Mendelian randomization of serum urate and Parkinson’s disease progression

Abstract

Objective

Higher serum urate concentrations predict more favorable prognosis in individuals with Parkinson disease (PD). The purpose of this study was to test the causality of this association using a mendelian randomization approach.

Methods

The study was conducted among participants in DATATOP and PRECEPT, two randomized trials among patients with early PD. The 808 patients with available DNA were genotyped for three SLC2A9 single nucleotide polymorphisms that identify an allele associated with lower urate concentrations, and for selected SNPs in other genes encoding urate transporters that have modest or no effect on serum urate levels. A SLC2A9 score was created based on the total number of minor alleles at the three SLC2A9 loci . Primary outcome was disability requiring dopaminergic treatment.

Results

Serum urate concentrations were 0.69mg/dL lower among individuals with 4 or more SLC2A9 minor alleles as compared to those with two or less (p = 0.0002). The hazard ratio (HR) for progression to disability requiring dopaminergic treatment increased with increasing SLC2A9 score (HR=1.16; 95% confidence interval 1.00 to 1.35; p=.056). In a comparative analysis, the HR was 1.27 (1.00 to 1.61; p =0.0497) for a 0.5 mg/dL genetically conferred decrease in serum urate, and 1.05 (1.01 to 1.10; p=0.0133) for a 0.5 mg/dL decrease in measured serum urate. No associations were found between polymorphisms in other genes associated with urate that do not affect serum urate and PD progression.

Interpretation

This Mendelian randomization analysis adds to the evidence of a causal protective effect of high urate levels.

Introduction

Previous longitudinal investigations have shown that individuals with higher serum urate levels ^1-3 or a diet that increases serum urate ⁴ have a lower risk of developing Parkinson disease (PD). Further, in individuals with early PD, higher urate predicts milder clinical and radiographic progression. ^{5, 6} Urate is a potent antioxidant ⁷ and several lines of evidence support a role for oxidative stress in the neurodegenerative process of PD ⁸, but whether the inverse association between serum urate and PD progression reflects a neuroprotective effect remains uncertain due to the possibility of unmeasured confounders. Because urate levels are in part heritable -- the estimate of between person variation due to inherited genetic factors ranges from 25% to 70%, ⁹ -- we sought to use a mendelian randomization design ¹⁰ to investigate whether genetic polymorphisms that predict serum urate levels predict the rate of clinical progression among individuals with early PD. Although several genes are associated with serum urate, and a multiple genes score has been used in a previous study of PD risk ¹¹, we selected as an instrumental variable for this investigation only the gene for solute carrier family 2 [facilitated glucose transporter], member 9 (SLC2A9 , also known as GLUT9) ¹², which explains most of the genetically specified variability in serum urate. ^13-19 By using a single gene with a strong effect on serum urate, but no known direct effects in the central nervous system, we minimized the possibility of violating the assumption that there are no genetic effects on PD progression other than those mediated by urate levels. ¹⁰ Other genes encoding urate transporters that are known to have modest or no effects on serum urate, but could nevertheless modulate its biological effects, were included in exploratory analyses.

Methods

Study population

The source population for this study includes participants in two randomized clinical trials of PD: the Parkinson Research Examination of CEP-1347 (PRECEPT) and the Deprenyl and Tocopherol Antioxidative Therapy of Parkinsonism (DATATOP) trials. The details of these studies and their participants are described elsewhere. ^{20, 21} We have previously reported an inverse association between serum urate and rate of disease progression in 804 individuals enrolled in PRECEPT and 774 enrolled in DATATOP. ^{5, 6} The study population for this study comprises the subset of these individuals from whom DNA was also available. In DATATOP (two year study with enrollment from September 1987-November 1988), DNA was collected at the end of the extended follow-up in 1995. DNA was not collected during PRECEPT (a two year trial with enrollment from April 2002-April 2004), but DNA collection began during a follow-up investigation, known as POSTCEPT, in which all the surviving individuals previously enrolled in the original trial at participating sites were invited to participate. Overall, DNA was available for 808 individuals, of which 63 were excluded from the mendelian randomization analyses due to lack of serum urate levels or failure in genotyping of SLC2A9; further, we excluded 10 patients who reported use of allopurinol at baseline, leaving 735 patients (390 in DATATOP and 345 in PRECEPT). Exploratory analyses of other genes included between 759 and 783 patients, because we excluded only those patients missing the specific SNP of interest.

SNPs and genotyping

Numerous SNPs in SLC2A9 -- a urate transporter ²² -- have been identified in several genome-wide association studies as the strongest genetic predictors of serum urate levels and gout. ^13-19 Because these SNPs are in high linkage disequilibrium ¹⁵, and a single causal variant has not been identified, we selected three of the top SNPs for the present study. Specifically, the following SNPs in SLC2A9 were genotyped : rs6855911, an intronic SNP with minor allele frequency (MAF) of 0.31 (G allele); rs7442295 (intronic, MAF: 0.21 for G allele); and rs16890979 (missense mutation, MAF: 0.22 for T allele) (using HapMap data from Utah residents with ancestry from northern and western Europe, abbreviated CEU ²³), for which each minor allele has been associated with a 0.30-0.43 mg/dL decrease in serum urate in individuals of European descent . ^{13, 18} Because these three SNPs are in strong linkage disequilibrium (LD; pairwise r² range from 0.68-0.76 from Haploview ²⁴ with HapMap CEU data), we used information from these three SNPs to create an SLC2A9 score with values equal to 0 (<= 2 minor alleles; i.e., preponderance of wild-type alleles); 1 (3 minor alleles and 3 wild-type alleles); and 2 (>=4 minor alleles; i.e., preponderance of minor alleles).

Other genes of interest because of their role in the transport of urate include solute carrier family 22, member 12 (URAT1/SLC22A12) - that encodes a urate-anion exchanger, ²⁵ ATP-binding cassette sub-family G member 2 (ABCG2), and solute carrier family 19 (sodium phosphate), member 3 (SLC17A3). All genotyping was performed through the Harvard Partners Center for Genetics and Genomics (HPCGG) at the Harvard Partners Genotyping Facility using the OpenAssay SNP Genotyping System (BioTrove, Woburn, Massachusetts, USA). Concordance rates for blinded duplicates quality control samples were 100%. Test of Hardy-Weinberg equilibrium revealed no significant deviations (all p>0.05).

Serum urate and clinical outcomes

Serum urate was measured in PRECEPT and DATATOP participants at baseline prior to treatment assignment, as previously reported. ^{5, 6} The outcome evaluated in this study for both DATATOP and PRECEPT participants was the accumulation of disability sufficient to require dopaminergic therapy (this was also the primary outcome of the original studies). ^{20, 21} The mean duration of follow-up until endpoint or study termination was 13.6 months in DATATOP and 13.3 months in PRECEPT.

Statistical analysis

Initial analyses were conducted separately in DATATOP and PRECEPT. Because all tests of heterogeneity between studies were not significant (p>0.05), data from the two trials were pooled, and all models were adjusted for study group and treatment. Differences in serum urate according to genotype were assessed using generalized linear models. Primary analyses to assess the relation between genetic variants and PD progression assumed additive models (per unit increase in score for SLC2A9, per allele associations for other genes); secondary analyses used separate indicators for each genetic score category or genotype. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs) for reaching the primary endpoint according to number of minor alleles or genotype. Analyses were adjusted for study, treatment, gender, age and use of thiazide diuretics at baseline. We assessed potential effect modification by gender and, in DATATOP, by randomization to α-tocopherol supplementation, which was included in some of the treatment arms in DATATOP. These interactions were assessed by including in the regression models an interaction term that was the cross product of the number of minor alleles of each individual SNP by gender (M/F) or α-tocopherol (yes/no). The association between genetically determined serum urate and PD progression was estimated by two stage regression: first, we fitted a generalized linear regression model with serum urate as the dependent variable and the SLC2A9 score and potential confounders (study, gender, age and use of thiazide diuretics) as independent variables, then the predicted urate level from the first stage regression was used as a continuous independent variable to determine its association with PD progression in a Cox proportional hazard model, adjusting for potential confounders. Sensitivity analyses were conducted estimating the genetically predicted urate level in a generalized linear model with separate indicators for each SLC2A9 SNP.

Results

As expected, serum urate concentrations decrease with increasing number of minor SLC2A9 alleles, with a stronger association in women than in men (Figure 1). In an additive model, the rate of progression to a level of disability requiring dopaminergic treatment increased with the number of the minor alleles associated with lower serum UA (HR = 1.16 for each point increase in genetic score; 95% confidence interval: 1.00 to 1.35; p=0.056). As compared with individuals with <=2 minor SLC2A9 alleles, the HR was 1.12 (0.89 to 1.41) for individuals with 3 minor alleles, and 1.39 (0.98 to 1.96) for individuals with 4 or more minor alleles. The HR for a genetically determined lower serum urate was higher (HR for 0.5 mg/dL lower urate=1.27) than the corresponding HR for directly measured 0.5 mg/dL lower serum urate (HR=1.05) (figure 2). Results were not materially changed if each SLC2A9 SNP was used as an independent predictor of serum urate; in this analysis, the HR for 0.5 mg/dL genetically predicted lower urate was 1.24 (0.99 to 1.54). There was no significant effect modification by either gender or α-tocopherol supplementation (in DATATOP only) on the association between SLC2A9 score and initiation of dopaminergic therapy (all p for interaction > 0.05). Additionally, there was no evidence of interaction between SLC2A score and serum urate.

Figure 1.

Serum urate by SLC2A9 score. *p<.001, **p<.01 for comparison with score=0 (<=2 minor alleles).

Figure 2.

Hazard ratios and 95% confidence interval for initiation of dopaminergic therapy for a 0.5 mg/dL observed decrease in serum urate or for genetically conferred decrease in serum urate.

Overall, polymorphisms in genes other than SLC2A9 were not significantly associated with serum UA (Table 1) or subsequent initiation of dopaminergic therapy adjusting for age, gender and treatment (Table 2).

Table1.

Serum urate according to urate transport-related genotype.

SNP	Serum urate (Mean ± SD)	P-Value *
URAT1-rs11231825	N=764
TT	5.5 ± 1.4	0.18
TC	5.3 ± 1.4
CC	5.3 ± 1.4
URAT1-rs11602903	N=780
AA	5.5 ± 1.3	0.56
AT	5.3 ± 1.4
TT	5.3 ± 1.4
URAT1-rs3825016	N=780
CC	5.6 ± 1.3	0.36
CT	5.3 ± 1.4
TT	5.3 ± 1.4
URAT1-rs3825018	N=759
AA	5.3 ± 1.4	0.68
AG	5.4 ± 1.4
GG	5.4 ± 1.4
URAT1-rs475688	N=770
CC	5.3 ± 1.5	0.95
CT	5.3 ± 1.3
TT	5.4 ± 1.3
URAT1-rs476037	N=763
AA	6.1 ± 1.0	0.35
AG	5.4 ± 1.4
GG	5.3 ± 1.4
URAT1-rs7932775	N=783
CC	5.7 ± 1.5	0.86
CT	5.2 ± 1.3
TT	5.3 ± 1.4
URAT1-rs893006	N=761
AA	5.3 ± 1.4	0.46
AC	5.3 ± 1.4
CC	5.6 ± 1.4
ABCG2-rs2231142	N=779
GG	5.3 ± 1.4	0.07
GT	5.6 ± 1.5
TT	4.9 ± 1.1
SLC17A3-rs1165205	N=773
AA	5.3 ± 1.4	0.49
AT	5.3 ± 1.4
TT	5.4 ± 1.4

^*P-value for trend test, adjusted for study.

Table 2.

Hazard ratio for initiating dopaminergic therapy according to urate transport related genotype.

SNP	Genotypes	Risk Allele	Genotype Frequencies	HR (95% CI)*
URAT1-rs11231825	TT/TC/CC	C	89/313/375	1.10 (0.95, 1.27)
URAT1-rs11602903	AA/AT/TT	T	81/351/362	1.07 (0.92, 1.24)
URAT1-rs3825016	CC/CT/TT	T	78/365/351	1.10 (0.95, 1.28)
URAT1-rs3825018	AA/AG/GG	G	355/334/84	0.91 (0.78, 1.05)
URAT1-rs475688	CC/CT/TT	T	404/319/61	0.94 (0.81, 1.10)
URAT1-rs476037	AA/AG/GG	G	9/145/623	1.03 (0.83, 1.29)
URAT1-rs7932775	CC/CT/TT	T	34/258/505	1.05 (0.89, 1.25)
URAT1-rs893006	AA/AC/CC	C	358/332/85	0.92 (0.80, 1.07)
ABCG2-rs2231142	GG/GT/TT	T	614/169/10	1.04 (0.84, 1.29)
SLC17A3-rs1165205	AA/AT/TT	T	221/380/186	0.89 (0.78, 1.03)

^*HR for increasing number of risk alleles, adjusted for gender and age, stratified by a 4-level treatment by study variable.

Discussion

In this investigation we found that among individuals with early PD, SNPs in SLC2A9 predicted differences in serum urate that are similar to those previously reported in the general population. ^13-19 Further, the rate of progression to a level of disability requiring dopaminergic treatment was faster among those patients carrying the SLC2A9 genotypes associated with lower serum UA. Although the statistical significance was marginal according to conventional levels, these novel results suggest that among participants in DATATOP and PRECEPT the previously reported better prognosis of early PD patients with higher urate levels^5,6 is due to a protective effect of urate itself rather than to confounding by unknown factors.

A limitation of this study is that DNA was collected only several years after the trial completion and was only available for a subset of participants in DATATOP and PRECEPT, so that patients with more rapidly progressive disease may be underrepresented. It is unlikely, however, that this selection would result in a spurious inverse association between the SLC2A9 SNPs and the rate of PD progression during the trials. Further, as in the previous studies including all trial participants, baseline serum urate was inversely related to time to initiation of dopaminergic therapy.

A mendelian randomization approach has been used to investigate the causality of the association between serum urate and PD risk in three previous studies. In the first, conducted among individuals with PD in Italy, Croatia, and Germany, a SLC2A9 SNP predicting lower serum urate was associated with a younger age at onset of PD. ²⁶ In the second, a case-control study in Spain, individuals in the highest tertile of a genetic score predicting lower serum urate were found to have a 50% higher risk of PD. ¹¹ In the third study, only one of 12 genotyped SNPs in SLC2A9 was associated with a significantly increased PD risk in women, and none in men. ²⁷ In this last study, however, serum urate levels were not available, and it is thus possible that the association between the genotyped SNPs and serum urate in the study population was weaker than expected. An important limitation of these studies is that even for those SNPs with the most robust associations with serum urate, the expected effects on PD risk are small, and power to detect an association is thus modest. Considering these limitations, overall the results of these studies support the hypothesis that higher urate levels reduce PD risk.

The association between genetically decreased serum UA levels and PD prognosis was somewhat stronger than the comparable association for measured circulating UA, suggesting that the latter may have been attenuated by unmeasured confounding. The lower measurement error and long term stability of genetically determined changes in serum urate may have contributed to this difference, but it is also possible that the association between serum urate and PD progression is attenuated by unmeasured confounders and thus underestimate the true effect of urate on PD progression. Serum urate is associated with obesity and insulin resistance, which in some investigations has been associated with an increased risk of PD ^{28, 29} , and could be a marker of dysfunctional energy metabolism. ³⁰ In vitro, urate production is stimulated by compounds that lower ATP, including inhibitors of mitochondrial respiration ³⁰, which have been implicated in the pathogenesis of PD. ³¹ The observed association between serum urate and PD progression could thus reflect in part the protective effect of urate on neurodegeneration, and in part the adverse effects of the upstream metabolic dysregulation that results in elevated serum urate. Although we did not find a significant interaction between SLC2A9 genotype and α-tocopherol supplementation or gender, the power for these analyses was modest, and effect modification by these factors therefore cannot be excluded. Whereas among DATATOP and PRECEPT participants serum urate was found to be a stronger prognostic predictor in men than in women^5,6 it is noteworthy that the results of a recent phase 2 randomized trial in patients with early PD suggested that urate elevation may be more effective in women than in men. ³²

We a priori considered SLC2A9 the primary gene of interest in relation to serum urate levels, so we did not consider potential joint or synergistic effects of a combination of SNPs recently considered by other authors. ^{19, 33} Although a composite genetic score incorporating several loci could be used,^{11, 19, 33} the contribution of the additional genes to serum urate is small relative to SLC2A9. The inclusion of numerous genes with modest effects on serum UA could increase the possibility that at least one of these genes affects PD progression via mechanisms other than serum urate, thus violating a key assumption of the mendelian randomization method. ¹⁰ In particular, the second strongest genetic predictor of serum urate levels is the ATP-binding cassette, subfamily G, isoform 2 protein (ABCG2), which has been related to the clearance of neurotoxic polypeptides from the brain ³⁴ and neuroregeneration ³⁵, and whose expression in brain capillaries is altered in an animal model of PD. ³⁶ We cannot therefore exclude the possibility that variations in ABCG2 could affect PD progression through mechanisms independent from its effects on urate. The validity of the mendelian randomization approach in our study is supported by the fact that the genotype used as an instrumental variable (SLC2A9) is strongly associated with the exposure of interest (serum urate), and is most likely independent of the factors that confound the association between serum urate and PD progression. The fact that other genes involved in urate transport but without sizable effects on serum urate were not related to PD progression indirectly supports this conclusion. Because urate is also inversely associated with PD risk, one might expect that SNPs in SLC2A9 that predict lower urate levels should have been found to be associated with PD risk in large genome-wide association studies (GWAS). So its absence ³⁷ may appear to contradict the hypothesis of a genuine protective effect of urate. However, because of the stringent significance criteria imposed by the large number of tests performed, even large GWAS are underpowered to detect the small effects attributable to single SLC2A9 SNPs.

In summary, we found that patients in the early stages of PD who carry the variant SLC2A9 alleles associated with lower urate levels have a faster rate of disease progression than those homozygous for the wild-type alleles. This finding suggests that the previously reported inverse association between higher urate levels and rate of PD progression is not explained by unmeasured confounders and is thus likely to reflect a genuine neuroprotective effect of urate. Genotypic characterization may be useful in identifying those most likely to respond to urate-elevating interventions. These data raise the possibility that modulation of SLC2A9 might be an equally or even more effective approach to urate elevation, compared to urate precursor administration³⁸, as a candidate strategy for slowing PD progression.