and MGP genes, respectively), were considered as candidates (results not shown). to assess the existence of haplotypes. Supplementary Figure Finally, in order to evaluate the CKD predictive power of SNPs S2 shows the existence of haplotypes for the three pairs of in general population, Receiver Operating Characteristic (ROC) SNPs in linkage disequilibrium, but not for the pair of SNPs curves for the eight SNPs model, the diabetes + hypertension of chromosome 12. TABLE 3 | Multivariate model containing SNPs, classical risk factors and Association of Combination of SNPs confounding variables. With CKD _ To implement the model fittings, only those subjects with β (SE) p-value OR (95% CI) complete data in all of the twelve selected SNPs were considered Intercept −1.57(0.35) <0.00001 0.21 (0.1−0.41) (n = 1,585; 1,047 cases and 538 controls). Three of the 4,095 rs1126616 0.32(0.13) 0.02 1.37 (1.06−1.78) multivariate fitted models attained the minimum AIC value rs35068180 0.25(0.13) 0.06 1.29 (0.99−1.66) (2006.6; Supplementary Figure S3). This was because, under rs1800247 0.3(0.14) 0.03 1.35 (1.04−1.76) the reference and CKD risk genotype categories, rs1126616 and rs4236 0.31(0.14) 0.02 1.37 (1.04−1.79) rs9138 were perfectly correlated, and models with one of each rs2248359 −0.28(0.33) 0.4 0.76 (0.39−1.4) or both were equally considered. Arbitrarily, we selected the Hypertension (yes) 2.6(0.15) <0.00001 13.53 (10.06−18.35) model that contains the first SNP (rs1126616). The multivariate Diabetes (yes) 0.4(0.18) 0.03 1.49 (1.04−2.15) model consisted in the combination of eight SNPs (rs1126616, Sex (male) 0.11(0.13) 0.42 1.11 (0.85−1.45) rs35068180, rs2238135, rs1800247, rs385564, rs4236, rs2248359, Race (non-Caucas.) 0.8(0.43) 0.06 2.22 (1−5.37) and rs1564858) associated with CKD (p < 0.05) whose estimated Age −0.005(0.005) 0.37 1 (0.98−1.01) coefficients, standard errors, unadjusted and adjusted p-values Interaction: 0.99(0.41) 0.02 2.68 (1.23−6.15) and odds-ratios are shown in Supplementary Table S6. We hypertension with also generated a model using classical risk factors, adjusting by rs2248359 sex, race and age, and, as expected, hypertension and diabetes Complete summary of the model containing classical risk factors, SNPs and the _ were found significantly associated to the prevalence of CKD adjusting variables. β (SE) denotes the estimated coefficient (standard error) in the (Supplementary Table S7). logistic regression model, p-value denotes the z-test p-value for the coefficient, OR In order to evaluate the possible clinical utility of the proposed (95%) denotes the odds-ratio (95% confidence interval). model, we generated a multivariate model on which we included the eight SNPs, hypertension, diabetes and the adjusting variables TABLE 4 | Summary of the ROC curves of the different multivariate models (sex, age, and race). For both, adjusted and unadjusted models, explored for CKD risk prediction. the combination of the 8 SNP was statistically significant for CKD risk prediction, even when classical risk factors and adjusting 5-SNP model Classical risk Classical risk factors model factors and variables were considered (Supplementary Table S8). Going SNPs model further, we generated a model considering interactions of the eight SNPs with diabetes and with hypertension, which allowed Probability 0.65 0.43 0.70 threshold us a backward elimination of three SNPs (rs2238135, rs385564, Sensitivity 0.636 0.900 0.881 and rs1564858). On this process we identified an interaction of (0.606–0.665) (0.880–0.918) (0.860–0.900) hypertension with rs2248359, whose genotype had no effects on Specificity 0.565 0.645 0.669 CKD risk for non-hypertensive individuals, but for hypertensive (0.520–0.606) (0.603–0.685) (0.628–0.709) patients the TT genotype increased the risk of CKD [p = 0.017, Positive PV 0.740 0.832 0.839 OR = 2 (1.23,3.32); Supplementary Figure S4]. (0.710–0.776) (0.809–0.853) (0.816–0.860) Therefore, a multivariate model containing five SNPs, Negative PV 0.442 0.768 0.742 diabetes, hypertension and the interaction of hypertension (0.404–0.480) (0.726–0.806) (0.701–0.781) with rs2248359 was generated (Table 3). The predicted risk AUC 0.620 0.794 0.824 model is summarized in Figure 1. Thus, our model shows (0.591–0.649) (0.770–0.819) (0.802–0.847) that in a hypothetical high-risk profile patient (male, non- Summary of the ROC curves for each model. Probability threshold is the chosen Caucasian with diabetes and hypertension), the presence of probability threshold obtained from the optimal cut-off for each curve, chosen as the point that maximizes the distance to the diagonal line. Sensitivity and specificity, the five described SNPs increases the predicted risk of being positive and negative predicted values (PV) for that optimal threshold are displayed a CKD patient six times. In any case, this is a predictive tool with 95% confidence intervals. Finally, the respective area under the curve (AUC), that needs to be validated in a different cohort. Although the with 95% confidence intervals, is included. Frontiers in Genetics | www.frontiersin.org 5 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease FIGURE 1 | Prediction models of CKD risk. The height of each bar corresponds to the predicted odds-ratio for that category when compared to the reference category. The value of these odds-ratio, their 95% confidence interval and their p-value are also displayed inside each bar. Above bars, odds-ratios, 95% confidence intervals and p-values comparing each possible other pair of categories are also displayed, always taking the group with less CKD risk as reference. The vertical axis is plotted in logarithmic scale. For all models, taken age was the sample median age (60). FIGURE 2 | Receiver operating characteristic (ROC) curves of the different multivariate models explored for CKD risk prediction. ROC curves corresponding to explored multivariate models, including in all cases the adjusting variables: sex, race (Caucasian/Non-Caucasian) and age. The “SNPs” ROC curve (orange) corresponds to the model containing the 8-SNP combination and the adjusting variables. The “classical risk factors” ROC curve (green) corresponds to the model containing diabetes, hypertension and the adjusting variables. The ROC “classical risk factors and SNPs” (blue) corresponds to the model containing the 5-SNP combination, diabetes, hypertension, the interaction of hypertension with rs2248359 and the adjusting variables. For each curve, the filled diamond represents the optimal cut-off, chosen as the point that maximizes the distance to the diagonal line (dashed black line). model and the five SNPs + diabetes + hypertension model, were risk factors model, was statistically significant (p < 0.0001; calculated (Figure 2). The predictive power of the model with Table 4). Results of the bootstrap simulations confirmed the the five SNPs + diabetes + hypertension showed the best AUC, predictive power of the model, leading to a mean AUC of and this AUC increase, compared to the one of the classical 0.823 (0.79, 0.84). Frontiers in Genetics | www.frontiersin.org 6 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease DISCUSSION It is worth mentioning that previous GWAS analyses for genes involved in CKD (Köttgen et al., 2009, 2010) did not find any In the present study, 79 SNPs were genotyped in a large cohort of association of the SNPs described here; that could reflect different CKD patients and non-CKD subjects. After adjusting for age, sex frequencies of the polymorphisms among the United States based and race, we found twelve SNPs associated to CKD, in which three cohorts and our Spain based cohort. Validation of our results pairs showed LD. We also account for additive effects among the in another Spanish cohort would strengthen our results and, twelve selected SNPs, finding an eight SNP combination whose due to the changing dynamics of Spanish population, would additive effects were statistically significant, even when adjusting be desirable the inclusion of non-caucasic genetic background by potentially confounding variables. We also showed that the to better replicate the Spanish multicenter NEFRONA cohort. model including diabetes, hypertension, sex, age and race, and The fact that a small amount of patients presented the five the five selected SNPs, displays a statistically significant increase SNPs in our cohort also limits the validation of our model. in the AUC, compared with the same model without SNPs. It is also remarkable that our model needs a minimum of The five SNPs that improved CKD risk detection were five SNPs to improve clinical parameters prediction; a cost- rs1126616, rs35068180, rs1800247, rs4236, and rs2248359 that effectiveness analysis would be desirable to test feasibility of were located at the SPP1, MMP3, BGLAP, MGP, and CYP24A1 clinical implantation of our results. A second limitation is genes, respectively. that only 1603 patients presented complete data for all the SPP1 codes for Osteopontin (OPN), a protein constitutively genotypes, so the final multivariate model was fitted only in expressed in kidney (Hudkins et al., 1999) whose levels are those. Despite these limitations, the strength of this study is increased in advanced stages of CKD (Lorenzen et al., 2010) that we analyzed seventy-nine SNPs in relatively large cohort and in diabetic nephropathy, where it regulates macrophage of CKD patients, and after adjusting for potential confounding recruitment (Kelly et al., 2002). More recently, its levels have been variables, we found 5 SNPs related to CKD-MBD that might aid linked to nephritis in lupus patients (Salimi et al., 2016), and in determining CKD risk. OPN knock out mice showed resistance against high phosphate induced nephrocalcinosis (Paloian et al., 2016), a condition that can lead to CKD (Shavit et al., 2015). Despite these data ETHICS STATEMENT linking OPN to CKD onset and progression, currently there are no studies defining the effect of the rs1126616 SNP on OPN The protocol of the study was approved by the ethics committee expression or function. of the Hospital Universitari Arnau de Vilanova (Lleida) and The rs35068180 located in the gene that codes for the matrix all patients were included after signing informed consent. This metalloprotease 3 (also known as Stromelysin 1) has been related research followed the principles of the Declaration of Helsinki. to diabetic nephropathy (Kure et al., 2011), and its combination with the rs1799750 located in the MMP1 gene strongly associates to end stage renal disease (Cozzolino et al., 2009). Indeed AUTHOR CONTRIBUTIONS matrix metalloproteases are key factors in kidney physiology JV and CP-G performed statistical analyses. SC wrote the and its pathologies, as MMP inhibition improves lesions in manuscript. JMV, EF, IR, JV, and SC conceptualized the design glomerular disease, diabetic nephropathy and interstitial fibrosis of the study. EF, MB-L, MB, and ÀB provided valuable feedback (Tan and Liu, 2012). for manuscript writing and experimental design. All authors BGLAP and MGP genes code for Ostecalcin and Matrix contributed to manuscript revision, read and approved the Gla Protein, respectively, proteins whose levels vary with CKD submitted version. progression (Delmas et al., 1983; Thamratnopkoon et al., 2017). The rs4236 SNP in MGP gene has been associated with aortic calcification in Spanish men and functional studies have FUNDING demonstrated that the protein expressing the minor allele is less effective in inhibiting calcium deposition in vascular smooth The NEFRONA study was funded by a research grant from muscle cells (Tuñón-Le Poultel et al., 2014). The rs1800247 AbbVie, FEDER funds and the Instituto de Salud Carlos III has been previously related to Osteocalcin levels in women RETIC (RD16/0009), FIS PI16/01354, and PI10/00173. IR was (McGuigan et al., 2010; Ling et al., 2016), although we did not financially supported by Fundación para el Fomento en Asturias find any effect of this SNP on Osteocalcin levels in our population de la Investigación Científica Aplicada y la Tecnología (FICYT). (data not shown). CYP24A1, is a key enzyme of vitamin D metabolism, and has been extensively involved in CKD (Petkovich and Jones, 2011). ACKNOWLEDGMENTS In GWAS analyses, different SNPs of the CYP24A1 gene have been related to vitamin D levels (Manousaki et al., 2017; Jiang The authors would like to thank the NEFRONA team et al., 2018), to calcium levels (O’Seaghdha et al., 2013), and even (Eva Castro, Virtudes María, Teresa Molí, Teresa Vidal, to glomerular filtration rate (Mahajan et al., 2016; Pattaro et al., Meritxell Soria) and the Biobank of REDinREN for their 2016), but to the best of our knowledge, this is the first time that invaluable support. The NEFRONA study investigator the rs2248359 polymorphism has been associated to CKD. group is composed by the following: Aladrén Regidor, Frontiers in Genetics | www.frontiersin.org 7 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease Ma José. Hospital Comarcal Ernest Lluch (Calatayud); Almirall, (León); Munar Vila, Antonia. Hospital Universitari Son Espases Jaume; Ponz, Esther. Corporació Parc Taulí (Barcelona); (Palma de Mallorca); Muñoz Díaz, Ana Beatriz. Hospital Vithas Arteaga Coloma, Jesús. Hospital de Navarra (Pamplona); Bajo Nisa Virgen del Consuelo (Valencia); Navarro González, Juan F. Rubio, Ma Auxiliadora; Díaz, Raquel Raquel Hospital La Paz Hospital Universitario Nuestra Señora de Candelaria (Santa Cruz (Madrid); Belart Rodríguez, Montserrat. Sistemes Renals (Lleida); de Tenerife); Nieto, Javier; Carreño, Agustín. Hospital General Gascón, Antonio, Hospital Obispo Polanco (Teruel); Bover Universitario de Ciudad Real (Ciudad Real); Novoa Fernández, Sanjuan, Jordi. Fundació Puigvert. IIB Sant Pau (Barcelona); Enrique. Complexo Hospitalario de Ourense (Ourense); Bronsoms Artero, Josep. Clínica Girona (Girona); Cabezuelo Ortiz, Alberto; Fernandez, Beatriz. IIS-Fundación Jiménez Romero, Juan B; Muray Cases, Salomé. Hospital Reina Sofía Díaz (Madrid); Paraíso, Vicente. Hospital Universitario del (Murcia); Calviño Varela, Jesús. Hospital Universitario Lucus Henares (Madrid); Pérez Fontán, Miguel. Complejo Hospitalario Augusti (Lugo); Caro Acevedo, Pilar. Clínica Ruber (Madrid); Universitario A Coruña (A Coruña); Peris Domingo, Ana. Carreras Bassa, Jordi. Diaverum Baix Llobregat (Barcelona); Hospital Francesc de Borja (Valencia); Piñera Haces, Celestino. Cases Amenós, Aleix; Massó Jiménez, Elisabet. Hospital Clínic Hospital Universitario Marqués de Valdecilla (Santander); (Barcelona); Moreno López, Rosario. Hospital de la Defensa Prados Garrido, Ma Dolores. Hospital Universitario San Cecilio (Zaragoza); Cigarrán Guldris, Secundino; López Prieto, Saray. (Granada); Prieto Velasco, Mario. Hospital de León (León); Puig Hospital Da Costa (Lugo); Comas Mongay, Lourdes. Hospital Marí, Carmina. Hospital d’Igualada (Barcelona); Rivera Gorrín, General de Vic (Barcelona); Comerma, Isabel. Hospital General Maite. Hospital Universitario Ramón y Cajal (Madrid); Rubio, de Manresa (Barcelona); Compte Jové, Ma Teresa, Hospital Esther. Hospital Puerta de Hierro-Majadahonda (Madrid); Ruiz, de la Santa Creu de Jesús (Tarragona); Cuberes Izquierdo, Pilar. Hospital de Sant Joan Despí Moisès Broggi (Barcelona); Marta. Hospital Reina Sofía (Navarra); de Álvaro, Fernando; Salgueira Lazo, Mercedes; Martínez Puerto, Ana Isabel. Hospital Hevia Ojanguren, Covadonga. Hospital Infanta Sofía (Madrid); Universitario Virgen Macarena (Seville); Sánchez Tomero, de Arriba de la Fuente, Gabriel. Hospital Universitario de José Antonio. Hospital Universitario de la Princesa (Madrid); Guadalajara (Guadalajara); del Pino y Pino, Ma Dolores. Sánchez, José Emilio. Hospital Universitario Central de Asturias Complejo Hospitalario Torrecardenas de Almeria (Almería); (Oviedo); Sans Lorman, Ramon. Hospital de Figueres (Girona); Diaz-Tejeiro Izquierdo, Rafael; Ahijado Hormigos, Francisco Saracho, Ramon. Hospital de Santiago (Vitoria); Sarrias, Maria; Hospital Virgen de la Salud (Toledo); Dotori, Marta. USP Serón, Daniel. Hospital Universitari Vall d’Hebron (Barcelona); Marbella (Málaga); Duarte, Verónica. Hospital de Terrassa Soler, María José; Barrios, Clara. Hospital del Mar (Barcelona); (Barcelona); Estupiñan Torres, Sara. Hospital Universitario de Sousa, Fernando. Hospital Rio Carrión (Palencia); Toran, Canarias (Santa Cruz de Tenerife); Fernández Reyes, Ma José. Daniel. Hospital General de Jerez (Cadiz); Tornero Molina, Hospital de Segovia (Segovia); Fernández Rodríguez, Ma Loreto. Fernando. Hospital Universitario Del Sureste (Arganda del Hospital Príncipe de Asturias (Madrid); Fernández, Guillermina. Rey); Usón Carrasco, José Javier. Hospital Virgen de la Luz Clínica Santa Isabel (Seville); Galán Serrano, Antonio. Hospital (Cuenca); Valera Cortes, Ildefonso. Hospital Virgen de la Victoria General Universitari de València (Valencia); García Cantón, (Málaga); Vilaprinyo del Perugia, Ma Merce. Institut Catala Cesar. Hospital Universitario Insular de Gran Canaria (Las d’Urologia i Nefrologia (Barcelona); Virto Ruiz, Rafael C. Palmas); García Herrera, Antonio L. Hospital Universitario Hospital San Jorge (Huesca); Vicente Pallarés Carratalá Clinica Puerto Real (Cádiz); García Mena, Mercedes. Hospital San MEDEFIS (Vila-real. Castellón), Carlos Santos Altozano CS Juan de Dios (Zaragoza); Gil Sacaluga, Luis; Aguilar, Maria. Azuqueca de Henares (Guadalajara); Miguel Artigao Ródenas Hospital Virgen del Rocío (Seville); Górriz, José Luis. Hospital CS Zona III (Albacete); Inés Gil Gil Área Básica Sanitaria Universitario Doctor Peset (Valencia); Huarte Loza, Emma. de Arán. CAP Viella (Lleida); Francisco Adan Gil CS Alfaro Hospital San Pedro (Logroño); Lerma, José Luis. Hospital (La Rioja); Emilio García Criado Centro de Salud del Carpio Universitario de Salamanca (Salamanca); Liebana Cañada, (Córdoba); Rafael Durá Belinchón CS Godella (Valencia); Jose Antonio. Hospital de Jaén (Jaén); Marín Álvarez, Jesús Pedro. Ma Fernández Toro CS Zona Centro (Cáceres); Juan Antonio Hospital San Pedro de Alcántara (Cáceres); Martín Alemany, Divisón Garrote Centro de Salud de Casas Ibáñez. Consultorio Nàdia. Hospital Universitari Dr. Josep Trueta (Girona); Martín de Fuentealbilla (Albacete). García, Jesús. Hospital Nuestra Señora de Sonsoles (Ávila); Martínez Castelao, Alberto. Hospital Universitari de Bellvitge (Barcelona); Martínez Villaescusa, María. Complejo Hospitalario SUPPLEMENTARY MATERIAL Universitario de Albacete (Albacete); Martínez, Isabel. Hospital Galdakao (Bilbao); Moina Eguren, Iñigo. Hospital Basurto The Supplementary Material for this article can be found (Bilbao); Moreno Los Huertos, Silvia. Hospital Santa Bárbara online at: https://www.frontiersin.org/articles/10.3389/fgene. (Soria); Mouzo Mirco, Ricardo. Hospital El Bierzo, Ponferrada 2019.00118/full#supplementary-material REFERENCES Ammirati, E., Moroni, F., Norata, G. D., Magnoni, M., and Camici, P. G. (2015). Markers of inflammation associated with plaque progression and instability in 1000 Genomes Project Consortium, Auton, A., Brooks, L. D., Durbin, patients with carotid atherosclerosis. Med. Inflamm. 2015:718329. doi: 10.1155/ R. M., Garrison, E. P., Kang, H. M., et al. (2015). A global reference 2015/718329 for human genetic variation. Nature 526, 68–74. doi: 10.1038/nature Calleros-Basilio, L., Cortés, M. A., García-Jerez, A., Luengo-Rodríguez, A., 15393 Orozco-Agudo, A., Valdivielso, J. M., et al. (2016). Quality assurance of Frontiers in Genetics | www.frontiersin.org 8 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease samples and processes in the spanish renal research network (REDinREN) Lorenzen, J., Krämer, R., Kliem, V., Bode-Boeger, S. M., Veldink, H., Haller, H., biobank. Biopreserv. Biobank 14, 499–510. doi: 10.1089/bio.2015. et al. (2010). Circulating levels of osteopontin are closely related to glomerular 0095 filtration rate and cardiovascular risk markers in patients with chronic kidney Cozzolino, M., Biondi, M. L., Galassi, A., Turri, O., Brancaccio, D., and Gallieni, M. disease. Eur. J. Clin. Invest. 40, 294–300. doi: 10.1111/j.1365-2362.2010.02271.x (2009). Matrix metalloproteinase-1 and matrix metalloproteinase-3 gene Ma, J., Yang, Q., Hwang, S. J., Fox, C. S., and Chu, A. Y. (2017). Genetic risk promoter polymorphisms are associated with mortality in haemodialysis score and risk of stage 3 chronic kidney disease. BMC Nephrol. 18:32. doi: patients. Nephrol. Dial. Transplant. 24, 2207–2212. doi: 10.1093/ndt/ 10.1186/s12882-017-0439-3 gfp061 Mahajan, A., Rodan, A. R., Le, T. H., Gaulton, K. J., Haessler, J., Stilp, A. M., et al. Delmas, P. D., Wilson, D. M., Mann, K. G., and Riggs, B. L. (1983). Effect of renal (2016). Trans-ethnic fine mapping highlights kidney-function genes linked to function on plasma levels of bone Gla-protein. J. Clin. Endocrinol. Metab. 57, salt sensitivity. Am. J. Hum. Genet. 99, 636–646. doi: 10.1016/j.ajhg.2016.07.012 1028–1030. doi: 10.1210/jcem-57-5-1028 Manousaki, D., Dudding, T., Haworth, S., Hsu, Y. H., Liu, C. T., Medina- Devuyst, O., and Pattaro, C. (2018). The UMOD locus: insights into the Gómez, C., et al. (2017). Low-frequency synonymous coding variation in pathogenesis and prognosis of kidney disease. J. Am. Soc. Nephrol. 29, 713–726. CYP2R1 has large effects on vitamin D levels and risk of multiple sclerosis. Am. doi: 10.1681/ASN.2017070716 J. Hum. Genet. 101, 227–238. doi: 10.1016/j.ajhg.2017.06.014 GBD 2015 Disease and Injury Incidence and Prevalence Collaborators (2016). McGuigan, F., Kumar, J., Ivaska, K. K., Obrant, K. J., Gerdhem, P., and Akesson, K. Global, regional, and national incidence, prevalence, and years lived with (2010). Osteocalcin gene polymorphisms influence concentration of serum disability for 310 diseases and injuries, 1990-2015: a systematic analysis for osteocalcin and enhance fracture identification. J. Bone Miner. Res. 25, 1392– the global burden of disease study 2015. Lancet 388, 1545–1602. doi: 10.1016/ 1399. doi: 10.1002/jbmr.32 doi: 10.1002/jbmr.32 S0140-6736(16)31678-6 O’Seaghdha, C. M., Wu, H., Yang, Q., Kapur, K., Guessous, I., Zuber, A. M., et al. Go, A. S., Chertow, G. M., Fan, D., McCulloch, C. E., and Hsu, C. Y. (2004). Chronic (2013). Meta-analysis of genome-wide association studies identifies six new kidney disease and the risks of death, cardiovascular events, and hospitalization. Loci for serum calcium concentrations. PLoS Genet. 9:e1003796. doi: 10.1371/ N. Engl. J. Med. 351, 1296–1305. doi: 10.1056/NEJMoa041031 journal.pgen.1003796 Higgins, C. L., Isbilir, S., Basto, P., Chen, I. Y., Vaduganathan, M., Paloian, N. J., Leaf, E. M., and Giachelli, C. M. (2016). Osteopontin protects against Vaduganathan, P., et al. (2015). Distribution of alkaline phosphatase, high phosphate-induced nephrocalcinosis and vascular calcification. Kidney osteopontin, RANK ligand and osteoprotegerin in calcified human carotid Int. 89, 1027–1036. doi: 10.1016/j.kint.2015.12.046 atheroma. Protein J. 34, 315–328. doi: 10.1007/s10930-015-9620-3 Parsa, A., Kanetsky, P. A., Xiao, R., Gupta, J., Mitra, N., Limou, S., et al. (2017). Hudkins, K. L., Giachelli, C. M., Cui, Y., Couser, W. G., Johnson, R. J., and Alpers, Genome-wide association of CKD progression: the chronic renal insufficiency C. E. (1999). Osteopontin expression in fetal and mature human kidney. J. Am. cohort study. J. Am. Soc. Nephrol. 28, 923–934. doi: 10.1681/ASN.2015101152 Soc. Nephrol. 10, 444–457. Pattaro, C., Teumer, A., Gorski, M., Chu, A. Y., Li, M., Mijatovic, V., et al. Jiang, X., O’Reilly, P. F., Aschard, H., Hsu, Y. H., Richards, J. B., Dupuis, J., (2016). Genetic associations at 53 loci highlight cell types and biological et al. (2018). Genome-wide association study in 79,366 European-ancestry pathways relevant for kidney function. Nat. Commun. 7:10023. doi: 10.1038/ individuals informs the genetic architecture of 25-hydroxyvitamin D levels. Nat. ncomms10023 Commun. 9:260. doi: 10.1038/s41467-017-02662-2 Petkovich, M., and Jones, G. (2011). CYP24A1 and kidney disease. Curr. Opin. Junyent, M., Martínez, M., Borrás, M., Bertriu, A., Coll, B., Craver, L., et al. (2010). Nephrol. Hypertens. 20, 337–344. doi: 10.1097/MNH.0b013e3283477a7b Usefulness of imaging techniques and novel biomarkers in the prediction R Core Team (2014). R: A Language and Environment for Statistical Computing of cardiovascular risk in patients with chronic kidney disease in Spain: the [computer program]. Vienna: R Foundation for Statistical Computing. NEFRONA project. Nefrologia 30, 119–126. doi: 10.3265/Nefrologia.pre2010. Rong, S., Zhao, X., Jin, X., Zhang, Z., Chen, L., Zhu, Y., et al. (2014). Vascular Jan.10216 calcification in chronic kidney disease is induced by bone morphogenetic Kazancioğlu, R. (2013). Risk factors for chronic kidney disease: an update. Kidney protein-2 via a mechanism involving the Wnt/β-catenin pathway. Cell Physiol. Int. 3, 368–371. doi: 10.1038/kisup.2013.79 Biochem. 34, 2049–2060. doi: 10.1159/000366400 Kelly, D. J., Wilkinson-Berka, J. L., Ricardo, S. D., Cox, A. J., and Gilbert, Salimi, S., Noora, M., Nabizadeh, S., Rezaei, M., Shahraki, H., Milad, M. K., R. E. (2002). Progression of tubulointerstitial injury by osteopontin- et al. (2016). Association of the osteopontin rs1126616 polymorphism and a induced macrophage recruitment in advanced diabetic nephropathy higher serum osteopontin level with lupus nephritis. Biomed. Rep. 4, 355–360. of transgenic (mRen-2)27 rats. Nephrol. Dial. Transplant. 17, 985–991. doi: 10.3892/br.2016.589 doi: 10.1093/ndt/17.6.985 Schwarz, S., Trivedi, B. K., Kalantar-Zadeh, K., and Kovesdy, C. P. (2006). Köttgen, A., Glazer, N. L., Dehghan, A., Hwang, S. J., Katz, R., Li, M., et al. (2009). Association of disorders in mineral metabolism with progression of chronic Multiple loci associated with indices of renal function and chronic kidney kidney disease. Clin. J. Am. Soc. Nephrol. 1, 825–831. doi: 10.2215/CJN. disease. Nat. Genet. 41, 712–717. doi: 10.1038/ng.377 02101205 Köttgen, A., Pattaro, C., Böger, C. A., Fuchsberger, C., Olden, M., Glazer, N. L., et al. Shavit, L., Jaeger, P., and Unwin, R. J. (2015). What is nephrocalcinosis? Kidney Int. (2010). New loci associated with kidney function and chronic kidney disease. 88, 35–43. doi: 10.1038/ki.2015.76 Nat. Genet. 42, 376–384. doi: 10.1038/ng.568 Stevens, L. A., Coresh, J., Greene, T., and Levey, A. S. (2006). Assessing kidney Kure, M., Pezzolesi, M. G., Poznik, G. D., Katavetin, P., Skupien, J., Dunn, J. S., et al. function–measured and estimated glomerular filtration rate. N. Engl. J. Med. (2011). Genetic variation in the matrix metalloproteinase genes and diabetic 354, 2473–2483. doi: 10.1056/NEJMra054415 nephropathy in type 1 diabetes. Mol. Genet. Metab. 103, 60–65. doi: 10.1016/j. Tan, R. J., and Liu, Y. (2012). Matrix metalloproteinases in kidney homeostasis and ymgme.2011.01.001 diseases. Am. J. Physiol. Renal. Physiol. 302, F1351–F1361. doi: 10.1152/ajprenal. Levey, A. S., Bosch, J. P., Lewis, J. B., Greene, T., Rogers, N., and Roth, D. (1999). 00037.2012 A more accurate method to estimate glomerular filtration rate from serum Thamratnopkoon, S., Susantitaphong, P., Tumkosit, M., Katavetin, P., creatinine: a new prediction equation. Modification of diet in renal disease Tiranathanagul, K., Praditpornsilpa, K., et al. (2017). Correlations of study group. Ann. Intern. Med. 130, 461–470. doi: 10.7326/0003-4819-130-6- plasma desphosphorylated uncarboxylated matrix gla protein with vascular 199903160-00002 calcification and vascular stiffness in chronic kidney disease. Nephron 135, Limou, S., Vince, N., and Parsa, A. (2018). Lessons from CKD-related genetic 167–172. doi: 10.1159/000453368 association studies-moving forward. Clin. J. Am. Soc. Nephrol. 13, 140–152. Thio, C. H. L., van der Most, P. J., Nolte, I. M., van der Harst, P., Bültmann, U., doi: 10.2215/CJN.09030817 Gansevoort, R. T., et al. (2018). Evaluation of a genetic risk score based Ling, Y., Gao, X., Lin, H., Ma, H., Pan, B., and Gao, J. (2016). A common on creatinine-estimated glomerular filtration rate and its association with polymorphism rs1800247 in osteocalcin gene was associated with serum kidney outcomes. Nephrol. Dial. Transplant. 33, 1757–1764. doi: 10.1093/ndt/ osteocalcin levels, bone mineral density, and fracture: the shanghai gfx337 changfeng Study. Osteoporos Int. 27, 769–779. doi: 10.1007/s00198-015- Trudu, M., Janas, S., Lanzani, C., Debaix, H., Schaeffer, C., Ikehata, M., et al. (2013). 3244-5 Common noncoding UMOD gene variants induce salt-sensitive hypertension Frontiers in Genetics | www.frontiersin.org 9 February 2019 | Volume 10 | Article 118 Valls et al. Polymorphisms and Risk of Chronic Kidney Disease and kidney damage by increasing uromodulin expression. Nat. Med. 19, 1655– Wuttke, M., and Köttgen, A. (2016). Insights into kidney diseases from genome- 1660. doi: 10.1038/nm.3384 wide association studies. Nat. Rev. Nephrol. 12, 549–562. doi: 10.1038/nrneph. Tuñón-Le Poultel, D., Cannata-Andía, J. B., Román-García, P., Díaz-López, J. B., 2016.107 Coto, E., Gómez, C., et al. (2014). Association of matrix Gla protein gene Xu, Y., and Sun, Z. (2015). Molecular basis of Klotho: from gene to function in functional polymorphisms with loss of bone mineral density and progression aging. Endocr. Rev. 36, 174–193. doi: 10.1210/er.2013-1079 of aortic calcification. Osteoporos. Int. 25, 1237–1246. doi: 10.1007/s00198-013- Zhang, Q. L., and Rothenbacher, D. (2008). Prevalence of chronic kidney disease 2577-1 in population-based studies: systematic review. BMC Public Health 8:117. Valdivielso, J. M., Betriu, A., Martinez-Alonso, M., Arroyo, D., Bermudez- doi: 10.1186/1471-2458-8-117 doi: 10.1186/1471-2458-8-117 Lopez, M., Fernandez, E., et al. (2017). Factors predicting cardiovascular events in chronic kidney disease patients. Role of subclinical atheromatosis extent Conflict of Interest Statement: The authors declare that the research was assessed by vascular ultrasound. PLoS One 12:e0186665. doi: 10.1371/journal. conducted in the absence of any commercial or financial relationships that could pone.0186665 be construed as a potential conflict of interest. van der Velde, M., Matsushita, K., Coresh, J., Astor, B. C., Woodward, M., Levey, A., et al. (2011). Lower estimated glomerular filtration rate and higher albuminuria Copyright © 2019 Valls, Cambray, Pérez-Guallar, Bozic, Bermúdez-López, are associated with all-cause and cardiovascular mortality. A collaborative Fernández, Betriu, Rodríguez and Valdivielso. This is an open-access article meta-analysis of high-risk population cohorts. Kidney Int. 79, 1341–1352. distributed under the terms of the Creative Commons Attribution License (CC BY). doi: 10.1038/ki.2010.536 The use, distribution or reproduction in other forums is permitted, provided the Wolf, M. (2012). Update on fibroblast growth factor 23 in original author(s) and the copyright owner(s) are credited and that the original chronic kidney disease. Kidney Int. 82, 737–747. doi: 10.1038/ki. publication in this journal is cited, in accordance with accepted academic practice. No 2012.176 use, distribution or reproduction is permitted which does not comply with these terms. Frontiers in Genetics | www.frontiersin.org 10 February 2019 | Volume 10 | Article 118