Effects of urban living environments on mental health in adults

Abstract

Urban-living individuals are exposed to many environmental factors that may combine and interact to influence mental health. While individual factors of an urban environment have been investigated in isolation, no attempt has been made to model how complex, real-life exposure to living in the city relates to brain and mental health, and how this is moderated by genetic factors. Using the data of 156,075 participants from the UK Biobank, we carried out sparse canonical correlation analyses to investigate the relationships between urban environments and psychiatric symptoms. We found an environmental profile of social deprivation, air pollution, street network and urban land-use density that was positively correlated with an affective symptom group (r = 0.22, P_perm < 0.001), mediated by brain volume differences consistent with reward processing, and moderated by genes enriched for stress response, including CRHR1, explaining 2.01% of the variance in brain volume differences. Protective factors such as greenness and generous destination accessibility were negatively correlated with an anxiety symptom group (r = 0.10, P_perm < 0.001), mediated by brain regions necessary for emotion regulation and moderated by EXD3, explaining 1.65% of the variance. The third urban environmental profile was correlated with an emotional instability symptom group (r = 0.03, P_perm < 0.001). Our findings suggest that different environmental profiles of urban living may influence specific psychiatric symptom groups through distinct neurobiological pathways.

Fig. 1. Characterization of the study design.

In 141,087 UKB-non-NI participants, we identified urban environmental profiles correlated to psychiatric symptom groups using sCCA with a train–test dataset split design. Next, we carried out GWAS analyses of the symptom groups in 76,508 participants with complete genomic, urban environmental categories and psychiatric symptoms from the UKB-non-NI dataset. The UKB-NI dataset data (n = 14,988) was used for independent replication of the multivariate relationship between urban environmental profiles, genes and symptom groups, and for additional neuroimaging analyses. We analyzed the relationships between urban environmental profiles, regional brain volume and symptom groups using msCCA with a train–test dataset split design. Using a moderated mediation analysis, we then investigated the interaction effect between urban environmental profiles and genetics on psychiatric symptoms groups mediated by brain components in 8,705 participants with complete genomic, urban environmental categories, regional brain volume and symptoms of mental illness in the UKB-NI dataset.

Fig. 2. Distinct urban environmental profiles are correlated with specific psychiatric symptom groups.

a, Fifty-three urban environmental categories belonging to 13 areas (the dots on the right) and 21 psychiatric symptoms are included. UE, urban living environment. b, The sCCA model linking 53 urban environmental categories to 21 psychiatric symptoms identified three significant canonical correlates in the training datasets (red dot), including affective symptom (r = 0.20, P_perm < 0.001), anxiety symptom (r = 0.11, P_perm < 0.001) and emotional instability symptom (r = 0.05, P_perm < 0.001) groups. These results remained significant in the test datasets of affective (r = 0.22, P_perm < 0.001, P_FDR < 0.001), anxiety (r = 0.10, P_perm < 0.001, P_FDR < 0.001) and emotional instability (r = 0.03, P_perm < 0.001, P_FDR < 0.001) (orange square) symptom groups. P values were estimated using one-sided P_perm with FDR correction for multiple comparisons (P_FDR). c, A correlation map between the first urban living environmental profile and affective symptom group in the training (left) and test (right) datasets. d–f, In the first (d), second (e) and third (f) correlates, urban environmental categories contributing to this profile are shown in the first column. EV and fraction of EV of crossloadings of each urban environmental category on each of the three symptom groups are shown in the second and third columns. Symptoms of mental illness contributing to this group are shown on the right radar plots. The affective, anxiety and emotional instability symptom groups are shown in yellow, green and blue. OI, object of interest.

Fig. 3. Genome-wide significant associations of environmental psychiatric symptom groups.

a, The GWAS of the affective symptom group identified 3,436 significantly associated SNPs after Bonferroni correction P < 0.05. The lead SNP rs62062288 is located in intron 6 of the MAPT gene on chromosome 17q21.3 (two-sided P = 6.09 × 10⁻¹⁵). b, Locus zoom plots of 17q21.3 (left), 18q21.2 (middle) and 14q24.1 (right) in the GWAS analysis of the affective symptom group. The purple dots show the lead SNPs of each genomic region. c, GSEA of the affective symptom group-associated 22 genes revealed over-representation in molecular function (dark yellow) of CRH/CRF receptor activity (Q = 5.23 × 10⁻⁴), biological function (yellow) of cellular response to CRH stimulus (Q = 0.02) and cellular components (light yellow) in the axonal growth cone (Q = 0.002) after Bonferroni correction P < 0.05. d, CRHR1, MAPT, DCC and TCF4 gene normalized expression values in 12 brain regions (Human Protein Atlas). e, Left, Correlation r value between 22 gene scores and the affective symptom group in the replication UKB-NI dataset. The replicated 14 genes with two-sided P < 0.05 are marked with an asterisk. Right, Participants with lower CRHR1 scores (upper left), MAPT scores (upper right), DCC scores (lower left) and TCF4 scores (lower right) showed statistically smaller correlations of the first urban environmental profile with the affective symptom group compared to those with higher ones (two-sided P < 0.05). f, The GWAS of the anxiety symptom group identified 29 significantly associated SNPs after Bonferroni correction P < 0.05. The lead SNP rs77641763 is located in intron 15 of the EXD3 gene of chromosome 9 (two-sided P = 9.53 × 10⁻¹¹). g, EXD3 gene normalized expression values in 12 brain regions. h, Top, Of the nine genes scores, six gene scores with two-sided P < 0.05 (marked with an asterisk) repeatedly correlated with the anxiety symptom group. Bottom, Participants with lower EXD3 scores showed a statistically smaller correlation of the second urban environmental profile with the anxiety symptom group compared to higher ones (two-sided P < 0.05). i, The GWAS of the emotional instability symptom group identified ten significantly associated SNPs after Bonferroni correction P < 0.05. The lead SNP rs77786116 is located in chromosome 9 of the IFT74 gene (two-sided P = 4.16 × 10⁻¹⁰). j, IFT74 gene normalized expression value in 12 brain regions. k, Top, Replicated correlations between three gene scores with two-sided P < 0.05 (marked with an asterisk) and emotional instability symptom group from six genes. Bottom, Participants with lower IFT74 scores showed a statistically smaller correlation of the third urban environmental profile with the emotional instability symptom group compared to those with higher scores (two-sided P < 0.05). TPM, transcripts per million.

Fig. 4. Brain volume differences underlying environmental profiles and psychiatric symptom groups.

a–c, Left, Top urban environmental categories contributing to the first (a), second (b) and third (c) urban living environmental profile in the msCCA regression in the UKB-NI dataset. Right, Regional brain volume maps associated with the three urban living environmental profiles and affective (a), anxiety (b) and emotional instability (c) symptom groups. d, Top, Schematic diagram of moderated mediation analysis between genomics, urban environmental profile, brain components and psychiatric symptom groups. Bottom, Each dot shows an indirect effect in the moderated mediation analysis between urban environmental profiles, gene scores, brain components and psychiatric symptom groups. We found that the CRHR1 (EME = 2.01%), MAPT (EME = 1.72%), TCF4 (EME = 1.71%) and DCC (EME = 1.51%) genes moderate the mediation pathway from the first urban environmental profile to brain components of the affective symptom group. The EXD3 gene moderates the mediation pathway from the second urban environmental profile to brain components of the anxiety symptom group (EME = 1.65%). The IFT74 gene moderates the mediation pathway from the third urban environmental profile to brain components of emotional instability symptom group (EME = 1.52%).

Fig. 5. Schematic summary of main findings.

a, Distinct urban-living environmental profiles are correlated with three psychiatric symptom groups. b, GWAS associations and relevent replication analyses reveal that three environmental psychiatric symtom groups are invloved distinct biological pathways. c, msCCA analyses revealed that three environmental psychiatric symtom groups were invloved distinct neurobilogical substrate. d, Different environmental profiles of urban-living may influence specific psychiatric symptom groups through distinct neurocognitive pathways.

Extended Data Fig. 1. A schematic summary of the study design.

GWAS, genome-wide association analysis; msCCA, multiple sparse canonical-correlation analysis; PSY, psychiatric; sCCA, sparse canonical-correlation analysis; SG, symptom groups; UE, urban-environmental; UKB-non-NI dataset and UKB-NI dataset, participants from UK Biobank with complete urban-living environmental data and psychiatric symptoms (n = 156,075) were divided into datasets without neuroimaging data (UKB-non-NI dataset, n = 141,087) and with neuroimaging data (UKB-NI dataset, n = 14,988).

Extended Data Fig. 2. Histograms of demographic, urban-living environment category and psychiatric symptoms variables in the analytical sample and total sample in UK Biobank.

Histograms distributions of demographic variables of age (a) and gender (b), urban-living environment category variables with top weight including IMD score (c), air pollution (d), street network radius (e) and distance to service (f), psychiatric symptoms variables with top weight including frequency of unenthusiasm (g), anxious feelings (h), grief and stress (i) in the analytical sample and total sample in UK Biobank.

Extended Data Fig. 3. Urban-environmental categories construction.

53 urban-living environment categories composed of 128 variables were included in the study. Among these, 34 categories had one independent environmental variable. In the remaining categories, redundancy between related environmental variables was avoided by collapsing the information into 19 latent environmental categories using ten-fold cross-validation confirmatory-factor-analysis (CFA).

Extended Data Fig. 4. Robustness assessment.

a. We used bootstrapping to resample the training data (with replacement) 1000 times, each containing 10% to 150% of the training dataset in 10% increments. Box and whiskers graphs showed the correlation coefficient r value in each resampling. The bounds of box demonstrated data extend from the 25th to 75th percentiles. The centre line in the box was plotted at the median. The whiskers went down to the smallest and up to the largest value. Stability in correlation coefficient after about 40% of the sample size were observed. b. To estimate the stability of the findings across subsamples, we resampled the same proportion 90% of original sample size as train dataset for 1000 times, reran the sCCA algorithm and calculated the correlation between the resulting feature in the remaining 10% test dataset.

Extended Data Fig. 5. The sCCA-regression between urban-living environment categories and psychiatric symptoms in 122,516 participants with different households.

a. Top left: A total of 53 categories of urban-living environment belonged to 13 areas are included. Top right: Each dot demonstrates the 13 areas of urban-living environment. Bottom: A total of 21 psychiatric symptoms are included; b. The sCCA-regression model linking 53 urban-living environment categories to 21 psychiatric symptoms identified three significant correlates in train datasets (red dot), including affective symptoms group (r = 0.20, P_perm < 0.001), anxiety symptoms group (r = 0.12, P_perm < 0.001) and emotional instability symptoms group (r = 0.06, P_perm < 0.001). These results were still significant in test datasets of affective (r = 0.20, P_perm < 0.001, P_FDR < 0.001), anxiety (r = 0.10, P_perm < 0.001, P_FDR < 0.001) and emotional instability symptom-groups (r = 0.03, P_perm < 0.001, P_FDR < 0.001) (orange square). P values here were estimated using two-sided permutation tests (P_perm) and FDR correction have been made for multiple comparisons (P_FDR); c–e. In the first (c), second (d) and third (e) correlates, urban-living environment profiles contributing to this relationship were shown on the top, psychiatric symptoms contributing to this relationship were shown on the bottom radar plots. Dist., Distance; LD, landuse density; IMD, Index of Multiple Deprivation; NDVI, normalized difference vegetation index; SG, symptoms of group; SN, street network; STD, standard deviation; UE, urban-living environment.

Extended Data Fig. 6. Correlations between urban-living environmental profile and psychiatric symptom groups.

Correlation maps between the second urban-living environmental profile and anxiety symptom-group (a) as well as the third urban-living environmental profile and emotional instability symptom-group (b).