Alterations in gut microbiome and fecal metabolome in functional dyspepsia patients: insights into pathophysiological mechanisms Authors: Honghua Zhou, Ronghua Lu, You Lv, Yanbing Ding Published: 2025 ABSTRACT AbstractBackgroundGastrointestinal microbiota dysregulation is recognized as a key factor in the pathogenesis of functional dyspepsia (FD). Previous investigations have focused predominantly on the microbiota associated with the oral, gastric, and duodenal mucosa. However, the intestinal microbiome and fecal metabolome in FD patients remain poorly characterized. This study aims to evaluate their alterations, thereby providing novel insights into the pathophysiological mechanisms of FD.MethodsFecal samples from 30 FD patients and 19 healthy controls (HCs) were subjected to 16S rRNA sequencing and metabolomics to characterize microbial–metabolic profiles, followed by Spearman correlation analysis to explore the associations between differentially abundant taxa and metabolites.ResultsCompared with the HCs, the FD group presented greater microbial richness but comparable diversity, along with increased abundances of Akkermansia and Ruminococcus_gnavus_group and decreased levels of Dorea, Collinsella, and Agathobacter (p < 0.05). AUC analysis revealed Akkermansia as a potential biomarker for FD (AUC = 0.848, 95% CI: 0.708–0.988). Furthermore, the levels of the fecal metabolites phosphatidylglycerol (PG) and phosphatidylcholine (PC) were significantly reduced in the FD group (p < 0.05). KEGG enrichment analysis revealed that these metabolites were involved in glycerophospholipid metabolism. Spearman correlation analysis revealed negative associations between Ruminococcus_gnavus_group, Fusobacterium, Erysipelatoclostridium, and PG/PC (p < 0.05).ConclusionsThe FD group presented concurrent alterations in the gut microbiome and fecal metabolites. Specifically, gut microbiota-derived metabolites, such as PG and PC, may disrupt host glycerophospholipid metabolism, contributing to FD pathogenesis. Given that this is an exploratory study with a small sample size and cross-sectional design, these results require validation in larger, longitudinal cohorts and should be interpreted with caution.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40001-025-03743-5.Keywords: Functional dyspepsia, Gut microbiome, Fecal metabolome, Glycerophospholipid metabolism, Pathophysiological mechanisms Introduction Functional dyspepsia (FD), a disorder of disrupted gut‒brain interactions, is considered one of the most prevalent functional gastrointestinal conditions and can lead to substantial socioeconomic burdens by increasing health care costs and adversely affecting individual mental health [1]. In accordance with the Rome IV criteria [2], FD is characterized by the presence of at least one of the following specific symptoms: epigastric pain, epigastric burning sensation, postprandial fullness, and early satiety. Importantly, these symptoms should not be attributable to systemic, organic, or metabolic ailments. The clinical manifestations must have persisted for at least 6 months before the confirmed diagnosis and must have met the above criteria in the past 3 months. Clinically, FD is categorized into three primary forms: epigastric pain syndrome (EPS), postprandial distress syndrome (PDS), and the EPS‒PDS overlap group, delineated by specific symptom profiles [3]. The global prevalence of FD is approximately 8.4%, with a higher prevalence in developing countries and a significantly greater female predominance [4, 5]. Current evidence indicates that the pathogenesis of FD is governed by multiple factors, including gastrointestinal motility disorders, visceral hypersensitivity, low-grade duodenal inflammation, Helicobacter pylori (H. pylori) infection, gut‒brain axis dysfunction, dietary influences, genetic susceptibility, and psychosocial stressors [3, 6]. These factors synergistically contribute to the heterogeneous clinical phenotypes of FD, yet the underlying mechanisms remain unclear. Notably, conventional FD therapies, such as the use of prokinetics to improve motility or antidepressants to alleviate visceral sensitivity, primarily target symptomatic relief. However, their efficacy is often modest and transient, highlighting the urgent need to identify novel pathogenic pathways and therapeutic approaches. The gastrointestinal microbiota, a key regulator of intestinal barrier integrity, immune homeostasis, and metabolic balance, has emerged as a pivotal contributor to human diseases, including inflammatory bowel disease and metabolic syndrome [7]. In the context of FD, accumulating evidence implicates gastrointestinal dysbiosis as a key driver, with studies reporting altered microbiota compositions in the oral cavity, gastric mucosa, and duodenum of FD patients [8–10]. Additionally, the fecal metabolome in FD remains underexplored, with limited existing research [11, 12]. Critical knowledge gaps persist, particularly regarding the role of the distal intestinal microbiome in FD, the mechanistic links between microbiota-derived metabolites and FD symptoms, and the development of microbiota-targeted therapeutic strategies. Given the limitations of current treatments for FD in targeting underlying etiological mechanisms and considering that the gastrointestinal microbiota is a modifiable determinant with systemic effects, investigating microbiota‒metabolite‒host interactions in FD is crucial. Fecal samples offer unique advantages for FD research, including noninvasive collection, clinical accessibility, and the ability to robustly reflect the characteristics of the distal intestinal microbiota and metabolites. Therefore, the present study aims to preliminarily explore FD-related pathophysiological pathways from a distinct perspective by analyzing alterations in the fecal microbiota and metabolites of FD patients and their interrelationships, thereby providing experimental evidence for an in-depth understanding of disease pathogenesis. Materials and methods Enrollment of participants and formation of the study cohort We recruited a total of 30 FD patients from the outpatient Department of Gastroenterology, Affiliated Hospital of Yangzhou University, from April 2024 to October 2024. All patients fulfilled the following inclusion criteria: (i) aged 18 to 70 years, without sex restrictions; (ii) met the Rome IV diagnostic criteria; (iii) had no obvious abnormalities on upper gastrointestinal endoscopy, including esophagitis, gastric ulcers, and duodenal bulb ulcers; and (iv) showed negative results on the 13C-urea breath test (13C‒UBT). During the same period, 20 healthy controls who had normal gastroscopy results and negative results on the 13C‒UBT were recruited from the Health Screening Center. The HCs were matched to the FD group for age, sex, and body mass index (BMI). The exclusion criteria for both groups were as follows: (i) had a history of major abdominal surgery; (ii) were pregnant or lactating; (iii) had neuropsychiatric disease or severe diseases of the liver, kidney, or respiratory system; (iv) had a history of taking proton pump inhibitors (PPIs), antianxiety medications, antidepressants, gastrointestinal prokinetics, antibiotics or probiotics within the month preceding stool sampling; (v) had inflammatory bowel disease, including Crohn's disease, and ulcerative colitis; and (vi) were unwilling to participate in this study. Fecal sample collection On the morning after an overnight fast (more than 8 h), the research subjects used a sterile spoon to collect approximately 10 g of feces during their first bowel movement of the day and divided the fecal sample into two cryostorage tubes. The samples were subsequently preserved at − 80 °C in a refrigerator. Importantly, throughout the process, strict aseptic procedures were followed to ensure the reliability and accuracy of the assay results derived from the stool samples. DNA extraction and sequencing Genomic DNA was extracted from the fecal samples via the Fecal Genome DNA Extraction Kit (AU46111-96; Beijing Biotech Co., Ltd., China). The DNA concentration was quantified with a Qubit fluorometer (Invitrogen, USA). PCR amplification of the DNA samples was performed via specific primers targeting the V3‒V4 hypervariable region of the 16S rRNA gene: 341 F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) [13]. The PCR products were subsequently purified with AMPure XT magnetic beads (Beckman Coulter Genomics, Danvers, MA, USA) and quantified via a Qubit fluorometer. Sequencing was conducted on an Illumina NovaSeq 6000 platform (PE250 mode) by LC-Bio Technology Co., Ltd. (Hangzhou, Zhejiang, China). Metabolite extraction and LC‒MS/MS analysis A 50 mg sample was extracted with 0.5 mL of precooled 80% methanol. The supernatant obtained after centrifugation was transferred to a new tube for vacuum drying, followed by reconstitution in 100 µL of 80% methanol for subsequent liquid chromatography‒tandem mass spectrometry (LC‒MS/MS) analysis. Quality control (QC) samples were prepared by pooling aliquots of the supernatants from all the samples. The UltiMate 3000 UPLC system (Thermo Fisher Scientific, Bremen, Germany) in conjunction with the ACQUITY UPLC T3 column (100 mm × 2.1 mm, 1.8 µm, Waters, Milford, USA) enabled rapid and efficient separation of sample components via ultrahigh-performance liquid chromatography. The mobile phase comprised solvent A (aqueous solution of 5 mM ammonium acetate and 5 mM acetic acid) and solvent B (acetonitrile). The flow rate was 0.3 mL/min. The elution gradient was programmed as follows: 2% B (0–0.8 min), 2–70% B (0.8–2.8 min), 70–90% B (2.8–5.6 min), 90–100% B (5.6–6.4 min), 100% B (6.4–8.0 min), 100–2% B (8.0–8.1 min), and 2% B (8.1–10 min). Metabolite detection was performed using a Q-Exactive mass spectrometer (Thermo Scientific), which was operated in both positive and negative ion modes. Precursor ions were scanned at a resolution of 70,000 over m/z 70–1050 with an automatic gain control (AGC) target of 3 × 106 and a maximum injection time of 100 ms. In data-dependent acquisition (DDA) mode, a top 3 strategy was applied for fragment ion collection at 17,500 resolution, with an AGC target of 1 × 105 and an 80 ms injection time. We conducted QC detection on every ten samples. Bioinformatics analysis For microbiome analysis, taxonomic annotation was performed using the feature-classifier plugin in QIIME2. The sequences were annotated against the SILVA and NT-16S databases. The bacterial community alpha and beta diversities were calculated using QIIME2. Principal coordinate analysis (PCoA) was performed using the unweighted UniFrac distance, and the p value was determined using an analysis of similarities (ANOSIM) permutation test. Linear discriminant analysis (LDA) effect size (LEfSe) analysis was conducted using LEfSe software (LDA > 3.0, p < 0.05). The functional prediction of 16S rRNA sequencing data was performed using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2), and metabolic pathway annotation and intergroup difference analysis were performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Other graphs were generated using R (version 3.4.4). For metabolomics analysis, raw LC–MS/MS data were processed using R toolboxes to generate a three-dimensional matrix. Metabolites were annotated and validated by matching accurate mass data to the KEGG and Human Metabolome Database (HMDB) databases (mass deviation < 10 ppm) and further verified with a custom-built fragment spectrum library. Orthogonal projections to latent structures discriminant analysis (OPLS‒DA) and variable importance in the projection (VIP) value calculations were completed using the ropls package. Metabolites with significant differences were screened using a combination of t-tests (p < 0.05, FC > 1.2) and VIP thresholds (VIP > 1). KEGG pathway enrichment analysis was conducted on the differentially abundant metabolites using R (version 4.0.0). The Benjamini–Hochberg (BH) method was applied to calculate false discovery rate (FDR)-corrected p values (Q values). Statistical analyses The sample size was estimated using G-power software (version 3.1). Normally distributed continuous variables were expressed as the means (± SD). The enumeration data were presented as percentages. Student's t test or the Mann–Whitney U test was used to compare two groups. For the correlation analysis, Spearman's rank correlation coefficient was used. All the statistical analyses in this research were conducted using SPSS software (version 26.0) and R software on the Mac OS X system. Statistical significance was considered only when the p value was less than 0.05. Results Demographic characteristics The study enrolled 30 FD patients and 20 healthy controls and analyzed 50 samples via 16S rRNA gene sequencing. One subject was excluded from the subsequent analysis because of a low number of sequencing reads. Ultimately, 49 subjects (FD = 30 and HCs = 19) remained for the following statistical analyses (Fig. 1). The age profile of the 30 FD patients was characterized by a mean age of 49.47 ± 12.11 years, and 18 females (60.0%) were included. The HCs included 19 subjects, with a mean age of 44.95 ± 12.57 years, of whom 10 were female (52.6%). No significant intergroup differences were observed in age, sex, or BMI (p > 0.05). On the basis of these comparable characteristics, we investigated fecal samples from the two groups via 16S rRNA sequencing and untargeted metabolomic analyses of their intestinal microbiota and metabolites. The baseline characteristics of all the subjects, including a series of clinical features of the FD group, are presented in Table 1, and the detailed information is given in Additional file 1: Table S1. Enrollment and analysis workflow of the FD group and HCs. FD, functional dyspepsia; HCs, healthy controls Table 1Baseline characteristics of the FD group and HCsCharacteristicsFD (n = 30)HCs (n = 19)PSex (male/female)12/189/100.498aAge (years)49.47 ± 12.1144.95 ± 12.570.216bBMI (kg/m2)24.49 ± 2.3425.31 ± 2.400.242bEducation level0.962a• High school or less14 (46.7%)9 (47.4%)• College and above16 (53.3%)10 (52.6%)Place of residence0.535a• Urban area20 (66.7%)11 (57.9%)• Rural area10 (33.3%)8 (42.1%)Cigarette smoking0.990a• Current or former11 (36.7%)7 (36.8%)• Never or almost never19 (63.3%)12 (63.2%)Alcohol drinking0.411a• Current or former13 (43.3%)6 (31.6%)• Never or almost never17 (56.7%)13 (68.4%)Symptom duration/• 6–12 months6 (20.0%)/• More than 12 months24 (80.0%)/Type of FD/• PDS18 (60.0%)/• EPS3 (10.0%)/• PDS + EPS9 (30.0%)/FD: functional dyspepsia; HCs: healthy controls; BMI: body mass index; EPS: epigastric pain syndrome; PDS: postprandial distress syndromeaChi-square testbt-test. P < 0.05 was considered statistically significant Baseline characteristics of the FD group and HCs FD: functional dyspepsia; HCs: healthy controls; BMI: body mass index; EPS: epigastric pain syndrome; PDS: postprandial distress syndrome aChi-square test bt-test. P < 0.05 was considered statistically significant Overview of the 16S rRNA sequencing data Among the 49 fecal samples, 309,533 high-quality sequences were obtained for analysis. The intestinal microbiota within all the samples was categorized into 27 phyla, 77 classes, 171 orders, 300 families, 721 genera, and 1104 species according to the phylogenetic taxonomic level. Figure 2A presents a Venn diagram that analyzes the overlap of the fecal microbiota composition between the FD group and HCs, which provides a better evaluation of the shared richness. There were 1418 amplicon sequence variants (ASVs) shared between the two groups, while the FD group and HCs had 2748 and 1615 unique ASVs, respectively (Additional file 2). Gut microbiota composition and diversity comparisons between the two groups. A The numbers of ASVs that were shared and unique between the two groups, indicating partial overlap and distinct components in their microbial compositions. B The smooth rarefaction curves of each group suggest that the sequencing depth was adequate for a comprehensive analysis of microbial diversity. C, D α-Diversity of the gut microbiota in the two groups, with higher Chao1 and Shannon indices in the FD group than in the HCs. The Chao1 index reached statistical significance (*** p < 0.001). E, F PCoA of unweighted UniFrac data revealed distinct microbial community compositions between the two groups (Anosim, R = 0.5474, p = 0.001). ASVs, amplicon sequence variants; FD, functional dyspepsia; HCs, healthy controls. PCoA, principal coordinate analysis; Anosim, analysis of similarities Through the simulation of the resampling process, the ASV sequences of each sample were randomly selected at various sequencing depths. The alpha-diversity indices were subsequently calculated to construct rarefaction curves. The curves in each group approached saturation with increasing sequencing output, suggesting that the sequencing depth was adequate (Fig. 2B). Thus, the results could comprehensively represent the fecal microbiome. Different indices were used to evaluate alpha-diversity (Additional file 1: Table S2). The species richness (Chao1 index) in the FD group was significantly greater than that in the HC group (Fig. 2C, p< 0.001). The Shannon diversity index was used to quantify the species diversity in a community. As depicted in Fig. 2D, compared with the HCs, the FD group presented greater microbial diversity; however, the difference was not statistically significant (p > 0.05). To assess between-group diversity, beta-diversity was calculated with unweighted UniFrac distances and subjected to PCoA (Fig. 2E). The results revealed significant intergroup clustering. Similarly, as a nonparametric test, ANOSIM revealed that the differences between two groups were significantly greater than those within individual groups, validating the grouping strategy (Fig. 2F, R = 0.5474, p = 0.001). Differences in species abundance between groups The exploration of the bacterial communities and their relative abundance in the two groups was carried out at diverse taxonomic levels. As depicted in Fig. 3A (Additional file 3), the top 20 most abundant bacterial phyla are presented, with the main constituents being Firmicutes, Bacteroidota, Actinobacteriota, Proteobacteria, Verrucomicrobiota, and Desulfobacterota. The major phyla in the intestinal microbiota of the FD group were similar to those of the HCs: Firmicutes (59.07% vs. 58.42%), Bacteroidota (14.15% vs. 13.38%), and Proteobacteria (11.18% vs. 8.36%). However, the proportions of Actinobacteriota were lower in the FD group than in the HC group. The four phyla constituted 96.55% and 95.84% of the entire microbiota in the FD group and HC group, respectively. Furthermore, at the phylum level, we found that among the top 5 microorganisms, the abundances of Verrucomicrobiota, Fusobacteriota, Cyanobacteriota, Campylobacterota, and Deferribacterota in the intestinal microbiota of the FD group were markedly greater than those in the HCs (Fig. 3B, p < 0.05). Gut microbiota analysis and differential functional prediction between the two groups. A Distribution of the gut microbiota at the phylum level in 49 samples. B Distribution of the gut microbiota with significant differences in relative abundance between the two groups at the phylum level (p < 0.05). C Distribution of the gut microbiota at the genus level in 49 samples. D Distribution of the gut microbiota with significant differences in relative abundance between the two groups at the genus level (p < 0.05). E LEfSe analysis revealed the taxa with the greatest differences in abundance between the two groups at the phylum, genus, and family levels (LDA > 3.0, p < 0.05). F The area under the ROC curve (AUC) for Akkermansia as a predictor is depicted, with the 95% CI indicated. LEfSe, linear discriminant analysis effect size; LDA, linear discriminant analysis; ROC: receiver operating characteristic; AUC, area under the curve; CI, confidence interval At the genus level (Fig. 3C, Additional file 4), 20 predominant bacterial genera were identified, including Bifidobacterium (9.76% vs.13.56%), Bacteroides (7.21% vs. 5.31%), Faecalibacterium (6.36% vs. 6.13%), and Phascolarctobacterium (5.15% vs. 5.23%). Importantly, among the top 10 genera showing significant intergroup differences, Ruminococcus_gnavus_group, Akkermansia, Clostridium, Romboutsia, Fusobacterium, and Lachnospiraceae_NK4A136_group had increased abundances in the FD group (Fig. 3D, p < 0.05). The relative abundances of Agathobacter, Collinsella, Erysipelotrichaceae_UCG-003, and Dorea markedly increased in the HCs (p < 0.05). To identify characteristic discriminators between the two groups, an annotated cladogram of the differentially abundant taxa was generated (Additional file 1: Fig. S1). A total of 35 discriminative features were identified via LEfSe analysis (LDA > 3, p < 0.05) and classified into 3 phyla, 12 families, and 20 genera (Additional file 5, Fig. 3E). Compared with those in the HCs, the significantly enriched phyla included Verrucomicrobiota, Fusobacteriota, and Cyanobacteria. At the family level, Peptostreptococcaceae, Clostridiaceae, and Akkermansiaceae were markedly enriched. Additionally, the prominently enriched genera included Ruminococcus_gnavus_group, Clostridium, Lachnoclostridium, and Akkermansia. The key distinguishing units in the fecal microbiota of HCs were Coriobacteriaceae (family), Clostridium_sensu_stricto_1 (genus), Dorea (genus), Erysipelotrichaceae_UCG−003 (genus), Collinsella (genus), and Agathobacter (genus). Receiver operating characteristic (ROC) curve analysis with calculation of the area under the curve (AUC) was conducted to determine whether the gut microbiota could function as a potential biomarker for FD diagnosis and prediction. Specifically, we used the relative abundances of these candidate genera. ROC curve analysis revealed that the AUC value of Akkermansia was relatively high (0.848 [95% CI 0.708–0.988]) (Fig. 3F). The results suggest that classifiers based on Akkermansia abundance possess high specificity and offer potential for clinical translation. Gut microbiota functional alterations in FD patients To explore potential functional pathways associated with the pathogenesis of FD, we performed STAMP-based metabolic pathway differential analysis on 16S rRNA sequencing data using PICRUSt2 software. At levels 1 and 3 of the KEGG pathway hierarchy, PICRUSt2 predicted 2 and 30 pathways, respectively, with potential differential enrichment between the two groups (Additional file 1: Fig. S2, S3). At level 2, 11 pathways exhibited statistically significant differences in predicted enrichment (Fig. 4, p < 0.05). These predicted differential pathways included carbohydrate metabolism, folding, sorting and degradation, and signal transduction. Collectively, these inferred pathway profiles offer potential directional targets for further investigations of the functional mechanisms underlying microbiota‒metabolome interactions in FD. Visualization of the STAMP differential analysis results on the basis of the KEGG Level 2 database. STAMP, statistical analysis of metagenomic profiles; KEGG, kyoto encyclopedia of genes and genomes Differential fecal metabolite profiles between functional dyspepsia patients and HCs In our study, LC‒MS/MS analysis of all the fecal samples yielded 21,437 metabolite ions, of which 17,146 were annotated at the MS1 level and 3749 were annotated at the MS2 level (Additional files 6, 7). Additionally, the metabolites were categorized into 17 superclass classifications within the HMDB. Among these, the superclasses with the largest number of metabolite entries were lipids and lipid-like molecules (1331), organic acids and derivatives (566), organoheterocyclic compounds (374), and phenylpropanoids and polyketides (231) (Fig. 5A). The KEGG annotation revealed the top 20 metabolic pathways with the most metabolites at level 3, including metabolic pathways (601), biosynthesis of secondary metabolites (287), and glycerophospholipid metabolism (268) (Fig. 5B). Metabolomic analysis and pathway interpretation of differentially abundant metabolites between the two groups. A Statistical chart of the classification and annotation of the superclasses in the HMDB. B Top 20 Level 3 pathways ranked by metabolite count. C OPLS‒DA score plot revealing obvious separation between the two groups. D Permutation tests (200 permutations) validated the OPLS‒DA model without overfitting. HMDB, human metabolome database; OPLS‒DA, orthogonal projections to latent structures‒discriminant analysis For a preliminary visualization of differences among samples from the two groups, the OPLS‒DA model was constructed by employing a supervised dimensionality reduction technique. The two groups exhibited clear clustering, indicating that disparities in metabolites between the FD group and HCs were evident (Fig. 5C). Moreover, the permutation test on OPLS‒DA indicated that the Q2 regression line exhibited a negative intercept, and the R2 and Q2 values on the left side were consistently lower than those at the original rightmost point, which indicated that the model was stable and not overfit (intercept of Q2 = − 0.0929; Fig. 5D). We employed both multivariate and univariate statistical analyses to detect significantly differentially abundant metabolites between the two groups, which met the following criteria: fold change (FC) ≥ 1.2 or ≤ 1/1.2, p < 0.05, and VIP ≥ 1. A total of 307 significantly different metabolites were detected. Within the volcano plot, the highlighted dots in the upper-right corner (109) and upper-left corner (198) represent upregulated and downregulated metabolites, respectively (Additional file 1: Fig. S4). The box plot illustrates the top 6 differentially expressed metabolites, whose concentrations were lower in patients with FD (Additional file 8, Fig. 6). Notably, glycerophospholipid metabolites were the predominant differentially abundant metabolites, including PG [16:0/18:1 (11Z)]; PG O-14:1_15:0; PG [16:0/16:1 (9Z)]; PG (16:1/19:0); PG (TXB2/i-20:0); PC [15:0/18:2 (9Z,12Z)]; PC O-31:2; PC [18:3 (6Z,9Z,12Z)/18:1 (9Z)]; and PC [16:0/18:4 (6Z,9Z,12Z,15Z)]. These metabolites exhibited significant disparities between the FD group and HCs (p < 0.05), indicating the presence of glycerophospholipid metabolism disorders in the FD group. To assess the discriminatory potential of fecal metabolites between the FD group and HCs, we selected the top 5 differentially abundant metabolites with the smallest p values: PS [16:0/18:1 (9Z)]; PG (16:1/19:0); ST24:1; O3; T/18:3; PG O-17:0_18:2; and PG [16:0/18:1 (11Z)]. We then generated individual ROC curves and calculated the total ROC curve through the logistic regression model (Additional file 1: Fig. S5, Fig. 7A). The top 6 differentially abundant metabolites were further compared and visualized via a box plot Differentially abundant metabolite-based pathway enrichment analysis and predictive model evaluation. A Overall ROC curve of the top 5 differentially abundant metabolites (with the lowest P values) in a logistic regression model. B KEGG hierarchical bar chart showing the enrichment of differentially expressed metabolites across metabolic pathways. ROC, receiver operating characteristic; KEGG, kyoto encyclopedia of genes and genomes The differentially expressed metabolites were enriched predominantly in several metabolic pathways (Fig. 7B), including metabolic pathways, biosynthesis of secondary metabolites, glycerophospholipid metabolism, microbial metabolism in diverse environments, and biosynthesis of cofactors. The top 20 KEGG annotations of the differentially expressed metabolites identified between the two groups included glycerophospholipid metabolism, sphingolipid signaling pathway, metabolic pathways, retrograde endocannabinoid signaling, and adipocytokine signaling pathway, suggesting that these metabolic pathways may be involved in the pathogenesis or symptom presentation of FD. (Additional file 1: Fig. S6; Additional file 9). A regulatory network diagram depicting the relationships between metabolites and pathways was constructed, which visually illustrates their associations (Fig. 8). Network diagram of regulatory relationships between the top 30 differentially abundant metabolites with the lowest p values and metabolic pathways Correlation analysis of the intestinal microbiota with metabolites A correlation network was constructed to illustrate associations between 35 differential microbiota (3 phyla, 12 families, and 20 genera) and 30 differentially abundant metabolites (Spearman's correlation coefficient, |rho|> 0.50 and p < 0.05) (Additional file 1: Fig. S7, S8; Fig. 9; Additional file 10). At the genus level, FD-enriched Ruminococcus_gnavus_group, Fusobacterium, and Erysipelatoclostridium were negatively associated with PG O-17:0_18:2, PG O-14:1_15:0, and PC [18:3 (6Z,9Z,12Z)/18:1 (9Z)] (Ruminococcus_gnavus_group: rho = − 0.596 to − 0.518, p < 0.001; Fusobacterium: rho = − 0.572 to − 0.623, p < 0.001; Erysipelatoclostridium: rho = − 0.605 to − 0.686, p < 0.001) (Additional file 11). Network diagram of the correlation between differential genus-level microbiota and differentially abundant metabolites (Spearman's correlation analysis, | rho |> 0.50 and p < 0.05) Discussion In this study, we revealed significant differences and potential correlations in the fecal microbiota composition and metabolic profiles between FD patients and HCs through 16S rRNA sequencing and untargeted metabolomic analysis. H. pylori infection is recognized as a significant contributor to the development and progression of FD, possibly through a mechanism involving gastrointestinal mucosal inflammation and altered gut motility [14]. Research has indicated that H. pylori eradication therapy, although modestly beneficial, significantly improves FD symptoms [15]. Given the focus of this study on the fecal microbiota and metabolites in the lower gastrointestinal tract, we enrolled participants who tested negative for H. pylori to eliminate potential confounding effects of this pathogen. Previous work reported unchanged alpha diversity in the gastric or duodenal mucosal microbiota of FD patients [8, 16], whereas luminal aspirates presented reduced richness and evenness, mirroring IBS [17]. We now extend these findings to the fecal compartment: FD subjects harbored significantly more species (richness), yet Shannon diversity remained equivalent to that of HCs, implying the expansion of low-abundance taxa with redundant functions and possible ecological imbalance. Our results revealed a significant increase in the abundance of the genus Akkermansia in fecal samples from FD patients. However, due to the small cohort size, ROC analysis findings for Akkermansia as a potential biomarker should be interpreted with caution. Given that our sequencing data are limited to the genus level, we can only speculate that the observed association is driven primarily by Akkermansia muciniphila (A. muciniphila), the predominant and well-characterized mucin-degrading species within this genus. Consequently, the following discussion focuses on A. muciniphila as a working hypothesis. A. muciniphila is widely recognized for maintaining colonic mucus-layer homeostasis and strengthening gut barrier function [18], and its depletion has been linked to various metabolic and inflammatory diseases [19–22]. Conversely, under specific conditions such as a low-fiber diet, the mucinolytic activity of A. muciniphila may excessively degrade the mucus layer, potentially increasing intestinal permeability and promoting low-grade inflammation [23, 24], mechanisms theoretically implicated in FD pathophysiology. Furthermore, animal studies indicate that A. muciniphila may, in certain contexts, such as in the absence of innate lymphoid cells, amplify pathogen virulence [25]. This highlights its context-dependent role, which remains to be elucidated in human FD cohorts. Ruminococcus_gnavus_group, which may include Ruminococcus gnavus (R. gnavus) and other closely related species, was markedly enriched in the feces of the FD group. Previous studies have suggested that overgrowth of R. gnavus may be linked to inflammatory bowel disease (IBD), IBS, and colorectal cancer (CRC) [26]. Notably, in Crohn's disease patients, the relative abundance of R. gnavus correlates with disease severity, potentially via its secretion of glucorhamnan, a complex polysaccharide that induces the secretion of tumor necrosis factor-α (TNF-α) by dendritic cells [27, 28]. As a pleiotropic cytokine, TNF-α mediates intestinal inflammation and immune dysregulation, which may bridge microbial dysbiosis to mucosal injury. Similarly, duodenal microbial dysbiosis has been shown to promote FD by impairing mucosal integrity, suppressing immune function, and disrupting gastrointestinal motility [29]. FD is closely linked to psychological factors [30, 31]. The microbiota‒gut‒brain axis provides a plausible framework for this connection, as gut dysbiosis may mediate the interaction between psychological factors and FD pathogenesis [32]. Notably, recent evidence has indicated that R. gnavus can influence central nervous system function through bacterial enzymes and metabolites involved in the gut‒brain axis [33]. However, the correlation between R. gnavus and FD, along with the potential underlying mechanisms, requires further elucidation through laboratory and clinical studies. In addition to FD, accumulating evidence suggests that alterations in the gut microbiota may contribute to dyspeptic symptoms across different disease contexts. For example, Rossi et al. [34] reported that chronic inflammation and impaired gastric acid secretion associated with autoimmune gastritis could induce alterations in the gut microbiota, which may further contribute to the development of dyspeptic symptoms. This cross-disease commonality suggests that the gut microbiota may serve as a core shared mediator of dyspeptic symptoms. We employed LC–MS/MS to profile metabolites in stool samples from both groups. A key finding was that PG and PC—two core glycerophospholipid subclasses—were the most prominently altered metabolites, with significantly lower levels in FD patients than in controls. Glycerophospholipids, particularly PG and PC, play indispensable roles in maintaining intestinal epithelial integrity, mediating intracellular signaling, and regulating host metabolism [35]. Thus, the observed reduction in PG and PC in FD may disrupt the intestinal epithelial structure and function, impair the intestinal mucosal barrier, and dysregulate gut immunity—processes that are consistent with known FD pathophysiological features, although direct causality remains to be confirmed. Notably, PG and PC are key mediators of the KEGG glycerophospholipid metabolism pathway, a central branch of lipid metabolism. While previous studies have linked lipid metabolite alterations to gastrointestinal immunity and barrier permeability [36], their specific association with gut microbiota dysbiosis in FD has not been systematically explored. Our correlation analysis revealed that differentially abundant bacterial genera—including Ruminococcus_gnavus_group, Fusobacterium, and Erysipelatoclostridium—exhibited significant negative correlations with the PG and PC levels. These findings establish a novel link between the gut microbial composition and glycerophospholipid metabolism in FD, as these bacterial genera are known to exert proinflammatory effects [25, 37, 38], yet their regulatory role in glycerophospholipid homeostasis has not been previously reported in FD. In support of the biological plausibility of this interaction, gut microbiota dysbiosis has been shown to disrupt glycerophospholipid metabolism in feces, serum, and tissues in depressive mouse models [39]. However, it is critical to emphasize that our observations are correlational and that causal relationships cannot be inferred from the current cross-sectional design. Importantly, this study is the first to systematically characterize the coalterations of the gut microbiota and glycerophospholipid metabolism in FD, suggesting the potential involvement of a “microbiota–glycerophospholipid axis” in FD pathophysiology. This novel framework provides a new direction for future research to dissect the functional crosstalk between gut microbes and host lipid metabolism in FD. Several limitations should be considered when interpreting the results. First, the small sample size compromises statistical power and generalizability, precluding robust subgroup analysis and adequate control for confounders. Second, the lack of systematic phenotypic data [e.g., the Functional Dyspepsia Symptom Severity Scale (FD-SSS), 36-Item Short Form Health Survey (SF-36), and Generalized Anxiety Disorder-7 Scale (GAD-7)] hampers correlation analyses between microbiota, metabolites, and clinical symptoms, impeding in-depth elucidation of the gut‒brain axis. Third, the 4-week washout period for PPIs, probiotics, and antibiotics may be insufficient to achieve full normalization of the gut microbiota. The cross-sectional design precludes causal inference, whereas relative quantification metabolomics limits translational value. Although participants were recruited from the same region to reduce heterogeneity, residual confounding may persist because dietary, socioeconomic, and lifestyle factors were not systematically assessed. Moreover, key gastrointestinal hormones (e.g., motilin and gastrin) were not measured owing to sample constraints, further limiting mechanistic insight. Finally, as an exploratory study, the observed microbe‒metabolite associations (e.g., Ruminococcus_gnavus_group and PG/PC) are purely observational; their causal relationships and underlying mechanisms require validation in animal models and isotope-tracing experiments. Future large-scale longitudinal studies that integrate multiomics datasets and functional experiments are warranted to increase clinical and translational relevance. Conclusions Patients with FD exhibited dysbiosis in both the gut microbiota and fecal metabolome. The imbalanced levels of PG and PC, which are significantly correlated with differential bacterial genera, may contribute to FD pathogenesis by disrupting host glycerophospholipid metabolism. Consequently, our study revealed that the “gut microbiota‒glycerophospholipid axis” is a potential key regulatory component in FD pathophysiology. This study is a preliminary exploration with limitations, and its conclusions should be interpreted with caution. Further validation in large-scale longitudinal cohorts and mechanistic studies is warranted. Supplementary Information Supplementary Material 1: Table S1 Detailed information on the baseline characteristics of all the subjects. Table S2 Gut microbiota α-Diversity indices between the two groups. Fig. S1 Annotated cladogram of differential species displaying statistically significant taxa between the two groups. Fig. S2 Visualization of the STAMP differential analysis results on the basis of the KEGG Level 1 database. Fig. S3 Visualization of the STAMP differential analysis results on the basis of the KEGG Level 3 database. Fig. S4 Volcano plot showing the differentially expressed metabolites between the two groups. Fig. S5 ROC plots of the top 5 differentially abundant metabolites with the smallest p-values. Fig. S6 Bubble plot of the 20 pathways with the lowest p-values. Fig. S7 Network diagram of the correlation between differential phylum–level microbiota and differentially abundant metabolites. Fig. S8 Network diagram of the correlation between differential family-level microbiota and differentially abundant metabolites Supplementary Material 2: Determined number of ASVs in each sample Supplementary Material 3: Abundance values of the top 20 bacterial phyla in each sample Supplementary Material 4: Abundance values of the top 20 bacterial genera in each sample Supplementary Material 5: Abundance values of 35 differential bacterial taxa in each sample Supplementary Material 6: Table of primary metabolite identification results Supplementary Material 7: Table of secondary metabolite identification results Supplementary Material 8: Expression levels of the top 30 differentially abundant metabolites in each sample Supplementary Material 9: Thirty-two pathways associated with MS2 metabolites were significantly enriched Supplementary Material 10: Spearman's correlationvalues between 30 differentially abundant metabolites and 35 differential microbes Supplementary Material 11: Corrected p-values and effect sizes for genus-level microbial and metabolite correlations. Acknowledgements We would like to thank all participants of the study for their cooperation and support. Abbreviations Functional dyspepsia Epigastric pain syndrome Postprandial distress syndrome Helicobacter pylori 13C‒urea breath test Healthy controls Body mass index Proton pump inhibitors Liquid chromatography‒tandem mass spectrometry Quality control Data-dependent acquisition Automatic gain control Principal coordinate analysis Analysis of similarities Linear discriminant analysis Linear discriminant analysis effect size Phylogenetic investigation of communities by reconstruction of unobserved states 2 Kyoto encyclopedia of genes and genomes Human metabolome database Orthogonal projections to latent structures discriminant analysis Variable importance in the projection Benjamini‒Hochberg False discovery rate Standard deviation Amplicon sequence variants Receiver operating characteristic Area under the curve Confidence Interval Statistical analysis of metabolic pathways Fold change Phosphatidylglycerol Phosphatidylcholine Sterol lipids Phosphatidylserine Irritable bowel syndrome Akkermansia muciniphila Ruminococcus gnavus Inflammatory bowel disease Colorectal cancer Tumor necrosis factor‒alpha Functional Dyspepsia Symptom Severity Scale 36-Item Short Form Health Survey Generalized Anxiety Disorder-7 Scale Author contributions YD supervised the study and carried out the ethical procedures, HZ performed the analysis and drafted the manuscript, RL and YL proofread the paper. All authors read and approved the final manuscript. Funding None. Data availability All data generated or analysed during this study are included in this published article and its supplementary information files. Declarations Ethics approval and consent to participate This study was conducted with the approval of the Ethics Committee of Affiliated Hospital of Yangzhou University (Ethics approval number: 2024-YKL04-06). All participants provided written informed consent. Consent for publication Informed consent was obtained from all participants before the publication of this study. Besides, this article does not contain personal information (including personal details, images or videos) that allows the identification of patients. 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 Associated Data Supplementary Materials Supplementary Material 1: Table S1 Detailed information on the baseline characteristics of all the subjects. Table S2 Gut microbiota α-Diversity indices between the two groups. Fig. S1 Annotated cladogram of differential species displaying statistically significant taxa between the two groups. Fig. S2 Visualization of the STAMP differential analysis results on the basis of the KEGG Level 1 database. Fig. S3 Visualization of the STAMP differential analysis results on the basis of the KEGG Level 3 database. Fig. S4 Volcano plot showing the differentially expressed metabolites between the two groups. Fig. S5 ROC plots of the top 5 differentially abundant metabolites with the smallest p-values. Fig. S6 Bubble plot of the 20 pathways with the lowest p-values. Fig. S7 Network diagram of the correlation between differential phylum–level microbiota and differentially abundant metabolites. Fig. S8 Network diagram of the correlation between differential family-level microbiota and differentially abundant metabolites Supplementary Material 2: Determined number of ASVs in each sample Supplementary Material 3: Abundance values of the top 20 bacterial phyla in each sample Supplementary Material 4: Abundance values of the top 20 bacterial genera in each sample Supplementary Material 5: Abundance values of 35 differential bacterial taxa in each sample Supplementary Material 6: Table of primary metabolite identification results Supplementary Material 7: Table of secondary metabolite identification results Supplementary Material 8: Expression levels of the top 30 differentially abundant metabolites in each sample Supplementary Material 9: Thirty-two pathways associated with MS2 metabolites were significantly enriched Supplementary Material 10: Spearman's correlationvalues between 30 differentially abundant metabolites and 35 differential microbes Supplementary Material 11: Corrected p-values and effect sizes for genus-level microbial and metabolite correlations. Data Availability Statement All data generated or analysed during this study are included in this published article and its supplementary information files.