3.1 Genetic Links between Air Pollution and Psychiatric Disorders
Our investigation into the genetic interplay between air pollution and psychiatric disorders revealed considerable correlations. Specifically, out of 32 evaluated trait pairings, 14 and 21 pairs displayed significant genetic correlations in the LDSC and HDL analyses, respectively. Notably, the trait pairs identified in LDSC were also evident in the HDL outcomes, underscoring their significance (refer to Figure 2 and Supplementary Table 2 for detailed insights).
In more detail, the study identified genetic correlations between four air pollutants and six different psychiatric disorders. For instance, the association of NO2 with ADHD was positively genetically correlated, as indicated by an LDSC correlation of 0.12 and an adjusted P-value of 9.70E-01. HDL analysis further validated this correlation, showing an elevated significance with a correlation of 0.15 and an adjusted P-value of 1.09E-06. Remarkably, the correlation of NO2 with ASD was particularly strong, presenting a correlation coefficient of 0.33 and a highly significant adjusted P-value of 1.05E-16 in HDL analysis.
The study also highlighted the genetic susceptibility linked to NO2 exposure in BD and MDD. In LDSC analysis, the correlations were 0.15 for BD and 0.17 for MDD. Particularly for MDD, the HDL analysis revealed a notable correlation (adjusted P-value of 3.19E-08). However, the association with AN was not substantial in LDSC analysis and remained non-significant in HDL analysis.
Furthermore, in exploring the impact of particulate matter sizes, namely PM10 and PM2.5, ASD demonstrated a robust genetic correlation once again. The correlation with PM10 was especially strong, as evidenced by a correlation of 0.42 in the HDL analysis, along with a significantly adjusted P-value of 3.04E-05. Similarly, PM2.5 exhibited pronounced correlations with ADHD and MDD, with the adjusted P-values in the HDL analysis indicating significant genetic correlations (ADHD = 1.55E-15, MDD = 5.12E-18).
3.2 GenomicSEM Analysis of Genetic Factors
The application of GenomicSEM, integrating LDSC and HDL methodologies, provided critical insights into the genetic architecture underlying the relationship between air pollution and psychiatric disorders, as depicted in Figure 3. The LDSC model exhibited a chi-square value of 1001.069 at 34 degrees of freedom, reflecting an excellent model fit (p=3.20E-188). The Comparative Fit Index (CFI) stood at 0.983 and the Standardized Root Mean Square Residual (SRMR) at 0.105, both indicative of a robust model fit and satisfactory residual approximation.
The analysis identified two potential genetic factors, labeled F1 and F2. F1 was linked to components of air pollution (NO2, NOx, PM2.5, and PM10), while F2 was associated with a range of psychiatric disorders including ADHD, ASD, BD, MDD, PTSD, and SC. The relationship between F1 and F2 was notably significant, with a standardized estimate of 0.34, suggesting a moderate correlation between these factors. The most potent genetic pathways were evident between F1 and NOx (standardized estimate = 0.99, p=8.93E-109), and F2 and PTSD (standardized estimate = 0.75, p=1.49E-44), as detailed in Supplementary Table 3.
Contrastingly, the HDL method's chi-square statistic was markedly higher at 12455.647, with a p-value of zero, underscoring a more robust fit for the model. Correlations between the latent factors and their indicators remained significant. Notably, a pronounced genetic correlation was observed between PM2.5 and F1 (standardized estimate = 0.59, p=4.51E-05). Similarly, strong genetic pathways linked F2 to psychiatric disorders, particularly PTSD (standardized estimate = 0.84, p=7.22E-28) and MDD (standardized estimate = 0.79, p=2.14E-51). The ongoing significant genetic correlation between air pollution and mental illness was further corroborated by the relationship between F1 and F2 (standardized estimate = 0.53), as elaborated in Supplementary Table 4.
3.3 GWAS Meta-analysis and Multiple Trait Analysis
In our integrated analysis of GWAS, we examined summary data related to six psychiatric disorders associated with air pollution: ADHD, ASD, BD, MDD, PTSD, and SC. This comprehensive meta-analysis yielded a combined dataset, referred to as META PD, encompassing 6,095,927 Single Nucleotide Polymorphisms (SNPs) (see Supplementary Figure 1 for details). Additionally, a separate meta-analysis of GWAS data pertaining to four types of air pollution produced another dataset, META Air, comprising 9,837,128 SNPs (illustrated in Supplementary Figure 1).
Subsequently, we utilized the MTAG method to further process both the META PD dataset and the four GWAS datasets for air pollution. This advanced analysis resulted in the MTAG PD dataset, which includes a total of 5,183,765 SNPs. Within the MTAG PD dataset, we identified 142 loci of statistical significance. Notably, 43 of these loci were previously undiscovered in the META PD dataset. This novel discovery suggests that these additional loci may have potential pleiotropic connections with the psychiatric disorders under study, as well as with the four types of air pollutants examined (further details in Supplementary Figure 1).
3.4 Localized Genetic Correlations Between Air Pollution and Psychiatric Disorders
In our detailed analysis, we conducted a LAVA examination focusing on the META PD dataset in conjunction with data from four air pollution categories. This analysis centered on 142 loci regions identified as significantly variant within the MTAG PD dataset. Following the BH correction, we identified a total of 114 loci, which included 46 unique loci, as detailed in Supplementary Table 5. Notably, the 5q21.2 region, situated on chromosome 5, exhibited significant positive correlations across all examined trait combinations.
Subsequent colocalization analysis of these 114 loci revealed the identification of 13 loci, with four being unique, where there is a convergence between psychiatric disorders and air pollution. These findings are presented in Table 1 and Supplementary Table 6. The 5q21.2 locus, in particular, showed consistent correlations across different air pollutants, indicating that this region could be a key genetic intersection impacting the risk of psychiatric disorders in relation to air pollution exposure. Notably, the SNV rs30266 within the 5q21.2 region was identified as having the highest likelihood of being a causal variant across the four air pollution categories, thereby suggesting a potential causal link between air pollution and psychiatric disorders.
These findings collectively point towards certain genomic regions, especially 5q21.2, playing a pivotal role in the genetic response to air pollution and its subsequent effect on psychiatric disorders. SNVs within these key regions, particularly rs30266 in 5q21.2, may serve as crucial biomarkers for understanding the intricate interplay between environmental factors and genetic susceptibility in psychiatric disorders.
3.5 Identifying Common Genetic Loci Between Air Pollution and Psychiatric Disorders
In our initial stage of analysis, we utilized PLACO analysis on the META PD and META Air datasets, which led to the identification of 1461 SNVs, as depicted in Supplementary Figure 1. Further exploration using the FUMA tool enabled us to pinpoint 23 independent genomic risk loci. These loci are dispersed over 22 distinct chromosomal regions and are indicative of pleiotropic effects, as detailed in Supplementary Table 7.
Progressing further, we employed the GCTA-COJO tool's stepwise model selection method for a conditional and joint association analysis on the MTAG PD data. This approach facilitated the identification of 173 independent genomic risk loci spread across 131 independent chromosomal regions, as elaborated in Supplementary Table 8.
Among these findings, two SNVs stood out for their association with both air pollution and psychiatric disorders across different analytical methodologies. These are rs76416526, located in the 11p11.2 region, and rs30266 in the 5q21.2 region. The consistent associations observed with these two SNVs across different analyses underscore their potential role as key pleiotropic loci in mediating the relationship between exposure to air pollution and the development of psychiatric disorders.
3.6 Investigating Biological Mechanisms Shared by Multiple Efficacy Loci in Psychiatric Disorders and Air Pollution
3.6.1 Annotation of Pleiotropic Loci
Through the application of the ANNOVAR tool for genetic annotation of pleiotropic loci, we discerned seven SNVs classified as exonic variants. Notably, three of these (rs13107325, rs78648104, and rs3814883) were situated in the exonic regions of their respective mRNA. While rs3814883 (TAOK2) did not exhibit high CADD scores indicative of potential harmfulness, the other two SNVs did, as detailed in Supplementary Table 9. Interestingly, two critical loci in our study, rs76416526 and rs30266, were categorized differently: one as a downstream gene variant and the other as a noncoding RNA variant in an intronic region. Neither of these exceeded the threshold for CADD scores.
3.6.2 Analysis of Genes Influenced by Pleiotropic SNVs
Utilizing MAGMA, we identified 297 genes from 194 pleiotropic SNVs (refer to Supplementary Tables 10 and 11). Further refinement using the POPS tool led us to 38 genes with PoPS Scores above 1, which we classified as pleiotropic genes linking air pollution to psychiatric disorders (Supplementary Tables 10 and 11).
3.6.2.1 Tissue-Specific Enrichment in Pleiotropic Genes
Our tissue-specific enrichment analysis, conducted using deTS with reference to GTEx and ENCODE data, revealed a significant concentration of pleiotropic genes in brain-associated tissues. These tissues include the cerebral cortex, prefrontal cortex, hypothalamus, hippocampus, anterior cingulate cortex, and various regions within the basal ganglia, substantia nigra, and cerebellum. This finding underscores the central nervous system's pivotal role in psychiatric disorders and how it might be influenced by air pollution. Additionally, the observed enrichment in non-brain tissues, such as the salivary glands and esophageal muscles, hints at a broader, multi-systemic network. This suggests that air pollution's impact on mental health might extend through these tissues, with notable implications for ingestive behavior and digestion (detailed in Figure 4A; Supplementary Tables 12 and 13; Supplementary Figure 2 and 3).
3.6.2.2 Genomic Analysis Elucidating the Role of Pleiotropic Genes in Air Pollution and Psychiatric Disorders
Our genomic enrichment studies have illustrated a notable enrichment of pleiotropic genes at the intersection of air pollution and psychiatric disorders, especially in pathways critical to synaptic structure, organization, and neurogenesis - fundamental aspects of central nervous system regulation. Further examination revealed these genes' involvement in a variety of neural processes, including head development, learning, cellular morphogenesis, axon growth, trans-synaptic signaling, cell adhesion, and general nervous system functions. These elements collectively play vital roles in developmental growth regulation, synaptic plasticity, cellular structure, and guidance of neuronal projections.
Significantly, our data underscore the enrichment of specific pathways, such as the negative regulation of G protein-coupled receptor signaling, modulation of synaptic plasticity, cognitive processing, response modulation to negative stimuli, prepulse inhibition, and neurological control of systemic arterial blood pressure. These insights shed light on potential molecular mechanisms common to both air pollution and psychiatric disorders, highlighting the intricate nature of neuroplasticity alterations in response to environmental factors (as shown in Figure 4A and detailed in Supplementary Table 14).
Further cell type analysis revealed a significant enrichment of pleiotropic genes in midbrain cells and various neural types. This suggests a close linkage between genetic factors influencing susceptibility to psychiatric disorders and responses to air pollution, and the fundamental neurobiological processes. Notably, the enrichment in neural types like GABAergic, dopaminergic, and 5-hydroxytryptaminergic neurons in the midbrain underscores the neurochemical pathways potentially influenced by air pollution. These pathways are pivotal in managing emotions, reward processing, and cognitive functions, which are often disrupted in psychiatric disorders. The presence of pleiotropic genes in these neural populations indicates that air pollution might impact mental health by affecting the development, function, or survival of these neurons.
Furthermore, the identification of pleiotropic genes in fetal pulmonary visceral neurons and enteric nervous system neurons suggests that air pollution may influence central nervous system functions via peripheral pathways. This could involve interactions with the gut-brain axis or direct impacts on pulmonary signaling. The distribution of these genes across various cell types, including retinal ganglion cells and kidney cells, indicates a broad spectrum of effects that air pollution might have on different body systems. Such alterations could potentially exacerbate the risk of mental health deterioration (refer to Supplementary Figure 4 for details). These findings collectively offer vital insights into the complex interplay between environmental factors and genetic susceptibility, emphasizing the extensive and pervasive nature of environmental pollution's potential impact on mental health.
3.6.2.3 Comprehensive Analysis of Phenotypic Diversity Linked to Pleiotropic Genes in Air Pollution and Psychiatric Disorders
Our extensive parallel phenotypic enrichment analyses, utilizing the Mouse Genome Informatics platform, identified 19,326 genes with applicable phenotypic annotations. These pleiotropic genes were found to be significantly overrepresented in six key phenotypic categories, highlighting their diverse potential effects (as illustrated in Figure 4A and detailed in Supplementary Table 15).
Particularly notable is the range of possible impacts of air pollution on the nervous system, encompassing neurodevelopmental impairments, degenerative neurological conditions, and various neurobehavioral anomalies. These factors are intricately connected to the etiology of psychiatric disorders. Air pollution appears to exert a direct influence on the nervous system, potentially compromising mental health by affecting neuronal structures, synaptic organization and signaling, and overall neurodevelopment and plasticity.
The observed enrichment in extremity phenotypes suggests that environmental pollution might interfere with hormonal balances or growth factors, impacting physical development and overall body morphology. The emphasis on cardiovascular system phenotypes indicates the potential of air pollutants to harm cardiovascular health, possibly by inciting inflammatory responses or inducing oxidative stress, thereby increasing the risk of conditions like heart disease and hypertension.
Furthermore, the substantial enrichment in behavioral and neurological phenotypes highlights the possible impact of air pollution on behavioral and cognitive functions. This aligns with existing research suggesting links to neurobehavioral disorders, including anxiety, depression, and cognitive impairment. Meanwhile, the enrichment in mortality and aging phenotypes implies that prolonged exposure to air pollution may accelerate the biological aging process. This could hasten the onset of age-related conditions, underlining the broad and lasting impact of environmental factors on overall health and longevity.
3.6.3 Investigating the Role of Pleiotropic Genes in Air Pollution and Psychiatric Disorders
Employing SMR analysis, we explored the interrelations between pleiotropic genes, various air pollution elements, and psychiatric disorders (as depicted in Figure 4B and detailed in Supplementary Table 16). In our examination of blood samples, the INPP4B gene emerged as a noteworthy example. It exhibited a negative correlation with NO2 exposure, yet displayed a positive association with MDD and SC. In contrast, brain tissue analysis revealed that the ELAVL2 gene was negatively correlated with both PM10 exposure and MDD and SC. Similarly, the PPP2R2B gene showed a consistent negative correlation with PM10, NO2 exposure, and SC. Lung tissue analysis highlighted the DCC gene, which demonstrated positive associations with NO2, NOx, and PM2.5 exposures. Interestingly, in brain tissue samples, the same gene was positively associated with ADHD, MDD, and SC. These insights offer vital clues about the molecular pathways through which air pollution may influence the risk of psychiatric disorders.
3.6.4 Analysis of Pleiotropic Proteins
In our protein-focused exploration, we analyzed pQTL data using an SMR approach akin to eQTL analysis. This led to the identification of 298 proteins related to either air pollution or mental illness (refer to Supplementary Figure 5 and Supplementary Table 17). Notably, among these proteins, four (LSAMP, EFNA5, NCAM1, and HS6ST3) were also identified as pleiotropic genes. Although these four proteins were found in a single phenotypic study, there was no clear evidence of their simultaneous impact on both air pollution and psychiatric disorders. However, we did find 37 proteins among other pleiotropic proteins that were concurrently associated with at least one air pollutant and a psychiatric disorder (details in Supplementary Figure 6). These discoveries not only broaden our comprehension of the shared influences of air pollution and mental illness but also hint at potential biological pathways at the protein level.
3.7 Exploring the Mediating Role of Personality Traits in the Link Between Air Pollution and Psychiatric Disorders
In light of the strong correlations observed in the 5q21.2 region between air pollution and psychiatric disorders across multiple methodologies, we engaged in hyprcoloc analysis to investigate colocalization signals among psychiatric disorders, air pollution, and personality traits in this region. Our analysis identified colocalization between four air pollution factors, psychiatric disorders, and three personality traits (Fed-up, Loneliness, and Mood swings) within the 5q21.2 region (as outlined in Table 2). Notably, the pleiotropic locus SNV rs30266 demonstrated a high posterior probability for all examined phenotypes, indicating robust evidence of its potential role as a mediator in the association between these traits.
We then employed GenomicSEM to develop a mediation model focusing on the relationship of Air Pollution - Fed-up/Loneliness/Mood swings - Psychiatric Disorder. The findings suggested that air pollution might not directly lead to psychiatric disorders but instead may act through personality traits as mediators (details in Supplementary Table 18). For instance, we observed a significant positive correlation between psychiatric disorders and PM2.5, with a standardized genetic estimate of 0.24 (p=4.95E-06). Moreover, the genetic correlation between psychiatric disorders and the trait of feeling fed up (Fed-up) was 0.39 (p=3.29E-32), while the correlation between feeling fed up and PM2.5 was 0.30 (p=2.75E-10), hinting at the mediating role of personality traits in the impact of air pollution on psychiatric disorders.
Similarly, we noted patterns in the associations of various air pollutants with different personality traits. For example, the standardized genetic correlation of NOx with Loneliness was 0.25 (p=4.61E-07), indicating that loneliness might be a key mediator in the link between air pollution and psychiatric disorders. The correlation of Mood swings with PM2.5 was also significant, at 0.26 (p=2.53E-08), suggesting its importance as a mediator in the association between PM2.5 exposure and psychiatric disorders.
Further, our analysis showed mediating effects associated with PM10. The standardized genetic correlation between psychiatric disorders and PM10 was 0.18 (p=0.049), while the correlation between Fed-up and PM10 was 0.33 (p=0.0022), reinforcing the mediating role of personality traits. Additionally, the correlation between Mood swings and PM10 was 0.31 (p=0.0019), indicating a similar mediating role.
Collectively, these findings imply that the connection between air pollution and psychiatric disorders may be moderated by personality traits, with Fed-up, loneliness, and mood swings potentially serving as critical psychological mediators linking environmental factors to mental health outcomes.
3.8 Mendelian Randomization Analysis of Air Pollution and Psychiatric Disorders
In our MR analysis, we evaluated four air pollution variables as exposures and eight psychiatric disorders as outcomes. The initial analysis, based on raw p-values, identified a positive causal relationship between PM10 exposure and several psychiatric disorders: ADHD, BD, and MDD. A similar positive association was observed between PM2.5 exposure and ADHD. However, after applying the BH correction for multiple comparisons, the causal link between PM10 and ADHD was the only one that remained statistically significant (β = 0.83). This finding was further supported by robust results in the heterogeneity test (Q.pval = 0.84) and the Egger-intercept test (Egger-intercept = 0.76), as elaborated in Supplementary Table 19.
Furthermore, the causal relationship between PM10 exposure and ADHD was re-examined using GSMR analysis, which yielded a P-value of 0.00011. Interestingly, this analysis presented an inverse association (β=-0.81) compared to the initial TSMR findings. This contrasting result adds to our understanding of the complex and nuanced causal relationships between air pollution and psychiatric disorders.