Pn-Fi Wu ,Hui Lu ,Xiaotin Zhou ,Xuchn Lian ,Ruizhuo Li,Wan Zhan ,Danyan Li,Kun Xia

a Center for Medical Genetics&Hunan Provincial Key Laboratory of Medical Genetics,School of Life Sciences,Central South University,Changsha 410008,China

b Department of Neurology,Beth Israel Deaconess Medical Center&Harvard Medical School,Boston,MA 02115,USA

c Department of Neurology,Xuanwu Hospital,Capital Medical University,Beijing 100053,China

d Department of Neurology,Xiangya Hospital,Central South University,Changsha 410008,China

e School of Physical Education,Henan University,Kaifeng 475001,China

f School of Medicine,South China University of Technology,Guangzhou 510006,China

g Department of Biology,College of Arts&Sciences,Boston University,Boston,MA 02215,USA

h CAS Center for Excellence in Brain Science and Intelligence Technology,Shanghai 200031,China

Abstract Background:Physical activity has been hypothesized to play a protective role in neurodegenerative diseases.However,effect estimates previously derived from observational studies were prone to confounding or reverse causation.Methods:We performed a two-sample Mendelian randomization (MR) analysis to explore the causal association of accelerometer-measured physical activity with 3 common neurodegenerative diseases:Alzheimer’s disease(AD),Parkinson’s disease(PD),and amyotrophic lateral sclerosis(ALS).We selected genetic instrumental variants reaching genome-wide signif icance(p ＜5×10－8)from 2 largest meta-analyses of about 91,100 UK Biobank participants.Summary statistics for AD,PD,and ALS were retrieved from the up-to-date studies in European ancestry led by the international consortia.The random-effect,inverse-variance weighted MR was employed as the primary method,while MR pleiotropy residual sum and outlier (MR-PRESSO),weighted median,and MR-Egger were implemented as sensitivity tests.All statistical analyses were performed using the R programming language(Version 3.6.1;R Foundation for Statistical Computing,Vienna,Austria).Results:Primary MR analysis and replication analysis utilized 5 and 8 instrumental variables,which explained 0.2%and 0.4%variance in physical activity,respectively.In each set,one variant at 17q21 was signif icantly associated with PD,and MR sensitivity analyses indicated them it as an outlier and source of heterogeneity and pleiotropy.Primary results with the removal of outlier variants suggested odds ratios(ORs)of neurodegenerative diseases per unit increase in objectively measured physical activity were 1.52 for AD (95% conf idence interval (95%CI):0.88－2.63,p=0.13) and 3.35 for PD (95%CI:1.32－8.48,p=0.01),while inconsistent results were shown in the replication set for AD(OR=1.06,95%CI:1.01－1.12,p=0.02)and PD(OR=0.99,95%CI:0.88－0.12,p=0.97).Similarly,the benef icial effect of physical activity on ALS(OR=0.51,95%CI:0.29－0.91,p=0.02)was not conf irmed in the replication analysis(OR=0.96,95%CI:0.91－1.02,p=0.22).Conclusion:Genetically predicted physical activity was not robustly associated with risk of neurodegenerative disorders.Triangulating evidence across other studies is necessary in order to elucidate whether enhancing physical activity is an effective approach in preventing the onset of AD,PD,or ALS.

Keywords:Alzheimer’s disease;Amyotrophic lateral sclerosis;Genetic epidemiology;Mendelian randomization;Parkinson’s disease;Physical activity

1.Introduction

Neurodegenerative diseases have become a major health burden around the world.Despite numerous advances and progress in the field of molecular biology,genetics,and pharmaceutical research,there are still no effective drugs to halt or retard neurodegeneration.Most intervention trials targeting Alzheimer’s disease (AD),Parkinson’s disease (PD),and amyotrophic lateral sclerosis (ALS) have failed thus far.1－3Identifying modif iable risk factors and seeking potential treatments to slow down the progression or even prevent the onset of these diseases are essential.Physical activity is one of these factors and has recently aroused much research interest.Previous observational cohorts aiming to study the relationship between physical activity and neurodegenerative diseases have mostly yielded inconsistent results.Emerging studies have explored the relationship between physical activity,cognitive performance,and the risk of developing AD in the elderly,only to end in insuff icient evidence supporting its protective role.4,5Similarly,an increasing amount of research has investigated whether exercise has a negative or benef icial effect on the risk of PD6,7and ALS8,9and whether exercise affects motor performance,but thus far this research has failed to draw convincing conclusions.

AD is the most common neurodegenerative disease among the elderly.Emerging evidence demonstrates the potential protective effect that physical activity can have on AD.A large cohort study that included 404,840 cases found that physical inactivity was associated with increased incidence of different kinds of dementia,including AD.However,the study found that the relationship no longer existed when the measurement of physical activity occurred less than 10 years before the dementia diagnosis.10Another prospective cohort study that included 10,308 cases and an average 27-year follow-up found no evidence of neuroprotective effects of physical activity on the risk of AD.11

As early as 1992,a clinical report showed exercise was associated with a lower risk of PD diagnosis later in their life in a male university cohort,12and more recent evidence shows that exercise greatly reduces the risk of PD.The famous National Institutes of Health-American Association of Retired Persons Diet and Health Study cohort,which recruited 213,701 participants with a wide age range,showed that those who had moderate-to-vigorous physical activity had signif icantly reduced the risk of PD.6Likewise,a recent meta-analysis revealed an inverse dose－response association between physical activity and PD risk.13However,another study suggested that this neuroprotective effect of physical activity was more prominent in males than females.7

The relationship between physical activity and the risk of ALS has not been fully elucidated yet.In the beginning,people noticed a positive correlation between ALS and strenuous exercise for the reason that ALS is quite common in professional athletes.8Although this study found that ALS patients had slightly higher levels of physical exercise,there is no evidence that an extremely increased duration or level of physical activity is associated with ALS.8A recently published study using instrumental variables for self-reported physical activity showed that the risk of ALS was negatively correlated with light levels of physical activity,but it was positively correlated with more strenuous moderate levels of physical activity.14Inconsistent results from longitudinal studies8,14indicate that better methodologies are desperately needed in order to elucidate the relationship between physical activity and neurodegenerative diseases.

Traditional measurement of physical activity based on self-reported exercise patterns incurs bias to some extent.Self-reported questionnaires may represent what individuals should do in terms of exercise rather than how routinely they actually exercise.However,advancing technologies,such as wearable trackers and statistical machine learning,provide an opportunity to implement a relatively objective measure of physical activity in free-living environments.15－18Elucidating causal associations when using traditional observational designs,even in randomized clinical trials,is indeed prone to various biases due to restricted sample sizes and ethical and financial challenges.Therefore,high-quality evidence is still largely warranted.Recently,genome-wide association studies(GWAS)have been conducted by large international consortia to identify signif icant loci for objectively measured physical activity19,20and neurodegenerative diseases.21－25Meanwhile,Mendelian randomization (MR) design utilizing valid instrumental variants has been applied as a robust approach to making causal inference.26－30In our study,we implemented a two-sample MR design to investigate the role of physical activity in 3 common neurodegenerative diseases:AD,PD,and ALS.

2.Methods

2.1.Summary-level data for physical activity

We constructed 2 sets of instrumental variables for accelerometer-measured physical activity (Table 1 and Supplementary Table 1).For the primary analysis,summary statistics for physical activity were obtained from a recent GWAS,which identif ied 5 single nucleotide polymorphisms (SNPs) associated with objectively measured physical activity reaching genome-wide signif icance (p ＜5×10－8).19For the replication analysis,8 genome-wide signif icant SNPs from another GWAS (Klimentidis et al.20) were utilized.Two original GWAS were conducted independently with approximately 91,100 European participants in the UK Biobank.One of the 2 GWAS measured overall activity level as average acceleration per each 30-s epoch,while the other measured it in a 5-s window.18,20Both instrumental sets for genetically predicted physical activity have been validated and employed in several recent MR studies.16,18,31Raw data for physical activity were initially collected from activity sensors for a 7-day period;measurement on free-living ambient conditions should be well representative of daily exercise routine of individuals to some extent.A statistical machine-learning model23has been validated and employed to integrate sleep,sedentary behavior,and various levels of exercise intensity and yields an indicator for overall physical activity:average acceleration (milligravity).The GWAS results were adjusted for principal covariates like sex,age,age squared,and season;effect size was interpreted as unit change in standard deviation of physical activity per additional risk allele,which is roughly equal to 75 min of moderate activity(i.e.,fast walking)each day in place of sedentary behaviors.18,19

Table1Demographic information for summary-level datasets of physical activity and 3 common neurodegenerative diseases used in the Mendelian randomization study.

2.2.Instrumental variable selection

Selected instrumental SNPs were examined to determine whether they satisf ied 3 MR assumptions (Fig.1).Brief ly,these assumptions were that genetic variants should (1) be robustly associated with physical activity,(2) be unrelated to factors confounding the exposure-outcome relationship,and(3) inf luence the risk of neurodegenerative diseases only via its effects on physical activity.First,2 sets of SNPs were associated with accelerometer-measured physical activity at genome-wide signif icance,thus conforming to the relevance assumption.They were also validated to be independent and not in linkage disequilibrium(threshold set at r2=0.001 within the window of 10 mega base pairs).Second,compared to randomized controlled trails,a similar randomized allocation in MR studies is realized,mediated by genetic proxies in which a set of alleles is associated with higher or lower physical activity levels.Genetic variants,which are set in gamete formation,are an ideal tool for randomization of exposure of interest;and confounding factors and reverse causation,which are often found in traditional observational studies,hardly exists here.Additionally,our MR study only incorporated Europeanancestry samples,and population stratif ication could be excluded from violating the independence assumption.Lastly,we looked up these instrumental SNPs in the GWAS catalog.32Considering the limited number of instrumental SNPs in our study,we initially included all of them into the liberal analyses and then implemented several sensitivity analyses to investigate horizontal pleiotropic effects and the robustness of the MR findings.

2.3.Outcome data sources

Summary data on genetic associations with the 3 neurodegenerative diseases were extracted from 3 GWAS meta-analyses in which Europeans were the dominant participants.21,24,25AD summary statistics were retrieved from the most recent GWAS of clinically diagnosed,late-onset AD patients;the GWAS incorporated 21,982 cases and 41,944 controls.PD datasets were released by the International Parkinson’s Disease Genomics Consortium (https://pdgenetics.org);these datasets consisted of 33,674 cases and 449,056 controls.GWAS data on ALS were accessed from the ALS Variant Server (http://als.umassmed.edu)and included 20,806 cases and 59,804 controls.All summary statistics were accessed through publicly shared websites or through sources that were available on request from the consortia (Table 1).For instrumental SNPs whose equivalent variants were not present in the outcome GWAS datasets,proxy SNPs (r2＞0.9,according to the European panel,1000 Genomes Project Phase 3) were searched in LDlink33(https://ldlink.nci.nih.gov/?tab=home)and adopted for MR analyses (Supplementary Table 2).Several variants were palindromic,yet they were not excluded since their minor allele frequency was lower than 0.36,which wouldn’t incur ambiguity when inferring the strand.Finally,we harmonized the exposure and outcome datasets for subsequent analysis.

Fig.1.Schematic of the Mendelian randomization study and key assumptions.First,genetic instrumental variants associated with accelerometer-measured physical activity at genome-wide signif icance (p ＜5×10－8) all satisf ied the relevance assumption.Second,there scarcely existed any factors confounding the natural randomization in gamete formation to violate the independence assumption.Lastly,potential horizontal pleiotropic effects violating the exclusionrestriction assumption were inspected.AD=Alzheimer’s disease;ALS=amyotrophic lateral sclerosis;PD=Parkinson’s disease;SNPs=single-nucleotide polymorphisms.

2.4.Statistical analysis

We conducted a two-sample MR using the R programming language (Version 3.6.1;R Foundation for Statistical Computing,Vienna,Austria),with the TwoSampleMR (Version 0.4.26) and MR pleiotropy residual sum and outlier(MR-PRESSO;Version 1.0.0) packages.34,35First,the Wald ratio for each instrumental SNP was derived by dividing the outcome-association effect size (β per additional risk allele)by the corresponding exposure-association coeff icient.Then,the pooled estimates were calculated by 1 primary MR approach,the inverse-variance weighted method,and 3 sensitivity tests,MR-PRESSO,weighted median,and MR-Egger.The inverse-variance weighted estimate was based on a random-effect meta-analysis to integrate the causal effects of individual SNPs,whereas Cook’s distance (Cook’s D ＞4/number of SNPs indicates signif icant heterogeneity) and Cochran’s Q test were employed to evaluate heterogeneous effects within instrumental SNPs.The weighted median estimator pooled the effects of individual variants eff iciently under the prerequisite that more than 50%of the weight came from valid instrumental variables.MR-PRESSO was capable of identifying pleiotropic effects and outlier SNPs,which yielded overall estimates adjusted for them.The MR-Egger regression intercept and MR-PRESSO global test were also implemented to examine horizontal pleiotropy.Lastly,we used the web-tool mRnd35(https://shiny.cnsgenomics.com/mRnd/)to calculate a priori power for a given odds ratio(OR)scenario.

3.Results

For the primary analysis,instrumental variables from Doherty et al.19collectively explained about 0.2% of the variance in accelerometer-measured physical activity.For the 8 SNPs in the replication analysis,the variables from Klimentidis et al.20explained about 0.4% of the variance (Supplementary Table 1).The strength of instrumental SNPs was measured by the F-statistic (＜10,deemed as a weak instrument),and no weak strength bias (F-statistic ranged from 27 to 60) seemingly presented.Notably,rs2696625 in the primary instrument set and rs55657917 in the replication set both are located at 17q21.31,a previously identif ied locus for PD.36To minimize the distortion to MR estimates due to heterogeneity and pleiotropy,we also conducted analyses after removal of the outlier in each instrument set.Look-up in the GWAS catalog(Supplementary Table 3) indicated no pleiotropy for primary instrumental SNPs.Three SNPs in the replication analysis were associated with other traits;nevertheless,we didn’t preclude them from the instrument set.Instead,we examined their pleiotropic effects and effects on causal estimates,also through MR sensitivity analyses.After inputting required parameters in mRnd,35the minimum detectable OR was roughly estimated(Supplementary Table 4).Specif ically,our study would be underpowered to detect the OR intervals of 0.57－1.61 for AD,0.66－1.36 for PD,and 0.57－1.56 for ALS.

3.1.Physical activity and AD

Overall,there was no causal relationship between accelerometer-measured physical activity and AD.MR analysis(Table 2,Fig.2,and Supplementary Table 4) indicated that risk for AD was 1.03 (95% conf idence interval (95%CI):0.48－2.21,p=0.94) per one-unit increase in genetically predicted physical activity level in the primary analysis and 1.03(95%CI:0.96－1.10,p=0.32) in the replication analysis.Cochran’s Q test (pHet=0.04),MR-PRESSO global test(pRss=0.03),MR-Egger regression intercept (pInt=0.06),and Cook’s D(62.10,＞4/number of SNPs indicates heterogeneity)consistently showed that rs2696625 unproportionally exerted an inf luence on the overall MR estimate in the primary analysis (Supplementary Tables 5－8).After removal of the outlier SNP,heterogeneity within the remaining instrumental SNPs(pHet=0.80;pRss=0.80;and pInt=0.65) went down,but it didn’t support the effect of objectively measured physical activity on AD (OR=1.52,95%CI:0.88－2.63,p=0.13).Likewise,heterogeneity incurred by rs55657917 in the replication instrument set was investigated through sensitivity analyses (Supplementary Tables 5－8 and SupplementaryFigs.1－3),but it was not as signif icant (pHet=0.08;pRss=0.08;and pInt=0.58)as rs2696625 in the primary analysis.With the removal of rs55657917 in the replication analysis,a weak association between physical activity and AD(OR=1.06,95%CI:1.01－1.12,p=0.02)was shown.

Table2Mendelian randomization estimates for the effects of physical activity on 3 neurodegenerative diseases.

Fig.2.Mendelian randomization analysis of the effect of physical activity on AD.Scatter plots depict the genetic variant－physical activity effect(point and horizontal line)vs.the genetic variant－AD effect(point and horizontal line).The fitted line denotes the overall estimate given by the IVW method with all instrumental variants included (solid line) or after the removal of certain variant (dashed line).Regarding the 2 variants signif icantly associated with Parkinson’s disease,rs2696625 in the primary analysis(A)showed signif icant heterogeneity(Cook’s D=62.10;Cochran’s pHet=0.04),while rs55657917 in the replication set(B)presented minimal heterogeneity (Cook’s D=0.95;Cochran’s pHet=0.08).AD=Alzheimer’s disease;Cook’s D=Cook’s distance;IVW=inverse-variance weighted;OR=odds ratio;SNP=single-nucleotide polymorphism.

3.2.Physical activity and PD

Genetically predicted physical activity was not associated with PD either in the primary analysis(OR=0.81,95%CI:0.43－1.50,p=0.51)or replication analysis(OR=1.11,95%CI:0.88－1.40,p=0.37).Notably,rs2696625 and rs55657917 were strongly associated with PD (p=2.95 × 10－19and p=9.63 × 10－20,respectively) in the dataset (Table 2,Fig.3,and Supplementary Table 2).They showed evident heterogeneity both in the primary and replication analysis(pHet＜0.001 and pRss＜0.001,respectively).Cook’s D for rs2696625 and rs55657917(Supplementary Table 9) denoted that they are potential outliers (＞4/number of instrumental SNPs)as well.A causal association between physical activity and PD was present in the primary analysis after excluding rs2696625 (OR=3.35,95%CI:1.32－8.48,p=0.01),whereas the non-null effect was not shown in the replication analysis after excluding s55657917 (OR=0.99,95%CI:0.88－1.12,p=0.97).

3.3.Physical activity and ALS

On the whole,the association of genetically predicted physical activity with ALS was not consistently verif ied in our MR study (Table 2 and Fig.4).Accelerometer-measured physical activity was related to risk of ALS in the primary analysis(OR=0.45,95%CI:0.27－0.74,p ＜0.001),but this was not conf irmed in the replication analysis (OR=0.95,95%CI:0.90－1.00,p=0.06).Sensitivity analyses (Supplementary Tables 5－8) did not indicate heterogeneous or pleotropic effects of rs2696625 and rs55657917 in the primary and replication analysis(pHet,pRss,and pInt,all ＞0.29).Cook’s D was 0.33 (＜4/5) for rs2696625 and 0.32 (＜4/8) for rs55657917(Supplementary Table 10),which indicates that there were no outlying effects,even though some outlying effects were found in the analysis of physical activity on AD and PD.After the removal of rs2696625 in the primary analysis,the protective effect of physical activity on ALS existed(OR=0.51,95%CI:0.29－0.91,p=0.02).However,in the replication analysis without rs55657917,the association was not conf irmed(OR=0.96,95%CI:0.91－1.02,p=0.22).

4.Discussion

Neurodegenerative diseases,including AD,PD,and ALS,have largely unknown etiologies and thus are deemed as incurable chronic diseases.However,researchers are nonetheless dedicated to seeking potential disease-modifying treatments.Many observational studies5－12seeking to establish the relationship between physical activity and neurodegenerative diseases have yielded inconsistent results.Establishing associations using a clinical-trial design is indeed prone to various biases brought about by restricted sample sizes and ethical and financial challenges.To strengthen the causal inference,we therefore conducted an MR study.Our results suggest that physical activity might act as a protective factor for ALS,but no causal associations were found between physical activity and AD or PD.

Fig.3.Mendelian randomization analysis of the effect of physical activity on PD.Scatter plots depict the genetic variant－physical activity effect(point and horizontal line)vs.the genetic variant－PD effect(point and horizontal line).The fitted line denotes the overall estimate given by the IVW method with all instrumental variants included(solid line)or after the removal of certain variant(dashed line).Both in the primary(A)and replication(B)analysis,each variant at 17q21.3 manifested signif icant heterogeneity (rs2696625,Cook’s D=42.36;Cochran’s pHet ＜0.001;and rs55657917,Cook’s D=0.53;Cochran’s pHet ＜0.001).Cook’s D=Cook’s distance;IVW=inverse-variance weighted;OR=odds ratio;PD=Parkinson’s disease;SNP=single-nucleotide polymorphism.

Physical activity was first considered as a risk factor for ALS based on the observation of its higher incidence among top athletes.37Other studies have supported this hypothesis in that their results showed that more top athletes were diagnosed with ALS than healthy controls.For example,a study showed more weight loss and leaner figures during the premorbid stage among varsity athletes.38However,a population-based study involving 636 patients with sporadic ALS and 2166 controls failed to support the hypothesis that an excess of extreme exercise such as marathon running or that occupations requiring extreme energy use increased the incidence of ALS,although some participants with ALS had slightly higher levels of leisure-time physical activity.8

Fig.4.Mendelian randomization analysis of the effect of physical activity on ALS.Scatter plots depict the genetic variant－physical activity effect(point and horizontal line)vs.the genetic variant－ALS effect(point and horizontal line).The fitted line denotes the overall estimate given by the IVW method with all instrumental variants included (solid line) or after the removal of certain variant (dashed line).Either in the primary (A) or replication (B) instrumental set,the 17q21.3 variant showed negligible heterogeneity(rs2696625,Cook’s D=0.33;Cochran’s pHet=0.55;and rs55657917,Cook’s D=0.32;Cochran’s pHet=0.29).ALS=amyotrophic lateral sclerosis;Cook’s D=Cook’s distance;IVW=inverse-variance weighted;OR=odds ratio;SNP=single-nucleotide polymorphism.

Recall bias,of course,is one of the most important factors to consider.It is hard to conclude that exercise simply leads to a higher risk of ALS since multiple injuries,such as head trauma,are inevitable during strenuous exercise.One recent community-based study exploring the incidence of developing ALS among varsity high school football players from 1956 to 1970 failed to find an increased risk of ALS.39One recent MR study8exploring the causal relationship between different self-reported types of physical activity and ALS revealed that light-intensity physical activity was negatively associated with the risk of ALS.However,the subjective nature of the participants’self-reported physical activity was subject to heterogeneity and measurement errors due to the fact that the study did not use clearly def ined measures of exercise,like strength and frequency.In our study,we did find a protective effect on ALS risk from objectively measured physical activity based on measurements from wearable accelerometers and our evaluation system using 5 SNPs as instrumental variables.Our results are also supported by a recent prospective cohort study,which found that total physical activity was inversely associated with ALS mortality,with those who were physically active being 33% less likely to die from ALS compared to those who were inactive,although the results were borderline statistically signif icant across categories (p=0.042).9However,we failed to replicate these results using the 8 SNPs as instrumental variables.The conf licting results may have occurred because the 8 SNPs involved the use of devices that were worn by participants for only 72 h,which may be less accurate than devices worn for 7 days.Therefore,it is still not possible to firmly conclude that there is an association between physical activity and the risk of ALS.

Our study did not conf irm a causal relationship between physical activity and PD or AD.Although we used a primary MR method and sensitivity analyses,we found consistent evidence for a null association of physical activity with PD or AD,and signif icant heterogeneity and inadequate power were issues that undermined the findings of our study.Additional studies are warranted in order to elucidate whether an increase in physical activity is an effective approach to preventing the onset of PD or AD.

The SNPs rs2696625 and rs55657917 were identif ied as outlier SNPs that led to increased pleiotropy in the overall MR estimate.Both of the SNPs are located at 17q21.31,a previously reported signif icant locus for various kinds of neurodegenerative diseases.40The 17q21.31 locus was unusual because of its pattern of long-range linkage disequilibrium.This major genetic risk factor is an extended H1 haplotype on chromosome 17q21.31,which includes microtubule-associated protein tau (MAPT).41Recently,a large GWAS study revealed 3 independent regions associated with PD(p ＜5 × 10－8),one of which is the 17q21/MAPT.Another chromosome region has been strongly correlated with the 17q21/MAPT region and is associated with the onset of PD.42This may partially explain the role of rs2696625 and rs55657917 in neurodegenerative diseases.However,their biological functions are still largely unknown.

One major advantage of our study lies in its two-sample MR design,which helps circumvent measurement errors,residual confounding,and reverse causation.But our study also has several limitations.First,the exposure data from wrist-worn accelerometers used for only 1 week may not well represent regular or lifelong physical activity patterns.Second,there are only a few SNPs associated with physical activity.These were used as instrumental variables and only accounted for a relatively small proportion of variance,which may have caused distortion to the MR estimate.Third,several proxy SNPs were utilized,and the biological implications of these SNPs are largely unknown.Thus,the strength of evidence given by the MR analysis is undermined.Lastly,given that our findings are mainly restricted to participants having European ancestry,caution should be used in interpreting our results and generalizing them to other populations.

5.Conclusion

Genetically predicted physical activity was not robustly associated with risk of neurodegenerative disorders.Triangulating evidence across other studies is necessary to elucidate whether enhancing physical activity is an effective approach in preventing the onset of AD,PD,and ALS.

Acknowledgments

The authors gratefully thank Dr Aiden Doherty,Dr Yann Klimentidis,Dr Brian Kunkle,Dr Aude Nicolas,Dr Mike Nalls,and all investigators and related consortia for sharing GWAS summary statistics on physical activity,Alzheimer’s disease,Parkinson’s disease,and amyotrophic lateral sclerosis.

This work was supported by the Natural Science Foundation of China (81525007 and 81730036) and Key R&D Program of Hunan Province (2019SK2051).PFW and XZ received a visiting PhD scholarship from the China Scholarship Council.

Authors’contributions

PFW played a role in conceiving and designing the study,overseeing and performing data collection,project administration,formal analysis,and drafting and editing this manuscript;HL played a role in conceiving and designing the study,project administration,and funding acquisition;XZ played a role in conceiving and designing the study and drafting and editing this manuscript;XL played a role in performing data collection and drafting this manuscript;RL played a role in formal analysis and drafting this manuscript;WZ played a role in performing data collection and editing this manuscript;DL played a role in formal analysis and editing this manuscript;KX played a role in conceiving and designing the study,funding acquisition,and drafting and editing this manuscript.All authors have read and approved the final version of the manuscript,and agree with the order of presentation of the authors.

Competing interests

The authors declare that they have no competing interests.

Supplementary materials

Supplementary material associated with this article can be found in the online version at doi:10.1016/j.jshs.2021.01.008.