The brain structure, immunometabolic and genetic mechanisms underlying the association between lifestyle and depression

The study addresses the rising prevalence and public health burden of major depressive disorder by examining how multiple lifestyle factors collectively influence depression risk and the neurobiological pathways involved. Building on evidence that individual lifestyle components (physical activity, diet, sleep, smoking, alcohol use, sedentary behavior, social connection) and genetic risk contribute to depression, the authors hypothesized that adherence to a healthy lifestyle reduces depression risk across genetic risk strata and that shared neurobiological mechanisms (brain structure and immunometabolic function) modulated by genetic variants underlie this association. Using the UK Biobank’s behavioral, imaging, biochemical, and genetic data, the objectives were: (1) quantify composite lifestyle impact and combined lifestyle–polygenic risk effects on incident depression; (2) assess correlations between lifestyle and brain structural imaging plus peripheral immunometabolic markers; (3) model pathways among lifestyle, polygenic risk score (PRS), brain structure, immunometabolic function, and depression using structural equation modeling (SEM).

Prior work links single lifestyle factors to depression risk: inadequate social connection, higher sedentary time, lack of physical activity, smoking, unhealthy diet, and nonoptimal sleep duration (often U-shaped with risk). Genetic architecture of depression is polygenic, with brain structure considered an intermediate phenotype linking risk variants to psychiatric outcomes. Innate immune mechanisms and immunometabolic dysregulation have been implicated in depression pathophysiology. Previous studies often focused on single behaviors, whereas this study integrates seven lifestyle factors into a composite score to evaluate their combined effects and potential shared neurobiological pathways, including brain morphological changes and immune-metabolic markers.

Study population: UK Biobank prospective cohort (>500,000 participants, ages 37–73) with informed consent and ethical approvals. Data included demographics, behaviors, depression diagnoses, brain MRI, blood biochemistry, blood cell counts, NMR-based metabolic biomarkers, and genotypes. Lifestyle score: Constructed at baseline (2006–2010) from seven factors assessed by touchscreen questionnaire—smoking, physical activity (IPAQ-SF; meeting WHO/AHA recommendations), alcohol consumption (never or moderate: women 0–14 g/day; men 0–28 g/day), diet (meeting ≥4 of 7 food group criteria per DGA), sleep duration (7–9 h per AASM/SRS/NHS), sedentary behavior (screen-based 0–4 h/day as healthy), and social connection (least or moderately isolated per social isolation index). Each healthy factor scored 1; total score 0–7. Lifestyle classes: unfavorable (0–1), intermediate (2–4), favorable (5–7). Depression outcomes: Incident depression from linked hospital inpatient records (ICD-10 F32.x, F33.x), primary care read codes, and death registries; earliest recorded date used. PHQ-4 scores (2006–2010) used for symptom measures and SEM. Genetics and PRS: UKB v3 imputed genotype data (post-QC 8,239,652 SNPs; 337,151 British-ancestry participants). External GWAS summary statistics for depression (PGC meta-analysis excluding UKB and 23andMe to avoid overlap). PRS computed with PRSice-2 with clumping (r^2=0.1, 250 kb) and P-thresholding; PRS at P<0.05 used, categorized into low (lowest quintile), intermediate (quintiles 2–4), high (highest quintile). An alternative PRS based on the original GWAS (excluding 23andMe) showed high correlation with the primary PRS. Neuroimaging: T1-weighted structural MRI on Siemens Skyra 3T; FreeSurfer-derived cortical (68 regions) and subcortical (14 regions) volumes; Qoala-T quality control. Imaging–lifestyle analyses used n=32,839 participants. Peripheral markers: Baseline blood biochemistry (30 markers) and blood counts (31 markers; 59 used after excluding nucleated red cells) from ~480,000 participants; categorized into liver, renal, endocrine, immunometabolic, bone/joint; white/red cell, platelet. NMR metabolomics (168 direct markers) in ~120,000 participants, categorized into lipid-related classes, amino acids, glycolysis-related, ketone bodies, inflammation, etc. Statistical analyses: Multivariable Cox proportional hazards models estimated associations of individual lifestyle factors, lifestyle class, and lifestyle score with incident depression (n=287,282). Models adjusted for age, sex, BMI, Townsend deprivation index, and education; proportional hazards assumption checked via Schoenfeld residuals. Combined analyses of PRS category and lifestyle class (n=197,344) assessed joint effects and tested interaction. Sensitivity analyses for depression subtypes: single episode, recurrent, and treatment-resistant depression (TRD); analyses of smoking status (current vs previous vs never) and alternative reference groups. Correlation analyses: Pearson correlations of lifestyle with brain volumes (adjusted for age, sex, BMI, Townsend index, education, imaging site, intracranial volume; FDR correction) and with peripheral markers (Bonferroni correction). Replicated correlations using lifestyle and depression measured at neuroimaging timepoint (2014+); spatial correlation between lifestyle- and depression-associated brain maps assessed. Mendelian randomization (MR): Bidirectional two-sample MR using genetic instruments for lifestyle (GWAS on lifestyle score within UKB, adjusted for age, sex, 20 PCs) and for depression (PGC meta-analysis excluding UKB and 23andMe). Instruments selected at P<1×10^-7 with LD pruning (r^2<0.01, <1,000 kb). Primary method IVW (fixed effect); sensitivity: weighted median, simple median, weighted mode; tests for heterogeneity (Cochran’s Q, RadialMR), pleiotropy (MR-Egger, MR-PRESSO). Reverse MR tested depression liability on lifestyle. Structural equation model (SEM): Implemented in lavaan (R). Latent variables: depression (four PHQ-4 items), immunometabolic function (top four lifestyle-correlated markers: C-reactive protein, triglycerides, HbA1c, glucose), brain structure (top 20 lifestyle-correlated cortical/subcortical volumes). Paths estimated among lifestyle, PRS, brain structure, immunometabolic function, and depression (n=18,244). Inputs normalized; FDR correction across paths. Two alternative SEMs explored reverse directionality and mediation from depression to lifestyle via brain and immunometabolic latent variables. Covariates: All models adjusted for baseline age, sex, BMI, Townsend deprivation index, education; imaging analyses additionally for site and intracranial volume; genetic analyses additionally for top 20 ancestry principal components. Missing covariate data imputed by mean.

- Population: Among 287,282 participants (mean age 57.52; 50.70% female), 12,916 developed depression over median 9.01 years. Lifestyle score mean 4.75 (s.d. 1.36); 1.25% unfavorable, 38.90% intermediate, 59.85% favorable. - Individual lifestyle factors (healthy vs unhealthy) and depression risk (Cox HR, 95% CI): moderate alcohol consumption HR 0.89 (0.85–0.92); healthy diet 0.94 (0.90–0.97); regular physical activity 0.86 (0.83–0.90); never smoking 0.80 (0.78–0.83); healthy sleep 0.78 (0.75–0.81); low-to-moderate sedentary behavior 0.87 (0.84–0.90); frequent social connection 0.82 (0.78–0.86). All P<10^-3 to 10^-41. - Lifestyle class vs unfavorable: intermediate HR 0.59 (0.53–0.65); favorable HR 0.43 (0.38–0.47). - Lifestyle score (vs 0): HRs for 1–7 points: 0.83, 0.61, 0.53, 0.45, 0.39, 0.33, 0.28; per 1-point increment HR 0.85 (0.84–0.86). - PRS and lifestyle joint effects (n=197,344): Lower PRS associated with reduced risk (intermediate vs high PRS HR 0.87; low vs high PRS HR 0.75). Within each PRS stratum, more favorable lifestyle monotonically lowered risk. Compared to high PRS + unfavorable lifestyle, low PRS + favorable lifestyle HR 0.36 (0.28–0.46). No significant PRS × lifestyle interaction (Pinteraction=0.41). Favorable lifestyle mitigated genetic risk; high PRS + favorable lifestyle had lower risk than low/intermediate PRS + unfavorable lifestyle (permutation tests significant). - Mendelian randomization: One-point higher lifestyle score causally reduced depression risk: IVW OR 0.65 (0.55–0.76), P=1.0×10^-7; consistent in weighted median (0.65), simple median (0.63), and weighted mode (0.52). Reverse MR suggested depression liability associated with less healthy lifestyle (IVW OR 0.85, 0.78–0.92; P=5.9×10^-3), indicating bidirectionality. - Brain structure correlations (n=32,839): Higher lifestyle scores associated with larger volumes in multiple regions (e.g., superior prefrontal, orbitofrontal, precentral, insula; pallidum, thalamus, amygdala, hippocampus). These regions tended to be negatively associated with PHQ-4 depression symptoms. Spatial correlation between lifestyle- and depression-associated brain maps at imaging visit: −0.52 (P=5.4×10^-7). - Peripheral markers: 48 blood and 130 metabolic markers significant after correction. Notable associations: C-reactive protein r=−0.065; triglycerides r=−0.075; neutrophils r=−0.106; leukocytes r=−0.094; degree of unsaturation r=0.153; glycoprotein acetyls r=−0.109 (all P<~10^-240 to <10^-295). - SEM (n=18,244): Lifestyle predicted depression (β=−0.157, P<1×10^-20), immunometabolic function (β=−0.043, P=4.7×10^-7), and brain structure (β=0.038, P=2.9×10^-4/10^-5). Depression was also predicted by PRS (β=0.036, P=1.5×10^-5/10^-3), brain structure (β=−0.023, P=1.2×10^-2/10^-4), and immunometabolic function (β=0.020, P=3.5×10^-2/10^-4). PRS predicted lifestyle (β=−0.022, P=3.5×10^-3) and immunometabolic function (β=0.018, P=3.4×10^-2). Brain–PRS and brain–immunometabolic paths were not significant. Alternative SEMs supported bidirectional and mediated pathways from depression to lifestyle. - Sensitivity analyses: Patterns held for single episode, recurrent depression, and TRD; healthy sleep showed strongest protection for single episode and TRD; social connection strongest for recurrent depression. Smoking risk increased monotonically from previous to current vs never.

Findings demonstrate that a comprehensive healthy lifestyle substantially lowers incident depression risk in a dose–response fashion and that this protective effect is robust across genetic risk strata without significant gene–lifestyle interaction. Causal inference via MR supports that healthier lifestyle reduces depression risk, while reverse MR and alternative SEMs indicate bidirectional influences. Correlations between lifestyle and brain volumes in prefrontal, orbitofrontal, motor, limbic, and thalamic regions, along with broad immunometabolic associations (inflammation, lipids, glycemic measures), suggest lifestyle reflects integrated brain and systemic health. The SEM integrates these observations, proposing coherent pathways where lifestyle influences depression both directly and indirectly through brain structure and immunometabolic function, with PRS exerting additional effects. These results underscore the significance of lifestyle modification as a preventive strategy for depression and provide mechanistic targets involving neural and immunometabolic systems.

This large, multimodal UK Biobank study shows that adherence to a healthy lifestyle—spanning sleep, physical activity, diet, alcohol moderation, non-smoking, low sedentary time, and social connection—protects against depression, with risk decreasing monotonically as lifestyle improves and benefits observed across all genetic risk levels. Mendelian randomization supports a causal, likely bidirectional relationship between lifestyle and depression. Correlations with brain structures and peripheral immunometabolic markers, alongside SEM, reveal plausible pathways linking lifestyle, genetics, neurobiology, and depression. Future work should validate findings in more diverse, independent cohorts; incorporate longitudinal neuroimaging and biomarker data; leverage objective lifestyle measures; and experimentally probe brain–immune mechanisms to refine targeted interventions.

- Lifestyle factors were self-reported, introducing potential measurement error; objective measures (e.g., accelerometry) were not widely available, and their use would reduce sample size and follow-up. - Depression diagnosis timing may lag symptom onset; SEM used PHQ-4 symptoms rather than diagnoses due to sample-size constraints, which may affect generalizability of path estimates. - UK Biobank selection bias (healthier, less diverse participants) limits external validity; limited data on ethnic minorities. - Potential residual confounding given the breadth of variables and observational design. - Stratification schemes for lifestyle and PRS categories may not align with a single validated standard; however, continuous and categorical analyses were consistent. - Lack of longitudinal imaging and peripheral biomarker data to track temporal dynamics of neurobiological mediators. - Non-significant brain–immune path in SEM may reflect model structure or limited power to detect brain–immune interactions amid strong direct paths.