No causal relationship between serum urate and neurodegenerative diseases: A Mendelian randomization study

Objective: Observational studies have shown that increased serum urate is associated with a lower risk of neurodegenerative diseases (NDs), but the causality remains unclear. We employed a two-sample Mendelian randomization (MR) approach to assess the causal relationship between serum urate and four common subtypes of NDs, including Parkinson's disease (PD), Alzheimer's disease (AD), amyotrophic lateral sclerosis (ALS), and multiple sclerosis (MS). Methods: Serum urate data came from the CKDGen Consortium. GWAS data for PD, AD, ALS, and MS were obtained from four databases in the primary analysis and then acquired statistics from the FinnGen consortium for replication and meta-analysis. Inverse variance weighted (IVW), weighted median (WM), and MR-Egger regression methods were applied in the MR analyses. Pleiotropic effects, heterogeneity, and leave-one-out analyses were evaluated to validate the results. Results: There was no evidence for the effect of serum urate on PD (OR: 1.00, 95 % CI: 0.90 – 1.11, P = 0.97), AD (OR: 1.02, 95 % CI: 1.00 – 1.04, P = 0.06), ALS (OR: 1.05, 95 % CI: 0.97 – 1.13, P = 0.22), and MS (OR: 1.01, 95 % CI: 0.89 – 1.14, P = 0.90) risk when combined with the FinnGen consortium, neither was any evidence of plei-otropy detected between the instrumental variables (IVs). Conclusion: The MR analysis suggested that serum urate may not be causally associated with a risk of PD, AD, ALS, and MS.


Introduction
Neurodegenerative diseases (NDs), represented by Alzheimer's disease (AD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), and multiple sclerosis (MS), are characterized by the gradual decline of specific groups of neurons (Jucker and Walker, 2018).The World Health Organization (WHO) predicts that in the next twenty years, NDs will exceed cancer and become the second leading cause of death (Gammon, 2014).The start of NDs can manifest at any stage of human life, encompassing juvenile forms that emerge around birth (such as some types of spinal and muscular atrophy) and later stages like old age.A majority of NDs often appear after an individual's third decade of life, potentially indicating a correlation with the natural aging process.Various factors were postulated to play a role in the development of these conditions, encompassing genetic and environmental elements.Presently accessible treatments merely serve to alleviate symptoms or impede the disease's progression (Gonzales et al., 2022;Wyss-Coray, 2016).It has become a significant health burden around the world.Hence, it is imperative to ascertain the causes and develop therapeutic interventions for NDs.
As the anionic form of uric acid (2,6,8-trioxy-purine), urate is found in body fluids.It is a very effective scavenger of free radicals and makes up 60 % of the antioxidant capacity in plasma (Proctor, 1970).Urate is present in the human body's bloodstream at high levels, approaching the boundaries of its ability to dissolve.(Cipriani et al., 2010).The blood concentration of purine is influenced by dietary consumption, with foods such as meats, seafood, and beer containing high levels of purine (Torralba et al., 2012).Additionally, the excretion of purine depends on kidney function and variations in urate transporters due to genetic factors (Vitart et al., 2008).
Numerous studies have demonstrated the preventative impact of urate levels on the progression of NDs.Long-term studies tracking the same group of individuals have shown a connection between urate levels and the likelihood of developing PD in the future.Higher urate levels are linked to a lower risk of the disease (Davis et al., 1996;Chen et al., 2009;Uribe-San Martín et al., 2013).A history of gout is associated with a decreased likelihood of developing either PD (Alonso et al., 2007) or AD (Lu et al., 2016).Observational studies revealed that individuals diagnosed with PD (Annanmaki et al., 2007;Bogdanov et al., 2008) and ALS (Paganoni and Schwarzschild, 2017;Keizman et al., 2009) have decreased levels of plasma or serum urate compared to healthy controls.Furthermore, several meta-analyses have reported that serum urate levels in MS patients are far lower than in normal controls (Liu et al., 2012;Wang et al., 2016).Therefore, establishing the causality of the connection between urate and NDs might be crucial in identifying potential biomarkers and therapeutic targets for NDs.
Mendelian randomization (MR) is an approach that assesses the causal effects of exposure on an outcome by using genetic variations as instrumental variables of the exposure.Genome-wide association studies (GWAS) summary statistics are used in the two-sample MR analysis instead of examining data at the individual level (Davey Smith and Hemani, 2014).Since genetic variants are randomly assigned at conception according to Mendel's second law, MR can effectively eliminate the effect of confounding factors and identify causal determinants of a specific outcome.In genetic correlation, the direction of causality is determined.Genetic diversity leads to different phenotypes, and the reverse is not valid.Therefore, MR is a more reliable strategy than traditional observational epidemiological studies, which have been plagued by problems such as confounding (common causes of exposure X and outcome Y may distort the association between X and Y), reverse causation (Y or the disease process that causes Y affects X), and other forms of bias, leading to potentially misleading causal inferences (Davies et al., 2018;Richmond and Davey Smith, 2022).
This study systematically utilized GWAS statistics to evaluate the potential causality in a two-sample MR method, considering the limited understanding of the causality between serum urate and four common NDs.We then combined the same outcome of two MR estimates via meta-analysis.This research would aim to realize the pathophysiology of NDs, and also provide reliable evidence for establishing feasible strategies for NDs prevention and treatment in clinical.

Study design
We evaluated the causal relationship between serum urate and four common NDs using a two-sample MR design.Our MR design met three basic assumptions: 1.The genotype is related to exposure (relevance assumption); 2. The genotype affects the outcome solely through the studied exposure (exclusion restriction assumption); and 3.The genotype is unrelated to other factors influencing outcome (independence assumption) (Boef et al., 2015).For the purposes of primary analysis and replication analysis, genetic data were gathered from separate GWAS consortia, and a meta-analysis was then carried out (Fig. 1).

GWAS data for serum urate (instrument-exposure associations)
The serum urate summary statistics were derived from the GWAS for serum urate previously presented in the CKDGen Consortium which was open access.The primary analysis relied on genetic information from 74 studies that encompassed individuals of different ancestries, including Europeans (288,649 people), East Asians (125,725 people), African Americans (33,671 people), South Asians (9037 people), and Hispanics (608 people) (Tin et al., 2019).

GWAS data for neurodegenerative diseases (instrument-outcome associations)
In primary analysis, the International Parkinson's Disease Genomics Consortium (IPDGC) provided data on 33,674 cases and 449,056 controls for the PD summary statistics (Nalls et al., 2019).71,880 cases and 383,378 controls were included in the Psychiatric Genomics Consortium (PGC) data used to analyze AD (Jansen et al., 2019).A recent GWAS meta-analysis of 138,086 people (27,205 cases and 110,881 controls) yielded the ALS summary data (van Rheenen et al., 2021).For the analysis of MS, data from the International Multiple Sclerosis Genetics Consortium (IMSGC) containing 47,429 patients and 68,374 controls were included (International Multiple Sclerosis Genetics Consortium, 2019).All the datasets acquired in the primary analysis were openaccess.
To verify our findings, replication analysis, and meta-analysis were carried out.We also obtained data associated with PD, AD, ALS, and MS from FinnGen (https://www.finngen.fi/en),which compiles and analyzes genetic and clinical information from 500,000 Finnish biobank participants.The results can be browsed with the web browser and GWAS data can be downloaded from Google cloud storage free of charge.The replication analysis included 4235 cases of PD and 373,042 matched controls.Additionally, there were 13,393 cases of AD and 36,884 controls, as well as 483 cases of ALS and 170,667 controls.Furthermore, the analysis involved 2182 cases of MS along with a control group of 373,987 individuals.
In the GWAS, PD, AD, ALS, and MS were defined by International Classification of Disease (ICD) codes: ICD10 (G20) for PD cases; ICD-10 (G30) for AD cases; ICD-10 (G12.2) for ALS cases, and ICD-10 (G35) for MS cases.This study's subjects were entirely of European ancestry, which can reduce bias caused by confounding factors with an ethnic component.Each sample cohort should be distinct from the others.Since all the analyses were conducted using publicly accessible data, there was no need for an institutional review board to grant ethical approval for this study.

Instruments selection
First, single nucleotide polymorphisms (SNPs) significantly linked to the exposures (P-value < 5 × 10 − 8 ) were kept as IVs.Second, these SNPs must exhibit linkage disequilibrium (LD) r 2 < 0.001 and <1000 kB from the index variant.Third, in the presence of palindromic SNPs, the alleles on the forward strand were deduced based on the data of allele frequency.Finally, the potency of IVs was assessed by calculating the Fstatistic using the formula signifies the portion of the exposure's variance elucidated by the genetic variants, N represents the sample size, and K indicates the number of instruments.A lack of any significant weak instrumental bias was indicated by an F-statistic exceeding 10.In addition, SNPs associated with outcomes and P-value < 5 × 10 − 8 were removed.PhenoScanner (http://www.phenoscanner.medschl.cam.ac.uk/) is a meticulously curated repository of publicly available discoveries derived from extensive human genetic association research.To eliminate the influence of confounding factors, we used PhenoScanner to rule out any instruments associated with additional risk factors (Kamat et al., 2019).

Primary analysis
We utilized three different methods, including inverse-variance weighted (IVW), MR Egger, and the weighted median, to conduct MR analysis.When the instrumental variable assumptions are not met, MR-Egger can still be used as a statistical method as long as a less strict assumption is fulfilled (Burgess and Thompson, 2017).In addition, the weighted median method can precisely estimate the causal association impact even if fewer than 50 % of the genetic variations deviate from the fundamental assumptions of MR (Bowden et al., 2016).The IVW approach is the most frequently used and essential computation approach in MR studiesal references (Burgess et al., 2013).Therefore, we employed the IVW as our main analytical technique while utilizing the other two methods (MR-Egger and weighted median) as supplementary tools.
The results were reported as odds ratios (ORs) with 95 % confidence intervals (95 % CI).After Bonferroni adjustment, P-value < 0.05/4 = 0.013 was considered statistically significant, while suggestive evidence was considered for P-values between 0.013 and 0.05.

Sensitivity analysis
In addition, we performed additional sensitivity analyses to evaluate the robustness of the associations and the presence of pleiotropy.These included the utilization of the Cochran Q test, MR-Egger intercept method, MR-PRESSO method, and leave-one-out analyses.To evaluate compliance with the exclusion limitation assumption, we applied the systematic leave-one-out strategy to identify potential pleiotropy for each SNP.The difference between findings obtained with and without the SNP was then evaluated.Then, we looked for heterogeneity among the instrumental variables using the Cochran Q test.By computing the Fstatistic, we assessed the strength of our instrument.We also tested for pleiotropy using the MR-PRESSO and the MR-Egger regression based on its intercept terms.A P-value < 0.05 in sensitivity analyses signified considerable heterogeneity and pleiotropy (Pierce et al., 2011;Verbanck et al., 2018).Research documented that smoking cigarettes, excessive alcohol consumption, being overweight, chronic stress, and Vitamin D play a role in the development of NDs (Wallensten et al., 2023;Anwar et al., 2023).The causal effects were rechecked after eliminating the confounder-related SNPs mentioned above to see if they remained significant.

Replication and meta analysis
FinnGen is a large public-private GWAS that gathers and examines genetic and medical information from half a million individuals in the Finnish biobank.To confirm the reliability of the results, we repeated the MR analysis using another independent GWAS data of PD, AD, ALS, and MS from the FinnGen collaboration and then performed a metaanalysis.

Results
The selection of genetic instruments followed specific criteria mentioned above in the instruments selection section.A total of 83 different SNPs were chosen as the IVs of serum urate based on the criteria.All F statistics exceeded 10, suggesting a relatively minimal chance of weak instrument bias in MR analyses (baseline characteristics  1 and Table 2, detailed information on the IVs used in the primary and replication MR analyses showed in the Supplementary Table 1).

Sensitivity analysis
To prevent an overabundance of bias effects, several measures were implemented to assess the sensitivity analysis and identify any potential pleiotropy.We detected some outlier SNPs (rs2581817, rs141990161, and rs1359232) in MR analysis when we employed the MR-PRESSO global test (P = 2.00 × 10 − 4 for PD; P < 2 × 10 − 4 for MS).With three outliers removed, the MR results acquired through the utilization of the MR-PRESSO method with outlier correction exhibited similarities to the MR IVW results above (P = 0.14 for PD; P = 0.64 for MS).According to the Cochran's Q test, there was significant heterogeneity observed (P = 1.047 × 10 − 4 for PD; P = 4.12 × 10 − 2 for AD; P = 3.61 × 10 − 3 for ALS; P = 1.28 × 10 − 6 for MS).Despite the identification of diversity in particular outcomes, the random-effect IVW in the research did not render the MR estimates invalid, as it could potentially equalize the combined heterogeneity.Furthermore, the intercepts of MR-Egger regression exhibited no evidence of pleiotropy, as all intercepts had a P-value > 0.05.This implies that there was no introduction of pleiotropic bias to MR estimates in the presence of heterogeneity.We didn't find any instruments associated with other risk factors through the PhenoScanner website.To summarize, the outcomes of these sensitivity analyses demonstrate the strength and reliability of our MR analyses.

Risk factor analysis
Epidemiological studies and meta-analyses have provided evidence that the risk of NDs is significantly linked to smoking cigarettes, excessive alcohol consumption, and being overweight (Durazzo et al., 2014;Liu et al., 2020;Chandrashekar et al., 2023;Moody et al., 2021).We used the IVW method to examine the connection between serum urate and various risk factors for NDs, including smoking, alcohol consumption, and body mass index, in order to determine whether the MR association between genetically determined serum urate and NDs was influenced by pleiotropic pathways associated with NDs.The result shows no observed causal effects on potential risk factors for NDs from serum urate (details are presented in Table 3).

Discussion
To examine the causal impact of serum urate and four common NDs, we conducted a two-sample MR analysis in this study.Despite previous observational study findings that serum urate was strongly associated with the risk of incident NDs, our results found no causal effect of serum urate on the risk of NDs.Furthermore, we applied multiple sensitivity analyses, each based on distinct assumptions, that yielded no indication of pleiotropy and further bolstered the credibility of the findings.This implies that the previously reported connections between serum urate and the risk of NDs may be due to confounding factors or reverse causation (Paternoster et al., 2017).
Observational studies and meta-analyses generally indicated a negative correlation between urate and the incidence of NDs.However, reverse causality in these studies can't be ignored.A meta-analysis of typical case-control studies involving 1217 cases and 1276 controls revealed that PD cases had a standardized mean difference of − 0.52 (95 % CI, − 0.72 to − 0.31) in urate levels compared to healthy controls.Nevertheless, the results didn't differentiate between PD leading to a reduced urate or urate levels acting as a protective factor against PD  (Shen and Ji, 2013).Another meta-analysis involving 1308 individuals with MS and 908 controls found decreased serum urate levels in MS, which used urate measurements in prevalent MS cases.As a result, they could't judge whether the decrease in serum levels occurs before or after the onset of MS, making them vulnerable to bias from reverse causality (Wang et al., 2016).Several dietary and lifestyle differences are examples of confounding factors that might account for the possible neuroprotective effects of raising serum urate.For instance, it has been suggested that changes in microbiota impact circulating urate levels and PD risk, potentially confusing relationships reported in earlier observational studies (Scheperjans et al., 2015).Unmeasured confounding variables reduce the correlation between serum urate and the progression of PD, resulting in an underestimation of the true effect of urate on PD progression.Serum urate is correlated with obesity and insulin resistance, both of which have been found in some studies to an elevated risk of PD (Chen et al., 2004).Moreover, the findings of the previous two MR studies (Harroud et al., 2020;Kia et al., 2018) that addressed the causal impact of serum urate on PD and MS risk were consistent with ours.Nonetheless, the other two MR articles had given contradictory outcomes, both of which have found some positive associations between serum urate and PD, AD (Simon et al., 2014;S ¸anlı et al., 2022).Given that several previous MR studies have concluded inconsistent causal relationships, our MR research used the latest datasets from different GWAS data sources and performed meta-analyses to sum up the MR results.Therefore, the proof from our study is more convincing.
Though our findings indicated no causality between serum urate and NDs incidence, it was possible that serum urate could affect the risk and progression of NDs.Several underlying pathways between serum urate and NDs have been reported.Multiple studies have proposed that urate may have a wide-ranging neuroprotective impact on various populations of neurons within the central nervous system (CNS).In experimental settings using cultured nigral neurons, urate has been seen to inhibit the spontaneous degeneration of these neurons.Additionally, urate has been found to protect dopaminergic cells against death caused by oxidative stress and mitochondrial toxins in Parkinson's disease models (Cipriani et al., 2012).Improved phenotypic and histologic results in PD animal models were obtained in vivo by genetic modification of urate oxidase, which raised urate concentrations in the CNS (Chen et al., 2013;Gong et al., 2012).Urate provides defense in different neurotoxicity models.Studies have shown that urate has the ability to safeguard cultured spinal cord neurons against glutamate toxicity (Du et al., 2007).Additionally, it has been found to provide neuroprotection in various models of spinal cord injury, brain injury, and MS (Kean et al., 2000).On the other hand, oxidative damage plays a role in the progression of NDs and injury to neural tissue.The presumed mechanism is linked to the antioxidant qualities of urate, which could potentially provide neuroprotection by eliminating reactive oxygen and nitrogen compounds and functioning as an iron binder, thereby saving cells from oxidative stress (Hink et al., 2002).Then, the IL-6/signal transducer and activator of the transcription 3 (STAT3) signaling pathway was reported to be implicated in the connection of urate to neuroprotection.Aliena-Valero et al. demonstrated that exogenous administration of urate increases IL-6 levels and plays a neuroprotective role through the activation of the IL-6/STAT3 signaling pathway, which in turn leads to modulation of relevant mediators of oxidative stress, neuroinflammation, apoptotic cell death in the brain, and prevent the disruption of blood-brain barrier integrity (Mijailovic et al., 2022;Aliena-Valero et al., 2021).
The advantages of our research include its extensive sample size with NDs, which enhances statistical strength and enables the utilization of various approaches to evaluate potential bias associated with pleiotropy.Furthermore, when conducting an MR study, utilizing a genetically predicted phenotype as the exposure reduces the likelihood of reverse causality and confounding bias compared to an observational study.Lastly, we performed a replication analysis on two separate datasets to validate the MR estimates' reliability, significantly bolstering our findings' trustworthiness.
Some limitations should also be considered.To minimize potential bias related to ethnic variations, we limited the study participants exclusively to individuals of European descent.Nevertheless, the confirmation of our findings in other populations is still required.Furthermore, utilizing GWAS summary-level data presents difficulty in performing stratified analyses based on age, gender, and other relevant characteristics.Again, despite the expansion of liberal studies, which introduced additional instrumental variables and enhanced the ability to identify and address pleiotropy, there is a possibility of remaining confounding factors.Hence, further research is required to validate our findings once a more extensive GWAS is accessible.

Conclusion
In conclusion, our findings showed no evidence of a causal effect of serum urate on risk PD, AD, ALS, and MS.This result contributes to understanding the function of serum urate in the development of NDs and suggests that strategies to raise serum urate levels would not be effective in disease prevention.Further investigation of other biomarkers or pathways in NDs' etiology would be a fruitful direction for future MR research.

Fig. 2 .
Fig. 2. Forest plot for the causal effect of serum urate on the risk of four common NDs derived from inverse variance weighted (IVW).OR, odds ratio; CI, confidence interval.

Table 1
Characteristics of data sources used in primary analysis.
GWAS: genome-wide association study.M.Wang and Z. Tangof the study cohort are summarized in Table

Table 2
Characteristics of data sources used in replication analysis.

Table 3
MR estimates of the associations from serum urate on potential NDs risk factors.