Skip to main content
BMC Psychiatry logoLink to BMC Psychiatry
. 2026 Aug 7;26:870. doi: 10.1186/s12888-026-08503-1

Discovery of serum metabolomics characterization in patients with generalized anxiety disorder based on LC-MS

Xianhui Guo 1,2,✉, Jing Zhang 1, Sheng Dong 2, Weibiao Li 2, Tingting Han 1, Xuexia Wang 2, Ye Xing 1, Yanan Sang 2
PMCID: PMC13746563  PMID: 43000343

Abstract

Background

The pressure on people’s lives and work is increasing with the continuous development of modern society, and generalized anxiety disorder (GAD) has become a relatively common psychological problem. Nevertheless, the pathological mechanisms underlying anxiety disorders remain poorly understood, and specific biomarkers available for identification and clinical diagnosis are still lacking.

Methods

Serum samples were collected from 20 patients with GAD and 20 healthy controls. Untargeted metabolomics based on liquid chromatography-mass spectrometry (LC-MS) was used for detection and analysis to screen differential metabolites and perform correlation analyses.

Results

Compared with those in healthy controls, a total of 1271 differential metabolites were identified in patients with GAD, including 986 upregulated metabolites and 285 downregulated metabolites. Among all the identifiable metabolite classes, the top six categories were lipids and lipid-like molecules (32.53%), organic acids and derivatives (15.12%), organoheterocyclic compounds (13.95%), benzenoids (5.81%), fatty acids (4.65%), and amino acids and peptides (4.65%). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed 11 significantly different metabolic pathways, including those related to choline metabolism in cancer, glutamatergic synapse, and GABAergic synapses. We performed receiver operating characteristic (ROC) curve analysis and correlation analysis between differential metabolites and disease duration, sex, age, and scale scores. The results revealed that 16-hydroxypalmitic acid and 2-hydroxypalmitic acid were closely correlated with Self-Rating Anxiety Scale(SAS) scores.

Conclusion

This study preliminarily investigated the differences in serum metabolites between patients with GAD and healthy controls, providing a reference for the clinical diagnosis and relevant research of GAD.

Trial registration

This study is registered in the International Traditional Medicine Clinical Trial Registry (Registration number: ITMCTR2025000970; Registration date: January 12, 2025).

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12888-026-08503-1.

Keywords: Generalized anxiety disorder, Metabolomics, Biomarkers

Introduction

Generalized anxiety disorder (GAD) is a prevalent mental disorder characterized by excessive anxiety and uncontrollable worry, accompanied by symptoms such as restlessness and/or muscle tension, with a minimum duration of six months [1]. In recent years, its incidence has been rising gradually. The incidence rate is higher in females than in males, with an overall trend towards younger onset [2]. Adolescence is a unique developmental stage that is vulnerable to mental health disturbances; anxiety disorders and depressive disorders are among the leading causes of illness and disability among adolescents [3]. Research has shown that the widespread occurrence of anxiety disorders among adolescents and children has reached 27% in countries with lower and middle economic statuses in recent years [4]. A psychological survey conducted in a cohort study of a large population in Europe revealed that 2.2% of the respondents had been diagnosed with GAD. During the COVID-19 pandemic, anxiety disorders and depressive disorders were both prevalent. Among the respondents who participated in the epidemiological surveys in various countries, 25.6% of respondents experienced anxiety and 23.1% experienced depression, and anxiety had a high incidence rate [5]. The disorder has an insidious onset and a protracted, intractable clinical course. It impairs patients’ mental well-being, daily life and occupational functioning and is frequently complicated by numerous comorbidities, such as insomnia and abdominal distension. These somatic complaints further aggravate emotional distress and exacerbate disease severity [6]. Symptoms of GAD often affect the patient’s cardiovascular system, respiratory system and digestive system [7–11] and are associated with the occurrence of chest pain [12], gastroesophageal reflux disease [13], chronic fatigue syndrome [14], and chronic diseases such as hypertension and heart disease [15]. First-line therapy for anxiety disorders involves medication, such as norepinephrine reuptake inhibitors and 5-hydroxytryptamine reuptake inhibitors [16], and can also include repetitive transcranial magnetic stimulation and mindfulness therapy [17, 18] to alleviate patients’ anxiety and somatic symptoms. The progression of anxiety disorders may lead to comorbid depressive symptoms or full-blown depressive disorder, rendering patients incapable of performing normal daily and occupational activities. Severe cases may present with suicidal ideation. Therefore, more attention should be given to the generation and pathogenesis of GAD, which needs to be suppressed from its source to reduce its incidence and alleviate distress.

Metabolomics is the discipline of identifying and quantifying small-molecule metabolites and is mainly used to identify biomarkers and characterize biological mechanisms of action [19]. A major advantage of untargeted metabolomics is that it can perform a holistic analysis of sample metabolites [20], and it is also used for biomarker discovery and holistic metabolite analysis. Liquid chromatography-mass spectrometry (LC-MS) is the preferred analytical platform for untargeted metabolomics. It enables the profiling of all detectable metabolites, both known and unknown, facilitates the identification of altered metabolites and supports their structural characterization. Untargeted metabolomics involves foundational data analysis and advanced data analysis. Univariate analysis and multivariate analysis are performed on the foundational data, and the qualitative and quantitative results of the metabolome, respectively, are analysed, after which the metabolites whose abundance significantly differ are screened out. Advanced data analysis is a battery of bioinformatics analyses that involve visualizing and exploiting biological functions for differential metabolites on the basis of foundational data analysis. Untargeted metabolomics represents a discovery-based workflow [21].

Therefore, untargeted metabolomics was adopted to identify the changes in the metabolic spectrogram of patients with GAD in this study, which can provide a basis for explaining the pathogenesis of GAD and provide reliable targets for clinical treatment.

Methods

Study population

In this study, patients who were treated at Henan University of Traditional Chinese Medicine were recruited, with a combined total of 20 patients who experienced symptoms of GAD and 20 healthy subjects included between July 2023 and November 2023.

GAD patient inclusion criteria

The inclusion criteria were as follows: Meeting the diagnostic criteria for GAD in the U.S. Diagnostic and Statistical Manual of Mental Disorders (DSM-V): Show excessive worry and anxiety about things and activities that happen at work or in life, for at least 6 months; finding it difficult to control this kind of anxiety or worry; with at least three of the following symptoms: ① feeling restless, agitated, or continuously nervous; ② being easily tired; ③ having difficulty concentrating; ④ being irritable; ⑤ having muscle tension; and ⑥ experiencing sleep disturbance; these symptoms cause clinical pain or damage to other social functions; are not caused by medicine, drugs, or other physical diseases; and are not explained by other mental disorders. Aged between 18 and 60 years regardless of sex; do not take any anti-anxiety drugs or other psychotropic drugs or received other related treatments 4 weeks before treatment; do not take any medication during the treatment period; voluntarily cooperate to complete the treatment and research; and sign the informed consent form.

Exclusion criteria

Patients with secondary anxiety caused by diseases such as hyperthyroidism, hypertension, or coronary heart disease; those with withdrawal reactions from excessive stimulant drugs, hypnotic sedatives, or antianxiety drugs; and those with anxiety associated with obsessive-compulsive disorder, phobia, hypochondria, neurasthenia, mania, depression, or schizophrenia were excluded; pregnant women or breastfeeding women and females who needed to avoid menstruation were excluded; and patients who had severe skin diseases or who had severe damage to the treatment area were excluded; patients who could not tolerate treatment were excluded. Patients who smoked and drank alcohol and those who were excessively obese (BMI ≥ 30) were excluded.

The study was designed and conducted in accordance with the World Medical Association Declaration of Helsinki and was approved by the Ethics Committee of the Third Affiliated Hospital of Henan University of Traditional Chinese Medicine (approval No. 2023HL-010). All the participants signed informed consent forms.

Sample collection

Five mL of fasting blood was collected from the patient in the morning. The collected blood was settled into a centrifuge tube at 37 °C (or room temperature) for 1 h to coagulate and stratify and was then centrifuged at 3000 rpm for 10 min at room temperature. After completion, the liquid supernatant was placed inside a clean centrifuge tube and centrifuged at 12,000 rpm at 4 °C for 10 min. Finally, the liquid supernatant was divided into 1.5 mL centrifuge tubes, and each test tube was filled with 0.2 mL and frozen at -80 °C. In the final step, sufficient dry ice was provided for delivery of the sample. The sample volume requirement was 200 µL/sample.

Experimental information

The main experimental reagents used were methanol, acetonitrile, ammonium acetate, ammonium hydroxide, ddH2O, acetic acid, and 2-propanol. The main experimental instruments wre ultra high performance liquid chromatography, a high resolution mass spectrometer, a centrifuge, a balance, an ultrasonograph, a homogenizer, and a freeze dryer.

Experimental method [22, 23]

Metabolite extraction

One hundred mL of the sample was transferred into an EP tube using a pipette, and 400 µL of extraction solution (with a methanol to acetonitrile ratio of 1:1 v/v) was added. The extraction solution contained internal labels for isotope labelling, the sample was vortexed evenly for 30 s and placed in an ultrasonograph for at least 10 min (ice-water bath method). The sample was subsequently placed at -40 ℃ for at least 1 h and centrifuged at 4 ℃ at 12,000 rpm (centrifugal force 13800 ×g, and a radius of 8.6 cm) for 15 min. The supernatant was collected into the injection bottle and tested with the instrument. An equal volume of the supernatant from each of the samples was transferred, and the samples were mixed and processed as quality control samples for equipment testing.

On-board testing

To identify polar metabolites, the target compounds were separated by using a Vanquish (Thermo Fisher Scientific) ultrahigh-performance liquid chromatograph and a Waters ACQUITY UPLC BEH Amide (2.1 mm * 50 mm, 1.7 μm) liquid chromatography column. Liquid Chromatogram A was an aqueous phase that contained both 25 mmol/L ammonia water and 25 mmol/L ammonium acetate, and Liquid Chromatogram B was acetonitrile. The temperature of the sample plate was set at 4 ℃, and the injection volume was 2 µL.

The experimental mass spectrometry data of the primary and secondary samples were collected under the control of the software (Xcalibur, version: 4.4, Thermo) on an Orbitrap Exploris 120 mass spectrometer. The detailed parameters were as follows: 50 Arb for the sheath gas flow rate, 15 Arb for the aux gas flow rate, 320 ℃ for the capillary temperature, 60,000 for the full MS resolution, 15,000 for the MS/MS resolution, a SNCE of 20/30/40 for the collision energy, and a spray voltage of 3.8 kV (positive) or -3.4 kV (negative).

Quality control

Quality control was performed using the QC samples. Detailed quality control information, including sample information, process QC, and data QC, is provided in the Supplementary Material.

Data processing

The raw data were converted to mzXML format using ProteoWizard software. Metabolite identification was performed using a collaboratively developed R package with the BiotreeDB (V3.0) database, followed by visualization analysis using an in-house developed R package. First, the raw data were preprocessed for peak extraction. Deviant peaks were filtered on the basis of relative standard deviation (RSD), and only peak area data with ≤ 50% missing values in a single group or ≤ 50% missing values across all groups were retained. Missing values were imputed with half of the minimum detected value. Data normalization was performed using internal standards. Data were subjected to logarithmic (LOG) transformation and unit variance (UV) scaling using SIMCA software (V16.0.2, Sartorius Stedim Data Analytics AB, Umeå, Sweden). An orthogonal projections to latent structures discriminant analysis (OPLS-DA) model was constructed on the basis of the first principal component. Model quality was evaluated via 7-fold cross-validation. Model validity was assessed using R2Y (explained variance of the categorical variable Y) and Q2 (predictive ability of the model) derived from cross-validation. Permutation testing was further conducted to verify model validity: The order of the categorical variable Y was randomly permuted multiple times to generate random Q2 values. The relevant plots were generated using the R package ggplot2 (version 3.3.5). The homogeneity of variance was determined using Bartlett’s test. Two-tailed Student’s t tests (with equal or unequal variances, determined by Bartlett’s test results) were performed, and p values were corrected using the Benjamini-Hochberg (BH) method.Combining univariate and multivariate statistical analyses, metabolites meeting the criteria of p < 0.05 (Student’s t test) and variable importance in projection (VIP) > 1 (from the OPLS-DA model) were defined as differential metabolites. Volcano plots were generated using the R package ggplot2 (version 3.3.5). Euclidean distance matrices were calculated on the basis of the relative abundance of metabolites using the R package pheatmap (version 1.0.12). Metabolites were clustered by the complete linkage method, and corresponding heatmaps were constructed. Pathway enrichment analysis for Homo sapiens (human) was performed on the basis of the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database, with statistical significance determined by Fisher’s exact test. KEGG annotation plots were drawn using the R base package (version 3.6.3), and bubble plots were generated using the R package ggplot2 (version 3.3.5).

Results

Characteristics of this study population

The age range of the GAD group was 22–55 years (35.00 ± 11.02), and the group included 4 males and 16 females. The disease trajectory ranged from 6 to 90 months, with a mean value of 31.55 ± 20.618. The age range of the control group was 21–60 years (33.65 ± 12.35), and 9 males and 11 females were included. The general features are shown in Table 1.

Table 1.

Features of the study population

Gender (m/f) Age (years) Disease duration (months)
GAD group 4/16 35.00 ± 11.02 31.55 ± 20.62
Control group 9/11 33.65 ± 12.35 -

Metabolomic analysis of serum samples

OPLS-DA

The results of the OPLS-DA analysis of the serum metabolic data between the control group and pretreatment group are shown in Fig. 1A. The sample sets between the two groups were completely separated in accordance with the OPLS-DA model scatter plot, the distinction between the groups was obvious, and the serum metabolic spectrograms of the two groups of samples were significantly different.

Fig. 1.

Fig. 1

OPLS-DA results. A shows the scatter plot of the OPLS-DA model. The red circles represent the GAD group samples, and the purple diamonds represent the normal group samples. The samples of each group were approximately located within the same interval, and the sample sets between the two groups were completely separated, indicating that the serum metabolic spectrograms of the two groups of samples were significantly different. B shows the dot plots of the OPLS-DA model permutation results. C shows the bar chart of the OPLS-DA pattern replacement results. Both graphs were designed to examine the reliability and rationality of the results

To examine its reliability and rationality, 200 permutation tests were used to avoid model overfitting. The R2Y value of the OPLS-DA model permutation result dot plot was 0.81, the Q2 value was − 0.62, and the location of the sample points was lower than the location of the original point of the fitting line. The results showed that the application of the data analysis model was reasonable and that there was no model overfitting. These results can be applied for further screening of differential metabolites. (Fig. 1B).

The R2Y value of the OPLS-DA pattern replacement result bar chart was 0.82, the P value was > 0.05, the Q2 value was 0.246, and the P value was < 0.05. These findings revealed that the created data analysis model had excellent data processing and prediction ability. (Fig. 1C).

Cluster analysis

On the basis of the experimental results, a cluster heatmap was drawn by normalizing the data and performing a cluster heatmap analysis for all the samples, as shown in Fig. 2A. The figure shows that a discernible difference was apparent in the expression of serum metabolites between the two groups.

Fig. 2.

Fig. 2

A iCluster analysis heatmap. The abscissa in the figure represents different groups of experiments; purple represents the GAD group, and red represents healthy subjects. The ordinate represents the differentially expressed peptides, and the blocks of colour in different places represent the relative expression levels of the peptides in the corresponding places. Red indicates high expression, and blue low expression. The results indicate that there was an obvious grouping pattern of the differentially expressed peptides in the figure. B is a volcano map. Every point in the figure represents a detected peptide, the horizontal coordinates represent the multiple changes in every peptide in the group, and the vertical coordinates represent the P value of the Student’s t test; the colour of the scatter represents the result of the final screening. The significantly upregulated differentially expressed peptides are shown in red, the significantly downregulated differentially expressed peptides are shown in blue, and the nonsignificantly different peptides are shown in grey. C is a circular diagram illustrating the classification and proportion of metabolites

After screening the results of the control group and the GAD group, a total of 1271 differential metabolites were acquired, covering 986 upregulated differential substances and 285 downregulated differential substances. The volcano plot intuitively reflects the complete situation of the differential metabolites between groups (Fig. 2B).

The classification and proportion of metabolites are shown in a circular diagram in Fig. 2C. Among the identifiable categories, lipid and lipoid molecules (32.53%) were the dominant class.

Screening of differential metabolites

With a VIP >1, a Q-VALUE <0.1, and a LOG_FOLDCHANGE absolute value greater than 0.41 as the screening conditions, the differential were screened, and 8 significantly different metabolites were ultimately obtained, including 3 types of lipids and lipid-like molecules, 2 types of fatty acids, 2 types of organoheterocyclic compounds, and 1 type of organic acid and derivative; moreover, 6 types showed upregulation, and 2 types showed downregulation. The specific information is shown in Table 2.

Table 2.

Overall differential metabolites

MS2-name Super class VIP P value Q value LOG_FOLDCHANGE
rac-trans-4-Cotininecarboxylic acid Organoheterocyclic compounds 2.4387 0.0004 0.0463 -0.5290
LPC (13:0) Lipids and lipid-like molecules 2.2294 0.0006 0.0505 0.5343
Glycerophospho-N-palmitoylethanolamine Lipids and lipid-like molecules 2.2196 0.0005 0.0505 0.5266
4-(2-Methyl-1 H-imidazol-1-yl)aniline Organoheterocyclic compounds 1.9103 0.0011 0.0693 -0.6282
1-O-Hexadecyl-2-O-acetyl-sn-glyceryl-3-phosphorylcholine Lipids and lipid-like molecules 2.3539 0.0011 0.0706 0.9938
Met-Phe Organic acids and derivatives 1.9692 0.0021 0.0916 0.5131
2-Hydroxypalmitic acid Fatty acids 1.2555 0.0022 0.0953 0.4556
16-Hydroxypalmitic acid Fatty acids 1.2555 0.0022 0.0953 0.4556

Notes:

VIP: Variable Importance in the Projection score of the metabolite in the OPLS-DA model for the group comparison

P value: Raw probability values derived from Student’s t test

Q value: This metric corresponds to the outcome obtained after multiple hypothesis testing correction of raw P values

LOG_FOLDCHANGE: Values obtained by log₂ transformation of the raw fold changes in metabolite abundance between the two groups

KEGG enrichment pathway analysis

The results of the KEGG enrichment pathway analysis are presented in Fig. 3, with a total of 11 significantly different metabolic pathways identified, including those related to choline metabolism in cancer, glutamatergic synapses, and GABAergic synapses (see Table 3 for details). These results reflected the complexity of GAD, and its pathogenesis was related to multiple factors.

Fig. 3.

Fig. 3

Bubble chart of metabolic pathway enrichment analysis. The horizontal axis represents the rich factor of each pathway, and the vertical axis corresponds to the names of the KEGG metabolic pathways. The size of each dot indicates the number of differential metabolites enriched in the corresponding pathway. The colour gradient reflects the magnitude of the adjusted P value; smaller P values correspond to redder colours, indicating more significant enrichment

Table 3.

KEGG enrichment pathway analysis

Description #.compounds.(dem) Total Percent Rich_factor p_value
Choline metabolism in cancer 2 11 20 0.1818 0.0002
Glutamatergic synapse 1 8 10 0.1250 0.0178
GABAergic synapse 1 9 10 0.1111 0.0200
Proximal tubule bicarbonate reclamation 1 17 10 0.0588 0.0375
Retrograde endocannabinoid signalling 1 19 10 0.0526 0.0418
Nitrogen metabolism 1 20 10 0.050 0.0440
Efferocytosis 1 21 10 0.0476 0.0461
Glycerophospholipid metabolism 2 56 20 0.0357 0.0065
Nucleotide metabolism 2 58 20 0.0345 0.0070
Pyrimidine metabolism 2 64 20 0.0313 0.0085
ABC transporters 2 138 20 0.0145 0.0363

p_value <0.05, percent ≥10%, and #compounds (dem) ≥1 were used as filtering conditions

# compounds (dem): Number of differential metabolites

Total: Total number of total metabolites in the pathway

Percent: Proportion of differential metabolites

Rich_factor: The ratio of the number of differential metabolites annotated in a specific pathway to the total number of all metabolites contained within this pathway

Differential metabolite diagnostic model and correlation analysis

ROC analysis was performed on differential metabolites, and the results are shown in the figure 4A. After constructing a combined diagnostic model incorporating all differential metabolites, the diagnostic efficacy of the model was significantly improved, with an area under the curve of 0.925. This finding suggests that the multi-metabolite combined model has superior diagnostic value. Moreover, a correlation analysis between differentially expressed metabolites and disease duration, sex, age and scale scores (Fig. 4B) showed that 16-hydroxypalmitic acid and 2-hydroxypalmitic acid were closely related to the SAS scores.

Fig. 4.

Fig. 4

A presents the ROC analysis results for the differential metabolites, with curves in distinct colors representing the data for each individual metabolite. B displays the results of the correlation analysis, where red indicates a positive correlation and purple a negative correlation, with darker colours indicating stronger correlations. The asterisk (*) indicates statistically significant correlations

Discussion

The recent high incidence of GAD has severely affected patients’ daily life and work. Therefore, an in-depth exploration of the pathogenesis of anxiety disorders is urgently needed to provide a more favourable basis for clinical diagnosis and treatment.

Previous serum metabolomics studies focusing on patients with chronic obstructive pulmonary disease (COPD) complicated by anxiety and depression have revealed significant alterations in the levels of metabolites such as lipids and lipid-like molecules, organic acids and derivatives, and organoheterocyclic compounds [24]. Other studies have demonstrated that carbohydrate and lipid metabolism are involved in the pathogenesis of psychiatric disorders [25]. Elevated levels of triglycerides, total cholesterol, apolipoproteins and blood glucose were reported to increase the risk of subsequent anxiety, depression and other mental illnesses. Collectively, anxiety exerts extensive systemic effects on the human body, accompanied by perturbed abundances of diverse classes of metabolites during disease onset. In the present study, LC-MS-based metabolomics was performed to characterize shared metabolic signatures in patients with GAD, laying a foundation for understanding metabolic disturbances underlying anxiety disorders.

Our results revealed that the metabolites with the most pronounced differential expression in GAD patients fell into the classes of lipids and lipid-like molecules, followed by organic acids and derivatives, organoheterocyclic compounds, benzenoids, fatty acids, and amino acids and peptides. Lipid molecules may be involved in the pathogenesis of anxiety disorders through multiple pathways. Lipid components in the central nervous system play a vital role and can be divided into three main categories: Phospholipids, sphingolipids, and cholesterol. The lipid concentration of the central nervous system is second only to that of adipose tissue and plays a key role in maintaining system stability. Lipids such as phospholipids and polyunsaturated fatty acids are involved in the pathogenesis of GAD. Studies have shown that long-term lipid metabolism disorders can damage the blood-brain barrier, induce neuroinflammation, and lead to symptoms of anxiety and depression [26]. In addition, abnormal lipid metabolism may activate innate immune cells such as microglia, promote the catabolism of proinflammatory cytokines and biogenic amines, and induce hypothalamic inflammation [27] or participate in complex pro-oxidative and antioxidant processes, leading to the occurrence of mental illness [28, 29]. Glycerophosphoethanolamines can regulate processes such as neural stem cell ageing, oxidative stress, and microglial activation, thereby optimizing neuronal status [30, 31]. Glycerophosphocholines play important roles in regulating glucose homeostasis and affect systemic metabolism [32]. Glycerophosphoethanolamines can induce the release of phospholipids and neurotransmitters in brain cell membranes and reduce lipid peroxidation levels, thereby reducing damage to the hippocampus and improving the body’s learning and memory functions [33]. Some studies have explored the correlation between functional connectivity abnormalities and lipid metabolism profiles in patients with depression. Among them, the right anterior cingulate and paracingulate gyri are rich in glycerophosphocholines and glycerophosphoethanolamines [34]. Among them, LPC can maintain the integrity of neuronal membrane structure by participating in lecithin metabolism and affecting cell functional activities [35, 36]; by participating in the synthesis of acetylcholine, LPC can change neuronal excitability and synaptic function; affect the function of the central nervous system; and regulate the body’s memory, attention, stress response and other mental activities [37], and it can be used as a target for neuropathological research on GAD. Fatty acids, especially omega-3 fatty acids, can regulate synaptic plasticity, neurotransmission and neuroprotection and play a regulatory role in mood disorders [38]. Some studies have reported that long-chain polyunsaturated fatty acids are risk biomarkers for mood disorders [39].

The results of metabolic pathway enrichment analysis revealed that the differential metabolites of GAD were enriched mainly in choline metabolism in cancer, glutamatergic synapse, GABAergic synapse and other pathways. Although enrichment analysis revealed multiple dysregulated pathways, these findings should only be regarded as preliminary clues given the limited number of detectable metabolites annotated to each pathway, and further validation is needed in subsequent studies.

Disorders of choline metabolism can participate in anxiety-related neural regulation by affecting cell membrane fluidity and signal transduction [40]. Moreover, the cholinergic system controls the production and release of acetylcholine, which is involved in emotions, sleep, memory, and other processes [41]. γ-aminobutyric acid (GABA) is an important inhibitory transmitter in the central nervous system that can regulate the excitability of neurons and participate in the pathogenesis of anxiety [42]. Some studies have reported that low GABA receptor function in the amygdala-hippocampus circuit is the root cause of stress-induced anxiety [43]. Glutamate is an important excitatory transmitter in the central nervous system. It is closely related to monoaminergic neurons, peptidergic neurons and neurotrophic factors and plays an important role in emotion regulation [44]. GABA and glutamate work together to maintain the balance between excitation and inhibition in the nervous system. In addition, changes in choline levels can also regulate the conduction of glutamate excitatory signals and the release of GABA inhibitory signals, leading to changes in the sensitivity of the nervous system [45].

Differential metabolites with specific alterations were further screened out, namely, LPC (13:0), glycerophospho-N-palmitoylethanolamine, 1-O-hexadecyl-2-O-acetyl-sn-glyceryl-3-phosphorylcholine, Met-Phe, 16-hydroxypalmitic acid and 2-hydroxypalmitic acid. Among them, alterations in 16-hydroxypalmitic acid and 2-hydroxypalmitic acid were strongly correlated with SAS scores. ROC curve results confirm decent predictive power for the analytical model, while the combined diagnostic model shows far more promise in elevating diagnostic accuracy than any standalone index. Large-sample clinical trials are needed in the future to further verify whether these metabolites can serve as biological indicators for the prediction and diagnosis of GAD.

The present study had several limitations. First, the sample size was relatively small. Multicentre trials with enlarged sample cohorts are needed to further validate the accuracy of our findings. Second, stratified cohort analyses suffer from low reliability because of the limited sample size, which restricts comprehensive profiling of sex-specific metabolic signatures. The differential metabolites identified merely serve as preliminary exploratory observations. Follow-up investigations with optimized research designs will systematically characterize metabolites closely linked to the pathogenesis of GAD. Third, multiple confounding variables were not incorporated into the analytical framework. Body mass index and other relevant covariates will be adjusted for in subsequent studies to strengthen the scientific validity of the results. Fourth, this study focused on profiling serum metabolomic alterations in patients with GAD. The identified metabolites belonged to restricted chemical classes, and several pathway enrichment signals were driven by only a small number of pivotal metabolites. Multi-omicsMultiomics approaches should be integrated in future work to comprehensively decipher the biological mechanisms underlying these metabolic pathways.

Conclusions

Our findings preliminarily revealed metabolites closely associated with the pathogenesis of GAD. These findings indicate that clinical diagnosis and treatment should not rely merely on patients’ clinical manifestations and subjective scale scores; relevant objective biomarkers should also be referenced to assist in evaluating disease status and severity. Exploring specific biomarkers of GAD is highly important for clinical diagnosis and management. Additional investigations are still warranted to thoroughly elucidate the correlations between each differential metabolite and GAD.

Supplementary Information

Below is the link to the electronic supplementary material.

Acknowledgements

Not applicable.

Abbreviations

GAD

Generalized anxiety disorder.

LC-MS

Untargeted liquid chromatography-mass spectrometry.

DSM-V

Diagnostic and Statistical Manual of Mental Disorders.

OPLS-DA

Orthogonal projections to latent structures- discriminant analysis.

VIP

Variable Importance in the Projection.

KEGG

Kyoto Encyclopedia of Genes and Genomes.

ROC

Receiver Operating Characteristic curve.

Author contributions

Xianhui Guo: Conceptualization, Funding acquisition, Methodology, Wrote review & editing; Jing Zhang: Data curation; Methodology, Writing – original draft; Sheng Dong: Formal analysis, Writing – original draft; Weibiao Li: Investigation, Wrote review & editing; Tingting Han: Validation, Writing – original draft; Xuexia Wang: Methodology, Supervision, Wrote review & editing; Ye Xing: Data curation, Visualization; Yanan Sang: Supervision, Wrote review & editing.

Funding

This study was funded by the Henan Province “Double First-class” establishment Discipline of Chinese Medical Science Research Project (HSRP-DFCTCM-2023-1-09), the Henan Provincial Key Science and Technology Research Project (252102310469), and the General Program of Henan Provincial Natural Science Foundation (262300420155).

Data availability

The datasets used and analysed during the current sub-study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of the Third Affiliated Hospital of Henan University of Traditional Chinese Medicine (approval No. 2023HL-010). All the participants signed informed consent forms.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  • 1.American Psychiatric Association Diagnostic and Statistical Manual of Mental Disorders. 5th ed. American Psychiatric Publishing, Inc.; Arlington, VA, USA: 2013.
  • 2.Bandelow B, Michaelis S. Epidemiology of anxiety disorders in the 21st century. Dialogues Clin Neurosci. 2015;17(3):327–35. 10.31887/DCNS.2015.17.3/bbandelow. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Beckwith S, Chandra-Mouli V, Blum RW. Trends in Adolescent Health: Successes and Challenges From 2010 to the Present. J Adolesc Health. 2024;75(4S):S9–19. 10.1016/j.jadohealth.2024.04.015. [DOI] [PubMed] [Google Scholar]
  • 4.Alozkan-Sever C, Uppendahl JR, Cuijpers P, de Vries R, Rahman A, Mittendorfer-Rutz E, et al. Research Review: Psychological and psychosocial interventions for children and adolescents with depression, anxiety, and post-traumatic stress disorder in low- and middle-income countries - a systematic review and meta-analysis. J Child Psychol Psychiatry. 2023;64(12):1776–88. 10.1111/jcpp.13891. [DOI] [PubMed] [Google Scholar]
  • 5.Morin CM, Bjorvatn B, Chung F, Holzinger B, Partinen M, Penzel T, et al. Insomnia, anxiety, and depression during the COVID-19 pandemic: an international collaborative study. Sleep Med. 2021;87:38–45. 10.1016/j.sleep.2021.07.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Remes O, Wainwright N, Surtees P, Lafortune L, Khaw KT, Brayne C. Generalised anxiety disorder and hospital admissions: findings from a large, population cohort study. BMJ Open. 2018;8(10):e018539. 10.1136/bmjopen-2017-018539. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Chen X, Liang XY, Zhang GL, Wei SY, Zou JX, Liu H, et al. Relationships between heart failure, depression, and anxiety: A Mendelian randomization study. Med (Baltim). 2024;103(42):e40005. 10.1097/MD.0000000000040005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Liang KJ, Cheng CH, Liu CY, Hsu SC, von Leupoldt A, Jelinčić V, et al. Neural oscillations underlying the neural gating of respiratory sensations in generalized anxiety disorder. Respir Physiol Neurobiol. 2024;321:104215. 10.1016/j.resp.2024.104215. [DOI] [PubMed] [Google Scholar]
  • 9.Doyle T, Palmer S, Johnson J, Babyak MA, Smith P, Mabe S, et al. Association of anxiety and depression with pulmonary-specific symptoms in chronic obstructive pulmonary disease. Int J Psychiatry Med. 2013;45(2):189–202. 10.2190/PM.45.2.g. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Hernaez R, Kramer JR, Khan A, Phillips J, McCallister K, Chaffin K, et al. Depression and Anxiety Are Common Among Patients With Cirrhosis. Clin Gastroenterol Hepatol. 2022;20(1):194–e2031. 10.1016/j.cgh.2020.08.045. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Wu H, Liang G, Kong M, Zhang Y, Zhou Y, Han J, et al. The status and risk factors for anxiety/depression in patients with atrophic chronic gastritis: a cross-sectional study. Ann Palliat Med. 2022;11(10):3147–59. 10.21037/apm-22-730. [DOI] [PubMed] [Google Scholar]
  • 12.Hamel S, Denis I, Turcotte S, Fleet R, Archambault P, Dionne CE, et al. Anxiety disorders in patients with noncardiac chest pain: association with health-related quality of life and chest pain severity. Health Qual Life Outcomes. 2022;20(1):7. 10.1186/s12955-021-01912-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Mohammad S, Chandio B, Soomro AA, Lakho S, Ali Z, Ali Soomro Z, et al. Depression and Anxiety in Patients with Gastroesophageal Reflux Disorder With and Without Chest Pain. Cureus. 2019;11(11):e6103. 10.7759/cureus.6103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Leung P, Li SH, Graham BM. The relationship between repetitive negative thinking, sleep disturbance, and subjective fatigue in women with Generalized Anxiety Disorder. Br J Clin Psychol. 2022;61(3):666–79. 10.1111/bjc.12356. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Lim LF, Solmi M, Cortese S. Association between anxiety and hypertension in adults: A systematic review and meta-analysis. Neurosci Biobehav Rev. 2021;131:96–119. 10.1016/j.neubiorev.2021.08.031. [DOI] [PubMed] [Google Scholar]
  • 16.Murrough JW, Yaqubi S, Sayed S, Charney DS. Emerging drugs for the treatment of anxiety. Expert Opin Emerg Drugs. 2015;20(3):393–406. 10.1517/14728214.2015.1049996. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Parikh TK, Strawn JR, Walkup JT, Croarkin PE. Repetitive Transcranial Magnetic Stimulation for Generalized Anxiety Disorder: A Systematic Literature Review and Meta-Analysis. Int J Neuropsychopharmacol. 2022;25(2):144–6. 10.1093/ijnp/pyab077. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Newman MG, Basterfield C, Erickson TM, Caulley E, Przeworski A, Llera SJ. Psychotherapeutic treatments for generalized anxiety disorder: cognitive and behavioral therapies, enhancement strategies, and emerging efforts. Expert Rev Neurother. 2022;22(9):751–70. 10.1080/14737175.2022.2125800. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Guijas C, Montenegro-Burke JR, Warth B, Spilker ME, Siuzdak G. Metabolomics activity screening for identifying metabolites that modulate phenotype. Nat Biotechnol. 2018;36(4):316–20. 10.1038/nbt.4101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Schrimpe-Rutledge AC, Codreanu SG, Sherrod SD, McLean JA. Untargeted Metabolomics Strategies-Challenges and Emerging Directions. J Am Soc Mass Spectrom. 2016;27(12):1897–905. 10.1007/s13361-016-1469-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Gertsman I, Barshop BA. Promises and pitfalls of untargeted metabolomics. J Inherit Metab Dis. 2018;41(3):355–66. 10.1007/s10545-017-0130-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wang WC, Huang CH, Chung HH, Chen PL, Hu FR, Yang CH, et al. Metabolomics facilitates differential diagnosis in common inherited retinal degenerations by exploring their profiles of serum metabolites. Nat Commun. 2024;15(1):3562. 10.1038/s41467-024-47911-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Zhang Y, Yu B, Qi Q, Azarbarzin A, Chen H, Shah NA, et al. Metabolomic profiles of sleep-disordered breathing are associated with hypertension and diabetes mellitus development. Nat Commun. 2024;15(1):1845. 10.1038/s41467-024-46019-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Ye J, Li P, Liu P, Pei W, Wang R, Liu H, et al. Serum Metabolomics Analysis Revealed Metabolic Pathways Related to AECOPD Complicated with Anxiety and Depression. Int J Chron Obstruct Pulmon Dis. 2024;19:2135–51. 10.2147/COPD.S471817. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Chourpiliadis C, Zeng Y, Lovik A, Wei D, Valdimarsdóttir U, Song H, et al. Metabolic Profile and Long-Term Risk of Depression, Anxiety, and Stress-Related Disorders. JAMA Netw Open. 2024;7(4):e244525. 10.1001/jamanetworkopen.2024.4525. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Wang A, Wan X, Zhuang P, Jia W, Ao Y, Liu X, et al. High fried food consumption impacts anxiety and depression due to lipid metabolism disturbance and neuroinflammation. Proc Natl Acad Sci U S A. 2023;120(18):e2221097120. 10.1073/pnas.2221097120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Lainez NM, Coss D. Obesity,neuroinflammation,and reproductive function. Endocrinology. 2019;160(11):2719–36. 10.1210/en.2019-00487. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Wang H, He Y, Sun Z, Ren S, Liu M, Wang G, et al. Microglia in depression: an overview of microglia in the pathogenesis and treatment of depression. J Neuroinflammation. 2022;19(1):132. 10.1186/s12974-022-02492-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Wang YL, Han QQ, Gong WQ, Pan DH, Wang LZ, Hu W, et al. Microglial activation mediates chronic mild stress-induced depressive- and anxiety-like behavior in adult rats. J Neuroinflammation. 2018;15(1):21. 10.1186/s12974-018-1054-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Biggio G, Di Fazio I, Martini C. Cognitive function modulation during aging: a focus on L-alpha-GPE. Eur Rev Med Pharmacol Sci. 2021;25(7):3015–27. 10.26355/eurrev_202104_25555. [DOI] [PubMed] [Google Scholar]
  • 31.Zappelli E, Daniele S, Ceccarelli L, Vergassola M, Ragni L, Mangano G, et al. α-Glyceryl-phosphoryl-ethanolamine protects human hippocampal neurons from ageing-induced cellular alterations. Eur J Neurosci. 2022;56(5):4514–28. 10.1111/ejn.15783. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Cikes D, Leutner M, Cronin SJF, Novatchkova M, Pfleger L, Klepochová R, et al. Gpcpd1-GPC metabolic pathway is dysfunctional in aging and its deficiency severely perturbs glucose metabolism. Nat Aging. 2024;4(1):80–94. 10.1038/s43587-023-00551-6. [DOI] [PubMed] [Google Scholar]
  • 33.Puri BK. Proton and 31-phosphorus neurospectroscopy in the study of membrane phospholipids and fatty acid intervention in schizophrenia, depression, chronic fatigue syndrome (myalgic encephalomyelitis) and dyslexia. Int Rev Psychiatry. 2006;18(2):145–7. 10.1080/09540260600581852. [DOI] [PubMed] [Google Scholar]
  • 34.Wei J, Zhang Z, Yang X, Zhao L, Wang M, Dou Y, et al. Abnormal functional connectivity within the prefrontal cortex is associated with multiple plasma lipid species in major depressive disorder. J Affect Disord. 2024;350:713–20. 10.1016/j.jad.2023.12.072. [DOI] [PubMed] [Google Scholar]
  • 35.Law SH, Chan ML, Marathe GK, Parveen F, Chen CH, Ke LY. An Updated Review of Lysophosphatidylcholine Metabolism in Human Diseases. Int J Mol Sci. 2019;20(5):1149. 10.3390/ijms20051149. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Spanner S, Ansell GB. Phospholipid transport and turnover in nervous tissue. Biochem Soc Trans. 1979;7:338–41. 10.1042/bst0070338. [DOI] [PubMed] [Google Scholar]
  • 37.Świt P, Pollap A, Orzeł J. Spectroscopic Determination of Acetylcholine (ACh): A Representative Review. Top Curr Chem (Cham). 2023;381(4):16. 10.1007/s41061-023-00426-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Müller CP, Reichel M, Mühle C, Rhein C, Gulbins E, Kornhuber J. Brain membrane lipids in major depression and anxiety disorders. Biochim Biophys Acta. 2015;1851(8):1052–65. 10.1016/j.bbalip.2014.12.014. [DOI] [PubMed] [Google Scholar]
  • 39.Berger ME, Smesny S, Kim SW, Davey CG, Rice S, Sarnyai Z, et al. Omega-6 to omega-3 polyunsaturated fatty acid ratio and subsequent mood disorders in young people with at-risk mental states: a 7-year longitudinal study. Transl Psychiatry. 2017;7(8):e1220. 10.1038/tp.2017.190. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Naser PV, Kuner R, Molecular. Cellular and Circuit Basis of Cholinergic Modulation of Pain. Neuroscience. 2018;387:135–48. 10.1016/j.neuroscience.2017.08.049. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Riley CA, Renshaw PF. Brain choline in major depression: A review of the literature. Psychiatry Res Neuroimaging. 2018;271:142–53. 10.1016/j.pscychresns.2017.11.009. [DOI] [PubMed] [Google Scholar]
  • 42.Felice D, Cryan JF, O’Leary OF. GABA Receptors: Anxiety and Mood Disorders. Curr Top Behav Neurosci. 2022;52:241–65. 10.1007/7854_2020_171. [DOI] [PubMed] [Google Scholar]
  • 43.Qin X, Pan HQ, Huang SH, Zou JX, Zheng ZH, Liu XX, et al. GABAA(δ) receptor hypofunction in the amygdala-hippocampal circuit underlies stress-induced anxiety. Sci Bull (Beijing). 2022;67(1):97–110. 10.1016/j.scib.2021.09.007. [DOI] [PubMed] [Google Scholar]
  • 44.An XL, Tai FD. AVP and Glu systems interact to regulate levels of anxiety in BALB/cJ mice. Dongwuxue Yanjiu. 2014;35(4):319–25. 10.13918/j.issn.2095-8137.2014.4.319. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Hudhud L, Hauksson J, Haney M, Sparrman T, Eriksson J, Lindgren L. Choline levels in the pregenual anterior cingulate cortex associated with unpleasant pain experience and anxiety. NeuroImage. 2025;310:121153. 10.1016/j.neuroimage.2025.121153. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data Availability Statement

The datasets used and analysed during the current sub-study are available from the corresponding author upon reasonable request.


Articles from BMC Psychiatry are provided here courtesy of BMC

RESOURCES