Impact of outdoor nature-related activities on gut microbiota, fecal serotonin, and perceived stress in preschool children: the Play&Grow randomized controlled trial

Aggressive and stress-related behaviors in early childhood are prevalent and can predict later psychiatric problems. Concurrently, rapid urbanization has reduced children’s opportunities to interact with nature, which may adversely affect mental health. Emerging evidence implicates the gut–brain axis, where the gut microbiome can influence neuroactive compounds (e.g., GABA, serotonin) and behavioral phenotypes. Prior studies show stress can alter gut microbiota composition, and that exposure to natural environments associates with psychological well-being. The Play&Grow program aims to increase children’s connectedness to nature. This randomized controlled trial investigated whether structured outdoor nature-related activities in preschoolers affect gut microbiota composition and functions, fecal serotonin levels, and psychosocial outcomes, with the hypothesis that increased connectedness to nature would modulate gut microbiota and serotonin and improve perceived stress and behavior.

Previous research has linked stress exposure to shifts in gut microbiota in animal and human studies (e.g., reductions in Lactobacilli). Gut microbes can produce neuroactive substances including GABA, implicated in anxiety and depression. Short-chain fatty acids also have been associated with depressive phenotypes. Serotonin (5-HT) is a key modulator of physiological and psychological processes; the gut microbiota promotes colonic 5-HT biosynthesis. Natural outdoor environments and greenness correlate with better mental health outcomes and reduced psychiatric risk, and increased exposure to environmental microbes can shape gut, skin, and salivary microbiomes. Early environmental education interventions (including earlier Play&Grow pilots) have improved health behaviors such as vegetable intake. Despite these links, data in young children connecting natural exposure, microbiota functional changes, serotonin, and psychosocial outcomes remain sparse.

Design: Two-arm randomized controlled trial with masked outcome assessment. Trial registration: ClinicalTrials.gov NCT02715544. Ethics: Approved by the Human Research Ethics Committee of the University of Hong Kong; written informed consent obtained. Participants: Children aged 2–5 years with main caregivers recruited in early 2018 in Hong Kong. Exclusions: non-local families, antibiotic therapy within 2 months pre-baseline, chronic health conditions. Randomization and allocation: 54 families randomized by computer-generated number (>0.5 to intervention group [IG], <0.5 to control group [CG]); no stratification/blocking. Timeline: April–June 2018 baseline; 10-week intervention from June 2018; pre- and post-intervention assessments for both groups. Intervention: Weekly 1 session/week for 10 weeks in public parks led by research assistants, with guided hands-on nature activities (e.g., playing with leaves/soil), and at-home tasks (collecting natural objects, nature art, growing plants) designed to increase contact with natural environmental microbiota. Control: No nature program; both groups reminded of standard government healthy lifestyle resources; no specific dietary change recommendations provided. Outcomes: - Connectedness to Nature (CN) including factor scores (e.g., Responsibility towards Nature). - Psychosocial outcomes via Perceived Stress Scale for Children (PSS-C) including anger frequency and prosocial behavior items. - Fecal serotonin (5-HT) quantified by ELISA (pg/mg feces). - Gut microbiota profiled by 16S rDNA amplicon sequencing; QIIME pipeline; OTUs summarized at phylum/genus/species. Sample flow: 63 assessed; 54 randomized (IG n=30; CG n=24). IG: 3 discontinued (antibiotics), 27 analyzed for microbiome; CG: 2 lost to follow-up (moved), 4 discontinued (antibiotics), 18 analyzed for microbiome. Serotonin subsamples: IG n=23, CG n=15 (pre/post paired). Psychosocial subsamples varied per measure (e.g., PSS-C IG n=20, CG n=9; anger/prosocial IG n=24, CG n=12). Sequencing: Mean 49,577 quality-filtered reads/sample. Statistical analysis: Within-group pre/post differences via Wilcoxon signed-rank test for CN scores, alpha diversity (Chao1, Shannon, Simpson), fecal serotonin, stress levels. Trend tests via Jonckheere–Terpstra. Associations: Spearman correlations (serotonin vs species/alpha diversity), repeated-measures correlation (serotonin vs PSS-C), logistic regression (serotonin associations categorized by median), Canonical Correspondence Analysis (CCA) on species profiles per group to identify influences (intervention completion, serotonin, anger). Functional prediction: PICRUSt to KEGG level-3; associations of functional pathways with stress items via Spearman; characterization of pathways increased/decreased in IG vs CG.

Participation: 54 randomized (IG 30; CG 24); microbiome analyses IG n=27, CG n=18; questionnaires completed by 45; 84% provided fecal samples. Connectedness to Nature: IG total CN increased significantly post-intervention (p=0.04; 3.54 to 3.85), CG no significant change (p=0.20; 4.03 to 4.04). Responsibility towards Nature factor increased in IG (p=0.003; 3.33 to 3.67). Fecal serotonin and psychosocial outcomes: IG maintained stable fecal serotonin while CG showed a decreasing trend (between-group p=0.07; CG 249.36 to 220.60 pg/mg). PSS-C total score improved post-intervention (p=0.05), anger frequency decreased (p=0.01), and prosocial behavior increased (p=0.04) in IG compared with CG. Across participants, 65% of IG (n=13) vs 22% of CG (n=2) had decreased PSS-C (Fisher’s exact p=0.01). Fecal serotonin correlated negatively with PSS-C (repeated-measures r = -0.45, p=0.02), trended positively with adequate sleep frequency (JT = 832, p<0.01), and negatively with anger frequency (JT = 475.5, p=0.01); association with anger independent of age. Microbiota diversity and composition: Dominant phyla were Firmicutes (median 0.44), Bacteroidetes (0.27), Proteobacteria (0.11). In IG, Bacteroidetes richness (Chao1) decreased (p<0.01) and Proteobacteria richness increased (p=0.03); no significant IG changes in Actinobacteria and Firmicutes diversity, whereas CG showed increased diversity in these phyla post-period. Participants in IG with decreased PSS-C exhibited higher overall richness. CCA showed pre/post shifts in both groups; intervention completion, serotonin, and anger frequency associated with microbiota structure in IG but not CG. Species-level changes in IG included decreases in Bacteroides (unknown species), Parabacteroides distasonis, unclassified Acidaminococcus, Dialister, and Bilophila; unclassified Blautia increased. Roseburia abundance increased in CG but not in IG. Correlations: Roseburia (unknown species) and Roseburia faecis correlated negatively with fecal serotonin (both p=0.01; r=-0.31). Positive correlations with serotonin observed for Trabulsiella farmeri, Streptococcus anginosus, Eggerthella lenta, and Staphylococcus aureus. PSS-C total score negatively correlated with Bacteroidetes alpha diversity (Chao1 p=0.03, r=-0.27; Shannon p=0.03, r=-0.26; Simpson p=0.05, r=-0.24) and trended with Proteobacteria Shannon (p=0.07, r=-0.23). Higher anger frequency associated with lower Bacteroidetes diversity (Shannon JT p=0.02; Simpson JT p=0.02), more similar community composition (Bray–Curtis p<0.001), and higher abundances of Roseburia spp. (Roseburia other p=0.003; Roseburia faecis p=0.01). Predicted functions: 273 KEGG level-3 categories detected; within IG, 22 pathways decreased and 17 increased post-intervention; CG had only one pathway decrease. PSS-C correlated with multiple level-2 metabolic categories (e.g., positive with carbohydrate metabolisms such as propanoate, glyoxylate/dicarboxylate, pyruvate, glycolysis/gluconeogenesis; correlations with amino acid metabolisms). Indole alkaloid biosynthesis increased with intervention and was negatively correlated with PSS-C; glycosphingolipid biosynthesis-related functions decreased. No functional pathways showed significant overall correlation with fecal serotonin.

This randomized trial demonstrates that a 10-week nature-related early environmental education program increased preschoolers’ connectedness to nature, improved perceived stress and anger-related behaviors, and modulated gut microbiota composition and predicted functions. Notably, the intervention group maintained fecal serotonin levels while the control group trended downward, and fecal serotonin was inversely associated with perceived stress and anger and positively with sleep frequency. Microbiota changes included shifts in phylum-level richness (decreased Bacteroidetes and increased Proteobacteria richness in IG) and species-level alterations, including stabilization of Roseburia abundance compared with a rise in controls; Roseburia spp. were inversely associated with fecal serotonin and enriched among children with higher anger. Functional predictions suggested increases in indole alkaloid biosynthesis (negatively associated with stress) and decreases in glycosphingolipid biosynthesis, aligning with hypothesized microbiota–brain interactions. These findings support the biophilia hypothesis by suggesting that structured contact with natural environments can beneficially influence child psychosocial health potentially via microbiota–serotonin pathways. While causality and mechanisms require confirmation, the results indicate that microbiota may mediate the benefits of nature exposure on behavior in early life.

A 10-week Play&Grow nature-related intervention in preschoolers increased connectedness to nature, improved psychosocial measures (lower perceived stress and anger, higher prosocial behaviors), modulated gut microbiota (including Roseburia abundance and phylum-level richness), and prevented a decline in fecal serotonin observed in controls. Predicted functional changes, such as increased indole alkaloid biosynthesis, were associated with lower stress scores. These data suggest the gut microbiota as a potential target and pathway for behavioral and mental health interventions leveraging nature exposure in early childhood. Future research should employ larger samples, longer follow-up, and mechanistic approaches (e.g., metagenomics/metabolomics, longitudinal mediation analyses) to clarify causal links between environmental microbial exposure, microbiota–serotonin dynamics, and child psychosocial outcomes.

Fecal sample collection time was not standardized, potentially overlooking diurnal variation (although fecal serotonin diurnal fluctuation is unobserved). Small sample size and short intervention duration limit power and generalizability. 16S rRNA profiling constrained species-level resolution and did not permit direct assessment of serotonin metabolism pathways; database/software limitations may have obscured relevant functional differences. Inter-individual variability in exposure intensity may have influenced microbiota changes. This proof-of-principle study lacked a formal a priori power calculation for the primary outcomes analyzed here.