Objectives Twenty-eight genetic loci are associated with serum urate levels in Europeans. Evidence for association with gout at most loci is absent, equivocal or not replicated. Our aim was to test the loci for association with gout meeting the American College of Rheumatology gout classification criteria in New Zealand European and Polynesian case-control sample sets.
Methods 648 European cases and 1550 controls, and 888 Polynesian (Ma¯ori and Pacific) cases and 1095 controls were genotyped. Association with gout was tested by logistic regression adjusting for age and sex. Power was adequate (>0.7) to detect effects of OR>1.3.
Results We focused on 24 loci without previous consistent evidence for association with gout. In Europeans, we detected association at seven loci, one of which was the first report of association with gout (IGF1R). In Polynesian, association was detected at three loci. Meta-analysis revealed association at eight loci—two had not previously been associated with gout (PDZK1 and MAF). In participants with higher Polynesian ancestry, there was association in an opposing direction to Europeans at PRKAG2 and HLF (HLF is the first report of association with gout). There was obvious inconsistency of gout association at four loci (GCKR, INHBC, SLC22A11, SLC16A9) that display very similar effects on urate levels.
Conclusions We provide the first evidence for association with gout at four loci (IGF1R, PDZK1, MAF, HLF). Understanding why there is lack of correlation between urate and gout effect sizes will be important in understanding the aetiology of gout.
- Gene Polymorphism
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Given that hyperuricaemia is requisite for gout, factors that influence serum urate levels are candidate causal risk factors for gout. Genome-wide association studies (GWAS) have previously associated common genetic variation in 10 loci with serum urate (SLC2A9, ABCG2, SLC22A11, SLC22A12, SLC17A1, GCKR, SLC16A9, RREB1, INHBC).1 ,2 These loci are dominated by transporters that are involved in renal and gut excretion of uric acid, and most have been associated with gout in European, Asian and Polynesian sample sets.3–9
Notably, PDZK1 was not associated with gout (OR=1.03) in a European sample set including 3151 cases despite very strong evidence for association with serum urate levels.6 Association of other loci with gout is not supported by robust statistical evidence (SLC16A9, RREB1) or is yet to be replicated (INHBC). Therefore, it is important to test these loci for association with gout. The lack of association could indicate pleiotropic effects (eg, at PDKZ1), and understanding these effects should yield insights into the aetiology of gout. The population-specific effects at ABCG2, SLC22A11, SLC22A12 and LRP2 in gout3 ,7 ,9 ,10 also highlight the necessity for transancestral studies, especially in populations with earlier onset and more severe gout, and a higher prevalence of comorbidities, such as those of Filipino and New Zealand (NZ) Ma¯ori and Pacific ancestry who have a prevalence of gout double that of European Caucasian.11–13
Recently, an additional 18 loci of weaker effect than the original 101 ,2 have been associated with serum urate levels with the collective 28 explaining 7.0% of variance in serum urate in Europeans.6 None of the new loci contain obvious uric acid transporters, with loci containing genes involved in glycolysis being prominent. Of the 18 loci, only 10 were associated with gout in Europeans at a nominal level of significance (p<0.05) in >3000 cases. The effect size estimates of all 28 loci may not be accurate because the ascertainment of gout largely by self-report and allopurinol prescription means that people without gout will have been assigned as cases.14 Therefore, we tested the 28 loci for association with gout in 1536 cases with gout meeting the American College of Rheumatology (ACR) gout classification criteria and 2645 controls drawn from the European Caucasian, Ma¯ori and Pacific Island (Polynesian) populations of Aotearoa New Zealand.
Subjects and methods
Gout cases all fulfilled the ACR criteria15 and were recruited from the Auckland, Waikato and Canterbury regions of NZ. The comparison group self-reported no diagnosis of gout, and were convenience-sampled from the Auckland and Otago regions of NZ (n=1973), with an additional group (n=672) of European controls sampled from the Otago region of NZ.16 ,17 Recruitment occurred during the period 2006–2013. Demographic and clinical data are reported in online supplementary table S1. Subjects were divided into two ancestral groups: European (648 cases, 1550 controls), and Polynesian (NZ Ma¯ori and Pacific Island, 888 cases, 1095 controls). Of the controls, 376 of the Europeans, and all the Polynesians, were also determined not to have gout by the ACR criteria. Based on previous reports,4 ,7 the Polynesian sample set was subdivided into Eastern Polynesian (EP; Cook Island and NZ Ma¯ori) people of higher Polynesian ancestry (EP/N; 334 cases, 395 controls), Eastern Polynesian people of lower Polynesian ancestry (EP/Z; 157 cases, 313 controls), people of Western Polynesian (WP; Samoa, Tonga, Niue, Tuvalu and Tokelau) ancestry (368 cases, 315 controls), and people of mixed Eastern and Western Polynesian ancestry (EP/WP; 29 cases, 72 controls).4 For the analysis presented in table 2 the WP, EP/N and EP/WP sample sets were grouped into a higher Polynesian ancestry sample set. The New Zealand Multi-Region Ethics Committee (MEC/105/10/130) granted ethical approval, and all participants gave fully informed written consent.
Single nucleotide polymorphism selection and genotyping
The same single nucleotide polymorphisms (SNP) marking 18 novel loci identified by Köttgen et al6 (tables 1 and 2), but excluding those that did not meet both criteria predefined by Köttgen et al6 for replication, were genotyped using a Sequenom MassARRAY System, with two surrogate SNPs used for BAZ1B (rs7811265 and rs11983997). Genotypes of SNPs from the 10 previously reported loci1 ,2 were determined using a combination of Taqman SNP genotyping on a Roche LightCycler 480 and genotyping on the Sequenom MassARRAY System, and included data previously reported on smaller overlapping sample sets.3–5 ,7 At SLC22A12/NRXN2 SNP rs3825018 (surrogate for rs5058021) was genotyped, and the previously reported rs11942223 was genotyped at SLC2A9.5 ,18 Ninety-four samples were genotyped by both methods for seven SNPs (rs11942223, rs2231142, rs1967017, rs780094, rs1183201, rs3825018, rs1106766) with 100% concordance in genotype (>96.5% of the time a genotype was obtained by both methods for each SNP).
The Taqman SNP genotyping assays were autocalled by the Lightcycler 480 software, and the reporter dye signal plots were visually inspected for correct genotype clustering by a trained analyst. Genotype data from the Sequenom MassARRAY was processed and analysed by the MassARRAY RT Workstation (V.4.0) software. Genotype clusters for each SNP were then visually verified by an analyst. The 672 European controls recruited from the Otago region had been genotyped with the Affymetrix SNP array and SNP genotypes were taken from imputed data generated by IMPUTE2 using HapMap3 CEU (NCBI Build 36 (db126b)) as reference haplotypes. Surrogate SNP rs11983997 was used at BAZ1B owing to difficulties in imputation of rs1178977.
Statistical analysis and power
Genotype data were checked for deviation from Hardy–Weinberg equilibrium (see online supplementary table S2). Association of genetic variants with gout was tested using STATA V.8.0 software by logistic regression with adjustment for age and sex. The Polynesian sample sets were additionally adjusted by a genetic estimate of Polynesian ancestry, determined as previously described.4 Sample sets were combined by an inverse-variance weighted fixed-effect method, using STATA V.8.0 software. A Q-statistic was calculated to determine the heterogeneity between cohorts, and for SNPs showing heterogeneity (PHet <0.10) the fixed-effect model was replaced with a random-effect model.
Power calculations done as previously described19 for the separate European and Polynesian analyses are presented in online supplementary figure S1. In both datasets, power was adequate (>0.7) to detect effects of OR >1.3 at minor allele frequency >0.1. A p value of <0.05 indicated nominal evidence for association with gout. We did not apply a Bonferroni correction given the prior probability that the confirmed serum urate-associated loci are causal of gout.
Association with gout
The 28 serum urate loci were genotyped for all NZ gout cases and controls (figure 1; table 1). Genotypes are presented in online supplementary table S2, and multivariate adjusted data in online supplementary table S3. Of the 10 established loci (the 10 loci presented in table 1 under the heading ‘Kolz and Yang loci’), unequivocal association of GCKR, SLC2A9, ABCG2 and SLC17A1 with gout has previously been reported.4–7 ,9 Here (table 1), we report a strong association of GCKR with gout in Polynesian (OR=1.38, p=6.6×10−5) for the first time, with a magnitude of effect similar to NZ European (OR=1.31, p=1.1×10−4). Evidence for association with gout at SLC22A12 strengthened our previous observation. However, the additional samples genotyped here weakened evidence for association of SLC22A11 with gout in Polynesian3 and, consistent with our previous report,3 provided no evidence for association in NZ European (OR=1.04, p=0.59).
We next examined PDZK1, SLC16A9, INHBC, RREB1 and the 18 new loci reported by Köttgen et al6 (table 1; the former 4 loci are listed under ‘Kolz and Yang loci’ and the new 18 under ‘Köttgen loci’). Evidence for association with gout was observed at six loci in the NZ European sample set, five which had been previously associated with gout by Köttgen et al (RREB1, TRIM46, SFMBT1, PRKAG2, A1CF) and one novel association (IGF1R). For all, the same allele associated with increased serum urate and gout by Köttgen et al6 was also associated with gout in NZ Europeans. Two loci were associated in the Polynesian samples (SFMBT1, VEGFA), both replicating Köttgen et al's6 association with gout, with the same allele associated with gout as associated with increased serum urate in Europeans.6 Meta-analysis of the NZ European and Polynesian datasets revealed eight significant associations (PDZK1, RREB1, INHBC, TRIM46, SFMBT1, VEGFA, IGF1R, MAF), of which PDZK1 and MAF had not previously been associated with gout in the individual ancestral groups (table 1) nor by Köttgen et al.6
We selected eight loci for further population-specific analysis (table 2) because they showed evidence for heterogeneity (at PHet<0.1) in the meta-analysis of the NZ European and Polynesian sample sets (SLC17A1, SLC16A9, TRIM46, INHBB, VEGFA, PRKAG2, UBE2Q2, HLF), suggesting the existence of population-specific genetic effects. Two loci gave notable results: PRKAG2 and HLF. At PRKAG2, there was evidence for an opposing direction of association to NZ European. In the higher Polynesian ancestry group, the European risk allele (T) conferred protection to gout (OR=0.54, p=0.039), with no evidence for association in the lower Polynesian ancestry group (EP/Z) (OR=0.97, p=0.88). At HLF, there was evidence for association with gout in the higher Polynesian ancestry group (OR=0.73, p=0.02), with the urate-increasing allele in Europeans conferring protection. Again, there was no evidence for association in the lower Polynesian ancestry group (OR=1.22, p=0.32). The HLF locus is one of the four genes that showed no evidence for association with gout in our first stage analysis or in the Köttgen et al data. At VEGFA, evidence for association with gout was restricted to the higher Polynesian ancestry group (OR=1.28, p=0.006). At SLC16A9, association with gout was restricted to the lower Polynesian ancestry (EP/Z) group (OR=2.60, p=0.006).
Previously, Köttgen et al6 reported association with gout at 13 serum urate loci (excluding the established SLC2A9, ABCG2, SLC17A1 and GCKR loci). We have replicated seven of these (RREB1, TRIM46, SFMBT1, PRKAG2, A1CF, SFMBT1, VEGFA), and provide the first report of association with gout at four others (PDZK1, IGF1R, MAF, HLF). Of the five loci without nominal evidence for association with gout in either the Köttgen et al6 or our analyses (figure 1 and tables 1 and 2), we note a strong trend towards association at INHBB and HNF4G in Europeans (0.06>p>0.05), leaving three loci for which there is no evidence for association with gout from our or the Köttgen et al6 data: UBE2Q2, ACVR1B and B3GNT4.
It is important to demonstrate association of urate loci with gout. Elevated serum urate is the critical risk factor for development of gout.20 Therefore, it can be argued that genetic variants with a stronger effect on serum urate should have a greater effect on the risk of gout. While this is clearly the case for SLC2A9 and ABCG2, both of which have a very strong effect on serum urate and on risk of gout, there is a clear lack of correlation within the next tier of loci of more moderate effect. The risk alleles of GCKR, SLC16A9, SLC22A11 and INHBC are associated with an average increase in serum urate of 0.004 mmol/L6. Of these loci, the effect size in gout of GCKR is consistently higher—it is associated with gout in European, Chinese, Japanese and Polynesian sample sets (table 1)6 ,8 ,9 with OR=1.3–1.5 in sample sets where gout is clinically ascertained. INHBC is also consistently associated in European and Polynesian with OR=∼1.15, although with a lower effect size than GCKR (table 1).6 By contrast, SLC22A11 is not consistently associated with gout, with the statistically strong evidence for association reported in Köttgen et al6 in European (OR=1.14, p=2.3×10−5) not replicated by us (table 1; OR=1.04, p=0.59).3 The still weaker evidence for association of SLC16A9 with gout in Köttgen et al6 (OR=1.10, p=0.017) was also not replicated by us (table 1; OR=1.01, p=0.89). There are, clearly, inconsistent effects on association with gout between the four loci with very similar effects on serum urate. It is also notable that the effect of ABCG2 on gout is consistently larger than SLC2A9 in European, Japanese and Polynesian sample sets, despite SLC2A9 having a 72% greater effect on serum urate levels (table 1).6 ,9 These observations may result from a lack of independence between molecular pathways of serum urate control and clinical presentation of gout in the presence of hyperuricaemia (ie, pleiotropic effects of the urate-associated loci), and/or differential effects of serum urate loci between strata of serum urate concentrations and/or confounding of serum urate and risk of gout effect sizes by unmeasured or unaccounted-for environmental exposures such as alcohol and sugar-sweetened beverage consumption. Supporting the latter hypothesis is evidence for non-additive interaction of SLC2A9 with a prevalent environmental exposure (sugar-sweetened beverages) in regulation of urate levels and risk of gout,18 and for interaction of SLC2A9 and SLC22A11 (that encodes the renal OAT4 transporter) with diuretic use in determining the risk of gout.21 Consequently, it is important to continue to test urate-associated loci with gout and, if the inconsistent pattern of association with gout at loci such as SLC16A9 and SLC22A11 is evident in other datasets, investigate why this is the case in clinical studies. One variable that should be included, if possible, in such epidemiological studies is duration of hyperuricaemia preceding the onset of gout. Finally, we adjusted, as did Kottgen et al,6 by age and sex - it is possible that other clinical variables are important in the progression from hyperuricaemia to gout.
It is important to emphasise that the locus names used to identify loci will not necessarily reflect the causal gene in serum urate control and risk of gout, because extensive linkage disequilibrium at some loci means that association can span multiple candidate genes. One example of this is the BAZ1B locus in which the association signal is strongest over the BAZ1B gene. However, the association signal also encompasses the glycolytic gene MLXIPL that encodes the transcription factor carbohydrate-responsive element-binding protein, which activates triglyceride synthetic genes in response to glucose (see online supplementary figure S2 of ref 6) and which was identified as the strongest candidate at the locus.6 At other serum urate loci, the association signal is often clearly intergenic, as at MAF, VEGFA and INHBB.6 These association signals illustrate that the considerable majority of common genetic variants associated with human phenotypes map to functionally important regions of the genome that regulate gene expression,22 generating an additional degree of difficulty in identifying the causal gene.
To translate the genetic findings into clinical impact, it is necessary to confirm the causal gene and causal genetic variants at all loci, perhaps excepting ABCG2, where the rs2231142 SNP explains the regional association, and the Q141K variant it encodes decreases the expression and function of ABCG2 to transport urate.23 ,24 Of potential clinical impact, small molecules have been identified that ameliorate this effect on ABCG2.24 Genetic techniques for identifying the causal gene could include resequencing all genes within association signals in people with extreme hyperuricemia and hypouricemia, under the hypothesis that rare, predicted, functional variants would be present in the exonic sequence of the causal gene.25–27 The common causal variant responsible for the GWAS signal6 would be expected to be the maximally associated variant in meta-analysis over diverse ancestral groups (where a common signal exists). Assisting mapping would be the use of populations where there is evidence for a recombinant haplotype, such as at PRKAG2 and HLF, where the serum urate increasing allele in Europeans protects from gout in Polynesians. This suggests that the associated SNPs (rs10480300 and rs7224610) are not causal; the causal variant can be expected to map to surrounding DNA where the same allele of genetic variants consistently associate with risk of gout in both ancestral groups. Even at loci with clearly defined association over a single gene (eg, PRKAG2, RREB1, IGF1R, HLF), the example of the FTO locus in weight control, where the causal genetic variant controls expression of the neighbouring IRX3 gene,28 emphasises that additional functional experiments are required to confirm the causal gene. Resequencing of urate loci would also be expected to detect additional aetiological variants that explain more variance in phenotype. Resequencing of lipid loci to detect rare variants and comprehensive conditional analyses to test for multiple independent common variants provided evidence that the proportion of variance explained by the individual lipid loci is approximately doubled.25 ,29 The existence of multiple common variants at urate loci is suggested by strongly associated variants that exhibit little linkage disequilibrium with the most strongly associated SNP (eg, SLC16A9, TMEM171, SLC22A12) and is obvious at MAF with distinct signals apparent (see online supplementary figure S2).6
In conclusion, we provide the first evidence for association of four serum urate-associated loci with gout. Understanding the reasons why the strength of association with gout is comparatively inconsistent with strength of effect on urate will further illuminate the aetiology of urate control and risk of gout.
This work was supported by the Health Research Council of New Zealand, Arthritis New Zealand, New Zealand Lottery Health and the University of Otago. The authors would like to thank Jill Drake, Roddi Laurence, Meaghan House and Gabrielle Sexton for assistance in recruitment. Murray Cadzow and Tania Flynn are thanked for technical assistance. All participants who gifted samples and information for this study are sincerely thanked.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Handling editor Tore K Kvien
Contributors TRM designed the study, interpreted data and led the writing of the manuscript; AJP-G performed the data analysis; MEM, RT, SA, GWM, CF, GTJ, AMvR, DW, LKS and ND made substantive contributions to data acquisition. All authors approved were involved in drafting, and approved the final submitted manuscript.
Competing interests None.
Patient consent Obtained.
Ethics approval The New Zealand Multi-Region Ethics Committee granted ethical approval (MEC/105/10/130).
Provenance and peer review Not commissioned; externally peer reviewed.