Circulating Metabolite Abundances Associated With Risks of Bipolar Disorder, Schizophrenia, and Depression: A Mendelian Randomization Study

BACKGROUND: Preventive measures and treatments for psychiatric disorders are limited. Circulating metabolites are potential candidates for biomarker and therapeutic target identi ﬁ cation, given their measurability and essential roles in biological processes. METHODS: Leveraging large-scale genome-wide association studies, we conducted Mendelian randomization analyses to assess the associations between circulating metabolite abundances and the risks of bipolar disorder, schizophrenia, and depression. Genetic instruments were selected for 94 metabolites measured in the Canadian Longitudinal Study on Aging cohort ( N = 8299). We repeated Mendelian randomization analyses based on the UK Biobank, INTERVAL, and EPIC (European Prospective Investigation into Cancer) – Norfolk studies. RESULTS: After validating Mendelian randomization assumptions and colocalization evidence, we found that a 1 SD increase in genetically predicted circulating abundances of eicosapentaenoate and docosapentaenoate was associated with odds ratios of 0.72 (95% CI, 0.65 – 0.79) and 0.63 (95% CI, 0.55 – 0.72), respectively, for bipolar disorder. Genetically increased U -3 unsaturated fatty acids abundance and U -3-to-total fatty acids ratio, as well as genetically decreased U -6-to- U -3 ratio, were negatively associated with the risk of bipolar disorder in the UK Biobank. Genetically increased circulating abundances of 3 N -acetyl-amino acids were associated with an increased risk of schizophrenia with a maximum odds ratio of 1.31 (95% CI, 1.18 – 1.44) per 1 SD increase. Furthermore, a 1 SD increase in genetically predicted circulating abundance of hypotaurine was associated with an odds ratio of 0.85 (95% CI, 0.78 – 0.93) for depression. CONCLUSIONS: The biological mechanisms that underlie U -3 unsaturated fatty acids, NAT8-catalyzed N -acetyl- amino acids, and hypotaurine warrant exploration to identify new biomarkers and potential therapeutic targets

https://doi.org/10.1016/j.biopsych.2024.04.016Psychiatric disorders pose significant medical and socioeconomic challenges globally (1,2).Due to the complex nature of these diseases, preventive measures and treatments for psychiatric disorders have been limited in both quantity and effectiveness (3)(4)(5).New biomarkers and therapeutic targets are urgently required, and there is a need to better understand the underpinning biological mechanisms.
Molecules in circulation, such as proteins and metabolites, are involved in various biological processes and may influence risks of diseases.They are favorable candidate biomarkers or therapeutic targets because their abundances can be measured relatively easily and may be modulated.However, elucidating the roles of circulating molecules in complex diseases is challenging.Associations identified in observational studies of psychiatric disorders may be biased by confounders, such as socioeconomic status (6), and reverse causation, wherein the circulating molecule abundances can be altered as a consequence of disease onset or progression.
One way to systematically identify circulating molecules that may play a role in complex diseases is through Mendelian randomization (MR) (7,8).MR leverages genetic variants as instrumental variables to assess the potential effect of a risk factor (exposure) on a disease outcome.MR has 3 fundamental instrumental variable assumptions: 1) relevance-a genetic instrument should be strongly associated with the exposure; 2) independence-a genetic instrument should be independent of any factors that confound the exposureoutcome relationship; and 3) exclusion restriction-a genetic instrument should act on the outcome only through the exposure and not through alternative pathways, which is referred to as the assumption of no horizontal pleiotropy (7,8).The relevance assumption is fulfilled by selecting genetic variants significantly associated with an exposure in a genome-wide association study (GWAS).These variants are typically not associated with potential confounders due to the randomization at conception (7,8).However, the assumption of no horizontal pleiotropy requires careful examination of the biological mechanisms through which the genetic variants act.
MR has been adopted to identify molecular risk factors for various complex diseases, such as coronary artery disease (9), ischemic stroke (10), and severe COVID-19 outcomes (11,12).We have also utilized MR to identify proteins whose circulating abundances may influence risks of psychiatric disorders (13).On the other hand, circulating metabolites originate from various sources, such as dietary intake, gut microbiota, endogenous metabolism, and environmental exposures (14)(15)(16).They have been important biomarkers and therapeutic targets for many human diseases (17)(18)(19).However, their roles in psychiatric disorders remain insufficiently investigated.
Therefore, in this study, leveraging large-scale GWASs, we conducted MR analyses for bipolar disorder (20), schizophrenia (21), and depression (22).We implemented a unified framework to identify genetic instruments while minimizing the risk of horizontal pleiotropy from our recent study that characterized circulating metabolite quantitative trait loci (mQTL) in the CLSA (Canadian Longitudinal Study on Aging) (N = 8299) (19).We strengthened the validity of our findings through multiple sensitivity analyses, colocalization analyses, and repeating MR analyses based on the UK Biobank (23), IN-TERVAL, and EPIC (European Prospective Investigation into Cancer)-Norfolk studies (18).Our results shed light on the underlying biological mechanisms that may contribute to risks of psychiatric disorders through the regulation of circulating metabolite abundances and suggest circulating metabolites that may be explored as biomarkers or therapeutic targets.

GWASs and Selection of Genetic Instruments
Circulating metabolite abundances were measured in up to 8299 unrelated individuals of European ancestry in the CLSA cohort (24).Large-scale meta-analyses of psychiatric disorder GWASs were conducted by the Psychiatric Genomics Consortium (20)(21)(22).GWASs and the selection of genetic instruments that affect effector genes (19) are described in Supplement 1. Participants in these GWASs were predominantly of European ancestry.There was no overlap between the participants in the CLSA study and the psychiatric disorder GWASs.

MR and Sensitivity Analyses
For each psychiatric disorder (outcome), we performed MR analyses to assess the potential effect of the circulating abundance of each metabolite (exposure) that could be instrumented.If a genetic variant was selected as instrument but was not present in the psychiatric disorder GWAS, another variant in high linkage disequilibrium (LD) with the genetic instrument (r 2 .0.8) may be selected as a proxy using LDlink (25), based on the 1000 Genomes Project non-Finnish European ancestry reference panel (26).
For metabolites with only 1 genetic instrument, the metabolite-disease association was assessed by a Wald ratio estimate.For metabolites with 2 or more genetic instruments, the metabolite-disease association was assessed by an inverse variance-weighted random effects estimate.Of the 280 metabolite-disease associations tested, those with a Benjamini-Hochberg (false discovery rate)-corrected p , .05 were considered significant.If a metabolite had 3 or more genetic instruments, we also performed sensitivity analyses using weighted median, penalized weighted median, weighted mode, and MR-Egger approaches, where a significant MR-Egger intercept (p , .05) would indicate the existence of directional pleiotropy (27).We calculated the F statistic for each test to quantify instrument strength.An F statistic .10 was considered to indicate low risk of weak instrument bias (28).We implemented Steiger filtering to assess whether the effect directionality was correctly specified for each genetic instrument, with the assumption that valid genetic instruments should explain more variance in the exposure than in the outcome (29).MR analyses were conducted using the Two-SampleMR R package version 0.5.6 (30).

Colocalization Analyses
Associations detected in MR may be subject to confounding by LD.Therefore, for each significant metabolite-disease association, we assessed whether the metabolite and the psychiatric disorder shared the same causal genetic variants.Colocalization analyses were performed based on GWAS summary statistics including all variants in a 6500-kb window around each genetic instrument.Default priors of coloc (p 1 = 1 3 10 24 , p 2 = 1 3 10 24 , and p 12 = 1 3 10 25 ) were used (31).Additionally, we adopted pairwise conditional analysis and colocalization, which relaxes the single causal variant assumption of coloc to assess colocalization between conditionally independent genetic associations with the same default priors (9).A posterior colocalization probability .0.8 was considered strong evidence of colocalization, while a posterior colocalization probability .0.5 was considered suggestive evidence of colocalization.

Assessment of Reverse Causation
Furthermore, for each significant metabolite-disease association supported by colocalization evidence, we assessed possible reverse causation by performing reverse MR, where the genetic liability to the psychiatric disorder was considered as the exposure, and the circulating metabolite abundance was considered as the outcome.Independent genome-wide significant variants were selected as genetic instruments based on the psychiatric disorder GWAS [p , 5 3 10 28 and LD r 2 , 0.001 based on the 1000 Genomes Project non-Finnish European ancestry reference panel (26)].Inverse varianceweighted random effects estimates were derived with sensitivity analyses performed using weighted median, penalized weighted median, weighted mode, and MR-Egger approaches, respectively.

MR Based on INTERVAL and EPIC-Norfolk Studies
We further validated our findings using Metabolon HD4 platform profiles of plasma metabolites in the INTERVAL and EPIC-Norfolk studies (18).Details of these studies have been described previously (18).Independent genome-wide Circulating Metabolites and 3 Psychiatric Disorders Biological Psychiatry --, 2024; -:---www.sobp.org/journalBiological Psychiatry significant variants associated with circulating metabolite abundances derived from conditional and joint analyses and their GWAS test statistics were retrieved from the INTERVAL cohort (N = 8455), the EPIC-Norfolk discovery cohort (N = 5841) and the EPIC-Norfolk validation cohort (N = 5698), respectively.MR analyses were repeated for significant metabolite-disease associations, using potentially cohortspecific genetic instruments when the metabolites were also measured and had genetic instruments identified in these studies.

Phenome-Wide Association Studies
For significant metabolite-disease associations supported by colocalization evidence, we assessed whether the genetic instruments had been associated with other human traits or diseases, which may suggest various functions of the metabolites or the underlying effector genes.We queried the Phe-noScanner (version 2) database, which included over 65 billion associations (32).

Additional Sensitivity Analyses
We performed additional sensitivity analyses based on UK Biobank participants of European ancestry for significant metabolite-disease associations supported by colocalization evidence.These analyses are described in Supplement 1, including MR analyses to further characterize the potential role of unsaturated fatty acids in bipolar disorder and genetic risk score analyses to assess the associations between genetically predicted metabolite abundances and lifetime risks of bipolar disorder, schizophrenia, and depression.We included sex and age as negative control outcomes, as well as measures of socioeconomic status and educational attainment, to evaluate potential bias due to unobserved confounding (33).

Associations Between Circulating Metabolite Abundances and Risks of Psychiatric Disorders
An overview of the MR study design is shown in Figure 1.In the CLSA cohort, we identified 94 effector genes that demonstrated biological relevance to the associated metabolites and whose gene expression was regulated by the same genetic variants of circulating metabolite abundances (Methods and Materials; Supplement 1; Tables S1-S3 in Supplement 2).
After data harmonization, we performed MR analyses to evaluate 94 metabolite-bipolar disorder associations, 93 metabolite-depression associations, and 93 metaboliteschizophrenia associations.Of these 280 tests, 243 were based on a Wald ratio estimate where the circulating metabolite abundance was instrumented by 1 genetic variant (Table S4 in Supplement 2), while 37 were based on an inverse variance-weighted random effects estimate where 2 or more genetic instruments were identified.Only 14 tests utilized LD proxies of independent genome-wide significant variants as genetic instruments (Table S4 in Supplement 2).All proxy variants were in high LD with the lead variants (minimum r 2 = 0.96).In total, 50 metabolite-disease associations were significant at a false discovery rate , .05, 14 of which reached the more stringent Bonferroni-corrected significance threshold (p , 1.8 3 10 24 ).Detailed MR results are summarized in Table S5 in Supplement 2. All of these tests had an F statistic .10,  S1 in Supplement 2; mQTL-cis-eQTL colocalization results derived using coloc are available in Table S2 in Supplement 2; mQTL-cis-sQTL colocalization results derived using coloc are available in Table S3  We performed colocalization analyses to assess potential bias due to confounding effects of LD.As a result, 11 significant associations (false discovery rate , .05) were supported by evidence of colocalization with a posterior colocalization probability .0.5 that the circulating metabolite abundance and the psychiatric disorder shared the same causal variants (Figure 2; Table S6 in Supplement 2).Colocalization evidence was strengthened for these 11 associations with pairwise conditional analysis and colocalization, which allows for multiple causal variants in the same locus (Table S7 in Supplement 2).
We did not find evidence of reverse causation affecting these associations in reverse MR analyses in which psychiatric disorders were considered as exposures and circulating metabolite abundances were outcomes (p ..05) (Methods and Materials; Table S8 in Supplement 2).

Circulating Metabolite Abundances Associated With the Risk of Bipolar Disorder
Of the 11 associations supported by colocalization evidence, increased circulating abundances of unsaturated fatty acids appeared to be strongly associated with reduced risk of bipolar disorder (Figure 2).For example, a 1 SD increase in genetically predicted circulating eicosapentaenoate abundance was associated with an odds ratio (OR) of 0.72 (95% CI, 0.65-0.79;p = 2.2 3 10 212 ; based on rs174546, minor allele frequency [MAF] = 0.339) for bipolar disorder, while a 1 SD increase in genetically predicted docosapentaenoate (n3 DPA) abundance was associated with an OR of 0.63 (95% CI, 0.55-0.72;p = 1.5 3 10 211 ; based on rs4246215, MAF = 0.353).Five associations between the risk of bipolar disorder and circulating abundances of unsaturated fatty acids, including 3 U-3 unsaturated fatty acids and 2 U-6 unsaturated fatty acids, were supported by strong evidence of colocalization with a minimum posterior colocalization probability of 0.92 (Figure 3A; Figure S1 in Supplement 1; Tables S5-S7 in Supplement 2).
The effector gene underlying genetic associations with the 5 unsaturated fatty acids was the FADS gene cluster coding for fatty acid desaturases, which play a crucial role in the biosynthesis of long-chain polyunsaturated fatty acids from their dietary precursors (34) (Figure S2 in Supplement 1; Table S4 in Supplement 2).As expected, for each genetic instrument, the allele that increased FADS gene expression in multiple tissues also increased the circulating abundances of the investigated unsaturated fatty acids (Table S2 in Supplement 2).
All metabolite-bipolar disorder associations were detected using genetic instruments identified in the INTERVAL and EPIC-Norfolk cohorts with highly consistent MR estimates (Figure 3C; Table S9 in Supplement 2).
Notably, the genetic instruments for the 5 unsaturated fatty acids demonstrated high pleiotropy and had been associated with various human traits and diseases, including other lipid biomarkers (Table S10 in Supplement 2).To better assess the potential effects of U-3 unsaturated fatty acids and U-6 unsaturated fatty acids, we performed additional MR analyses based on GWASs of U-3 unsaturated fatty acids, U-6 unsaturated fatty acids, total fatty acids, and their relative abundances conducted in the UK Biobank (Supplement 1).MR estimates suggested that increased circulating abundance of U-3 unsaturated fatty acids, increased U-3-to-total fatty acids ratio, and decreased U-6-to-U-3 ratio were significantly associated with decreased odds of bipolar disorder (Figure 3D; Table S11 in Supplement 2).For example, a 1 SD increase in the genetically predicted U-3-to-total fatty acids ratio was associated with an OR of 0.89 (95% CI, 0.83-0.95;p = 3.8 3 10 24 ; based on 29 genetic instruments) for bipolar disorder.Although these analyses may be subject to directional pleiotropy as indicated by significant MR-Egger intercepts (p , .05), the MR estimates obtained based on different approaches of sensitivity analysis were highly consistent (Figure 3D; Table S11 in Supplement 2).In contrast, genetically predicted circulating abundances of U-6 unsaturated fatty acids and total fatty acids and their relative abundance did not demonstrate associations with the risk of bipolar disorder (p ..05) (Figure 3D; Table S11 in Supplement 2).

Amino Acid Abundances Associated With the Risk of Schizophrenia
MR estimates suggested that increased circulating abundances of 3 N-acetyl-amino acids were associated with increased risk of schizophrenia, with suggestive evidence of colocalization (Figure 4A; Tables S5-S7 in Supplement 2).In These N-acetyl-amino acids may be products of NAT8 (N-acetyltransferase 8)-catalyzed reactions (Figure 4B), wherein the protein-coding gene NAT8 was the effector gene underlying the genetic associations with the circulating metabolite abundances.Notably, the genetic instruments for the circulating abundances of these 3 metabolites were in almost perfect LD (r 2 .0.99) (Figure 4C).For each genetic instrument, the allele that increased NAT8 gene expression in multiple tissues also increased the circulating metabolite abundance (Table S2 in Supplement 2).One of these genetic instruments was a missense variant, rs13538, affecting NAT8 (Phe143Ser), which did not seem to significantly alter the protein structure of NAT8 (Figure 4D) but may have a moderate impact on the enzymatic activity (36).This variant had been associated with biomarkers of kidney function and chronic kidney disease (Table S10 in Supplement 2).In addition, increased circulating abundances of these 3 metabolites may be marginally associated with increased risk of bipolar disorder, with inconclusive evidence of colocalization (posterior probability ranged between 0.4 and 0.5) (Tables S6 and S7 in Supplement 2).No genetic instrument was available for these 3 metabolites in the INTERVAL or EPIC-Norfolk studies.

Hypotaurine Associated With the Risk of Depression
We also found that a 1 SD increase in genetically predicted circulating abundance of hypotaurine was associated with an OR of 0.85 (95% CI, 0.78-0.93;p = 3.7 3 10 24 ; based on rs1366463, MAF = 0.470) for depression.This association was supported by a posterior probability of colocalization of 0.67 (Figure 5A; Tables S5-S7 in Supplement 2).The genetic instrument, rs1366463, is an intergenic variant and did not have known horizontal pleiotropic effects (Table S10 in Supplement 2).Hypotaurine is a downstream product of cysteine dioxygenase type 1 (coded by the underlying effector gene CDO1)catalyzed reaction derived from cysteine, and precursor of taurine (Figure 5B).The allele of the genetic instrument that increased CDO1 gene expression in multiple tissues also increased the circulating abundance of hypotaurine (Table S2 in Supplement 2).The circulating abundance of taurine could not be instrumented due to lack of genetic associations detected in GWASs.No genetic instrument was available for hypotaurine in the INTERVAL or EPIC-Norfolk studies.

Additional Sensitivity Analyses
To further verify the above metabolite-disease associations, genetic risk score analyses were conducted in the UK Biobank, including 1340 cases of bipolar disorder and 408,352 controls, 702 cases of schizophrenia and 408,990 controls, and 21,966 cases of depression and 387,726 controls based on clinical diagnoses (Supplement 1).As expected, due to limited statistical power, the associations between the genetically predicted circulating metabolite abundances and lifetime risks of the respective disease outcomes were not significant (Table S12 in Supplement 2).Nevertheless, notably, all of these tested associations exhibited the same effect direction as identified in MR analyses, with highly consistent effect size estimates and largely overlapping confidence intervals (Figure S3 in Supplement 1).
Importantly, although the negative control analyses were well powered, with 221,500 females and 188,192 males, we did not detect any significant associations between the genetically predicted circulating metabolite abundances and sex or age (p ..05) (Table S13 in Supplement 2), indicating a Circulating Metabolites and 3 Psychiatric Disorders Biological Psychiatry --, 2024; -:---www.sobp.org/journalBiological Psychiatry low risk of unobserved confounding.However, the genetically predicted circulating abundances of the 3 schizophreniaassociated N-acetyl-amino acids were marginally associated with household income and age at full-time education completion (Table S13 in Supplement 2).

DISCUSSION
Psychiatric disorders incur a high disease burden worldwide and necessitate identification of new biomarkers and therapeutic targets (1)(2)(3)(4)37).In this work, we leveraged large-scale GWASs of circulating metabolite abundances and psychiatric disorders to systematically identify metabolite-disease associations through MR analyses.These associations may implicate biological mechanisms that underlie disease pathogenesis.
Consistent with previous epidemiological studies (38)(39)(40), our MR analyses consistently demonstrated that an increased abundance of circulating U-3 unsaturated fatty acids, a higher U-3-to-total fatty acids ratio, and a lower U-6-to-U-3 ratio were associated with a reduced risk of bipolar disorder.The observed negative associations may be attributed to the potential neurotransmission regulatory effects (41), antioxidant properties against oxidative stress (42), and anti-inflammatory effects of U-3 unsaturated fatty acids (43).Importantly, randomized controlled trials have shown that U-3 unsaturated fatty acid supplementation can be well tolerated and may contribute to mood stabilization in individuals with bipolar disorder (41,44).Our results provide additional evidence supporting a potential role of U-3 unsaturated fatty acids in bipolar disorder, thereby warranting further investigations to understand the relevant biological pathways and explore possible therapeutic strategies.
Our analyses suggested that circulating abundances of 2 steroid hormones, androsterone sulfate and epiandrosterone sulfate, were associated with the risk of bipolar disorder.Importantly, the underlying effector gene, CYP3A7, has crucial roles in the metabolism of various endogenous and exogenous compounds, such as drugs, xenobiotics, and other steroids (35).Given the risk of horizontal pleiotropy and the weak genetic associations with bipolar disorder detected in the CYP3A7 locus, the associations between these steroid hormones and bipolar disorder should be interpreted with caution.
Associations between increased circulating abundances of 3 N-acetyl-amino acids and an increased risk of schizophrenia implicated the effector gene NAT8 (36), whose biological functions have not been fully characterized.However, the 3 Nacetyl-amino acids tested in MR do not have known functions as neurotransmitters or signaling molecules.Given the marginal associations between their genetic instruments and potential confounders, even if these metabolites have potential as predictive or diagnostic biomarkers, they may not directly feature causal biological pathways.The associations between the genetic instruments and biomarkers of kidney function, such as serum creatinine level (45), serum cystatin C level (45), and estimated glomerular filtration rate, (46) may suggest connections through renal comorbidities.Nonetheless, the precise mechanisms by which the genetic variants may influence the risk of schizophrenia remain unclear and may be explored with improved characterization of the enzymatic activity of NAT8.For example, the schizophrenia riskincreasing allele of the genetic instrument, while it increased the circulating abundance of N-alpha-acetylornithine in our study, has also been associated with decreased abundance of N-delta-acetylornithine in cerebrospinal fluid (47), which suggests the possibility of contrasting effects of these isomers.
Moreover, we found that an increased circulating abundance of hypotaurine was associated with a decreased risk of depression.As an important intermediate in the biosynthesis of taurine, hypotaurine can act on glycine receptors as an endogenous neurotransmitter and possesses antioxidant properties as an osmolyte (48), although additional studies are needed to fully understand its neurochemical effects.On the other hand, taurine, the product of hypotaurine dehydrogenation, has demonstrated various neuroprotective effects with characterized roles in modulation of neurogenesis, neuroinflammation, endoplasmic reticulum stress, apoptosis, energy metabolism, calcium homeostasis, and osmosis (49,50).However, the potential effect of circulating abundance of taurine could not be assessed in the current MR analyses because no genetic instrument could be identified.Integrative investigations into the taurine biosynthesis pathway may offer additional opportunities to identify biomarkers and therapeutic targets.
Our study has several strengths.The use of MR analyses reduced the risk of confounding and reverse causation that may bias observational studies.Notably, it has been recognized that population-based GWASs may be biased by confounding factors such as uncontrolled population stratification, assortative mating, and indirect genetic effects via relative phenotypes and shared environment (51).However, it has been shown that genetic associations with molecular traits are unlikely to be subject to such biases (51).Our analyses using negative control outcomes further mitigated the risk of unobserved confounding, with the exception that the genetic instruments for the 3 schizophrenia-associated N-acetyl-amino acids demonstrated marginal associations with potential confounders.While MR requires the assumption of no horizontal pleiotropy, the identification of effector genes established explicit and direct biological connections between genetic instruments and metabolite abundances, which effectively mitigated bias due to horizontal pleiotropy and provided new insights into the metabolite abundance-regulating enzymatic activities.Furthermore, our findings were supported by various sensitivity analyses and colocalization analyses.All metabolite-bipolar disorder associations were consistently detected using mQTL data from the INTERVAL and EPIC-Norfolk studies, although no genetic instrument was available in these cohorts for metabolites that were associated with the risk of schizophrenia or depression.
Our work has some important limitations.First, our MR analyses were restricted to metabolites that had been measured and had genome-wide significant loci and effector genes identified.Although this strict inclusion criterion could ensure a low false-positive rate, we acknowledge that a large number of causal biological pathways likely remain undiscovered.Novel findings may be rendered by improved profiling of the circulating metabolome in the future as well as identification of core metabolites in biologically relevant cascades (34).Second, both mQTL studies and GWASs of Circulating Metabolites and 3 Psychiatric Disorders psychiatric disorders utilized in this study were based on populations predominantly of European ancestry.Although genetic associations with molecular traits that are linked to explicit biological processes may be consistent across populations of different genetic ancestries, the genetic architecture of most psychiatric disorders has not been thoroughly investigated in non-European ancestry populations, and our results have not been replicated in independent studies of European ancestry populations.Meanwhile, there was sample overlap between the psychiatric disorder GWASs used in this study.While such sample overlap does not bias the MR estimates of the metabolite-disease associations, it may further limit the generalizability of our findings.These challenges underscore the necessity to aggregate large-scale genetic epidemiological resources for diverse populations.Last, it is important to note that our findings have not been validated by functional experiments and do not equate to randomized controlled trials in assessing the effectiveness of interventions.In particular, because the genetic instruments are not subject to modification throughout a lifetime and have modest effects on the circulating metabolite abundances, MR estimates reflect the potential effects of long-term exposure to risk factors in the general population.Additionally, as with any other epidemiological study design, it is unlikely that MR can fully eliminate confounding effects.Therefore, quantifying the direct clinical impact of the candidate circulating metabolites on the disease outcomes based on the magnitudes of MR estimates is difficult.Exploration of these circulating metabolites, especially as therapeutic targets, requires consideration of multiple lines of evidence.Nonetheless, we anticipate that future investigations employing advanced multiomics and epidemiological approaches will provide a more comprehensive understanding of the roles of these circulating metabolites and the underlying enzymatic activities in psychiatric disorders and other complex diseases.

Conclusions
Through mQTL-facilitated MR, we validated the negative associations between increased abundances of U-3 unsaturated fatty acids and the risk of bipolar disorder.We identified NAT8catalyzed N-acetyl-amino acids as candidate biomarkers for schizophrenia and hypotaurine as a candidate biomarker and possible therapeutic target for depression.These findings encourage further explorations of these metabolites in the context of psychiatric disorders.

Figure 1 .
Figure 1.Overview of study design.Mendelian randomization analyses were conducted based on genetic instruments affecting effector genes.Validity of prioritized circulating metabolites was evaluated through multiple sensitivity analyses assessing assumptions of Mendelian randomization, colocalization analyses, and repeating Mendelian randomization analyses based on external resources.Biological relevance between each pair of metabolite and gene based on public databases is available in TableS1in Supplement 2; mQTL-cis-eQTL colocalization results derived using coloc are available in TableS2in Supplement 2; mQTL-cis-sQTL colocalization results derived using coloc are available in TableS3in Supplement 2. CLSA, Canadian Longitudinal Study on Aging; EPIC, European Prospective Investigation into Cancer; eQTL, expression quantitative trait locus; FDR, false discovery rate; IVW, increase variance-weighted random effects estimate; GTEx, Genotype-Tissue Expression; GWAS, genome-wide association study; HMDB, Human Metabolome Database; KEGG, Kyoto Encyclopedia of Genes and Genomes; mQTL, metabolite QTL; sQTL, splicing QTL.

Figure 3 .
Figure 3. Circulating metabolite abundances associated with risk of bipolar disorder.(A) Colocalization of bipolar disorder GWAS signal and genetic associations with 5 unsaturated fatty acids.Relationships between the illustrated fatty acids are shown in Figure S2 in Supplement 1. (B) Colocalization of bipolar disorder GWAS signal and genetic associations with 2 steroid hormones.Genetic instruments are indicated.Genetic variants located in a 6500-kb window centered around each genetic instrument are colored by the magnitude of correlation (linkage disequilibrium r 2 ) with the corresponding instrument.The significance of genetic associations with different circulating metabolite abundances does not necessarily indicate the direction of the biological cascades.(C) MR results based on genetic instruments identified in the INTERVAL and EPIC-Norfolk studies.(D) MR scatterplots demonstrating estimated effects of U-3 unsaturated fatty acids, U-6 unsaturated fatty acids, total fatty acids, and their relative abundances on bipolar disorder.chr, chromosome; CLSA, Canadian Longitudinal Study on Aging; EPIC, European Prospective Investigation into Cancer; GWAS, genome-wide association study; MR, Mendelian randomization; SNP, single nucleotide polymorphism.

Figure 4 .
Figure 4. Circulating metabolite abundances associated with risk of schizophrenia.(A) Colocalization of schizophrenia GWAS signal and genetic associations with 3 N-acetyl-amino acids.Genetic instruments are indicated.Genetic variants located in a 6500-kb window centered around each genetic instrument are colored by the magnitude of correlation (linkage disequilibrium r 2 ) with the corresponding instrument.(B) Illustration of NAT8-catalyzed reaction.(C) Linkage disequilibrium between 3 genetic instruments based on individuals of European ancestry in the Canadian Longitudinal Study on Aging cohort.Linkage disequilibrium r 2 values are indicated.(D) Impact of missense variant rs13538 (Phe143Ser) on protein structure of NAT8.Protein structures were computationally predicted using default settings of AlphaFold2 (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb).The asterisk (*) denotes a putative identification of the metabolite.chr, chromosome; GWAS, genome-wide association study; SNP, single nucleotide polymorphism.

Figure 5 .
Figure 5. Circulating hypotaurine abundance associated with risk of depression.(A) Colocalization of depression GWAS signal and genetic associations with hypotaurine.Genetic instruments are indicated.Genetic variants located in a 6500-kb window centered around each genetic instrument are colored by the magnitude of correlation (linkage disequilibrium, r 2 ) with the corresponding instrument.(B) Illustration of biosynthesis of taurine.Cysteine dioxygenase, coded by the effector gene CDO1, and hypotaurine, the target metabolite, are highlighted.chr, chromosome; GWAS, genome-wide association study; SNP, single nucleotide polymorphism.