Untargeted metabolomics analysis of serum metabolic signatures as novel biomarkers for gastric carcinoma

Abstract

BACKGROUND

In recent years, metabolomics has emerged as a novel platform for biomarker discovery. However, the metabolic profiles associated with gastric carcinoma (GC) remain insufficiently explored.

AIM

To examine the differences in metabolites between patients with GC and healthy controls, with the objective of identifying potential serum biomarkers for GC diagnosis through a non-targeted metabolomics approach.

METHODS

An untargeted metabolic analysis was conducted on serum samples from 6 patients with GC and 6 healthy controls. Subsequently, the differential metabolites identified were further validated in serum samples from an expanded cohort of 50 patients with GC and 50 healthy controls. The discriminative capacity of differential metabolites in distinguishing patients with GC from healthy controls was assessed utilizing the receiver operating characteristic curve analysis. The association between the serum levels of differential metabolites and the disease severity, as determined by the tumor-node-metastasis staging system, was evaluated using Spearman’s rank correlation coefficient.

RESULTS

Our findings revealed a significant alteration in the metabolic profile, characterized by 111 up-regulated and 55 down-regulated metabolites in patients with GC compared to healthy controls. Among the top 10 up-regulated metabolites, the serum concentrations of eight metabolites including fenpiclonil, methyclothiazide, 5-hydroxyindoleacetate, 3-pyridinecarboxylic acid, guanabenz, 2,2-dichloro-N-(3-chloro-1,4-dioxo-2-naphthyl) acetamide, epigallocatechin gallate, and dimethenamid, were further validated to be significantly elevated in a cohort of 50 patients diagnosed with GC compared to 50 healthy control subjects (P < 0.001). With the exception of 3-pyridinecarboxylic acid, the area under the curve values for the remaining seven metabolites exceeded 0.7, suggesting that these metabolites possess substantial diagnostic potential for distinguishing patients with GC from healthy individuals. Additionally, the serum concentrations of methyclothiazide (r = 0.615, P < 0.001), epigallocatechin gallate (r = 0.482, P = 0.004), and dimethenamid (r = 0.634, P < 0.001) demonstrated a significant positive correlation with the T stage in patients with GC. The serum concentrations of methyclothiazide (r = 0.438, P = 0.008) and epigallocatechin gallate (r = 0.383, P = 0.023) exhibited a significant positive correlation with the N stage in these patients.

CONCLUSION

This study provides insights into the metabolic alterations associated with GC, and the identification of these biomarkers may enhance the clinical detection and management of the disease.

Key Words: Gastric carcinoma; Untargeted metabolomics; Serum; Biomarker; Diagnosis

Core Tip: We applied untargeted metabolomics to explore serum metabolic profile changes to identify potential serum biomarkers for the diagnosis of gastric carcinoma (GC). Firstly, a substantial alteration in the metabolic profile was observed. Secondly, the serum concentrations of eight metabolites were significantly elevated in 50 GC patients compared to 50 healthy controls. Thirdly, the area under the curve values for eight metabolites exceeded 0.7, indicating their effectiveness in distinguishing GC patients. Fourthly, the serum concentrations of methyclothiazide and epigallocatechin gallate demonstrated a positive correlation with T and N stage, while the serum concentrations of dimethenamid showed a positive correlation with T stage in GC patients.

INTRODUCTION

Gastric carcinoma (GC) ranks as the second most prevalent cancer in China and stands as the third leading cause of cancer-related fatalities globally[1]. Owing to its non-specific symptoms, the diagnosis of this disease is frequently delayed. Moreover, a large number of patients with advanced GC are unable to undergo surgical treatment, leading to a relatively low 5-year survival rate of approximately 25%[2,3]. Early detection and treatment are crucial for improving outcomes. Endoscopic mucosal biopsy is the diagnostic gold standard but is costly and invasive, limiting its use in China. Current tumor markers like carcinoembryonic antigen, carbohydrate antigen 19-9, and carbohydrate antigen 72-4 are not sufficiently sensitive or specific for GC diagnosis[4]. Other serum biomarkers such as microRNAs, long non-coding RNAs, and circular RNAs exhibit limited specificity as GC diagnostic markers, as evidenced by their dysregulation in multiple cancers; additionally, their expression variability due to tumor heterogeneity and patient-related factors, along with technical detection challenges and insufficient large-scale clinical validation, hinder their reliable clinical application[5]. Therefore, there is an urgent requirement to discover new and reliable non-invasive biomarkers for the detection of GC. This endeavor represents a critical measure to facilitate early intervention and curb mortality rates.

Metabolomics technology, which has developed following the advancements in genomics and proteomics[6,7], represents a significant component of systems biology. This technology involves the qualitative and quantitative analysis of a diverse array of small molecule metabolites, with a specific emphasis on energy metabolites within organisms[8]. The metabolome influences cellular physiology by modulating various other “omics” layers, including the genome, epigenome, transcriptome, and proteome[9]. The increasing application of metabolomic technology in biomedicine can be attributed to recent advancements in the field. The utilization of metabolomics presents a significant opportunity to elucidate disease mechanisms, identify biomarkers, and develop innovative therapeutic strategies[9,10].

In recent years, metabolomics has emerged as a pivotal platform for biomarker discovery[11,12]. The metabolic profiling facilitated by liquid chromatography (LC) coupled to tandem mass spectrometry (MS) (LC-MS/MS) enables the simultaneous measurement of multiple metabolic alterations during pathological processes and the identification of dynamic metabolic responses within critical intermediary biochemical pathways. Metabolomics detection has been utilized across various cancer types, including esophageal[13], breast[14], bladder[15], lung[16], and thyroid cancer[17]. A recent study investigated the disparities in plasma metabolic profiles between patients with esophageal squamous cell carcinoma and healthy controls, identifying a panel of eight metabolites as potential diagnostic biomarkers[18]. Another study identified a panel of ten serum metabolic biomarkers exhibiting exceptional discriminative power for non-small-cell lung cancer, achieving a combined area under the curve (AUC) of 0.95 in the validation cohort[19]. However, there is a paucity of research examining concurrent changes in metabolic profiles among patients with GC.

In this study, we utilized untargeted metabolomics to investigate alterations in serum metabolic profiles among GC patients, with the aim to: (1) Screen for potential serum biomarkers applicable to GC diagnosis; and (2) Analyze the correlation between the serum concentrations of differentially expressed metabolites and the disease severity in GC patients.

MATERIALS AND METHODS

Subjects

A cohort of 50 patients diagnosed with GC and admitted to the Second People's Hospital of Hefei, China, between November 2023 and October 2024, were recruited for this study. Each patient presented with upper abdominal discomfort and received a GC diagnosis confirmed through pathological examination. Concurrently, a control group comprising healthy individuals was established. These control subjects were volunteers who participated in a complimentary health screening designed to identify any potential organic lesions in the stomach. Clinical data were extracted from the medical records of the participants. The exclusion criteria were: (1) Individuals undergoing radiotherapy or chemotherapy; (2) Individuals diagnosed with other forms of cancer or major organ diseases, including those affecting the heart, liver, kidneys, or lungs; (3) Individuals with severe active infectious diseases; and (4) Individuals with severe hematological disorders, those who have undergone bone marrow transplantation, or those with severe trauma or immune disorders. The control group included 50 healthy volunteers, matched according to age, gender, and body mass index. This study received approval from the Ethics Committee of The Second People's Hospital of Hefei, approval No. 2023-keyan-123. Informed consent was obtained from all participants.

Sample collection and preparation

The blood sample was collected from the participant’s vein following a 10-11 hours overnight fasting period, between 8:00 am and 9:00 am. Subsequently, the samples were immediately centrifuged at 12000 g for 5 minutes at 4 ℃, and the resulting supernatant was used as serum samples. These serum samples were stored at -80 ℃ until further analysis.

Metabolite extraction and LC-MS/MS analysis

Metabolomics analysis was conducted on the samples using LC/MS by Gene Denovo Co. Ltd, Guangzhou, China. Following the addition of 1000 μL of extraction solvent (a mixture of acetonitrile, methanol, and water in a 2:2:1 ratio, containing an internal standard), the samples underwent vortexing for 30 seconds, homogenization at 45 Hz for 4 minutes, and sonication for 5 minutes in an ice-water bath. This process of homogenization and sonication was repeated three times. Subsequently, the samples were incubated at -20 °C for 1 hour and centrifuged at 12000 rpm and 4 °C for 15 minutes. The supernatants obtained were transferred to LC-MS vials and stored at -80 °C until analysis by ultra-high-performance LC-Quadrupole-Orbitrap/MS. A quality control sample was prepared by combining equal aliquots of the supernatants.

LC-MS/MS analyses were conducted utilizing an ultra-high-performance system (model 1290, Agilent Technologies) equipped with a UPLC HSS T3 column (dimensions: 2.1 mm × 100 mm, particle size: 1.8 μm) and interfaced with a Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific). The mobile phase A consisted of 0.1% formic acid in water for positive ion mode and 5 mmol/L ammonium acetate in water for negative ion mode, whereas mobile phase B comprised acetonitrile. The elution gradient was programmed as follows: Initial conditions at 0 minutes with 1% B, maintained at 1% B until 1 minute, increased to 99% B by 8 minutes, held at 99% B until 10 minutes, returned to 1% B at 10.1 minutes, and maintained at 1% B until 12 minutes. The flow rate was set at 0.5 mL/minute, and the injection volume was 2 μL. The QE mass spectrometer was employed due to its capability to acquire MS/MS spectra in an information-dependent acquisition mode during LC/MS experiments. In this configuration, the acquisition software (Xcalibur 4.0.27, Thermo) continuously analyzes the full scan survey MS data in real-time and initiates the acquisition of MS/MS spectra based on predefined criteria. The electrospray ionization source parameters were configured as follows: Sheath gas flow rate at 45 arbitrary units, auxiliary gas flow rate at 15 arbitrary units, capillary temperature at 320 °C, full MS resolution set to 70000, MS/MS resolution set to 17500, and collision energy set to 20/40/60 electron volts in the normalized collision energy mode. The spray voltage was adjusted to 3.8 kV for positive ion mode and -3.1 kV for negative ion mode.

Data reprocessing and annotation

The MS raw data files were converted to the mzML format utilizing ProteoWizard and subsequently processed using the R package XCMS (version 3.2). This processing included retention time (RT) alignment, peak detection, and peak matching. The dataset was then subjected to a filtering criterion whereby metabolites present in fewer than 50% of the samples within any group (with quality control samples considered as a separate group) were excluded. Following this, normalization was conducted using an internal standard for each sample. Missing values were imputed by substituting them with half of the minimum value observed in the dataset by default. Following preprocessing, a data matrix was produced comprising RT, mass-to-charge ratio values, and peak intensity. Peak annotation was subsequently conducted using OSI-SMMS (version 1.0, Dalian Chem Data Solution Information Technology Co. Ltd.). Material annotation was conducted utilizing data from MassBank, Metlin, Mona, and other publicly accessible databases, in addition to a proprietary secondary MS database developed by Gene Denovo Biotechnology Co. The structural elucidation of metabolites within biological samples was achieved by correlating metabolite RT, molecular mass (with a mass error margin within 25 ppm), secondary fragmentation spectra, collision energy, and additional relevant data from the local database. The identification results underwent rigorous verification and manual confirmation. The identification level achieved was above level 2.

Multivariate statistical analysis

Following data preprocessing, principal component analysis, partial least squares discrimination analysis (PLS-DA), and orthogonal PLS-DA were conducted using R package models (http://www.r-project.org/).

Differential metabolites analysis

To identify and quantify differential metabolites between the two groups, fold change analysis and t-tests were employed. Univariate analysis further assessed metabolite variations between the groups to discern significant differences. Metabolites exhibiting a P value of < 0.05, false discovery rate (FDR) value < 0.05 and variable importance projection > 1 were classified as differentially expressed between the groups.

Kyoto Encyclopedia of Genes and Genomes pathway analysis

Metabolites were annotated to the metabolic pathways within the Kyoto Encyclopedia of Genes and Genomes database for the purpose of conducting pathway and enrichment analyses. The pathway enrichment analysis revealed metabolic or signal transduction pathways that were significantly enriched among the differential metabolites relative to the entire background set. The computed P values were adjusted using the FDR correction, with a threshold set at FDR ≤ 0.05. Pathways that satisfied this criterion were designated as significantly enriched pathways within the differential metabolites.

Statistical analysis

Data analysis was performed utilizing SPSS version 17.0 (IBM Corporation, Armonk, NY, United States). The Kolmogorov-Smirnov one-sample test was applied to assess the normality of distributions. Descriptive statistics were conducted, with normally distributed quantitative variables reported as means ± SD, and non-normally distributed quantitative variables presented as medians and interquartile ranges (25^th-75^th percentile). The independent samples t-test was used to compare variables that followed a normal distribution between the two groups, while the Mann-Whitney U test was employed for the comparison of variables that did not conform to a normal distribution. Receiver operating characteristic (ROC) curves, in conjunction with AUC values, were employed to evaluate the discriminatory capacity of differential metabolites in differentiating patients with GC from healthy controls. Spearman’s rank correlation coefficient was utilized to examine potential associations. A threshold of P < 0.05 was established to determine statistical significance in the results obtained.

RESULTS

Differential metabolites associated with GC

Figure 1A illustrates the study design of our prospective case-control investigation. The PLS-DA score plot reveals a clear separation in metabolic profiles between patients with GC and healthy controls (Figure 1B). As depicted in Figure 1C, univariate analysis identified 168 metabolites with significant alterations (P < 0.05), comprising 111 up-regulated and 55 down-regulated metabolites. These significantly different metabolites were further analyzed using Kyoto Encyclopedia of Genes and Genomes pathway analysis via Metabo Analyst, with the top 10 pathways presented in Figure 1D. Based on an FDR value of less than 0.05 and a variable importance projection score greater than 1, the top 10 up-regulated differential metabolites were presented in Table 1. Subsequently, these differential metabolites were further validated using serum samples from 50 patients with GC and 50 healthy controls. As illustrated in Figure 1E, among the top 10 up-regulated differential metabolites, the serum levels of fenpiclonil (t = -4.920, P < 0.001), methyclothiazide (t = -7.369, P < 0.001), 5-hydroxyindoleacetate (t = -5.729, P < 0.001), 3-pyridinecarboxylic acid (t = -3.825, P < 0.001), guanabenz (t = -4.107, P < 0.001), 2,2-dichloro-n-(3-chloro-1,4-dioxo-2-naphthyl) acetamide (t = -3.673, P < 0.001), epigallocatechin gallate (t = -7.749, P < 0.001), and dimethenamid (t = -13.988, P < 0.001) were found to be significantly elevated in the cohort of 50 patients diagnosed with GC compared to the 50 healthy control subjects. No significant differences were observed between the groups for other metabolites, including benzonitrile (t = 0.270, P = 0.788) and methylene blue (t = 0.143, P = 0.887).

Figure 1 Analyses of serum metabolic profiling in patients with gastric carcinoma and healthy controls. ^aP < 0.01. A: Schematic of the study design; B: Partial least squares discrimination analysis score plots show discrimination between the patients with gastric carcinoma (GC) and healthy controls; C: Volcano map of differentially metabolites between the patients with GC and healthy controls; D: Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis of significantly different metabolites; E: Comparative analysis of the relative levels of the top 10 up-regulated differential metabolites in the serum of 50 patients diagnosed with GC vs 50 healthy control subjects. NS: No significance; GC: Gastric carcinoma; ROC: Receiver operating characteristic; PLS-DA: Partial least squares discrimination analysis; KEGG: Kyoto Encyclopedia of Genes and Genomes.

Table 1 Detailed information of top 10 up-regulated differential metabolites.

Metabolites	m/z	RT	P value	FDR	log2FC	VIP
Fenpiclonil	474.97	3.834	1.52E-07	0.00173	1.395	2.690
Methyclothiazide	359.96	6.578	1.03E-06	0.00252	1.308	1.408
5-hydroxyindoleacetate	192.02	6.579	1.92E-06	0.00252	1.578	2.216
3-pyridinecarboxylic acid	294.06	4.380	2.83E-06	0.00252	1.416	1.009
Guanabenz	214.00	6.587	3.56E-06	0.00252	1.488	2.214
2,2-dichloro-n-(3-chloro-1,4-dioxo-2-naphthyl) acetamide	319.97	6.586	4.59E-06	0.00274	1.382	1.639
Benzonitrile	346.99	3.798	7.38E-06	0.00338	1.998	4.299
Epigallocatechin gallate	459.00	3.798	9.76E-06	0.00338	1.545	1.778
Dimethenamid	244.08	4.327	1.01E-05	0.00338	1.549	2.517
Methylene blue	241.00	3.796	1.08E-05	0.00338	0.984	2.417

RT: Retention time; m/z: Mass-to-charge ratio; FDR: False discovery rate; FC: Fold change; VIP: Variable importance projection.

Diagnostic value of differential metabolites in discriminating patients with GC from healthy controls

The diagnostic accuracy of these eight differential metabolites in distinguishing patients with GC from healthy controls was assessed using ROC curve analysis (Figure 2). The results indicated that the AUC values of fenpiclonil, methyclothiazide, 5-hydroxyindoleacetate, 3-pyridinecarboxylic acid, guanabenz, 2,2-dichloro-n-(3-chloro-1,4-dioxo-2-naphthyl) acetamide, epigallocatechin gallate, and dimethenamid were 0.746 [95% confidence interval (CI): 0.651-0.841, P < 0.001], 0.833 (95%CI: 0.756-0.909, P < 0.001), 0.783 (95%CI: 0.695-0.871, P < 0.001), 0.698 (95%CI: 0.594-0.802, P = 0.001), 0.712 (95%CI: 0.612-0.812, P < 0.001), 0.701 (95%CI: 0.598-0.804, P = 0.001), 0.864 (95%CI: 0.793-0.935, P < 0.001), and 0.989 (95%CI: 0.000-1.000, P < 0.001), respectively. An AUC value exceeding 0.7 is generally considered suitable for clinical application. Among these differential metabolites, the AUCs of fenpiclonil, methyclothiazide, 5-hydroxyindoleacetate, guanabenz, 2,2-dichloro-n-(3-chloro-1,4-dioxo-2-naphthyl) acetamide, epigallocatechin gallate, and dimethenamid were greater than 0.7, suggesting that these seven metabolites possess substantial diagnostic value in distinguishing patients with GC from healthy individuals.

Figure 2 Receiver operating characteristic curves of differential metabolites in identification of patients with gastric carcinoma from healthy volunteers. A: Receiver operating characteristic (ROC) curve of fenpiclonil; B: ROC curve of methyclothiazide. C: ROC curve of 5-hydroxyindoleacetate; D: ROC curve of 3-pyridinecarboxylic acid; E: ROC curve of guanabenz; F: ROC curve of 2,2-dichloro-n-(3-chloro-1,4-dioxo-2-naphthyl) acetamide; G: ROC curve of epigallocatechin gallate; H: ROC curve of dimethenamid. AUC: Area under the curve.

Relationship between the serum levels of differential metabolites and the disease severity in patients with GC

The tumor-node-metastasis (TNM) staging system is globally recognized as the standard for evaluating the progression and spread of GC. As shown in Figure 3, a correlation diagram illustrates the associations between the serum levels of differentially expressed metabolites and the TNM staging in GC patients. It was observed that the serum levels of methyclothiazide (r = 0.615, P < 0.001), epigallocatechin gallate (r = 0.482, P = 0.004), and dimethenamid (r = 0.634, P < 0.001) exhibited a positive correlation with the T stage. Moreover, the serum levels of methyclothiazide (r = 0.438, P = 0.008) and epigallocatechin gallate (r = 0.383, P = 0.023) were found to be positively related to N stage.

Figure 3 Correlation between the serum levels of differential metabolites and the disease severity in patients with gastric carcinoma. ×: No significance.

DISCUSSION

In the current study, untargeted metabolomics was employed to investigate alterations in the serum metabolic profiles of patients with GC, with the objectives of identifying potential serum biomarkers for GC diagnosis and evaluating the correlation between serum levels of differential metabolites and disease severity in GC patients. The study yielded four principal findings. Firstly, a substantial alteration in the metabolic profile was observed, characterized by the up-regulation of 111 metabolites and the down-regulation of 55 metabolites in GC patients compared to healthy controls. Secondly, the serum concentrations of eight metabolites including fenpiclonil, methyclothiazide, 5-hydroxyindoleacetate, 3-pyridinecarboxylic acid, guanabenz, 2,2-dichloro-N-(3-chloro-1,4-dioxo-2-naphthyl) acetamide, epigallocatechin gallate, and dimethenamid were confirmed to be significantly elevated in a cohort of 50 patients diagnosed with GC compared to 50 healthy control subjects. Thirdly, the AUC values for fenpiclonil, methyclothiazide, 5-hydroxyindoleacetate, guanabenz, 2,2-dichloro-n-(3-chloro-1,4-dioxo-2-naphthyl) acetamide, epigallocatechin gallate, and dimethenamid exceeded 0.7, indicating their effectiveness in distinguishing patients with GC from healthy controls. Fourthly, the serum concentrations of methyclothiazide and epigallocatechin gallate demonstrated a positive correlation with both T stage and N stage, while the serum concentrations of dimethenamid showed a positive correlation with T stage in patients with GC.

Metabolic reprogramming constitutes a fundamental characteristic of cancer, wherein cancer cells autonomously modify various metabolic pathways to enhance their survival and proliferation[20]. Increasing evidence suggests that altered cellular metabolism underpins tumorigenesis, progression, metastasis, and resistance to therapeutic interventions[21]. Recent advancements in scientific research have intensified the focus on the interplay between metabolism and tumor biology[22]. Neoplastic cells adjust their metabolic processes to facilitate the biological mechanisms underlying neoplasia, with particular metabolic pathways playing a direct role in cellular transformation and tumor progression[23]. The inhibition or restoration of altered metabolic pathways has emerged as a promising therapeutic strategy[24,25]. Metabolomics aims to leverage the metabolic signatures associated with cancer to evaluate disease risk and facilitate early detection, diagnosis of specific disease subtypes, and monitoring of treatment efficacy. This approach yields critical insights into the metabolic profiles of distinct tumors and aids in the identification of key biomarkers for cancer diagnosis and prognosis. Additionally, blood-based markers offer several advantages compared to tissue-based markers. The prevailing standard for diagnosing GC involves the histological examination of tumor tissue obtained through biopsies or surgical specimens, all of which are invasive procedures that carry a risk of complications. In contrast, blood-based tests are minimally invasive, causing only limited discomfort and posing no significant complications. These tests can be easily repeated, facilitating close monitoring of the disease to assess treatment response or detect recurrence at an early stage. Thus, in this study, untargeted metabolomics was employed to investigate alterations in the serum metabolic profiles of patients with GC, with the objective of identifying potential serum biomarkers for GC diagnosis. The results demonstrated a notable modification in the metabolic profile, with 111 metabolites being up-regulated and 55 down-regulated in GC patients compared to healthy controls. Similarly, another study conducted comprehensive metabolite profiling in plasma samples, utilizing untargeted metabolomics to identify altered metabolites in a cohort of 15 patients with early-stage GC and 15 healthy controls[26]. A total of 19 metabolites were identified as dysregulated between early GC patients and healthy controls[26]. Moreover, Cao et al[27] developed a prognostic prediction model for GC patients, utilizing a metabolism-associated signature to elucidate the distinctive role of metabolites in GC prognosis. In total, 23 metabolites associated with GC survival were identified[27]. These altered metabolites implicated in early GC offer novel insights into the pathogenesis of GC. The modulation and regulation of these metabolic pathways, or the development of targeted metabolic therapeutics, hold promise for arresting cancer progression and devising potential strategies to address the challenges associated with cancer advancement. However, substantial clinical data are necessary to substantiate these findings.

Biomarker discovery studies employing metabolomics encounter significant challenges in terms of reproducibility, likely attributable to signal drifts in cross-batch or cross-platform analyses and the limited integration of data from diverse laboratory samples[28,29]. Consequently, the top ten up-regulated metabolites identified were further validated using serum samples from an expanded cohort comprising 50 patients with GC and 50 healthy controls. The findings indicated that the serum concentrations of fenpiclonil, methyclothiazide, 5-hydroxyindoleacetate, 3-pyridinecarboxylic acid, guanabenz, 2,2-dichloro-N-(3-chloro-1,4-dioxo-2-naphthyl) acetamide, epigallocatechin gallate, and dimethenamid were significantly elevated in the cohort of patients diagnosed with GC in comparison to the healthy control group. Subsequently, the diagnostic accuracy of these 8 differential metabolites in differentiating patients with GC from healthy controls was evaluated through ROC curve analysis. An AUC value exceeding 0.7 is typically deemed appropriate for clinical application[30,31]. Of these differential metabolites, fenpiclonil, methyclothiazide, 5-hydroxyindoleacetate, guanabenz, 2,2-dichloro-n-(3-chloro-1,4-dioxo-2-naphthyl) acetamide, epigallocatechin gallate, and dimethenamid exhibited AUC values greater than 0.7, indicating that these 7 metabolites have significant diagnostic potential in distinguishing patients with GC from healthy individuals.

TNM represents a globally recognized classification system used to characterize the spread of cancer[32]. To investigate the correlation between serum levels of differential metabolites and disease severity in GC patients, we applied Spearman’s rank correlation coefficient to analyze the association between these eight differential metabolites and the TNM staging of GC patients. The results indicated that serum concentrations of methyclothiazide and epigallocatechin gallate exhibited a positive correlation with both T stage and N stage, whereas serum concentrations of dimethenamid were positively correlated with T stage in patients with GC. Given the cross-sectional nature of this study's design, additional research is required to clarify the causal link between these metabolites and the TNM staging in GC patients.

Recent evidence underscores novel mechanisms underlying metabolite-driven pathogenesis in GC. One-carbon metabolism, which plays a pivotal role in DNA and histone methylation, is found to be dysregulated in GC[33,34]. Alterations in metabolites within this pathway lead to the hypermethylation of tumor suppressor gene promoters, resulting in the silencing of these genes and allowing cancer cells to evade growth control and apoptosis. Furthermore, the dysregulation of one-carbon metabolism contributes to the accumulation of reactive oxygen species (ROS), which inflict damage on cellular components and promote genomic instability, thereby facilitating tumor progression[35,36]. Additionally, metabolites derived from gut microbiota have a significant impact on GC development. In patients with GC, dysbiosis of the gut microbiota alters the production of short-chain fatty acids, such as butyrate. Butyrate, known as a histone deacetylase inhibitor, typically induces cell cycle arrest and apoptosis. However, reduced levels of butyrate due to changes in the microbiota disrupt epigenetic regulation, thereby promoting tumorigenesis[37]. Nucleotide metabolism is also disrupted in GC, with altered levels of purine and pyrimidine metabolites affecting DNA and RNA synthesis, which are crucial for the proliferation of cancer cells. Elevated levels of purine metabolites augment the activity of enzymes involved in the nucleotide salvage pathway, thereby enabling cancer cells to efficiently synthesize nucleic acids necessary for sustained proliferation[38,39]. These observations highlight the intricate relationship between metabolic alterations and the pathogenesis of GC, suggesting potential novel targets for therapeutic intervention.

Epigallocatechin gallate, major constituent of green tea, possesses antioxidant, antiviral, and anticancer activities[40]. The exact mechanism of elevated epigallocatechin gallate in the serum of GC patients still needs to be fully elucidated. Several factors may contribute to the elevated serum levels of epigallocatechin gallate in GC patients. Firstly, the physiological stress response to cancer may elevate epigallocatechin gallate levels as a protective mechanism against oxidative stress and inflammation induced by tumor cells. Secondly, the aberrant metabolic processes in GC cells could interfere with the normal absorption, distribution, metabolism, and excretion of epigallocatechin gallate, resulting in its accumulation in the serum. Thirdly, an increased dietary intake of epigallocatechin gallate-rich foods, such as green tea, by patients seeking potential anti-cancer effects may contribute to elevated serum levels. Lastly, the administration of medications or supplements containing epigallocatechin gallate for therapeutic or health-enhancing purposes may directly lead to increased concentrations of epigallocatechin gallate in the bloodstream.

The increased serum concentrations of dimethenamid observed in patients with GC may be attributable to environmental exposure. As dimethenamid is a commonly utilized herbicide, individuals with heightened exposure through contaminated food, water, or air may exhibit elevated serum levels. The relationship between dimethenamid and GC represents a burgeoning field of research characterized by limited yet compelling evidence. Although dimethenamid is predominantly acknowledged for its low acute toxicity in mammals, concerns regarding its long-term effects, particularly its potential carcinogenicity, remain prevalent. A study examining pesticide exposures, including chloroacetamide herbicides, provided suggestive evidence linking certain herbicides to an increased risk of GC; however, dimethenamid was not specifically identified in their analysis[41]. Mechanistically, certain chloroacetamide compounds may induce oxidative stress or DNA damage, which could play a role in carcinogenesis. Nonetheless, direct evidence supporting a causal relationship between dimethenamid and cancer remains insufficient.

The increased serum concentrations of methyclothiazide observed in patients with GC may be attributed to a variety of factors. GC frequently compromises hepatic and renal functions, which are essential for drug metabolism and excretion, thereby leading to decreased elimination of methyclothiazide and its subsequent accumulation. Moreover, the altered pathophysiological state associated with cancer can disrupt normal pharmacokinetic processes, such as drug-protein binding, thereby affecting the distribution and clearance of methyclothiazide. Nonetheless, definitive evidence supporting these associations remains insufficient and necessitates further investigation.

These newly identified biomarkers for GC hold significant potential for seamless integration into current clinical workflows: (1) Integration into existing clinical practice. These identified biomarkers for GC can be incorporated into multi-step screening programs. For high-risk populations, such as those with Helicobacter pylori infection or a family history of GC, biomarker testing can be used as a pre-endoscopic screening tool. Additionally, during the diagnostic phase, these biomarkers can serve as an adjunct to endoscopic biopsy results, especially in cases where histological diagnosis is ambiguous. In the treatment and follow-up stages, biomarker levels can be monitored regularly to predict treatment response and detect early signs of recurrence; and (2) Translation strategies. To facilitate clinical translation of these findings, two key strategies are proposed. Firstly, large-scale multicenter validation studies should be conducted to verify biomarker performance across diverse patient populations and establish standardized testing protocols. Secondly, development of point-of-care diagnostic systems for these biomarkers would enable rapid clinical testing, thereby supporting timely therapeutic interventions.

Our study has several limitations. Firstly, a major limitation of this study is its reliance on a single Chinese hospital and population. This narrow sampling frame likely introduces selection bias and severely restricts the generalizability of our findings, as the identified GC-related biomarkers and metabolic patterns may not be applicable to other ethnic groups, geographical regions, or healthcare settings. Further validation through multicenter, cross-ethnic prospective studies is urgently needed. Secondly, the relatively small number of samples (6 GC patients vs 6 controls) in the initial untargeted screening phase may give rise to concerns regarding generalizability and the potential for overfitting. A larger cohort would strengthen the robustness of the findings. Thirdly, the single time-point sampling represents a methodological constraint, precluding analysis of trajectory patterns or cause-effect relationships; multi-timepoint investigations would strengthen these observations. Fourthly, a significant limitation of this study is the exclusive reliance on a single LC-MS/MS platform without orthogonal validation. This lack of validation using alternative analytical platforms or techniques means that platform-specific biases, calibration differences, or instrument-related variations may have influenced the results. Without cross-platform validation, the generalizability and robustness of the identified GC-associated metabolite profiles could be compromised, potentially leading to over- or under- estimation of metabolite abundances and inaccurate biological interpretations. Fifthly, a key limitation of this study is the omission of lifestyle factors (including diet, smoking, and physical activity) and comorbidities (such as diabetes and obesity). These factors can substantially modify the serum metabolome, potentially confounding the identification of GC-specific metabolite signatures and limiting the generalizability of our finding.

CONCLUSION

In conclusion, the findings based on this study demonstrated significant discrepancy in circulating blood metabolome profiles in patients with GC when compared with healthy controls. These differential metabolites were capable to discriminate patients with GC from healthy controls with good performance and associated with the disease severity.