Causality of genetically determined metabolites on anxiety disorders: a two-sample Mendelian randomization study
Journal of Translational Medicine volume 20, Article number: 475 (2022)
Although anxiety disorders are one of the most prevalent mental disorders, their underlying biological mechanisms have not yet been fully elucidated. In recent years, genetically determined metabolites (GDMs) have been used to reveal the biological mechanisms of mental disorders. However, this strategy has not been applied to anxiety disorders. Herein, we explored the causality of GDMs on anxiety disorders through Mendelian randomization study, with the overarching goal of unraveling the biological mechanisms.
A two-sample Mendelian randomization (MR) analysis was implemented to assess the causality of GDMs on anxiety disorders. A genome-wide association study (GWAS) of 486 metabolites was used as the exposure, whereas four different GWAS datasets of anxiety disorders were the outcomes. Notably, all datasets were acquired from publicly available databases. A genetic instrumental variable (IV) was used to explore the causality between the metabolite and anxiety disorders for each metabolite. The MR Steiger filtering method was implemented to examine the causality between metabolites and anxiety disorders. The standard inverse variance weighted (IVW) method was first used for the causality analysis, followed by three additional MR methods (the MR-Egger, weighted median, and MR-PRESSO (pleiotropy residual sum and outlier) methods) for sensitivity analyses in MR analysis. MR-Egger intercept, and Cochran’s Q statistical analysis were used to evaluate possible heterogeneity and pleiotropy. Bonferroni correction was used to determine the causative association features (P < 1.03 × 10–4). Furthermore, metabolic pathways analysis was performed using the web-based MetaboAnalyst 5.0 software. All statistical analysis were performed in R software. The STROBE-MR checklist for the reporting of MR studies was used in this study.
In MR analysis, 85 significant causative relationship GDMs were identified. Among them, 11 metabolites were overlapped in the four different datasets of anxiety disorders. Bonferroni correction showing1-linoleoylglycerophosphoethanolamine (ORfixed-effect IVW = 1.04; 95% CI 1.021–1.06; Pfixed-effect IVW = 4.3 × 10–5) was the most reliable causal metabolite. Our results were robust even without a single SNP because of a “leave-one-out” analysis. The MR-Egger intercept test indicated that genetic pleiotropy had no effect on the results (intercept = − 0.0013, SE = 0.0006, P = 0.06). No heterogeneity was detected by Cochran’s Q test (MR-Egger. Q = 7.68, P = 0.742; IVW. Q = 12.12, P = 0.436). A directionality test conducted by MR Steiger confirmed our estimation of potential causal direction (P < 0.001). In addition, two significant pathways, the “primary bile acid biosynthesis” pathway (P = 0.008) and the “valine, leucine, and isoleucine biosynthesis” pathway (P = 0.03), were identified through metabolic pathway analysis.
This study provides new insights into the causal effects of GDMs on anxiety disorders by integrating genomics and metabolomics. The metabolites that drive anxiety disorders may be suited to serve as biomarkers and also will help to unravel the biological mechanisms of anxiety disorders.
Anxiety disorders are a significant health problem widespread and are the leading psychiatric causes of the global burden of diseases . The World Health Organization (WHO) has also ranked anxiety disorders as one of the largest causes of disability worldwide largely due to their high prevalence, chronicity, and comorbidity [2, 3]. Effective prevention and treatment of anxiety disorders are critical to reduce the morbidity and disability. Notably, the exploration of the biological mechanism is the basis for the prevention and therapy of anxiety disorders . Multiple factors, such as psychological and genetic factors, are thought to be involved in the biological mechanisms of anxiety disorders [4, 5]. However, anxiety disorders are complex conditions, and their biological mechanisms are not fully understood. Although progress in genetics (particularly genome-wide association studies (GWASs)) have largely improved the development of etiology research for mental disorders [6,7,8], there is still a great barrier translating these genetic findings into biological mechanisms.
In recent years, modern omics-based technologies, including metabolomics, have made a positive contribution to the exploration of disease mechanisms. Specifically, metabolomics can provide novel information into the biological mechanisms of diseases by revealing the intermediate metabolites and altered metabolic pathways [9, 10]. A recent robust study of GWAS of metabolites has identified the disease-relevant loci and suggest mechanisms for diseases and disease-related traits . Several studies have suggested that metabolites are functional intermediates that can be used to illustrate the potential biological mechanisms related to the genetics of mental disorders [12,13,14]. It is worth noting that metabolites are the final products or the intermediate of metabolism that can play important in human. The database of genotype dependent metabolic phenotypes (also known as genetically determined metabolites (GDMs)) has recently been established using a GWAS involving non-targeted metabolomics [15, 16]. The developed GDMs can promote insight of the underlying relationship of human serum metabolites and associative genetic variants in the biological mechanisms of mental disorders by providing functional intermediates [17,18,19]. Studies have shown the significance of GDMs in the biological mechanism of major depression, bipolar disorders, autism spectrum disorders and hyperactivity disorders . However, GDMs and pathway analysis geared toward exploring the biological mechanisms of anxiety disorders are still lacking, which calls for a deep analysis to determine the role played by the effects between genetic variation and metabolites in the biological mechanisms of anxiety disorders.
Mendelian randomization (MR) analysis is a useful epidemiological research strategy in which genetic variants are used to connect exposure with outcome as instrumental variables (IV) for assessing causal relationships. Compared to other epidemiological research strategies, MR can provide unbiased estimates on how genotypes are decided at conception, and are commonly not susceptible to confounding factors and reverse causation . Given this huge advantage, MR has been widely applied in the past decade to infer causality of related risk exposure to disease using publicly available GWAS summary statistics [21, 22]. Recently, GWAS have extended the metabolic spectrum, from which an atlas of GDMs was developed.
Herein, we speculated that this GDMs atlas could be used to infer the causality of GDMs on anxiety disorders. Consequently, we implemented a two-sample MR approach to: (1) assess the causal effects of human serum metabolites on anxiety disorders; (2) identify the GDMs that have causal effects on four different GWASs of anxiety disorders; and (3) identify potential metabolic pathways which might help to understand the mechanism of anxiety disorders.
Materials and methods
MR design and data source
The general design of this MR research was stated in Additional file 1: Fig. S1. The study methods were compliant with the STROBE-MR checklist . Genetic association data for serum metabolites were obtained from the metabolomics GWAS server (http://metabolomics.helmholtz-muenchen.de/gwas/). Notably, Shin et al.  reported the most comprehensive exploration of genetic influences on human metabolism so far, by performing a GWAS of non-targeted metabolomics, which successfully screened out 486 metabolites with genetic influences on human serum metabolites. In more detail, a total of 7824 participants were recruited from two European population cohorts, including 1768 participants from the KORA F4 study in Germany and 6056 from the UK Twin Study. Both studies were approved by local ethics committees, and all participants voluntarily signed informed consent before the study. Fasting serum was analyzed using non-targeted mass spectrometry analysis. For metabolic analyses, standardized processes of identification and relative qualification were conducted using Metabolon, Inc. (https://www.metabolon.com/). After strict quality control, 486 metabolites in total were analyzed, including 309 known metabolites and 177 unknown metabolites. Moreover, the 309 known metabolites were further classified into eight biochemical classes (peptides, nucleotides, amino acids, energy, cofactors and vitamins, lipids, carbohydrates, and xenobiotics) in conformity to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Notably, genotyping information for the two cohorts are described detailly in previous research [24, 25]. Finally, there were approximately 2.1 million single nucleotide polymorphisms (SNPs) in the GWAS meta-analysis.
Selection of instrumental variables (IVs)
In principle, the instrumental variables used in MR analysis must satisfy the following three assumptions: (1) IVs must be relevant to exposure (i.e.,: metabolites); (2) IVs must be associated with outcome (i.e., anxiety disorders) only via exposure (i.e., metabolites); and (3) IVs must be independent of any confounder . To determine the IVs for the 486 metabolites, some procedures were done to ensure that the first assumption was true. First, the genetic variants were extracted with association thresholds at P < 1 × 10–5, which are mostly used in MR analysis to elucidate a greater variation when few SNPs are available for exposure. Second, independent variants were identified using a clumping procedure implemented in R software, in which a linkage-disequilibrium threshold of r2 < 0.5 within a 5000 kb window in the European 1000 Genomes Project Phase 3 reference panel was set. Instrumental SNPs were selected by removing palindromic SNPs with middle allele frequency (MAF). Palindromic SNPs are those with the A/T or G/C allele, whereas the MAF is from 0.01 to 0.30. SNPs with the incorrect causal direction are excluded by MR Steiger filters. SNPs with an MAF less than 0.01 were also excluded from the original GWAS because of their low confidence level. Next, we used the explained variance (R2) and F statistic parameters to determine whether the identified IVs were powerful enough to represent metabolite levels. The above design formula is presented in Additional file 2: Table S1. Typically, a threshold of F > 10 is suggested for MR analysis. The mRnd was used to calculate the statistical power for Mendelian randomization (https://cnsgenomics.shinyapps.io/mRnd/).
GWAS of anxiety disorders
This study aimed at evaluating the potential causal relationship between metabolites and anxiety disorders with a wide range of impact in both males and females from MRC-IEU. Anxiety disorders are typically diagnosed on the basis of structured clinical interviews for the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) and the tenth edition of the International Classification of Diseases (ICD-10). Anxiety disorders also can be screened with self-report questionnaires. Moreover, providing care for mental health problems including anxiety disorders concerns general practitioners, and psychiatrists [27, 28]. Therefore, given the completeness of the data and diagnosis of anxiety disorders, we analyzed the data obtained from different sources, including patients diagnosed by psychiatrists for anxiety (53,414 cases and 407,288 controls); unspecified anxiety disorders diagnosed by secondary ICD-10 (1523 cases and 461,487 controls); patients diagnosed by general practitioners for anxiety (158,565 cases and 300,995 controls); and self-reported anxiety (6410 cases and 462,933 controls). Additional file 2: Table S2 shows the detailed information. Notably, all IVs were extracted from MR-Base (http://app.mrbase.org/) database (20).
The IVW method (when there is heterogeneity, a random effect model of IVW is used, and if there is no heterogeneity, a fixed effect model of IVW is used) was used to assess causal effects for two-sample MR analyses. Notably, the IVW method can provide a consistent assessment of the causality of the exposure when each variant satisfies all three assumptions of valid instrumental variables. An estimate of IVW can be obtained by calculating the slope of the weighted linear regression . Next, sensitivity analysis was performed using MR-Egger approach, which can provide consistent estimates with invalid instruments . The MR-Egger can discover the violations of the IVs assumption and provide estimates of effects unaffected by these violations. Moreover, the weighted median method also provided consistent estimates, with up to 50% of the variants being thought noneffective instruments. MR-PRESSO is another novel MR method that can check and rectify horizontal pleiotropic outliers, thereby providing a right estimate . To assess the possibility of horizontal pleiotropy and bias caused by ineffective IVs, MR-Egger intercepts were also calculated . Moreover, this study employed the MR-PRESSO global test to assess the existence of horizontal pleiotropy. We calculated the odds ratios (OR) to measure causal effects, as well as Cochran’s Q statistics, to estimate heterogeneity among SNPs . A “leave-one-out” sensitivity analysis was implemented to determine whether results were affected by a single SNP . Additionally, we performed the MR Steiger directionality test to ascertain whether our results supported our hypothesis. Notably, all analyses were conducted using Two Sample MR 0.5.6 and MR-PRESSO packages in R software (version 3.6.0). P < 0.05 was considered statistically significant. We adopted a multiple-testing-adjusted threshold of P < 1.03 × 10−4 (0.05/486) using the Bonferroni correction to declare a statistically significant, causal relationship . We also reported metabolites that had a P < 0.05, but were above the Bonferroni-corrected threshold, as suggestive risk predictors for anxiety disorders.
Metabolic pathway analysis
Metabolic pathway analysis was performed using the web-based Metaconflict 5.0. (https://www.metaboanalyst.ca/) . Functional enrichment analyses and the pathway analyses module were used to identify underlying metabolite groups or pathways which may be relevant to the biological process of anxiety disorders. In all, 25 serum metabolic pathways were screened out from two metabolite databases, including 24 from both the Small Molecule Pathway Database (SMPD)  and the KEGG database, and one from SMPD alone. Notably, this study only analyzed the metabolites that passed the advised threshold of association by IVW (P IVW < 0.05).
Strength of the instrumental variables
We performed a two-sample MR analysis to evaluate the causality of GDMs on anxiety disorders using four different pairs of GWAS summary data. The generated IVs of 486 metabolites are ranging from 3 to 56 SNPs (glycodeoxycholate generated the least IVs: 3 SNPs; and p-acetamidophenylglucuronide generated the most IVs: 56 SNPs); These generated IVs could explain 0.011–0.225% of the variance of their corresponding metabolites (Additional file 2: Tables S3–S11). In addition, the minimum F statistic of these IVs was 10.23, suggesting that all IVs were sufficiently effective for the MR analysis of the 486 metabolites (F statistic > 10).
Causality of genetically determined metabolites on anxiety disorders
The IVW method was employed to confirm the causality among the 486 metabolites and anxiety disorders using four pairs of GWAS summary data. 103 remarkable causative association features (matching with 85 unique metabolites) were conformed at PIVW < 0.05 in total, including 57 known metabolites and 28 unknown metabolites. Additional file 2: Table S12 shows the known metabolites that were significantly associated with the GWAS datasets of anxiety disorders. Additional file 2: Table S13 shows the characteristics of SNPs and their genetic associations with known metabolites and the anxiety disorders. And Fig. 1 shows the known metabolites significantly associated with anxiety disorders (as well as the subgroup analysis of metabolites significantly associated with the four different GWAS datasets of anxiety disorders). Table 1 shows that there were 11 overlapped metabolites in the four different GWAS datasets of anxiety disorders (there were no metabolites that overlapped with the other three anxiety disorders GWAS in the unspecified anxiety disorders diagnosed by secondary ICD-10). It is worth noting that there may be some common molecular mechanisms between four different GWAS datasets of anxiety disorders. Next, we performed the Bonferroni correction to determine the causative association features (P < 1.03 × 10–4). Results found one causal effect feature of metabolites associated with anxiety disorders diagnosed by psychiatrists, namely 1-linoleoylglycerophosphoethanolamine (P fixed-effect IVW = 4.31 × 10–5) (Additional file 2: Table S14). The statistical power of 1-linoleoylglycerophosphoethanolamine for anxiety diagnosed by psychiatrists was 47%. In particular, fixed-effect IVW estimates demonstrated that 1-linoleoylglycerophosphoethanolamine increased the risk of anxiety disorders (OR 1.04; 95% CI 1.021–1.06; P fixed-effect IVW = 4.3 × 10–5) (Table 1). Moreover, as for the additional methods, weighted median analysis (OR 1.049, 95% CI 1.022–1.077, P = 0.0003), and MR-Egger analysis (OR 1.092, 95% CI 1.040–1.147, P = 0.005) indicated consistent results (Additional file 2: Table S7). As indicated by the MR Steiger directionality test results, our estimate of causal direction was accurate (P < 0.001). In addition, we performed a reverse MR analysis, which found no causal relationship between anxiety disorders (exposure) and 1-linoleoylglycerophosphoethanolamine (outcome) (Additional file 2: Table S15); Finally, to verify our results, we performed MR analysis with two additional GWAS data of anxiety disorders (finn-b-KRA_PSY_ANXIETY: 20,992 cases and 197,800 controls; finn-b-KRA_PSY_ANXIETY_EXMORE: 20,992 cases and 166,584 controls), and also found that the 1-linoleoylglycerophosphoethanolamine might have causality with anxiety disorders (Additional file 2: Table S16). Consequently, we discovered that the 1-linoleoylglycerophosphoethanolamine might be causally associated with anxiety disorders, and the result is reliable.
While the IVW method is extremely effective for inferring causality between an exposure and a disease outcome, it is known to be susceptible to weak instrument bias. Therefore, we further conducted sensitivity and pleiotropy analysis to assess the robustness of the causality. The sensitivity analyses result for 1-linoleoylglycerophosphoethanolamine on the anxiety disorders diagnosed by psychiatrists are shown in Fig. 2. Notably, the causal relationship was reliable if three extra MR tests were performed, with results showing that there was no proof of horizontal pleiotropy for the 1-linoleoylglycerophosphoethanolamine presented in Additional file 1: Figs. S2, S3. And based on the results of “leave-one-out” method, MR analysis was responsible, and single SNPs did not affect the results (Additional file 1: Fig. S4). Moreover, for the IVW method, Cochran's Q statistic was 12.12 (P = 0.436). The Cochran's Q statistic was 7.68 (P = 0.742) for the MR-Egger method. The results of Cochran's Q test suggested little heterogeneity. MR-Egger regression was also conducted to examine the horizontal pleiotropy between IVs and results, but no remarkable intercept was discovered (intercept = − 0.0013, SE = 0.0006, P = 0.06). Furthermore, MR-PRESSO results indicated no horizontal pleiotropy in the MR study (P = 0.42). A funnel plot (Additional file 1: Fig. S5) displays neither horizontal pleiotropy nor heterogeneity in our MR study. All the results shown that the causal effect of 1-linoleoylglycerophosphoethanolamine on the anxiety disorders diagnosed by psychiatrists appears to be reliable. Furthermore, Table 2 shows that five association metabolites passed all the sensitivity analyses (P < 0.05) without horizontal pleiotropy in the 11 overlapped metabolites. MR-Egger regression was also conducted to examine the horizontal pleiotropy of the five association metabolites, no remarkable intercept was discovered. And the results of Cochran's Q test suggested little heterogeneity (Table 2). The genetic variants that explain the relationship between the five metabolites and anxiety disorders are presented in Additional file 2: Tables S17–S21 and Additional file 1: Figs. S6–S9, separately. Moreover, we performed the reverse MR analysis, which found no causal relationship between anxiety disorders (exposure) and the five association metabolites (outcome) (Additional file 2: Table S15). The P-value distribution of the Pfixed-effect IVW < 0.05 metabolites on anxiety disorders are presented in Additional file 1: Fig. S10.
Metabolic pathway analyses
Two important metabolic pathways which mainly involved in the anxiety disorders were identified in the metabolic pathways analyses (Additional file 2: Table S22). Results showed that the “primary bile acid biosynthesis” pathway might be relevant to the development of self-reported anxiety disorders (P = 0.008), whereas “valine, leucine, and isoleucine biosynthesis” pathway was found to be associated with anxiety disorders diagnosed by general practitioners (P = 0.03).
This Mendelian randomization study provides unbiased evaluation of the causal relationship between GDMs and anxiety disorders using four different GWAS datasets, including anxiety disorders diagnosed by psychiatrists, unspecified anxiety disorders diagnosed by secondary ICD-10, anxiety diagnosed by general practitioners, and self-reported anxiety disorders. We identified 85 metabolites relevant to the risk of anxiety disorders after using genetic variants as probes. Among them, 11 metabolites were overlapped in the four different datasets of anxiety disorders, with Bonferroni correction showing that 1-linoleoylglycerophosphoethanolamine had the most reliable causal relationship. Moreover, pathway enrichment analysis identified two significant metabolic pathways, the “primary bile acid biosynthesis” pathway and the “valine, leucine, and isoleucine biosynthesis” pathway, which are mainly involved in the anxiety disorders.
As far as we know, this is the first MR study that has combined genomics with metabolomics to evaluate the causality of GDMs on anxiety disorders. Herein, results identified a cluster of metabolites in serum showing relationship with anxiety disorders, among which 1-linoleoylglycerophosphoethanolamine had a robust effect on anxiety disorders diagnosed by psychiatrists and anxiety disorders diagnosed by general practitioners. In a previous study, researchers conducted a delivering preterm with or without preeclampsia population-based birth cohort, and found that higher 1-linoleoylglycerophosphoethanolamine density decreased odds of preeclampsia . The result is corresponding with previous research which revealed that women who developed preeclampsia had lower levels of lysophospholipids compared to healthy term pregnancies [37,38,39]. In addition, preeclampsia women were more likely to be diagnosed with anxiety disorders. 1-linoleoylglycerophosphoethanolamine is an important member of the phosphatidylethanolamine (PE) family, whose components mainly include fatty acids, ethanolamine, phosphoric acid, and glycerol . As a lipid chaperone, PE can assist in the folding of certain membrane proteins, and is closely associated with anxiety disorders. A previous study found that stress-prone Wistar-Kyoto rats had lower PE in the anxiety state compared to the non-stress prone rats, suggesting that PE may be associated with the anxiety disorders . Studies have also found that alcohol-dependent patients are often accompanied by anxiety disorders when they quit drinking, and the concentration of PE in plasma is higher after alcohol-dependent patients quit drinking . Notably, Yang et al.  explored the relationship between metabolism and some psychiatric disorders, and found that the genetic associations of 1-linoleoylglycerophosphoethanolamine were involved in the risk of major depression. Clinically, the comorbidity of depression and anxiety is common, and the two often influence each other. Although 1-linoleoylglycerophosphoethanolamine seems like a hopeful biomarker for anxiety disorders, further studies should be conducted to clarify the relevant mechanism.
This MR analysis also identified certain metabolites, some of which had been reported in previous research. Androsterone sulfate whose chemical formula is 3α-hydroxy-5α-androstan-17-one 3α-sulfate, is one of the primary urinary metabolites of androgens. Previous studies reported that metabolite networks enriched in androsterone sulfate, tyrosine, indoxyl sulfate, or caffeine are associated with a negative personality . Negative personality manifested as inhibition in social situations is often described as distress or anxiety. Moreover, sleep-related impairment was inversely correlated with androgen deprivation therapy-induced reduction in androsterone sulfate . Given that people with sleep disorders are often accompanied by anxiety disorders, androsterone sulfate might play a role in neurodevelopment of anxiety disorders. Epiandrosterone sulfate is a vital endogenous androstane steroid produced by the adrenal cortex. Notably, the steroid is an important neurosteroid and neurotrophin, which plays an important physiological function in human body. A recent omics investigation into chronic widespread musculoskeletal pain revealed that epiandrosterone sulfate is an important biomarker . Considering that people with chronic pain are often accompanied by anxiety disorders, epiandrosterone sulfate might play a role in neurodevelopment of anxiety disorders. In addition, p-acetamidophenylglucuronide showed robust association with intelligence, and thus it can be used to predict important health outcomes that may affect anxiety disorders . It is worth mentioning that our results are consistent with above results and emphasize the significance of genetics in the progression of mental disorders. It is worth noting that the use of drugs also has an effect on metabolite profiles. The baseline characteristics and the drug use of the participants involving in the study are listed in Additional file 2: Table S23–S25. The effect of drugs (such as antihypertensive and lipid lowering drugs) on the human metabolite concentrations is still not fully understood, and further studies are needed.
In this study, the metabolic pathway analysis showed that “primary bile acid biosynthesis”, and “valine, leucine, and isoleucine biosynthesis” pathways are mainly associated with anxiety disorders. Notably, primary bile acids are synthesized in liver cells by cytochrome P450-mediated oxidation of cholesterol, resulting in the synthesis of primary bile acids (such as deoxycholic acids). Primary bile acids have rich functions and are important physiological factors for intestinal nutrient absorption, bile secretion of lipids, toxic metabolites and exotic organisms. At the same time, primary bile acids are signaling molecules and metabolic regulators, which play an important role in activating nuclear receptor and G-protein-coupled receptor (GPCR) signals, regulating liver lipid, glucose and energy homeostasis and maintaining metabolic homeostasis. Studies have proved that primary bile acid metabolites/pathways are involved in several aspects of brain function and behavior . Moreover, it has been reported that changes in the gut microbiota composition may be associated with alterations in the primary bile acid metabolism that are involved in the biological process of psychological disorders in Crohn’s disease . A traditional Chinese medicine cohort revealed that amino acid metabolism, such as cysteine and methionine metabolism, might be involved in brain health disorders characterized by alterations in evaluation of bile acid biosynthesis . Altogether, these findings suggest that biosynthesis of primary bile acids might play an important role in the biological mechanism of anxiety disorders.
Valine, leucine, and isoleucine are structurally associated with branched-chain fatty acids and are important members of the family of 9 essential amino acids. Metabolism of valine, leucine and isoleucine have been proved to be associated with cancer progression in numerous studies, and key proteins in metabolic pathways may act as potentially prognostic and diagnostic biomarkers in human cancers . Given that people with cancers are often accompanied by anxiety disorders, valine, leucine, and isoleucine might play a role in neurodevelopment of anxiety disorders. Previous studies have proved that plasma concentrations of valine, leucine, and isoleucine are increased significantly in conditions associated with insulin resistance, such as obesity and diabetes mellitus . Research has demonstrated that anxiety symptoms are prevalent in people with diabetes, and may affect diabetes management and glycemic control [51, 52]. Chen et al.  established a rat model to explore the potential mechanisms of antidepressant effects, with the mainly enriched pathways being valine, leucine, and isoleucine degradation. It should be noted that this model is built on liver tissue, suggesting that the liver may potentially be connected with mental disease through some metabolites. To sum up, the biosynthesis of valine, leucine, and isoleucine might be relevant to the biological mechanism of anxiety disorders.
However, this study had several limitations. First, given the classification of the original data, we could not further subdivide the pressure type of anxiety disorders in combination with the ICD classification standard, and thus we could only analyze the anxiety disorders as a whole. Second, the power of the IVs depends largely on the sample size of GWASs, therefore, more data are needed to improve the accuracy of the generated GDMs. Third, although Mendelian randomization has been shown to be a powerful method to assess the causality between human blood metabolites and anxiety disorders, the results should be verified by further studies based on experimental data. Fourth, the veracity of the MR analysis relies largely on the explanation of the instrumental variables on exposure. Therefore, it is necessary to expand the sample size to provide a more accurate assessment of the genetic impact on metabolites. Fifth, due to insufficient data, we used metabolites with uncorrected P values for pathway analysis. At last, although this study identified multiple metabolites that contribute to the risk of anxiety disorders, further studies are needed to reveal their roles in the pathogenic mechanism of anxiety disorders.
In conclusion, this MR research totally identified 85 metabolites that may have causality on the pathogenesis of anxiety disorders, including 11 known metabolites having causality on more than one type of anxiety disorders. And 1-linoleoylglycerophosphoethanolamine had a robust effect on anxiety disorders among them. Moreover, this study identified two important metabolic pathways that may be relevant to the pathology of anxiety disorders. Collectively, our findings will provide valuable insights on using some metabolites as potential biomarkers to explore the targeted drugs for treating human diseases, but more studies are required to validate the results.
Genetically determined metabolites
Genome wide association study
MR pleiotropy residual sum and outlier
Kyoto encyclopedia of genes and genomes
Single nucleotide polymorphisms
Small molecule pathway database
World health organization
The fifth edition of the diagnostic and statistical manual of mental disorders
The tenth edition of the international classification of diseases
Penninx B, Pine DS, Holmes EA, Reif A. Benzodiazepines for the long-term treatment of anxiety disorders?—authors’ reply. Lancet. 2021;398(10295):120.
Diseases GBD, Injuries C. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2020;396(10258):1204–22.
Thibaut F. Anxiety disorders: a review of current literature. Dialogues Clin Neurosci. 2017;19(2):87–8.
Foa EB, McLean CP. The efficacy of exposure therapy for anxiety-related disorders and its underlying mechanisms: the case of OCD and PTSD. Annu Rev Clin Psychol. 2016;12:1–28.
Bandelow B, Michaelis S. Epidemiology of anxiety disorders in the 21st century. Dialogues Clin Neurosci. 2015;17(3):327–35.
Levey DF, Gelernter J, Polimanti R, Zhou H, Cheng Z, Aslan M, et al. Reproducible genetic risk loci for anxiety: results from approximately 200,000 participants in the Million Veteran Program. Am J Psychiatry. 2020;177(3):223–32.
Smoller JW, Block SR, Young MM. Genetics of anxiety disorders: the complex road from DSM to DNA. Depress Anxiety. 2009;26(11):965–75.
Gatt JM, Burton KL, Williams LM, Schofield PR. Specific and common genes implicated across major mental disorders: a review of meta-analysis studies. J Psychiatr Res. 2015;60:1–13.
Johnson CH, Ivanisevic J, Siuzdak G. Metabolomics: beyond biomarkers and towards mechanisms. Nat Rev Mol Cell Biol. 2016;17(7):451–9.
Lains I, Gantner M, Murinello S, Lasky-Su JA, Miller JW, Friedlander M, et al. Metabolomics in the study of retinal health and disease. Prog Retin Eye Res. 2019;69:57–79.
Yin X, Chan LS, Bose D, Jackson AU, VandeHaar P, Locke AE, et al. Genome-wide association studies of metabolites in Finnish men identify disease-relevant loci. Nat Commun. 2022;13(1):1644.
Humer E, Pieh C, Brandmayr G. Metabolomics in sleep, insomnia and sleep apnea. Int J Mol Sci. 2020. https://doi.org/10.3390/ijms21197244.
Quintero M, Stanisic D, Cruz G, Pontes JGM, Costa T, Tasic L. Metabolomic biomarkers in mental disorders: bipolar disorder and schizophrenia. Adv Exp Med Biol. 2019;1118:271–93.
Yuan X, Chen B, Duan Z, Xia Z, Ding Y, Chen T, et al. Depression and anxiety in patients with active ulcerative colitis: crosstalk of gut microbiota, metabolomics and proteomics. Gut Microbes. 2021;13(1):1987779.
Yang J, Zhao B, Qian L, Gao F, Fan Y, He X, et al. Causal relationships between genetically determined metabolites and human intelligence: a Mendelian randomization study. Mol Brain. 2021;14(1):29.
Sun S, Jiao M, Han C, Zhang Q, Shi W, Shi J, et al. Causal effects of genetically determined metabolites on risk of polycystic ovary syndrome: a Mendelian randomization study. Front Endocrinol. 2020;11:621.
Yang J, Yan B, Zhao B, Fan Y, He X, Yang L, et al. Assessing the causal effects of human serum metabolites on 5 major psychiatric disorders. Schizophr Bull. 2020;46(4):804–13.
van de Luitgaarden IAT, van Oort S, Bouman EJ, Schoonmade LJ, Schrieks IC, Grobbee DE, et al. Alcohol consumption in relation to cardiovascular diseases and mortality: a systematic review of Mendelian randomization studies. Eur J Epidemiol. 2021. https://doi.org/10.1007/s10654-021-00799-5.
Maukonen M, Havulinna AS, Mannisto S, Kanerva N, Salomaa V, Partonen T. Genetic associations of chronotype in the Finnish general population. J Biol Rhythms. 2020;35(5):501–11.
Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife. 2018. https://doi.org/10.7554/eLife.34408.
Zeng HL, Ge JY, Xu WJ, Ma H, Chen LY, Xia MF, et al. Type 2 diabetes is causally associated with reduced serum osteocalcin: a genomewide association and Mendelian randomization study. J Bone Miner Res. 2021;36(9):1694–707.
Wang J, Liu D, Tian E, Guo ZQ, Chen JY, Kong WJ, et al. Is hearing impairment causally associated with falls? evidence from a two-sample Mendelian randomization study. Front Neurol. 2022;13: 876165.
Skrivankova VW, Richmond RC, Woolf BAR, Yarmolinsky J, Davies NM, Swanson SA, et al. Strengthening the reporting of observational studies in epidemiology using Mendelian randomization: the STROBE-MR statement. JAMA. 2021;326(16):1614–21.
Shin SY, Fauman EB, Petersen AK, Krumsiek J, Santos R, Huang J, et al. An atlas of genetic influences on human blood metabolites. Nat Genet. 2014;46(6):543–50.
Suhre K, Shin SY, Petersen AK, Mohney RP, Meredith D, Wagele B, et al. Human metabolic individuality in biomedical and pharmaceutical research. Nature. 2011;477(7362):54–60.
Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37(7):658–65.
Penninx BW, Pine DS, Holmes EA, Reif A. Anxiety disorders. Lancet. 2021;397(10277):914–27.
Younes N, Hardy-Bayle MC, Falissard B, Kovess V, Chaillet MP, Gasquet I. Differing mental health practice among general practitioners, private psychiatrists and public psychiatrists. BMC Public Health. 2005;5:104.
Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015;44(2):512–25.
Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet. 2018;50(5):693–8.
Burgess S, Thompson SG. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol. 2017;32(5):377–89.
Cohen JF, Chalumeau M, Cohen R, Korevaar DA, Khoshnood B, Bossuyt PM. Cochran’s Q test was useful to assess heterogeneity in likelihood ratios in studies of diagnostic accuracy. J Clin Epidemiol. 2015;68(3):299–306.
Curtin F, Schulz P. Multiple correlations and Bonferroni’s correction. Biol Psychiatry. 1998;44(8):775–7.
Chong J, Soufan O, Li C, Caraus I, Li S, Bourque G, et al. MetaboAnalyst 4.0: towards more transparent and integrative metabolomics analysis. Nucleic Acids Res. 2018;46(W1):W486–94.
Jewison T, Su Y, Disfany FM, Liang Y, Knox C, Maciejewski A, et al. SMPDB 2.0: big improvements to the small molecule pathway database. Nucleic Acids Res. 2014;42(Database issue):D478-84.
Ross KM, Baer RJ, Ryckman K, Feuer SK, Bandoli G, Chambers C, et al. Second trimester inflammatory and metabolic markers in women delivering preterm with and without preeclampsia. J Perinatol. 2019;39(2):314–20.
Karumanchi SA, Granger JP. Preeclampsia and pregnancy-related hypertensive disorders. Hypertension. 2016;67(2):238–42.
Chappell LC, Cluver CA, Kingdom J, Tong S. Pre-eclampsia. Lancet. 2021;398(10297):341–54.
Nagamatsu T, Iwasawa-Kawai Y, Ichikawa M, Kawana K, Yamashita T, Osuga Y, et al. Emerging roles for lysophospholipid mediators in pregnancy. Am J Reprod Immunol. 2014;72(2):182–91.
Calzada E, Onguka O, Claypool SM. Phosphatidylethanolamine metabolism in health and disease. Int Rev Cell Mol Biol. 2016;321:29–88.
Dalziel JE, Fraser K, Young W, McKenzie CM, Bassett SA, Roy NC. Gastroparesis and lipid metabolism-associated dysbiosis in Wistar-Kyoto rats. Am J Physiol Gastrointest Liver Physiol. 2017;313(1):G62–72.
Reichel M, Honig S, Liebisch G, Luth A, Kleuser B, Gulbins E, et al. Alterations of plasma glycerophospholipid and sphingolipid species in male alcohol-dependent patients. Biochim Biophys Acta. 2015;1851(11):1501–10.
Altmaier E, Emeny RT, Krumsiek J, Lacruz ME, Lukaschek K, Hafner S, et al. Metabolomic profiles in individuals with negative affectivity and social inhibition: a population-based study of Type D personality. Psychoneuroendocrinology. 2013;38(8):1299–309.
Feng LR, Barb JJ, Allen H, Regan J, Saligan L. Steroid hormone biosynthesis metabolism is associated with fatigue related to androgen deprivation therapy for prostate cancer. Front Cell Dev Biol. 2021;9: 642307.
Livshits G, Macgregor AJ, Gieger C, Malkin I, Moayyeri A, Grallert H, et al. An omics investigation into chronic widespread musculoskeletal pain reveals epiandrosterone sulfate as a potential biomarker. Pain. 2015;156(10):1845–51.
Spichak S, Bastiaanssen TFS, Berding K, Vlckova K, Clarke G, Dinan TG, et al. Mining microbes for mental health: determining the role of microbial metabolic pathways in human brain health and disease. Neurosci Biobehav Rev. 2021;125:698–761.
Feng L, Zhou N, Li Z, Fu D, Guo Y, Gao X, et al. Co-occurrence of gut microbiota dysbiosis and bile acid metabolism alteration is associated with psychological disorders in Crohn’s disease. FASEB J. 2022;36(1): e22100.
Qu W, Liu S, Zhang W, Zhu H, Tao Q, Wang H, et al. Impact of traditional Chinese medicine treatment on chronic unpredictable mild stress-induced depression-like behaviors: intestinal microbiota and gut microbiome function. Food Funct. 2019;10(9):5886–97.
Peng H, Wang Y, Luo W. Multifaceted role of branched-chain amino acid metabolism in cancer. Oncogene. 2020;39(44):6747–56.
Adeva-Andany MM, Lopez-Maside L, Donapetry-Garcia C, Fernandez-Fernandez C, Sixto-Leal C. Enzymes involved in branched-chain amino acid metabolism in humans. Amino Acids. 2017;49(6):1005–28.
Buchberger B, Huppertz H, Krabbe L, Lux B, Mattivi JT, Siafarikas A. Symptoms of depression and anxiety in youth with type 1 diabetes: a systematic review and meta-analysis. Psychoneuroendocrinology. 2016;70:70–84.
OuYang H, Chen B, Abdulrahman AM, Li L, Wu N. Associations between gestational diabetes and anxiety or depression: a systematic review. J Diabetes Res. 2021;2021:9959779.
Chen Y, Wang W, Fu X, Sun Y, Lv S, Liu L, et al. Investigation of the antidepressant mechanism of combined Radix Bupleuri and Radix Paeoniae Alba treatment using proteomics analysis of liver tissue. J Chromatogr B Analyt Technol Biomed Life Sci. 2021;1179: 122858.
We would like to thank the Xiangya Center for Evidence-Based Practice & Healthcare Innovation Joanna Briggs Institute Affiliated Group of central south university for statistical assistance.
This work was supported by the National Natural Science Foundation of China Grant Number ; the Philosophy and Social Science Foundation of Hunan province Grant Number [19YBA351]; and the Key R & D plan of Hunan Province Grant Number [2020SK2089]. The funding sources played no role in conducting the research and preparing the article.
Ethics approval and consent to participate
This article contains human participants collected by several studies performed by previous studies. All participants gave informed consent in all the corresponding original studies, as described in the Methods. Here, our study is based on the large-scale GWAS datasets, and not the individual-level data. Hence, ethical approval was not applicable.
Consent for publication
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
: Figure S1. The rationale of Mendelian randomization; Figure S2. Scatter plots of the causal association of 1-linoleoylglycerophosphoethanolamine on the risk of anxiety disorders diagnosed by psychiatrists. Figure S3. Forest plots for 1-linoleoylglycerophosphoethanolamine on anxiety disorders diagnosed by psychiatrists; Figure S4. Leave-one-out plots for 1-linoleoylglycerophosphoethanolamine on anxiety disorders diagnosed by psychiatrists; Figure S5. Funnel plots for 1-linoleoylglycerophosphoethanolamine on anxiety disorders diagnosed by psychiatrists; Figure S6. Scatter plots of the genetic association of six metabolites on the risk of anxiety disorders. Figure S7. Forest plots for six potential metabolites on anxiety disorders; Figure S8. Leave-one-out plots for the six potential metabolites on anxiety disorders; Figure S9. Funnel plots for six potential metabolites on anxiety disorders; Figure S10. The p-value distribution of the Pfixed IVW < 0.05 metabolites on anxiety disorder.
The formula used to calculate R2 and F statistic between exposure and outcome; Table S2. Summary statistics of the anxiety disorders; Table S3–S12. Mendelian Randomization estimation for Amino acid(Table S3); Carbohydrate (Table S4); Co-factor or vitamin(Table S5);Energy(Table S6); Lipid(Table S7); Nucleotide (Table S8); Peptide (Table S9); Xenobiotic(Table S10);X(unknow metabolites) (Table S11);known metabolites (Table S12) on the risk of anxiety disorders; Table S13. The characteristics of SNPs and their genetic associations with known metabolites and the anxiety disorders; Table S15. The reverse MR analysis of 1-linoleoylglycerophosphoethanolamine and anxiety disorders using various methods; Table S16. Association of 1linoleoylglycerophosphoethanolamine and additional anxiety disorders GWAS using various methods; Table S14, S17–S21. Summary statistics for the associations of the five metabolites (passed all the sensitivity analyses)-associated SNPs with these exposures and the anxiety disorders from MRC-IEU Consortium; Table S22. Metabolic Pathway Associated with self-reported anxiety; Table S23. Clinical characteristics of participants involved in the study; Table S24. Characteristics of the KORA population related to antihypertensives and lipid-lowering drugs; Table S25. Characteristics and patterns of anti-hyperglycemic treatment of participants with self-reported type 2 diabetes in the KORA F4 study.
About this article
Cite this article
Xiao, G., He, Q., Liu, L. et al. Causality of genetically determined metabolites on anxiety disorders: a two-sample Mendelian randomization study. J Transl Med 20, 475 (2022). https://doi.org/10.1186/s12967-022-03691-2
- Anxiety disorders
- Genetically determined metabolites
- Mendelian randomization
- Serum metabolite