Advertisement

Gut microbiome in Parkinson's disease: New insights from meta-analysis

  • Tzi Shin Toh
    Affiliations
    Division of Neurology, Department of Medicine, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia

    Mah Pooi Soo & Tan Chin Nam Centre for Parkinson's & Related Disorders, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia
    Search for articles by this author
  • Chun Wie Chong
    Correspondence
    Corresponding author. Building 2, Level 5, Room 31, School of Pharmacy, Monash University Malaysia, Jalan Lagoon Selatan, 47500, Bandar Sunway, Selangor, Malaysia.
    Affiliations
    School of Pharmacy, Monash University Malaysia, Selangor, Malaysia
    Search for articles by this author
  • Shen-Yang Lim
    Affiliations
    Division of Neurology, Department of Medicine, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia

    Mah Pooi Soo & Tan Chin Nam Centre for Parkinson's & Related Disorders, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia
    Search for articles by this author
  • Jeff Bowman
    Affiliations
    Integrative Oceanography Division, Scripps Institution of Oceanography, University of California, California, USA

    Center for Microbiome Innovation, UC San Diego, California, USA
    Search for articles by this author
  • Mihai Cirstea
    Affiliations
    Department of Microbiology and Immunology, University of British Columbia, Vancouver, British Columbia, Canada

    Michael Smith Laboratories, UBC, Vancouver, British Columbia, Canada
    Search for articles by this author
  • Chin-Hsien Lin
    Affiliations
    Department of Neurology, National Taiwan University Hospital, College of Medicine, National Taiwan University, Taipei, Taiwan
    Search for articles by this author
  • Chieh-Chang Chen
    Affiliations
    Division of Gastroenterology and Hepatology, Department of Internal Medicine, National Taiwan University Hospital, College of Medicine, National Taiwan University, Taipei, Taiwan

    Graduate Institute of Clinical Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan
    Search for articles by this author
  • Silke Appel-Cresswell
    Affiliations
    Pacific Parkinson's Research Centre and Djavad Mowafaghian Centre for Brain Health, UBC, Vancouver, British Columbia, Canada

    Division of Neurology, Faculty of Medicine, UBC, Vancouver, British Columbia, Canada
    Search for articles by this author
  • B. Brett Finlay
    Affiliations
    Department of Microbiology and Immunology, University of British Columbia, Vancouver, British Columbia, Canada

    Michael Smith Laboratories, UBC, Vancouver, British Columbia, Canada

    Department of Biochemistry and Molecular Biology, UBC, Vancouver, British Columbia, Canada
    Search for articles by this author
  • Ai Huey Tan
    Correspondence
    Corresponding author. Neurology Laboratory, 6th Floor, South Tower, University of Malaya Medical Centre, 50603, Kuala Lumpur, Malaysia.
    Affiliations
    Division of Neurology, Department of Medicine, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia

    Mah Pooi Soo & Tan Chin Nam Centre for Parkinson's & Related Disorders, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia
    Search for articles by this author

      Highlights

      • Study methodology and geography have larger effect than PD status on gut microbiome.
      • After confounders analyses, gut microbiome alterations in PD remained significant.
      • Gut microbiome of Caucasian PD patients and controls differs from non-Caucasians.
      • Increased Akkermansia and reduced Roseburia were consistently found in PD.
      • Several bacterial taxa correlated with motor and cognitive function in PD.

      Abstract

      Background

      Gut microbiome alterations have been reported in Parkinson's disease (PD), but with heterogenous findings, likely due to differences in study methodology and population. We investigated the main microbiome alterations in PD, their correlations with disease severity, and the impact of study and geographical differences.

      Methods

      After systematic screening, raw 16S rRNA gene sequences were obtained from ten case-control studies totaling 1703 subjects (969 PD, 734 non-PD controls; seven predominantly Caucasian and three predominantly non-Caucasian cohorts). Quality-filtered gene sequences were analyzed using a phylogenetic placement approach, which precludes the need for the sequences to be sourced from similar regions in the 16S rRNA gene, thus allowing a direct comparison between studies. Differences in microbiome composition and correlations with clinical variables were analyzed using multivariate statistics.

      Results

      Study and geography accounted for the largest variations in gut microbiome composition. Microbiome composition was more similar for subjects from the same study than those from different studies with the same disease status. Microbiome composition significantly differed between Caucasian and non-Caucasian populations. After accounting for study differences, microbiome composition was significantly different in PD vs. controls (albeit with a marginal effect size), with several distinctive features including increased abundances of Megasphaera and Akkermansia, and reduced Roseburia. Several bacterial genera correlated with PD motor severity, motor response complications and cognitive function.

      Conclusion

      Consistent microbial features in PD merit further investigation. The large variations in microbiome findings of PD patients underscore the need for greater harmonization of future research, and personalized approaches in designing microbial-directed therapeutics.

      Keywords

      1. Introduction

      Emerging evidence suggests a bidirectional link between the gut and brain, with gut microbes believed to play a central role in this connection [
      • Lubomski M.
      • Tan A.H.
      • Lim S.Y.
      • Holmes A.J.
      • Davis R.L.
      • Sue C.M.
      Parkinson's disease and the gastrointestinal microbiome.
      ]. Recent studies have linked alterations in the gut microbiome to Parkinson's disease (PD), although findings have been quite heterogenous [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ,
      • Romano S.
      • Savva G.M.
      • Bedarf J.R.
      • Charles I.G.
      • Hildebrand F.
      • Narbad A.
      Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
      ]. Most studies were limited in sample size and geographically restricted [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ], thus limiting the generalizability of results. Further complicating interpretation across studies were differences in study methodology and populations. Cross-study comparisons using a standardized workflow are crucial to unveil consistent features of gut microbiome that may play important roles in PD pathogenesis and have potential implications for therapeutic development.
      Notably, meta-analysis of gut microbiome studies can be methodologically challenging. Original studies sequenced different regions of the bacterial 16S rRNA gene for bacterial identification, making it impossible to directly compare sequences using the conventional operational taxonomic units (OTUs)-based method. In attempting to overcome this limitation, previous OTUs-based meta-analyses therefore either statistically controlled for differences in sequencing regions [
      • Romano S.
      • Savva G.M.
      • Bedarf J.R.
      • Charles I.G.
      • Hildebrand F.
      • Narbad A.
      Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
      ] or excluded bacterial taxa that were found to be highly different among studies [
      • Nishiwaki H.
      • Ito M.
      • Ishida T.
      • Hamaguchi T.
      • Maeda T.
      • Kashihara K.
      • Tsuboi Y.
      • Ueyama J.
      • Shimamura T.
      • Mori H.
      Meta‐analysis of gut dysbiosis in Parkinson's disease.
      ]. Furthermore, clinical correlations were not explored in these studies.
      To enable more direct comparisons between microbiome studies, we systematically reviewed and meta-analyzed case-control 16S rRNA gene sequencing studies using a relatively novel ‘phylogenetic placement method’ [
      • Barbera P.
      • Kozlov A.M.
      • Czech L.
      • Morel B.
      • Darriba D.
      • Flouri T.
      • Stamatakis A.
      EPA-ng: massively parallel evolutionary placement of genetic sequences.
      ] which precludes the need for sequences to be sourced from similar regions in the 16S rRNA gene. We aimed to discover the most consistent gut microbiome features in PD, and to investigate the impact of study and geographical differences on PD gut microbiome. Besides including a larger number of studies, this meta-analysis encompassed a substantial number of non-Caucasian subjects underrepresented in previous work. Additional clinical data from three large studies were also obtained to evaluate associations between microbial features and PD severity.

      2. Methods

      2.1 Systematic review and screening

      A PubMed search was conducted with no language restriction for publications from January 1, 2000 to May 1, 2020, using the search terms: (“Microbiota” OR “Microbiome” OR “Microflora” OR “Dysbiosis”) AND (“Parkinson” OR “Parkinsonism”). Study inclusion criteria were: (1) human case-control studies involving PD patients; (2) microbial DNA extraction from stool; and (3) sequencing of the 16S rRNA gene using next-generation sequencing. The article screening process is summarized in a PRISMA flow diagram (Supplementary Fig. S1).

      2.2 Sequence acquisition and data processing

      For eligible studies that have deposited sequencing data in the NCBI GenBank database, raw 16S rRNA gene sequences and subjects’ information were retrieved using the BioProject accession numbers provided in the publications. For eligible studies with sequencing data not publicly available, written requests were made to the respective corresponding authors, and raw 16S rRNA gene sequences were acquired directly. The raw 16S rRNA gene sequences from different studies were quality-filtered separately using standardized parameters and subsequently classified taxonomically into bacterial features using paprica [
      • Bowman J.S.
      • Ducklow H.W.
      Microbial communities can be described by metabolic structure: a general framework and application to a seasonally variable, depth-stratified microbial community from the coastal West Antarctic Peninsula.
      ].

      2.3 Statistical analysis

      All statistical analyses were conducted with R version 3.6.3. Using the microbiome R package [
      • Lahti L.
      • Shetty S.
      Introduction to the Microbiome R Package.
      ], alpha diversity (i.e., the diversity within a particular ecosystem) was inferred using three indices: (1) Shannon diversity index which takes into account both richness (i.e., number of species in a community) and evenness (i.e., the extent to which species within a community are held in even abundance with one another); (2) Chao1 diversity index which measures the richness; and (3) Pielou's evenness index which measures the evenness; higher index correlates with higher diversity. Beta diversity (i.e., a comparison of diversity between ecosystems; gut microbiome compositional differences between different groups of subjects) was assessed using principal coordinate analysis (PCoA) and Aitchison distance-based permutational multivariate ANOVA (PERMANOVA). PCoA plot reduces the dimensionality of microbiome data and enables easy visualization to compare the beta diversity relationships between groups [
      • Goodrich J.K.
      • Di Rienzi S.C.
      • Poole A.C.
      • Koren O.
      • Walters W.A.
      • Caporaso J.G.
      • Knight R.
      • Ley R.E.
      Conducting a microbiome study.
      ], while PERMANOVA statistically determines whether the between-group differences (beta diversity) are significant. Bar plots and heat maps were used to illustrate the differences (and similarities) in distribution of bacterial taxa across groups. Differentially abundant bacterial taxa/pathways were first identified using the differential gene expression analysis based on the negative binomial distribution (DESeq2) R package [
      • Love M.I.
      • Huber W.
      • Anders S.
      Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2.
      ], and further confirmed using the analysis of compositions of microbiomes with bias correction (ANCOM-BC) [
      • Lin H.
      • Peddada S.D.
      Analysis of compositions of microbiomes with bias correction.
      ] and Wilcoxon-Mann-Whitney (WMW) [
      • Hothorn T.
      • Hornik K.
      • Van De Wiel M.A.
      • Zeileis A.
      A lego system for conditional inference.
      ] approaches. Differences in “study” was controlled as a confounder in all three methods. The DESeq2 method identifies differentially abundant features using the Wald test and generalized linear models, while the ANCOM-BC approach is based on the linear regression framework and corrects the bias induced by the differences among samples. Correlations between bacterial genera and severity scores in the PD subgroup were evaluated based on Spearman rank correlation with false-discovery rate (FDR)-adjusted p-value < 0.05. Detailed description of the methodology is provided in Supplementary Methods.

      3. Results

      After screening and filtering, ten studies (Fig. 1) comprising 1703 subjects (969 PD patients and 734 non-PD controls) from nine countries (seven studies from predominantly Caucasian populations [
      • Cirstea M.S.
      • Yu A.C.
      • Golz E.
      • Sundvick K.
      • Kliger D.
      • Radisavljevic N.
      • Foulger L.H.
      • Mackenzie M.
      • Huan T.
      • Finlay B.B.
      Microbiota composition and metabolism are associated with gut function in Parkinson's disease.
      ,
      • Heintz‐Buschart A.
      • Pandey U.
      • Wicke T.
      • Sixel‐Döring F.
      • Janzen A.
      • Sittig‐Wiegand E.
      • Trenkwalder C.
      • Oertel W.H.
      • Mollenhauer B.
      • Wilmes P.
      The nasal and gut microbiome in Parkinson's disease and idiopathic rapid eye movement sleep behavior disorder.
      ,
      • Hopfner F.
      • Künstner A.
      • Müller S.H.
      • Künzel S.
      • Zeuner K.E.
      • Margraf N.G.
      • Deuschl G.
      • Baines J.F.
      • Kuhlenbäumer G.
      Gut microbiota in Parkinson disease in a northern German cohort.
      ,
      • Hill‐Burns E.M.
      • Debelius J.W.
      • Morton J.T.
      • Wissemann W.T.
      • Lewis M.R.
      • Wallen Z.D.
      • Peddada S.D.
      • Factor S.A.
      • Molho E.
      • Zabetian C.P.
      Parkinson's disease and Parkinson's disease medications have distinct signatures of the gut microbiome.
      ,
      • Aho V.T.
      • Pereira P.A.
      • Voutilainen S.
      • Paulin L.
      • Pekkonen E.
      • Auvinen P.
      • Scheperjans F.
      Gut microbiota in Parkinson's disease: temporal stability and relations to disease progression.
      ,
      • Pietrucci D.
      • Cerroni R.
      • Unida V.
      • Farcomeni A.
      • Pierantozzi M.
      • Mercuri N.B.
      • Biocca S.
      • Stefani A.
      • Desideri A.
      Dysbiosis of gut microbiota in a selected population of Parkinson's patients.
      ,
      • Petrov V.
      • Saltykova I.
      • Zhukova I.
      • Alifirova V.
      • Zhukova N.
      • Dorofeeva Y.B.
      • Tyakht A.
      • Kovarsky B.
      • Alekseev D.
      • Kostryukova E.
      Analysis of gut microbiota in patients with Parkinson's disease.
      ], three from predominantly non-Caucasian [Asian] populations [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ,
      • Lin C.H.
      • Chen C.C.
      • Chiang H.L.
      • Liou J.M.
      • Chang C.M.
      • Lu T.P.
      • Chuang E.Y.
      • Tai Y.C.
      • Cheng C.
      • Lin H.Y.
      Altered gut microbiota and inflammatory cytokine responses in patients with Parkinson's disease.
      ,
      • Jin M.
      • Li J.
      • Liu F.
      • Lyu N.
      • Wang K.
      • Wang L.
      • Liang S.
      • Tao H.
      • Zhu B.
      • Alkasir R.
      Analysis of the gut microflora in Patients with Parkinson's disease.
      ]) were included in this meta-analysis. These included raw sequencing data from four studies (n = 817 subjects) which were not included in the two previous meta-analyses [
      • Romano S.
      • Savva G.M.
      • Bedarf J.R.
      • Charles I.G.
      • Hildebrand F.
      • Narbad A.
      Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
      ,
      • Nishiwaki H.
      • Ito M.
      • Ishida T.
      • Hamaguchi T.
      • Maeda T.
      • Kashihara K.
      • Tsuboi Y.
      • Ueyama J.
      • Shimamura T.
      • Mori H.
      Meta‐analysis of gut dysbiosis in Parkinson's disease.
      ].
      Fig. 1
      Fig. 1Details of 16S rRNA case-control fecal microbiome studies included in the meta-analysis.
      Values are reported as mean ± standard deviation or median [IQR]. World map created and downloaded from mapchart.net.

      3.1 Comparison of alpha and beta diversities

      In the pooled analysis, we found significantly increased alpha diversity in PD patients vs. non-PD controls (p < 0.001 for Shannon's Index; p < 0.05 for both Chao1 and Pielou's Indices; Supplementary Fig. S2A). Additionally, significant differences in alpha diversity were also detected between cohorts from different countries (Supplementary Fig. S2B) and studies (Supplementary Fig. S2C) (Supplementary Table S1).
      Beta diversity comparison was explored based on the status of disease (PD vs. controls) (Fig. 2A), population (Caucasian vs. non-Caucasian populations) (Fig. 2B), continent (North America, Europe, Asia) (Fig. 2C), country (Fig. 2D), and study (i.e., the distribution of microbial composition according to different studies) (Fig. 2E). All five factors had significant effects on microbiome composition, but PERMANOVA revealed that differences in countries and studies had the largest effect sizes (R2 = 0.171 and 0.179, respectively, both p < 0.001). Clear separation in microbiome composition was also seen between Caucasian vs. non-Caucasian populations (R2 = 0.031, p < 0.001), and among continents (R2 = 0.049, p < 0.001). Since each country was represented by only one study (except for two studies from Germany [
      • Heintz‐Buschart A.
      • Pandey U.
      • Wicke T.
      • Sixel‐Döring F.
      • Janzen A.
      • Sittig‐Wiegand E.
      • Trenkwalder C.
      • Oertel W.H.
      • Mollenhauer B.
      • Wilmes P.
      The nasal and gut microbiome in Parkinson's disease and idiopathic rapid eye movement sleep behavior disorder.
      ,
      • Hopfner F.
      • Künstner A.
      • Müller S.H.
      • Künzel S.
      • Zeuner K.E.
      • Margraf N.G.
      • Deuschl G.
      • Baines J.F.
      • Kuhlenbäumer G.
      Gut microbiota in Parkinson disease in a northern German cohort.
      ]), country and study-related factors likely overlapped in accounting for variances. We further explored this effect by comparing the two German studies and found a significant difference in microbiome composition between both studies (R2 = 0.123, p < 0.001) (Supplementary Fig. S3A), suggesting that differences in study methodology have a strong influence in the investigation of gut microbiome composition.
      Fig. 2
      Fig. 2Principal coordinate analysis showing gut microbial compositional differences in subjects.
      Each data point represents a single sample. PCO1 and PCO2 reflect the magnitude of differences captured along the x- and y-axis, respectively. The distance between data points shows how different the samples are from one another. Statistical analyses were conducted using permutational multivariate analysis of variance (PERMANOVA). The R2 value shows the effect size, while the p-value shows the significance level. Differences with p < 0.05 were considered significant. Gut microbial composition was significantly different between PD patients vs. non-PD controls (A), predominant Caucasian vs. predominant non-Caucasian populations (B), subjects from different continents (i.e., North America, Europe, and Asia) (C), subjects from different countries (D) and subjects from different studies (E).
      Meanwhile, PD status had a significant but very small effect size on overall microbiome composition, explaining only 0.4% of the variances (R2 = 0.004, p < 0.001). These differences remained significant after controlling for study status (R2 = 0.003, p < 0.001). In the pooled analysis of PD patients, we found a significant separation in gut microbiome composition between Caucasian (n = 703) vs. non-Caucasian (n = 266) patients (Supplementary Fig. S4A). Microbiome compositional differences according to sex and age (65 years cutoff) had very marginal effects only (Supplementary Figs. S4B and S4C).
      We further examined these between-group differences by using bar plots to depict the distribution of bacterial taxa at different taxonomic levels across disease status, studies, and populations (Supplementary Figs. S5–S7); and heat maps, to cluster groups based on the similarity of gut microbiome compositional patterns (Supplementary Fig. S8), where leaves/clades of the dendrograms are arranged according to how similar the samples are with each other. Notably, the taxonomic distribution was largely more similar for samples from the same population (Supplementary Fig. S8A) or study (Supplementary Fig. S8B).

      3.2 Differential abundance analysis

      Differentially abundant bacterial taxa that were detected in DESeq2 and at least one other statistical method (ANCOM-BC or WMW) are summarized in Supplementary Table S2. Those with log2 fold change |>0.5| are depicted in Fig. 3A (for genera) and Supplementary Fig. S9 (for phyla and families). After controlling for study, the highest log2 fold changes were seen in the increase of Megasphaera and Akkermansia, and the reduction of Roseburia in PD patients (all of which were consistently detected across three statistical approaches).
      Fig. 3
      Fig. 3Bar plots showing differentially abundant bacterial genera in PD patients vs. controls.
      Pink bars represent differentially abundant bacterial genera that are increased in PD, while blue bars represent differentially abundant bacterial genera that are reduced in PD (adjusted p < 0.05, log2 fold change |>0.5|). Only bacterial genera that are identified to be differentially abundant in DESeq2 and at least one other statistical method (ANCOM-BC or WMW) are demonstrated (the superscript letter ‘a’ indicates genera that are consistently detected in all three methods). All analyses were controlled for study differences. (A) Pooled analyses from all ten studies included in this meta-analysis (n = 1703); (B) Pooled analyses from studies with predominant Caucasian population (n = 1202); (C) Pooled analyses from studies with predominant non-Caucasian population (n = 501). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
      A summary comparing our methods, study cohorts, and findings with two previous OTU-based meta-analyses by Romano et al., 2021 [
      • Romano S.
      • Savva G.M.
      • Bedarf J.R.
      • Charles I.G.
      • Hildebrand F.
      • Narbad A.
      Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
      ] (n = 1211; 681 PD, 530 Controls) and Nishiwaki et al., 2020 [
      • Nishiwaki H.
      • Ito M.
      • Ishida T.
      • Hamaguchi T.
      • Maeda T.
      • Kashihara K.
      • Tsuboi Y.
      • Ueyama J.
      • Shimamura T.
      • Mori H.
      Meta‐analysis of gut dysbiosis in Parkinson's disease.
      ] (n = 1034; 599 PD, 435 Controls) is provided in Supplementary Table S3. Increased Akkermansia and Hungatella, and reduced Roseburia and Faecalibacterium were consistently found in PD across three meta-analyses; while 80.5% of bacterial taxa were found to be differentially abundant in only one, but not in either of the other two meta-analyses.
      We further examined differences between PD and controls in Caucasian (Fig. 3B) and non-Caucasian (Fig. 3C) populations separately. A higher number of differentially abundant bacterial genera were identified in the non-Caucasian case-control cohort compared to the Caucasians. Notably, reduced abundance of Roseburia and unclassified Lachnospiraceae were found in both Caucasian and non-Caucasian PD cohorts vs. their respective controls, across three statistical approaches. Additionally, 47 bacterial genera were found to be differentially abundant between Caucasian and non-Caucasian PD patients (Supplementary Fig. S10).

      3.3 Correlation with disease severity

      After controlling for age and disease duration, several bacterial taxa found to be differentially increased in PD in the pooled analysis correlated with worse motor function (i.e., higher International Parkinson Disease and Movement Disorder Society–Unified Parkinson's Disease Rating Scale [MDS-UPDRS] Part III scores) (with the largest effect size seen with Escherichia [r = 0.222], Desulfovibrio [r = 0.211], Lactobacillus [r = 0.199], and Megasphaera [r = 0.190]); and worse cognitive function (i.e., lower Montreal Cognitive Assessment [MoCA] scores) (including Escherichia [r = −0.132], Lactobacillus [r = −0.137] and Megasphaera [r = −0.157]) (Fig. 4). Other non-differentially abundant taxa (e.g., Hungatella and Megamonas) also had a positive correlation with MDS-UPDRS Part III score but with a smaller effect size (r = 0.105–0.143). On the other hand, Butyrivibrio (decreased in PD) correlated negatively with MDS-UPDRS part III (r = −0.224) and Part IV scores (r = −0.233); and several taxa found to be differentially decreased in PD in the pooled analysis correlated with better cognitive function, including unclassified Prevotellaceae (r = 0.169) and Roseburia (r = 0.141).
      Fig. 4
      Fig. 4Correlations between bacterial genera with clinical features in PD patients (n = 373).
      Correlation heat map showing bacterial genera that correlated significantly with MDS-UPDRS part III, MDS-UPDRS part IV, and MoCA scores. Bacterial genera not correlating with any of the clinical features are not displayed. Red indicates a significant positive correlation while blue indicates a significant negative correlation (Spearman correlation, adjusted p < 0.05). Asterisks labeled in the boxes indicate correlations that remained significant after covariates adjustment (i.e., analyses for MDS-UPDRS Part III and MoCA were adjusted for age and disease duration, while analyses for MDS-UPDRS Part IV were adjusted for age of PD onset and disease duration). Upward arrows indicate bacterial genera (adjusted p < 0.05, log2 fold change |>0.5|) that were differentially increased in PD vs. controls, while downward arrows indicate genera that were differentially decreased in PD vs. controls, in the pooled analysis. Higher MDS-UPDRS Part III and Part IV scores indicate worse motor function and motor response complications, respectively; lower MoCA score indicates worse cognitive function. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

      3.4 Functional analysis

      After controlling for study status, a total of 359 pathways were detected to be differentially abundant between PD patients vs. non-PD controls in DESeq2 and at least one other statistical method (ANCOM-BC or WMW). Those with log2 fold change |>0.5| are depicted in Supplementary Fig. S11. The majority of the differentially abundant pathways were related to amino acid degradation (l-histidine, d-arginine, l-glutamate, l-leucine, l-tryptophan, and l-cysteine), lipid biosynthesis (oleate, dodec-5-enoate, arachidonate and phosphatidylcholine), and degradation of inorganic nutrients (tetrathionate, sulfate and sulfoacetaldehyde), all of which were found to be upregulated in PD.

      4. Discussion

      In this meta-analysis, which had a substantially larger sample size compared to previous work [
      • Romano S.
      • Savva G.M.
      • Bedarf J.R.
      • Charles I.G.
      • Hildebrand F.
      • Narbad A.
      Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
      ,
      • Nishiwaki H.
      • Ito M.
      • Ishida T.
      • Hamaguchi T.
      • Maeda T.
      • Kashihara K.
      • Tsuboi Y.
      • Ueyama J.
      • Shimamura T.
      • Mori H.
      Meta‐analysis of gut dysbiosis in Parkinson's disease.
      ] and included a significant proportion of underrepresented non-Caucasian subjects (n = 501, 266 PD, 235 controls), we were able to replicate some important findings as well as provide new insights regarding the potential clinical relevance of gut microbiome changes in PD. Despite differences in sample size, study cohorts, and analytical pipelines (e.g., phylogenetic placement vs. conventional OTUs-based approach), we found, in keeping with previous results [
      • Romano S.
      • Savva G.M.
      • Bedarf J.R.
      • Charles I.G.
      • Hildebrand F.
      • Narbad A.
      Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
      ], that individual study differences contributed to much larger variances in gut microbiome composition findings compared to PD disease status. There were also consistent patterns in the altered abundances of several bacterial taxa (e.g., increased Akkermansia, reduced Roseburia). The importance of independent replication of scientific findings has been increasingly emphasized, and this is even more so in a complex field like microbiome science where numerous confounders exist [
      • Schmidt T.S.
      • Raes J.
      • Bork P.
      The human gut microbiome: from association to modulation.
      ]. Additionally, we found significant differences in gut microbiome composition comparing predominant Caucasian and predominant non-Caucasian populations, not only in the overall cohort but specifically in the PD subgroup, which to our knowledge has not previously been reported. Importantly (and not explored in previous meta-analyses), the abundances of several genera correlated with major clinical features of PD, including motor severity and cognitive function. Intriguingly, a general pattern emerged whereby bacterial genera found to be increased in PD were associated with worse clinical severity, whereas those less abundant were positively correlated with better clinical status.
      We found that the gut microbiome was more similar for subjects from the same study (regardless of disease status) compared to among subjects having the same disease status (i.e., PD patients from different studies). This suggests that interstudy variations persisted even after attempts to minimize the effect of disparities in the current meta-analysis design, including standardization of quality filter and implementation of a phylogenetic placement approach. The lack of agreement on microbial taxa differences between individual studies in PD is therefore confounded in large part by heterogeneity in study design employed by different research groups. Similarly, the heterogeneity of findings across the three meta-analyses (Supplementary Table S3) is likely to be a reflection of the inclusion of overlapping but different studies, and the use of different analytical pipelines. For example, higher Chao1 values were detected in the studies utilizing single-end sequencing technology in this meta-analysis [
      • Hill‐Burns E.M.
      • Debelius J.W.
      • Morton J.T.
      • Wissemann W.T.
      • Lewis M.R.
      • Wallen Z.D.
      • Peddada S.D.
      • Factor S.A.
      • Molho E.
      • Zabetian C.P.
      Parkinson's disease and Parkinson's disease medications have distinct signatures of the gut microbiome.
      ,
      • Petrov V.
      • Saltykova I.
      • Zhukova I.
      • Alifirova V.
      • Zhukova N.
      • Dorofeeva Y.B.
      • Tyakht A.
      • Kovarsky B.
      • Alekseev D.
      • Kostryukova E.
      Analysis of gut microbiota in patients with Parkinson's disease.
      ]. This is likely due to the higher error rate of single-end sequencing, resulting in higher variant call rates than paired-end sequencing. A similar observation was reported previously by Werner et al. [
      • Werner J.J.
      • Zhou D.
      • Caporaso J.G.
      • Knight R.
      • Angenent L.T.
      Comparison of Illumina paired-end and single-direction sequencing for microbial 16S rRNA gene amplicon surveys.
      ]. Taken together, protocols for sample collection, processing and storage, as well as for DNA extraction and sequencing should therefore be harmonized to achieve more definitive conclusions on the altered gut microbial community in PD.
      We demonstrated that gut microbiome composition was significantly different between predominant Caucasian vs. non-Caucasian populations, and among studies from different continents (North America, Europe, and Asia). These findings provide further evidence that geographic location and inherent differences in investigated populations, such as ethnicity, diet, and socioeconomic status are important to contextualize when studying the gut microbiome [
      • He Y.
      • Wu W.
      • Zheng H.M.
      • Li P.
      • McDonald D.
      • Sheng H.F.
      • Chen M.X.
      • Chen Z.H.
      • Ji G.Y.
      • Mujagond P.
      Regional variation limits applications of healthy gut microbiome reference ranges and disease models.
      ]. In this study, gut microbiome alterations in non-Caucasians appeared to be more complex with a large number of differentially abundant genera detected between PD vs. controls. This may be due to larger variations in the gut microbial community because of the heterogenous nature of some non-Caucasian populations studied (e.g., one study had participation of three major ethnic groups – Chinese, Malays and Indians [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ]). Previous research revealed that ethnicity remained the strongest determinant of gut microbiome profiles even after the participants had shared the same living environment for a prolonged period [
      • Deschasaux M.
      • Bouter K.E.
      • Prodan A.
      • Levin E.
      • Groen A.K.
      • Herrema H.
      • Tremaroli V.
      • Bakker G.J.
      • Attaye I.
      • Pinto-Sietsma S.J.
      Depicting the composition of gut microbiota in a population with varied ethnic origins but shared geography.
      ]. Ethnicity-related factors comprise many different aspects, including genetics, lifestyle and cultural habits, socioeconomic status, healthcare utilization (including antibiotic usage), and early-life environment - all of which contribute to shaping the gut microbiota [
      • Deschasaux M.
      • Bouter K.E.
      • Prodan A.
      • Levin E.
      • Groen A.K.
      • Herrema H.
      • Tremaroli V.
      • Bakker G.J.
      • Attaye I.
      • Pinto-Sietsma S.J.
      Depicting the composition of gut microbiota in a population with varied ethnic origins but shared geography.
      ]. Interestingly, a recent large genotype-microbiome study revealed that while environmental factors played a more dominant role in determining the gut microbiome compared to host genetics, combining microbiome and host genetics data substantially improved the prediction accuracy of human phenotypes [
      • Rothschild D.
      • Weissbrod O.
      • Barkan E.
      • Kurilshikov A.
      • Korem T.
      • Zeevi D.
      • Costea P.I.
      • Godneva A.
      • Kalka I.N.
      • Bar N.
      Environment dominates over host genetics in shaping human gut microbiota.
      ]. The large variations of gut microbiome and the potentially more complex host-microbial interactions suggest that a “personalized” - rather than a “one-size-fits-all” - approach in microbial-directed therapeutics will be needed in PD. This paradigm has shown success in other chronic diseases with complex gut microbiomes [
      • Harkins C.P.
      • Kong H.H.
      • Segre J.A.
      Manipulating the human microbiome to manage disease.
      ].
      Despite having a limited effect size, the effect of disease status on the gut microbiome remained significant after controlling for study, implying that the microbiome in PD is truly different from subjects without PD. Increased abundance of Akkermansia, a mucin-degrading bacterium, has emerged as a consistent finding in PD [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ,
      • Romano S.
      • Savva G.M.
      • Bedarf J.R.
      • Charles I.G.
      • Hildebrand F.
      • Narbad A.
      Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
      ,
      • Nishiwaki H.
      • Ito M.
      • Ishida T.
      • Hamaguchi T.
      • Maeda T.
      • Kashihara K.
      • Tsuboi Y.
      • Ueyama J.
      • Shimamura T.
      • Mori H.
      Meta‐analysis of gut dysbiosis in Parkinson's disease.
      ]. Disruption of the gut mucus barrier could result in gut inflammation and subsequent neurodegeneration [
      • Rolli-Derkinderen M.
      • Leclair-Visonneau L.
      • Bourreille A.
      • Coron E.
      • Neunlist M.
      • Derkinderen P.
      Is Parkinson's disease a chronic low-grade inflammatory bowel disease?.
      ], and increased Akkermansia in the gut has also been reported in several other diseases including multiple system atrophy, progressive supranuclear palsy, Alzheimer's disease, and multiple sclerosis [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ]. Conversely, A. muciniphila also has the ability to stimulate mucin production, and supplementation with some strains ameliorated age-related loss of mucin and had health benefits, including an extended lifespan in progeroid mice [
      • DeJong E.N.
      • Surette M.G.
      • Bowdish D.M.
      The gut microbiota and unhealthy aging: disentangling cause from consequence.
      ]. Observational studies also found increased A. muciniphila abundance in centenarians [
      • DeJong E.N.
      • Surette M.G.
      • Bowdish D.M.
      The gut microbiota and unhealthy aging: disentangling cause from consequence.
      ]. The role of Akkermansia, therefore, deserves further investigation, particularly in view of the fact that it is now one of several newly explored microorganisms being considered as probiotic therapy for metabolic and inflammatory conditions [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ].
      Among other bacterial taxa found to be increased in PD, Escherichia correlated with worse motor and cognitive functions. There is evidence in animal models that E. coli-produced curli (cell surface amyloid proteins) can promote alpha-synuclein aggregation and inflammation in the gut and brain, with corresponding motor deficits [
      • Sampson T.R.
      • Challis C.
      • Jain N.
      • Moiseyenko A.
      • Ladinsky M.S.
      • Shastri G.G.
      • Thron T.
      • Needham B.D.
      • Horvath I.
      • Debelius J.W.
      A gut bacterial amyloid promotes α-synuclein aggregation and motor impairment in mice.
      ,
      • Chen S.G.
      • Stribinskis V.
      • Rane M.J.
      • Demuth D.R.
      • Gozal E.
      • Roberts A.M.
      • Jagadapillai R.
      • Liu R.
      • Choe K.
      • Shivakumar B.
      Exposure to the functional bacterial amyloid protein curli enhances alpha-synuclein aggregation in aged Fischer 344 rats and Caenorhabditis elegans.
      ]. Gram-negative bacterial lipopolysaccharide can also result in toll-like receptor (TLR) activation, which is believed to contribute to leaky gut, and intestinal and brain inflammation in PD [
      • Forsyth C.B.
      • Shannon K.M.
      • Kordower J.H.
      • Voigt R.M.
      • Shaikh M.
      • Jaglin J.A.
      • Estes J.D.
      • Dodiya H.B.
      • Keshavarzian A.
      Increased intestinal permeability correlates with sigmoid mucosa alpha-synuclein staining and endotoxin exposure markers in early Parkinson's disease.
      ,
      • Perez-Pardo P.
      • Dodiya H.B.
      • Engen P.A.
      • Forsyth C.B.
      • Huschens A.M.
      • Shaikh M.
      • Voigt R.M.
      • Naqib A.
      • Green S.J.
      • Kordower J.H.
      Role of TLR4 in the gut-brain axis in Parkinson's disease: a translational study from men to mice.
      ]. Altogether, these findings highlight the need for further studies to interrogate the role of Escherichia in PD pathogenesis.
      Desulfovibrio also correlated with worse motor function, consistent with a previous study showing correlation with worse Hoehn and Yahr score [
      • Murros K.E.
      • Huynh V.A.
      • Takala T.M.
      • Saris P.E.
      Desulfovibrio bacteria are associated with Parkinson's disease.
      ]. Species of Desulfovibrio are dominant among intestinal sulfate-reducing bacteria, with a capacity for producing hydrogen sulfide, which is potentially toxic in humans [
      • Murros K.E.
      • Huynh V.A.
      • Takala T.M.
      • Saris P.E.
      Desulfovibrio bacteria are associated with Parkinson's disease.
      ]. Higher sulfate-reducing bacterial activities and consequent higher hydrogen sulfide production are believed to play a role in the development of inflammatory bowel disease [
      • Cao X.
      • Cao L.
      • Ding L.
      • Bian J.S.
      A new hope for a devastating disease: hydrogen sulfide in Parkinson's disease.
      ], which has parallels with changes in the PD gut [
      • Rolli-Derkinderen M.
      • Leclair-Visonneau L.
      • Bourreille A.
      • Coron E.
      • Neunlist M.
      • Derkinderen P.
      Is Parkinson's disease a chronic low-grade inflammatory bowel disease?.
      ]. Megasphaera and Acidaminococcus, from the family of Veillonellaceae (known for its ability to produce succinate [
      • Zeng X.
      • Gao X.
      • Peng Y.
      • Wu Q.
      • Zhu J.
      • Tan C.
      • Xia G.
      • You C.
      • Xu R.
      • Pan S.
      Higher risk of stroke is correlated with increased opportunistic pathogen load and reduced levels of butyrate-producing bacteria in the gut.
      ], a key driver of reactive-oxygen species production that has been implicated in tissue inflammation and injury [
      • Zeng X.
      • Gao X.
      • Peng Y.
      • Wu Q.
      • Zhu J.
      • Tan C.
      • Xia G.
      • You C.
      • Xu R.
      • Pan S.
      Higher risk of stroke is correlated with increased opportunistic pathogen load and reduced levels of butyrate-producing bacteria in the gut.
      ]), also correlated with worse motor function (Megasphaera additionally correlated with worse cognitive function). The literature regarding these two bacteria remains scarce; both are recognized to be potentially pathogenic [
      • Zeng X.
      • Gao X.
      • Peng Y.
      • Wu Q.
      • Zhu J.
      • Tan C.
      • Xia G.
      • You C.
      • Xu R.
      • Pan S.
      Higher risk of stroke is correlated with increased opportunistic pathogen load and reduced levels of butyrate-producing bacteria in the gut.
      ] and were found to be increased in individuals with elevated cardiovascular risk [
      • Zeng X.
      • Gao X.
      • Peng Y.
      • Wu Q.
      • Zhu J.
      • Tan C.
      • Xia G.
      • You C.
      • Xu R.
      • Pan S.
      Higher risk of stroke is correlated with increased opportunistic pathogen load and reduced levels of butyrate-producing bacteria in the gut.
      ].
      Conversely, reduced abundances of Roseburia and unclassified Prevotellaceae were observed in PD; furthermore, reductions in the abundances of these bacteria correlated with worse cognitive function. Meanwhile, reduced Butyrivibrio abundance correlated with worse motor function and motor response complications. These bacterial genera are known to be short-chain fatty acid (SCFA) producers, and a growing literature on SCFAs seem to indicate that they have neuroprotective properties through their ability, among others, to reduce oxidative stress and neuroinflammation, and by regulating gut epithelial and blood-brain barrier integrity [
      • Dalile B.
      • Van Oudenhove L.
      • Vervliet B.
      • Verbeke K.
      The role of short-chain fatty acids in microbiota-gut-brain communication.
      ]. Interestingly, reduced fecal SCFAs (i.e., butyrate, acetate and propionate) has been demonstrated earlier in PD case-control studies, using a targeted gas chromatography platform [
      • Unger M.M.
      • Spiegel J.
      • Dillmann K.U.
      • Grundmann D.
      • Philippeit H.
      • Bürmann J.
      • Faßbender K.
      • Schwiertz A.
      • Schäfer K.H.
      Short chain fatty acids and gut microbiota differ between patients with Parkinson's disease and age-matched controls.
      ], as well as in our recent study using an untargeted nuclear magnetic resonance (NMR) metabolomics platform [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ]. Notably, reductions in fecal SCFAs were observed in patients with lower cognitive scores [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ]. Taken together, these findings point to a clinically relevant role of SCFAs and SCFA-producing bacteria in PD, that warrants further investigation as potential therapeutic targets.
      At the functional level, numerous pathways were predicted (based on in silico analyses) to be altered in PD. These included pathways associated with the degradation of amino acids and inorganic nutrients, as well as the biosynthesis of lipids. It is important to note the significant limitations of attempting to infer potential roles of bacteria based on genus or even species-level data since bacterial function is likely to be strain-specific [
      • Shanahan F.
      • Ghosh T.S.
      • O'Toole P.W.
      The healthy microbiome - what is the definition of a healthy gut microbiome?.
      ]. Furthermore, the majority of genes in the human gut microbiome cannot yet be functionally assigned, and their dynamic transcriptional and translational activities have yet to be elucidated [
      • Lynch S.V.
      • Pedersen O.
      The human intestinal microbiome in health and disease.
      ,
      • Almeida A.
      • Nayfach S.
      • Boland M.
      • Strozzi F.
      • Beracochea M.
      • Shi Z.J.
      • Pollard K.S.
      • Sakharova E.
      • Parks D.H.
      • Hugenholtz P.
      A unified catalog of 204,938 reference genomes from the human gut microbiome.
      ]. Therefore, direct measurements of bacterial activities, using a variety of platforms such as fecal metatranscriptomics, metaproteomics, and metabolomics, are likely to be more informative in understanding bacterial functions in health and disease. As an example, we have previously demonstrated, using a combined microbiome-metabolomics approach, reduced fecal levels of butyrate (a major SCFA), amino acids (including glutamate), and choline (a precursor for phosphatidylcholine) in PD [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ]. These findings appear to align well with some of the results from the predictive functional profiling obtained in this meta-analysis (Supplementary Fig. S11); their potential implications have been discussed elsewhere [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ].
      Our study had some limitations, chiefly, that it did not include all relevant microbiome studies in PD, primarily because sequencing data were not available (either publicly or despite approaching study investigators). Furthermore, for the included studies, missing clinico-demographic information (e.g., dietary habits, disease severity, comorbid disorders, and medication use) did not allow proper control for some potentially important confounders. However, by sourcing clinical datasets from three large studies [
      • Tan A.H.
      • Chong C.W.
      • Lim S.Y.
      • Yap I.K.S.
      • Teh C.S.J.
      • Loke M.F.
      • Song S.L.
      • Tan J.Y.
      • Ang B.H.
      • Tan Y.Q.
      Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
      ,
      • Cirstea M.S.
      • Yu A.C.
      • Golz E.
      • Sundvick K.
      • Kliger D.
      • Radisavljevic N.
      • Foulger L.H.
      • Mackenzie M.
      • Huan T.
      • Finlay B.B.
      Microbiota composition and metabolism are associated with gut function in Parkinson's disease.
      ,
      • Lin C.H.
      • Chen C.C.
      • Chiang H.L.
      • Liou J.M.
      • Chang C.M.
      • Lu T.P.
      • Chuang E.Y.
      • Tai Y.C.
      • Cheng C.
      • Lin H.Y.
      Altered gut microbiota and inflammatory cytokine responses in patients with Parkinson's disease.
      ], we were able to explore associations between microbiome alterations in PD and important indices of disease severity. Finally, since we focused on studies utilizing 16S rRNA gene sequencing (by far the most common microbiome analytical platform used to date), bacterial changes at strain or species level could not be analyzed; future studies with higher taxonomic resolution (e.g., using shotgun metagenomics) will help to fill this gap.

      5. Conclusion

      We show that the investigation of the gut microbiome in PD is highly complex, with many important confounding factors including study methodology and geographical/population differences that need to be carefully considered in future research. Although the correlation of PD status with the gut microbiome appears to be modest, several consistent microbiome features could be discerned with important clinical correlations. This study highlights an urgent need for harmonized methods in microbiome research and systematic archiving of ’omics and clinical datasets in repositories, to unravel the gut microbiome puzzle in PD. The complexity of gut microbiome alterations in PD also underscores the need to consider personalized, precision medicine approaches in designing microbial-directed therapies to combat this disease.

      Author’s contribution

      CWC and AHT conceptualized and designed the study. TST conducted the systematic review, processing of sequencing data and statistical analyses. AHT, SYL, MC, SAC, BBF, CHL, and CCC acquired and provided clinical data. AHT supervised the systematic review while CWC and JB supervised the statistical analyses. TST, AHT, CWC and SYL planned the statistical analyses and interpreted the analyzed data. TST wrote the first draft of the manuscript, and all authors contributed to revising the manuscript. All authors approved the final manuscript for publication.

      Availability of data and materials

      Most datasets (eight out of ten studies) analyzed in the current study are available in the NCBI GenBank.

      Declaration of competing interest

      None.

      Acknowledgements

      This work was supported by the Ministry of Education , Malaysia Fundamental Research Grant Scheme (FRGS) FRGS/1/2018/SKK02/UM/02/1 (to Ai Huey Tan). Dr. Silke Appel-Cresswell is supported by the Marg Meikle Professorship for Parkinson's disease by the Pacific Parkinson's Research Institute (PPRI), the Vancouver sample collection is supported by PPRI, Parkinson Canada, and Parkinson Society British Columbia. We would like to express our gratitude to all authors who made their 16S rRNA gene sequences available and responded to our queries.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Lubomski M.
        • Tan A.H.
        • Lim S.Y.
        • Holmes A.J.
        • Davis R.L.
        • Sue C.M.
        Parkinson's disease and the gastrointestinal microbiome.
        J. Neurol. 2019; : 1-17
        • Tan A.H.
        • Chong C.W.
        • Lim S.Y.
        • Yap I.K.S.
        • Teh C.S.J.
        • Loke M.F.
        • Song S.L.
        • Tan J.Y.
        • Ang B.H.
        • Tan Y.Q.
        Gut microbial ecosystem in Parkinson’s disease: new clinico‐biological insights from multi‐omics.
        Ann. Neurol. 2020; 89: 546-559
        • Romano S.
        • Savva G.M.
        • Bedarf J.R.
        • Charles I.G.
        • Hildebrand F.
        • Narbad A.
        Meta-analysis of the Parkinson's disease gut microbiome suggests alterations linked to intestinal inflammation.
        NPJ Parkinsons. Dis. 2021; 7: 1-13
        • Nishiwaki H.
        • Ito M.
        • Ishida T.
        • Hamaguchi T.
        • Maeda T.
        • Kashihara K.
        • Tsuboi Y.
        • Ueyama J.
        • Shimamura T.
        • Mori H.
        Meta‐analysis of gut dysbiosis in Parkinson's disease.
        Mov. Disord. 2020; 35: 1626-1635
        • Barbera P.
        • Kozlov A.M.
        • Czech L.
        • Morel B.
        • Darriba D.
        • Flouri T.
        • Stamatakis A.
        EPA-ng: massively parallel evolutionary placement of genetic sequences.
        Syst. Biol. 2019; 68: 365-369
        • Bowman J.S.
        • Ducklow H.W.
        Microbial communities can be described by metabolic structure: a general framework and application to a seasonally variable, depth-stratified microbial community from the coastal West Antarctic Peninsula.
        PLoS One. 2015; 10e0135868
        • Lahti L.
        • Shetty S.
        Introduction to the Microbiome R Package.
        2018
        • Goodrich J.K.
        • Di Rienzi S.C.
        • Poole A.C.
        • Koren O.
        • Walters W.A.
        • Caporaso J.G.
        • Knight R.
        • Ley R.E.
        Conducting a microbiome study.
        Cell. 2014; 158: 250-262
        • Love M.I.
        • Huber W.
        • Anders S.
        Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2.
        Genome Biol. 2014; 15: 1-21
        • Lin H.
        • Peddada S.D.
        Analysis of compositions of microbiomes with bias correction.
        Nat. Commun. 2020; 11: 1-11
        • Hothorn T.
        • Hornik K.
        • Van De Wiel M.A.
        • Zeileis A.
        A lego system for conditional inference.
        Am. Statistician. 2006; 60: 257-263
        • Cirstea M.S.
        • Yu A.C.
        • Golz E.
        • Sundvick K.
        • Kliger D.
        • Radisavljevic N.
        • Foulger L.H.
        • Mackenzie M.
        • Huan T.
        • Finlay B.B.
        Microbiota composition and metabolism are associated with gut function in Parkinson's disease.
        Mov. Disord. 2020; 35: 1208-1217
        • Heintz‐Buschart A.
        • Pandey U.
        • Wicke T.
        • Sixel‐Döring F.
        • Janzen A.
        • Sittig‐Wiegand E.
        • Trenkwalder C.
        • Oertel W.H.
        • Mollenhauer B.
        • Wilmes P.
        The nasal and gut microbiome in Parkinson's disease and idiopathic rapid eye movement sleep behavior disorder.
        Mov. Disord. 2018; 33: 88-98
        • Hopfner F.
        • Künstner A.
        • Müller S.H.
        • Künzel S.
        • Zeuner K.E.
        • Margraf N.G.
        • Deuschl G.
        • Baines J.F.
        • Kuhlenbäumer G.
        Gut microbiota in Parkinson disease in a northern German cohort.
        Brain Res. 2017; 1667: 41-45
        • Hill‐Burns E.M.
        • Debelius J.W.
        • Morton J.T.
        • Wissemann W.T.
        • Lewis M.R.
        • Wallen Z.D.
        • Peddada S.D.
        • Factor S.A.
        • Molho E.
        • Zabetian C.P.
        Parkinson's disease and Parkinson's disease medications have distinct signatures of the gut microbiome.
        Mov. Disord. 2017; 32: 739-749
        • Aho V.T.
        • Pereira P.A.
        • Voutilainen S.
        • Paulin L.
        • Pekkonen E.
        • Auvinen P.
        • Scheperjans F.
        Gut microbiota in Parkinson's disease: temporal stability and relations to disease progression.
        EBioMedicine. 2019; 44: 691-707
        • Pietrucci D.
        • Cerroni R.
        • Unida V.
        • Farcomeni A.
        • Pierantozzi M.
        • Mercuri N.B.
        • Biocca S.
        • Stefani A.
        • Desideri A.
        Dysbiosis of gut microbiota in a selected population of Parkinson's patients.
        Park. Relat. Disord. 2019; 65: 124-130
        • Petrov V.
        • Saltykova I.
        • Zhukova I.
        • Alifirova V.
        • Zhukova N.
        • Dorofeeva Y.B.
        • Tyakht A.
        • Kovarsky B.
        • Alekseev D.
        • Kostryukova E.
        Analysis of gut microbiota in patients with Parkinson's disease.
        Bull. Exp. Biol. Med. 2017; 162: 734-737
        • Lin C.H.
        • Chen C.C.
        • Chiang H.L.
        • Liou J.M.
        • Chang C.M.
        • Lu T.P.
        • Chuang E.Y.
        • Tai Y.C.
        • Cheng C.
        • Lin H.Y.
        Altered gut microbiota and inflammatory cytokine responses in patients with Parkinson's disease.
        J. Neuroinflammation. 2019; 16: 1-9
        • Jin M.
        • Li J.
        • Liu F.
        • Lyu N.
        • Wang K.
        • Wang L.
        • Liang S.
        • Tao H.
        • Zhu B.
        • Alkasir R.
        Analysis of the gut microflora in Patients with Parkinson's disease.
        Front. Neurosci. 2019; 13: 1184
        • Schmidt T.S.
        • Raes J.
        • Bork P.
        The human gut microbiome: from association to modulation.
        Cell. 2018; 172: 1198-1215
        • Werner J.J.
        • Zhou D.
        • Caporaso J.G.
        • Knight R.
        • Angenent L.T.
        Comparison of Illumina paired-end and single-direction sequencing for microbial 16S rRNA gene amplicon surveys.
        ISME J. 2012; 6: 1273-1276
        • He Y.
        • Wu W.
        • Zheng H.M.
        • Li P.
        • McDonald D.
        • Sheng H.F.
        • Chen M.X.
        • Chen Z.H.
        • Ji G.Y.
        • Mujagond P.
        Regional variation limits applications of healthy gut microbiome reference ranges and disease models.
        Nat. Med. 2018; 24: 1532-1535
        • Deschasaux M.
        • Bouter K.E.
        • Prodan A.
        • Levin E.
        • Groen A.K.
        • Herrema H.
        • Tremaroli V.
        • Bakker G.J.
        • Attaye I.
        • Pinto-Sietsma S.J.
        Depicting the composition of gut microbiota in a population with varied ethnic origins but shared geography.
        Nat. Med. 2018; 24: 1526-1531
        • Rothschild D.
        • Weissbrod O.
        • Barkan E.
        • Kurilshikov A.
        • Korem T.
        • Zeevi D.
        • Costea P.I.
        • Godneva A.
        • Kalka I.N.
        • Bar N.
        Environment dominates over host genetics in shaping human gut microbiota.
        Nature. 2018; 555: 210-215
        • Harkins C.P.
        • Kong H.H.
        • Segre J.A.
        Manipulating the human microbiome to manage disease.
        J. Am. Med. Assoc. 2020; 323: 303-304
        • Rolli-Derkinderen M.
        • Leclair-Visonneau L.
        • Bourreille A.
        • Coron E.
        • Neunlist M.
        • Derkinderen P.
        Is Parkinson's disease a chronic low-grade inflammatory bowel disease?.
        J. Neurol. 2020; 267: 2207-2213
        • DeJong E.N.
        • Surette M.G.
        • Bowdish D.M.
        The gut microbiota and unhealthy aging: disentangling cause from consequence.
        Cell Host Microbe. 2020; 28: 180-189
        • Sampson T.R.
        • Challis C.
        • Jain N.
        • Moiseyenko A.
        • Ladinsky M.S.
        • Shastri G.G.
        • Thron T.
        • Needham B.D.
        • Horvath I.
        • Debelius J.W.
        A gut bacterial amyloid promotes α-synuclein aggregation and motor impairment in mice.
        Elife. 2020; 9e53111
        • Chen S.G.
        • Stribinskis V.
        • Rane M.J.
        • Demuth D.R.
        • Gozal E.
        • Roberts A.M.
        • Jagadapillai R.
        • Liu R.
        • Choe K.
        • Shivakumar B.
        Exposure to the functional bacterial amyloid protein curli enhances alpha-synuclein aggregation in aged Fischer 344 rats and Caenorhabditis elegans.
        Sci. Rep. 2016; 6: 1-10
        • Forsyth C.B.
        • Shannon K.M.
        • Kordower J.H.
        • Voigt R.M.
        • Shaikh M.
        • Jaglin J.A.
        • Estes J.D.
        • Dodiya H.B.
        • Keshavarzian A.
        Increased intestinal permeability correlates with sigmoid mucosa alpha-synuclein staining and endotoxin exposure markers in early Parkinson's disease.
        PLoS One. 2011; 6e28032
        • Perez-Pardo P.
        • Dodiya H.B.
        • Engen P.A.
        • Forsyth C.B.
        • Huschens A.M.
        • Shaikh M.
        • Voigt R.M.
        • Naqib A.
        • Green S.J.
        • Kordower J.H.
        Role of TLR4 in the gut-brain axis in Parkinson's disease: a translational study from men to mice.
        Gut. 2019; 68: 829-843
        • Murros K.E.
        • Huynh V.A.
        • Takala T.M.
        • Saris P.E.
        Desulfovibrio bacteria are associated with Parkinson's disease.
        Front. Cell. Infect. Microbiol. 2021; 11: 378
        • Cao X.
        • Cao L.
        • Ding L.
        • Bian J.S.
        A new hope for a devastating disease: hydrogen sulfide in Parkinson's disease.
        Mol. Neurobiol. 2018; 55: 3789-3799
        • Zeng X.
        • Gao X.
        • Peng Y.
        • Wu Q.
        • Zhu J.
        • Tan C.
        • Xia G.
        • You C.
        • Xu R.
        • Pan S.
        Higher risk of stroke is correlated with increased opportunistic pathogen load and reduced levels of butyrate-producing bacteria in the gut.
        Front. Cell. Infect. Microbiol. 2019; 9: 4
        • Dalile B.
        • Van Oudenhove L.
        • Vervliet B.
        • Verbeke K.
        The role of short-chain fatty acids in microbiota-gut-brain communication.
        Nat. Rev. Gastroenterol. Hepatol. 2019; 16: 461-478
        • Unger M.M.
        • Spiegel J.
        • Dillmann K.U.
        • Grundmann D.
        • Philippeit H.
        • Bürmann J.
        • Faßbender K.
        • Schwiertz A.
        • Schäfer K.H.
        Short chain fatty acids and gut microbiota differ between patients with Parkinson's disease and age-matched controls.
        Park. Relat. Disord. 2016; 32: 66-72
        • Shanahan F.
        • Ghosh T.S.
        • O'Toole P.W.
        The healthy microbiome - what is the definition of a healthy gut microbiome?.
        Gastroenterology. 2021; 160: 483-494
        • Lynch S.V.
        • Pedersen O.
        The human intestinal microbiome in health and disease.
        N. Engl. J. Med. 2016; 375: 2369-2379
        • Almeida A.
        • Nayfach S.
        • Boland M.
        • Strozzi F.
        • Beracochea M.
        • Shi Z.J.
        • Pollard K.S.
        • Sakharova E.
        • Parks D.H.
        • Hugenholtz P.
        A unified catalog of 204,938 reference genomes from the human gut microbiome.
        Nat. Biotechnol. 2021; 39: 105-114