Gut microbiome dysbiosis across early Parkinson's disease, REM sleep behavior disorder and their first-degree relatives

Alpha-synucleinopathies such as Parkinson’s disease (PD) feature abnormal alpha-synuclein (α-syn) aggregation in the CNS, but evidence indicates α-syn pathology can begin in the enteric nervous system before the brain, supporting a gut-to-brain propagation model. Gut dysbiosis—particularly depletion of short-chain fatty acid (SCFA)-producing bacteria and enrichment of pathobionts—has been linked to intestinal hyperpermeability, immune activation, and pathological α-syn aggregation. Because autonomic and enteric dysfunctions (e.g., constipation) may precede PD motor onset by decades, characterizing gut microbiota in prodromal stages is critical. REM sleep behavior disorder (RBD) is a specific prodromal marker for PD, associated with increased constipation and enteric α-syn histopathology. First-degree relatives of RBD patients (RBD-FDR) show increased constipation and a spectrum from isolated RBD symptoms to v-PSG-confirmed RBD, suggesting an even earlier at-risk stage. Prior RBD microbiome studies suggested PD-like changes but were small and potentially underpowered. This study used a large cross-sectional, quasi-staging design across controls, RBD-FDR, v-PSG–confirmed RBD, and early PD to test whether PD-like gut dysbiosis emerges in prodromal α-synucleinopathy and to assess host–microbiome interactions.

Multiple studies report gut microbiota alterations in PD, including consistent depletion of SCFA-producing taxa and enrichment of potential pathobionts, with proposed links to intestinal barrier dysfunction, inflammation, and α-syn aggregation. Early clinical signs such as constipation and ENS α-syn have been documented years before PD onset. In RBD, two small v-PSG–confirmed studies and a questionnaire-based study suggested microbial changes resembling PD, but limited sample sizes restricted comprehensive assessment. RBD-FDR populations exhibit increased constipation and RBD features, indicating an earlier prodromal window. Meta-analyses in PD corroborate dysbiosis patterns, but the emergence and host-factor contributions at prodromal stages remain unclear, motivating the present large, staged cross-sectional analysis.

Design: Cross-sectional study emulating α-synucleinopathy (Braak) staging with four groups: controls (stage 0–1), RBD-FDR (0–2), v-PSG–confirmed idiopathic RBD (2–3), and early PD with premotor RBD features (3–4). Participants: 452 recruited; 441 stool samples analyzed after excluding 11 for low read count (<1000). Group sizes analyzed: control n=108, RBD-FDR n=127, RBD n=170, early PD n=36. Controls were neurologically healthy with no probable/v-PSG RBD or family history of RBD. RBD diagnosed by v-PSG, probable RBD by structured interview (DISP). Early PD: neurologist-confirmed PD, disease duration <5 years, no dementia, v-PSG–diagnosed RBD, RBD preceding parkinsonism. Exclusions: antibiotics within 1 month; major GI diseases (e.g., IBD, liver cirrhosis); neurodegenerative diseases (except early PD group). RBD-FDR with v-PSG RBD or neurodegenerative disease excluded. Clinical assessment: Sociodemographics, lifestyle, ESS, SCOPA-AUT for autonomic symptoms, Rome IV for GI disorders (functional constipation, IBS, functional diarrhea), bowel movement frequency (BMF; 1=>1/day to 6=≤1/week), stool consistency (inverse BSFS), UPDRS-III, orthostatic hypotension, Olfactory Identification Test, psychiatric evaluation (M.I.N.I.), cognition (HK-MOCA), and MDS research criteria for prodromal PD (total likelihood ratio, LR). Medication usage (including PD-specific drugs) and probiotic/prebiotic use recorded. Stool collection and 16S sequencing: Fresh stool collected and transported on ice within 4 h, aliquoted, stored at −80°C. DNA extracted (Qiagen DNeasy PowerSoil Pro). 16S rRNA V3–V4 amplified (primers 341F/806R), sequenced on Illumina MiSeq PE300. Denoising with DADA2 in QIIME2 (v2021.4); reads truncated at 288/272; samples with total reads <1000 removed. Taxonomy assigned via q2-feature-classifier (sklearn) against SILVA v138 99% database. ASVs collapsed at genus and family levels. Community analyses: Alpha diversity (Chao1, Gini Simpson, Shannon) via vegan R; group comparisons with Kruskal–Wallis. Beta-diversity: Bray–Curtis at genus level, PERMANOVA (adonis2, 99,999 permutations) with age and sex adjustment; homogeneity of dispersion (betadisper). PCoA visualization. Differential abundance: Filtered taxa with prevalence ≥10% and abundance ≥0.05%, retaining 88 genera and 36 families. Read counts CLR-transformed (pseudocount=1). Trend across stages assessed by Kendall’s tau-b. Group differences by Kruskal–Wallis with FDR correction. Multivariable associations with MaAsLin 2 including stage as fixed effect and family ID as random effect; versions with and without age/sex included; BH FDR applied. Host–microbiome interactions: PERMANOVA including covariates (age, sex, BMF, antidepressants, benzodiazepines, osmotic laxatives, PPIs; PD drugs in models involving early PD). MaAsLin 2 tested associations between taxa (and pathways) and host factors with all covariates as fixed effects, family as random effect; early PD modeled separately for PD drugs. Machine learning: Random forest classifier to distinguish RBD vs control and RBD vs RBD-FDR using CLR of 88 genera. Data split 80% train/20% test with stratified sampling. Recursive feature elimination (caret, 25 repeats of 10-fold CV) to select features; whole pipeline (split, RFE, fit, evaluation) repeated 25 times. ROC and AUC computed. Mediation analysis: Using R ‘mediation’ package, tested whether BMF mediates effect of microbiota on prodromal PD LR (excluding constipation and RBD items) in controls, RBD-FDR, and RBD (n=405). Exposure: microbiota (PCoA1); mediator: BMF; outcome: LR. ACME (indirect), ADE (direct) via 10,000 bootstrap resamples; proportion mediated reported. Functional prediction: PICRUSt2 to infer MetaCyc pathway abundances from 16S data. Pathways filtered (prevalence ≥10%, abundance ≥0.05%), CLR-transformed. Kruskal–Wallis for differential pathways; MaAsLin 2 for multivariable associations with stages and covariates. Statistics: Multiple testing controlled with Benjamini–Hochberg. GEE used for some clinical comparisons to adjust family clustering. Sample size pre-specified; investigators blinded to group during data collection and processing.

• Cohort: 441 analyzed stool samples: control n=108, RBD-FDR n=127, RBD n=170, early PD n=36.
• GI symptoms: Functional constipation prevalence increased across control → RBD-FDR → RBD → early PD: 8.3% → 9.4% → 45.3% → 69.4% (P<0.001). Straining with defecation: 8.8% → 15.8% → 45.4% → 68.6% (P<0.001). BMF score and harder stools (inverse BSFS) increased across groups (both P<0.001). IBS and functional diarrhea did not differ.
• Medications: >50% of RBD and early PD used benzodiazepines. Early PD: osmotic laxatives 30.6%, antidepressants 13.9%. RBD: osmotic laxatives 5.3%, antidepressants 25.3%. PD-specific drugs in early PD: carbidopa/levodopa 47.2%, MAO-B inhibitors 41.7%, dopamine agonist 8.3%, benzhexol 5.6%, COMT inhibitors 2.8%.
• Microbiota composition: Alpha diversity comparable across groups. Beta-diversity (Bray–Curtis, genus level, PERMANOVA adjusted for age/sex): early PD distinct from control (R²=0.035, q<0.001). RBD differed from control and RBD-FDR (all q<0.001) and was similar to early PD (R²=0.008, q=0.066). No significant difference between control and RBD-FDR. RBD and early PD showed greater dispersion.
• Differential taxa across stages: After filtering, 31/88 genera (35.2%) correlated with stage (Kendall’s tau-b q<0.05). Strongest negative correlations with progression: Butyricicoccus (τb=−0.204, q<0.001) and Faecalibacterium (τb=−0.198, q<0.001). Kruskal–Wallis identified 16 families and 26 genera differing among groups (q<0.05). MaAsLin 2 showed 19/42 differential taxa similarly altered in RBD and early PD versus control (unadjusted q<0.05), including decreases in butyrate producers (Roseburia, Lachnospiraceae_ND3007_group, Lachnospira, [Eubacterium]_ventriosum_group, Butyricicoccus, Faecalibacterium; family Lachnospiraceae) and increases in Desulfovibrio, Akkermansia, Collinsella, Oscillospiraceae_UCG-002/-005; results largely robust after adjusting for age/sex (q<0.1).
• Early prodromal signals in RBD-FDR: Collinsella enriched already in RBD-FDR (β=0.58, q=0.035 adjusted; β=0.49, q=0.038 unadjusted). [Eubacterium]_ventriosum_group showed marginal decrease (β=−0.54, q=0.069). RBD-FDR with probable RBD (n=11) had lower [Eubacterium]_ventriosum_group than those without (CLR −0.36±1.8 vs 0.35±1.7; q=0.028).
• Host–microbiome interactions: In PERMANOVA models with covariates, group effects (RBD and early PD) remained significant vs control (all q<0.001). Covariates strongly associated with overall composition included BMF score, sex, osmotic laxatives, and PPIs; age, statin, antidepressants, benzodiazepines, and PD-specific drugs had minimal overall effects. MaAsLin 2 confirmed disease-associated depletion of SCFA producers and increase of Collinsella (including in RBD-FDR). BMF correlated with taxa similarly to disease progression, notably Butyricicoccus (β=−0.34, q=0.009) and Oscillospiraceae_UCG-005 (β=0.32, q=0.003). Antidepressant use co-occurred with enrichment of Akkermansia and UBA1819.
• Mediation: Effect of microbiota (PCoA1) on prodromal PD likelihood ratio was partially mediated by BMF (proportion mediated=0.31, P=0.0004).
• Functional predictions (PICRUSt2/MetaCyc): 18 pathways differed among groups (q<0.05). At prodromal and early stages (RBD, early PD), pathways for SCFA-related fermentation (e.g., Bifidobacterium shunt, heterolactic fermentation) and carbohydrate biosynthesis were enriched; cofactor/vitamin biosynthesis (B1, B2, B12) decreased. After covariate adjustment, SCFA-to-lactate/ethanol metabolism and preQ1 (7-deazapurine) biosynthesis remained altered; B12 salvage and de novo pathways were higher in controls vs RBD-FDR.
• Classification: Random forest distinguished RBD from controls with mean AUC 0.79 (95% CI 0.78–0.80) in training; in the test set, accuracy 0.68 (95% CI 0.66–0.70), AUC 0.75 (95% CI 0.73–0.78). Twelve genera selected in ≥60% of models; seven (Butyricicoccus, UBA1819, Lachnoclostridium, Oscillospiraceae_UCG-002, Uncultured_Oscillospiraceae_g061, [Ruminococcus]_torques_group, [Eubacterium]_ventriosum_group) appeared in all trained models. Classifier also distinguished RBD from RBD-FDR: test accuracy 0.67 (95% CI 0.66–0.69), AUC 0.72 (95% CI 0.69–0.74).

The study demonstrates that PD-like gut dysbiosis emerges at prodromal stages of α-synucleinopathy. RBD microbiota composition closely resembles early PD, marked by depletion of SCFA-producing taxa (e.g., Butyricicoccus, Faecalibacterium, [Eubacterium]_ventriosum_group) and enrichment of putative pro-inflammatory and barrier-disrupting taxa (e.g., Collinsella, Desulfovibrio, Akkermansia, Oscillospiraceae UCG-005). Importantly, early signatures (e.g., increased Collinsella and decreased [Eubacterium]_ventriosum_group) are detectable even in younger, earlier-stage RBD-FDR, suggesting dysbiosis precedes overt RBD and PD. Functional predictions indicate shifts toward increased fermentation to lactate/ethanol and decreased vitamin/cofactor biosynthesis, aligning with previously reported metabolic alterations in PD. Host factors, particularly bowel movement frequency, influence microbiota, but core disease-associated changes remain after adjustments, and mediation analysis suggests a pathway where dysbiosis contributes to constipation, which in turn increases prodromal PD risk. The identification of a robust microbial signature enabling classification of RBD from controls (AUC ~0.75 in testing) supports diagnostic potential. Collectively, findings support the gut-to-brain hypothesis, implicating early microbiome alterations in α-synucleinopathy pathogenesis and highlighting potential targets for early intervention and risk stratification.

Gut microbiome dysbiosis characteristic of PD is already present in v-PSG–confirmed RBD and begins to emerge in RBD first-degree relatives, preceding motor PD onset. Key features include depletion of butyrate-producing taxa and enrichment of pro-inflammatory/pathobiont-associated genera. Predicted functional shifts toward altered fermentation and reduced vitamin biosynthesis further support a perturbed gut environment. Bowel movement frequency partially mediates the microbiota’s association with prodromal PD risk, underscoring gut–motility–brain interactions. A microbial marker panel can aid in distinguishing RBD from controls and RBD-FDR, suggesting clinical utility for early detection. Future work should include prospective longitudinal studies to establish causality, multi-omics (metagenomics/metabolomics) to resolve species/strain and metabolic outputs, and integration with inflammatory markers and enteric α-syn pathology. Interventional studies targeting constipation, diet (e.g., fiber enrichment), and specific microbial taxa in prodromal stages may offer preventive or disease-modifying strategies.

• Cross-sectional design limits causal inference; longitudinal validation is needed.
• Early PD sample size was modest, though consistent PD-like changes were detected.
• Findings derived within a “body-first” (gut-to-brain) PD staging framework; generalizability to other PD subtypes (e.g., brain-first) is uncertain.
• RBD-FDR group was younger and more female; although adjusted analyses retained key findings, unmeasured factors (genetics, shared diet, early-life exposures) may contribute.
• 16S rRNA gene sequencing provides compositional profiles with limited taxonomic resolution and inferred functions; species/strain-level and metabolomic confirmation via metagenomics/metabolomics are warranted.