Identification and impact of microbiota-derived metabolites in ascites of ovarian and gastrointestinal cancer

Abstract

Background

Malignant ascites is a common complication of advanced ovarian cancer (OC) and gastrointestinal cancer (GI), significantly impacting metastasis, quality of life, and survival. Increased intestinal permeability can lead to blood or lymphatic infiltration and microbial translocation from the gastrointestinal or uterine tract. This study aimed to identify microbiota-derived metabolites in ascites from OC (stages II-III and IV) and GI patients, assessing their roles in tumor progression.

Methods

Malignant ascites samples from 18 OC and GI patients were analyzed using a four-dimensional (4D) untargeted metabolomics approach combining reversed-phase (RP) and hydrophilic interaction liquid chromatography (HILIC) with trapped ion mobility spectrometry time-of-flight mass spectrometry (timsTOF-MS). Additonally, a targeted flow cytometry-based cytokine panel was used to screen for inflammatory markers. Non-endogenous, microbiota-derived metabolites were identified through the Human Microbial Metabolome Database (MiMeDB).

Results

OC stage IV exhibited metabolic profiles similar to GI cancers, while OC stage II-III differed significantly. Stage IV OC patients exhibited higher levels of 11 typically microbiome-derived metabolites, including 1-methylhistidine, 3-hydroxyanthranilic acid, 4-pyridoxic acid, biliverdin, butyryl-L-carnitine, hydroxypropionic acid, indole, lysophosphatidylinositol 18:1 (LPI 18:1), mevalonic acid, N-acetyl-L-phenylalanine, and nudifloramide, and lower levels of 5 metabolites, including benzyl alcohol, naringenin, o-cresol, octadecanedioic acid, and phenol, compared to stage II–III. Correlation analysis revealed positive associations between IL-10 and metabolites such as glucosamine and LPCs, while MCP-1 positively correlated with benzyl alcohol and phenol.

Conclusion

4D metabolomics revealed distinct metabolic signatures in OC and GI ascites, highlighting microbiota-derived metabolites involved in lipid metabolism and inflammation. Metabolites like 3-hydroxyanthranilic acid, indole, and naringenin may serve as markers of disease progression and underscore the microbiota’s role in shaping malignant ascites and tumor biology.

Statement of translational relevance.

Ascites is a hallmark of advanced gynecological and gastrointestinal cancers with peritoneal metastasis, often associated with poor overall survival. Intraperitoneal chemotherapy (IPC), including hyperthermic intraperitoneal chemotherapy (HIPEC) and pressurized intraperitoneal aerosol chemotherapy (PIPAC), is widely used in clinical practice, showing promising outcomes such as 63.3% pain relief and 60% ascites resolution. However, enhancing IPC efficacy while minimizing toxicity remains a critical challenge. Emerging evidence highlights the role of gut microbiota modulation—through fecal microbiota transplantation (FMT), prebiotics, probiotics, antibiotics, and dietary interventions—in improving chemotherapy and Immune Checkpoint Inhibitors (ICI) sensitivity and overcoming drug resistance. Profiling microbiota-derived metabolites in malignant ascites across cancer stages and types could reveal tumor regression-associated changes and provide insights into targeted combination therapies. Integrating FMT, ICI with IPC has the potential to amplify therapeutic effects, offering a novel translational approach to optimize treatment outcomes for patients with advanced cancers and ascites.

Introduction

Ovarian cancer (OC) is a leading cause of cancer-related mortality among women and is often termed the “silent killer” due to its asymptomatic nature in the early stages and lack of effective screening tools [1, 2]. Consequently, most OC cases are diagnosed at advanced stages when the disease has metastasized, limiting treatment options and resulting in a poor prognosis [3]. A common feature of advanced OC is the abnormal accumulation of fluid in the abdominal cavity, known as ascites, which significantly contributes to patient mortality [4]. Ascites not only serves as a clinical hallmark of OC but is also observed in other malignancies such as gastrointestinal (GI) cancers and cirrhosis, and is rarely associated with non-cancerous conditions like heart failure and peritoneal tuberculosis [5, 6].

The pathogenesis of ascites is multifactorial, involving hypoalbuminemia, increased vascular permeability, impaired lymphatic drainage, and fluid retention due to activation of the renin–angiotensin–aldosterone system (RAAS) [7,8,9]. Recent research has highlighted the role of the gut microbiota in modulating these processes through its influence on immune and metabolic pathways. For instance, butyrate produced by gut bacteria can inhibit RAAS activity, reducing fluid retention [10,11,12]. Disruption of gut microbial balance in advanced disease may exacerbate intestinal permeability and promote bacterial translocation [13,14,15,16,17], contributing to ascites formation and creating a vicious cycle of inflammation and fluid accumulation.

One of the main complications of ascites is bacterial peritonitis (BP), which has been well-documented in patients with GI malignancies [18, 19]. BP is often spontaneous and linked to bacterial translocation from the GI tract to the mesenteric lymph nodes [20]. While less frequently reported in OC, the extensive metastasis and abdominal involvement in advanced OC may similarly compromise gastrointestinal integrity, increasing susceptibility to BP. The partially overlapping features of ascites in OC and GI cancers raise questions about the potential role of the microbiome in shaping the tumor-promoting properties of this fluid.

Moreover, ascites is not just a passive by-product of malignancy but may actively shape the tumor microenvironment, promoting metastasis and therapeutic resistance [21, 22]. Its heterogeneous nature — ranging from clear and free-flowing to viscous and loculated — indicates distinct underlying biological processes that could influence tumor behavior. While paracentesis provides temporary symptomatic relief, it is often palliative and requires repeated procedures, carrying risks such as infection and protein loss. Despite its clinical significance, the molecular and microbial landscape of OC-associated ascites remains poorly understood.

Reversed-phase (RP) and hydrophilic interaction liquid chromatography (HILIC) coupled with trapped ion mobility time-of-flight mass spectrometry (timsTOF) were employed for a fully untargeted four-dimensional (4D) metabolomics analysis of ascitic fluid. Flow cytometry was used to profile inflammatory cytokines. Microbiota-derived metabolites were identified through the Human Microbial Metabolome Database (MiMeDB). These approaches aimed to characterize the metabolic and cytokine profiles of ascitic fluid and evaluate the potential impact of microbiota-derived metabolites on OC and GI progression.

Materials and methods

This exploratory study included 10 malignant ascites specimens from patients undergoing ovarian cancer (OC) resection and 8 malignant ascites specimens from gastrointestinal (GI) cancers patients. Metabolomics and cytokine analysis were used to combine the results for in-depth phenotyping of malignant ascites.

Ethical background

This study was conducted in accordance with the principles outlined in the Declaration of Helsinki and approved by the Ethics Committee, Faculty of Medicine, University of Tübingen, Germany (Ref. Nr. 696/2016BO2 and 117/2020BO1). Written informed consent was obtained from all participating patients. The collection of samples did not interfere with or alter patient treatment plans. All data were anonymized in compliance with the European General Data Protection Regulation (GDPR) and applicable German data protection laws.

Collection and storage of ascitic fluid samples

Ascites samples were collected from patients undergoing surgery for ovarian cancer at the Department of General and Transplant Surgery and the Women’s Hospital, University Hospital Tübingen, Germany. Samples were obtained under sterile conditions in the operating room, stored in sterile tubes, and immediately transported to the laboratory using an icebox to maintain low temperatures. Upon arrival, the samples were centrifuged at 4 °C for 30 minutes at 1,200 rpm. The resulting supernatant was aliquoted into 2 mL tubes and stored at − 80 °C until further analysis. Relevant patient information, including demographic data, cancer histology, and the extent of peritoneal disease was recorded. All samples and associated patient data were anonymized prior to analysis.

Quantification of cytokines in malignant Ascites

To quantify cytokine levels in malignant ascites, 25 µL of ascites from ovarian cancer patients was mixed with 25 µL of assay buffer. Next, 25 µL of a 13-plex bead mix from the LEGENDplex™ Human Inflammation Panel 1 (13-plex, #740809, BioLegend, USA) was added to each well of a 96-well microplate. This multiplex bead-based assay is capable of quantifying 13 different cytokines/chemokines with the following minimum detectable concentrations (MDC) in parentheses: IL-1β (1.5 ± 0.6 pg/mL), IFN-α2 (2.1 ± 0.2 pg/mL), IFN-γ (1.3 ± 1.0 pg/mL), TNF-α (0.9 ± 0.8 pg/mL), MCP-1 (1.1 ± 1.2 pg/mL), IL-6 (1.5 ± 0.7 pg/mL), IL-8 (2.0 ± 0.5 pg/mL), IL-10 (2.0 ± 0.5 pg/mL), IL-12p70 (2.0 ± 0.2 pg/mL), IL-17 A (0.5 ± 0.0 pg/mL), IL-18 (2.0 ± 0.5 pg/mL), IL-23 (1.8 ± 0.1 pg/mL), and IL-33 (4.4 ± 1.5 pg/mL).

The plate was incubated and shaken at room temperature for 2 h, allowing the analytes (cytokines) to bind to the corresponding antibody-conjugated capture beads. Following incubation, the wells were washed to remove unbound analytes. Biotinylated detection antibodies (25 µL) were then added and allowed to bind to the analyte-bound beads. After 30 min of incubation, 25 µL of streptavidin–phycoerythrin (SA-PE) was added, which binds to the biotinylated detection antibodies and produces a fluorescent signal proportional to the amount of each cytokine.

After a further 1-hour incubation, the beads were washed, resuspended in wash buffer, and samples were acquired using a flow cytometer. Fluorescent signals were measured, and the concentrations of the analytes were determined based on standard curves generated using the LEGENDplex™ data analysis software (BioLegend, USA).

Sample Preparation for timsTOF LC-MS analysis

Frozen ascites samples were thawed at room temperature, and 1.2 mL of each sample was transferred into separate 1.5 mL Eppendorf tubes. The tubes were centrifuged at 14,000 rpm at 4 °C for 15 minutes. The supernatants (1 mL) were carefully collected and evaporated at room temperature using a Concentrator Plus (Eppendorf, Wesseling-Berzdorf, Germany) for 6 hours.

Reversed-Phase Liquid Chromatography (RPLC) sample Preparation

For reversed-phase (RP) liquid chromatography analysis, the dried samples were reconstituted in 100 µL of MilliQ water (MQ) and 300 µL of ice-cold (-20 °C) high-performance liquid chromatography (HPLC) grade acetonitrile (VWR Chemicals, Darmstadt, Germany). The mixture was vortexed for 1 minute, followed by incubation at -20 °C for 10 minutes. Samples were then centrifuged at 14,000 rpm at 4 °C for 15 min. From each tube, 300 µL of the supernatant was transferred to a new Eppendorf tube and evaporated to dryness for 1.5 hours at room temperature. The dried samples were then reconstituted in 60 µL of MQ/acetonitrile (9:1, v/v), vortexed for 10 s, and centrifuged again at 14,000 rpm at 4 °C for 15 min. Finally, 50 µL of the supernatant was transferred to HPLC vials (VWR, Leuven, Belgium) equipped with inserts for subsequent RPLC-MS analysis.

Hydrophilic Interaction Liquid Chromatography (HILIC) sample Preparation

For hydrophilic interaction liquid chromatography (HILIC) analysis, the dried samples were reconstituted in 100 µL of MQ and 300 µL of ice-cold (-20 °C) acetonitrile. The mixture was vortexed for 1 minute, followed by incubation at -20 °C for 10 minutes. The samples were centrifuged at 14,000 rpm at 4 °C for 15 minutes, and the resulting supernatant (300 µL) was transferred into HPLC vials with inserts for HILIC-MS measurements.

Liquid chromatography conditions

Analyte separation was performed using the Elute PLUS LC series (Bruker, Bremen, Germany).

RPLC conditions

RPLC separations were conducted on an Intensity Solo 2 C18 Column (100 Å; 2.0 μm; 2.1 mm × 100 mm; #BRHSC18022100, Bruker) using 0.1% formic acid in MilliQ water as mobile phase A and 0.1% formic acid in acetonitrile as mobile phase B. A 5 µL injection of each sample was used. The separation was carried out at a flow rate of 0.4 mL/min from 0 to 9 min and 10.6 to 13 min, and at 0.6 mL/min from 9.1 to 10.6 min, with a column temperature maintained at 50 °C using the following gradient: 0–1 min, 5% B; 1–7 min, 5–40% B; 7–9 min, 40–98% B; 9–10.6 min, 98% B; 10.6–10.7 min, 98 − 5% B; 10.7–13 min, 5% B.

HILIC conditions

HILIC separations were performed on an ACQUITY UPLC BEH Amide column (130 Å, 1.7 μm, 2.1 mm × 150 mm; #186004802, Waters) using 10 mM ammonium formate and 0.1% formic acid in MilliQ water as mobile phase A and 10 mM ammonium formate and 0.1% formic acid in acetonitrile as mobile phase B. A 5 µL injection was used, with separation conducted at a flow rate of 0.5 mL/min and a column temperature of 40 °C using the following gradient: 0–3 min, 100% B; 3–10 min, 100 −85% B; 10–14 min, 85 − 50% B; 14–15 min, 50% B; 15–15.1 min, 50–100% B; 15.1–23 min, 100% B.

Mass spectrometry analysis

The separated analytes were analyzed using a timsTOF fleX mass spectrometer (Bruker, Bremen, Germany) equipped with an Apollo II source for RP measurements, and a timsTOF Pro 2 (Bruker, Bremen, Germany) with a vacuum-insulated probe heated electrospray ionization (VIP-HESI) source for HILIC analysis. LC-MS/MS data were acquired in duplicate (technical replicates) using positive and negative dda-PASEF modes, with a TOF mass range of m/z 20-1300. Default Bruker PASEF acquisition parameters for MS/MS acquisition were used: 2 ramps (12 precursors each) per cycle; resulting cycle time 0.69 s; Intensity threshold 100 counts; Target Intensity 4000 counts (signals below that threshold will be scheduled for MS/MS fragmentation more often); Active Exclusion activated (0.1 min; reconsider if intensity increase is 2-fold or higher). The system was controlled using timsControl^® and Compass HyStar^® software, and data acquisition was managed using the same software. Quality control (QC) samples were run every ten injections, and blank samples were analyzed using H₂O for RP and acetonitrile for HILIC.

Data preprocessing and statistical analysis

Raw data processing was conducted using MetaboScape^® software (version 2024b, Bruker, RRID: SCR_026044) with four-dimensional (4D) feature extraction, capturing mass-to-charge ratio (m/z), isotopic pattern quality, retention times, MS/MS spectra, and collision cross-section (CCS) values. Feature extraction was performed using the T-ReX^® 4D algorithm (RRID: SCR_026044), followed by annotation through the Bruker Human Metabolome Database (HMDB, RRID: SCR_007712) and the NIST Mass Spectral Library (RRID: SCR_014668) Level 2 annotation according to Sumner et al. [23]. High-quality spectra were selected based on stringent criteria, including chromatogram and ion mobilogram quality, annotation scores, and CCS accuracy (Table S3). Potential microbiota-derived metabolites were identified and their origins traced to specific microbiota species using the Human Microbial Metabolome Database (MiMeDB, RRID: SCR_025108).

Data from RP and HILIC measurements were integrated for analysis. Sample intensities from omic data were normalized using probabilistic quotient normalization (PQN), log-transformed, and scaled using Pareto scaling to approximate normality. The final dataset was analyzed and visualized using MetaboAnalystR (RRID: SCR_016723), Pheatmap (RRID: SCR_016418), and ComplexHeatmap (RRID: SCR_017270) packages in R (version 4.3.2). Descriptive statistics and correlation analyses were performed in R. Comparative statistics included t-tests and one-way analysis of variance (ANOVA) for normally distributed data, and non-parametric tests for skewed data.

Results

Patients' clinical characteristics

Ascites samples were collected from 10 OC and 8 GI patients undergoing open surgery for malignancy removal. The clinicopathological characteristics of the patients are summarized in Table 1. None of the OC patients had received chemotherapeutic treatment before ascites collection. Notably, the OC patient with clear cell carcinoma did not have a history of endometriosis. Two of the GI patients had previously received neoadjuvant chemotherapy, and all had peritoneal metastases. Due to the limited sample size and the heterogeneity in histological subtypes and cancer origins, all available GI patient data were included to provide a general overview. Despite the treatment history in a subset of GI patients, the observed differences remain relevant and informative.

Table 1 Clinic-pathological characteristics of the explorative cohort

Metabolite differences between Ovarian Cancer (Stages II-III, IV) and Gastrointestinal Cancer groups

Distinct metabolite changes were observed between OC stages II-III and GI groups, while ovarian cancer stage IV samples displayed metabolic profiles equivalent to those of the GI cancer samples (Fig. 1A). Normalized intensity data from three groups—OC II-III (ovarian cancer stages II-III), OC IV (ovarian cancer stage IV), and GI (including appendiceal, colon, and gastric cancers)—were analyzed using one-way ANOVA with a false discovery rate (FDR, Benjamini-Hochberg) cutoff of 0.05 (Table S4.1 and S4.2). Twelve significant metabolites were identified out of the 696 measured. Approximately 92% of the metabolites showed distinct intensity distributions when comparing the GI group with the OC II-III group. However, when comparing the GI group with OC IV, around 83% of the metabolites exhibited similar intensity trends (Fig. 1A).

Fig. 1

Identification of Ascites Metabolic Signatures in OC II-III, OC IV, and GI Groups. (A) The heatmap displays significant metabolites identified through ANOVA (adjusted p-value cutoff of 0.05) across eight GI, seven OC II-III, and three OC IV biological replicates. Clustering was performed using Ward’s hierarchical method with Euclidean distance as the distance metric. (B) Raincloud plots (combining violin, box, and strip plots) illustrate the significant metabolites: (a) 2-tert-butyl-4-ethylphenol, (b) 3-methoxy-4-(2-methylpropoxy)benzoic acid, (c) 4-(2,5-dimethylphenyl)-4-oxobutanoic acid, (d) 4-isopropyl-3-methylphenol, (e) 4-prop-1-enylveratrole, (f) cuminaldehyde, (g) Glu-Gly-Arg, (h) phenylalanylphenylalanine. (i) PPA, (j) Propofol-β-D-glucuronide, (k) SM 36:3;O2, and (l) thymol, comparing GI, OC II-III, and OC IV groups. Post hoc analysis using Fisher’s LSD was applied for group comparisons (*p <.05, **p <.01, ***p <.001, ****p <.0001). Dots represent individual data points, the central line in the box indicates the median, and the box edges represent the upper and lower quartiles. Half-violins depict data distributions. Abbreviations: SM, sphingomyelin; Glu-Gly-Arg, glutamyl-glycyl-arginine; PPA, phenylpropionic acid

The top 12 significant metabolites identified in the heatmap were further visualized using individual raincloud plots (Fig. 1B). Ten metabolites exhibited similar trends, with propofol-β-D-glucuronide (j.) showing a stepwise and significant decrease across the OC II-III, OC IV, and GI groups (OC IV significantly decreased compared to OC II-III, and GI significantly decreased compared to both OC II-III and OC IV).

Several metabolites, including 2-tert-butyl-4-ethylphenol (a.), 3-methoxy-4-(2-methylpropoxy) benzoic acid (b.), 4-isopropyl-3-methylphenol (d.), 4-prop-1-enylveratrole (e.), cuminaldehyde (f.), PPA (i.), and thymol (l.), exhibited a similar trend: OC IV showed a significant decrease compared to OC II-III, while GI significantly decreased compared to OC II-III. Although the GI group also decreased compared to OC IV, the difference was not statistically significant. Additionally, 4-(2,5-dimethylphenyl)-4-oxobutanoic acid (c.) and Glu-Gly-Arg (g.) showed significant decreases only when comparing GI to OC II-III, with no significant differences observed between OC IV and OC II-III or GI and OC IV.

In contrast, phenylalanylphenylalanine (h.) and SM 36:3;O2 (k.) followed a different pattern. Phenylalanylphenylalanine (h.) significantly increased in OC IV compared to both OC II-III and GI. SM 36:3;O2 (k.) exhibited a similar trend, though the difference between OC II-III and OC IV was not statistically significant.

A Venn diagram was generated to visualize the shared and unique metabolites among the OC II–III, OC IV, and GI groups, providing an overview of group-specific and overlapping metabolic features (Figure S3). Specifically, 12 metabolites were shared between OC II–III and OC IV only, 87 between OC II–III and GI only, and 40 between OC IV and GI only, while 785 metabolites were common to all three groups. In terms of unique metabolites, 13 were specific to OC II–III, 46 to OC IV, and 76 to the GI group. Detailed information is provided in Supplementary Tables S9.1 and S9.2.

Potential Microbiota-derived metabolites in Ascites samples

Identification and grading of non-endogenous metabolites with the human microbial metabolome database

The Human Microbial Metabolome Database (MiMeDB) (https://mimedb.org) is a comprehensive multi-omics resource for microbiome research [24]. To explore the potential origins of metabolites that cannot be of human-origin and that were identified significant in OC II-III, OC IV, and GI ascites samples, a comparative analysis was first conducted between different stages of OC and between OC and GI. Using the MiMeDB database, significantly increased or decreased metabolites were then examined for their potential origin and relationship to microbiota species.

Differential general and microbiome-derived metabolites in Ascites of OC and GI

Identification of general metabolite changes between OC and GI ascites

Based on a fold change (FC) > 1.2 and p-value < 0.05, the analysis of normalized intensity data between the OC and GI groups identified 90 significant metabolites out of the 696 measured (Figure S1). Of these, 51 metabolites showed significantly lower intensity, and 39 showed significantly higher intensity in the GI group compared to the OC group. Due to the small sample size of ascites samples, the raw p-value was used for this comparison instead of the FDR-adjusted p-value (Fig. 2A). We also included the adjusted p-values in the Supplementary Table S5.

Fig. 2

Significant Metabolite Changes Between OC and GI Groups. (A) Volcano Plot: Red dots indicate metabolites upregulated in the GI group, while blue dots indicate those upregulated in the OC group. Thresholds: fold change ≥ 1.2 and raw p-value ≤ 0.05. (B) Circular Heatmap: Displays intensity differences of 17 significant microbiota-derived metabolites across 10 OC and 8 GI biological replicates. Analysis used t-tests (raw p-value ≤ 0.05, fold change ≥ 1.2), with clustering performed using Ward’s hierarchical method and Euclidean distance. (C) Raincloud Plots: Show significantly different metabolites between GI and OC groups: (a) 3-methylindole, (b) 3-methylxanthine, (c) benzamide, (d) caffeine, (e) D-glucurono-6,3-lactone, (f) D-tagatose, (g) glucosamine, (h) levulinic acid, (i) LPC 18:1, (j) LPC 20:1, (k) LPC 20:2, (l) LPC 22:1, (m) LPC 22:4, (n) phosphocholine, (o) sphinganine, (p) thymol, and (q) trimethylamine N-oxide. Unpaired t-tests were applied for comparisons (*p <.05, **p <.01, ***p <.001, ****p <.0001). Dots represent individual data points; boxplot lines show medians, quartiles are represented by box edges, and half-violins illustrate data distributions. (D) Sankey Plot: Depicts the structure of associated phyla and their respective kingdoms for differentiated microbiota-derived metabolites, highlighting the dominant role of bacteria in the ascites composition of both OC and GI. Abbreviations: LPC, lysophosphatidylcholine; CerP, ceramide phosphate; LPE, lysophosphatidylethanolamine; PE, phosphatidylethanolamine; SPB, sphingoid base; SM, sphingomyelin; PC, phosphatidylcholine; DG, diacylglycerol; Cer, ceramide; DGTS, diacylglyceryl-N, N,N-trimethylhomoserine; PPA, phenylpropionic acid; Glu-Gly-Arg, glutamyl-glycyl-arginine

Sphinganine and phosphocholine were significantly decreased in the GI group compared to the OC group. While LPC 18:1, LPC 20:2, LPC 22:1, and LPC 22:4 lipid compounds were significantly increased in the OC group compared to the GI group.

Identification of microbiota-derived metabolite changes between OC and GI Ascites

The MiMeDB database results identified a set of metabolites potentially produced or synthesized by the microbiome, including bacterial species, Eukaryota/Fungi, and Archaea. The associated phyla for the GI vs. OC comparison are listed in Table 2.

Table 2 Metabolites potentially derived from the microbiota identified in the comparison between ovarian Cancer (OC) and Gastrointestinal Cancer (GI) groups

Potentially microbiota-derived metabolites significantly increased in GI ascites compared to the OC group included 3-methylindole (a.), 3-methylxanthine (b.), caffeine (d.), D-glucurono-6,3-lactone (e.), D-tagatose (f.), glucosamine (g.), levulinic acid (h.), lysophosphatidylcholine 18:1 (LPC 18:1) (i.), LPC 20:1 (j.), LPC 20:2 (k.), LPC 22:1 (l.), LPC 22:4 (m.), and trimethylamine N-oxide (q.) (Fig. 2C). Conversely, benzamide (c.), phosphocholine (n.), sphinganine (o.), and thymol (p.) were significantly decreased in GI ascites compared to the OC group.

Figure 2B and D illustrate the overall variations in potential microbiota-derived metabolites across the 18 ascites samples from both OC and GI groups. The figures also depict the microbial sources and categories of metabolites that showed increases or decreases in each group, including their associated phyla and respective superkingdoms. In summary, the bacterial kingdom was the predominant source, followed by Archaea, with Eukaryota/Fungi contributing minimally.

Differential general and microbiome-derived metabolites in Ascites of OC stage II-III and stage IV

Identification of general metabolite changes between OC stage II-III and OC stage IV

A comparison between OC II-III (ovarian cancer stage II-III) and OC IV (stage IV) identified 84 significant metabolites out of 649, with 45 showing lower intensities and 39 higher intensities in the OC IV group relative to OC II-III (Fig. 3A). This analysis applied a fold change threshold of 1.2 and a raw p-value threshold of 0.05. Due to the limited sample size, raw p-values were used instead of FDR-adjusted p-values. However, FDR-adjusted p-values are provided in Supplementary Table S6 for reference. A heatmap of the 84 significant metabolites was generated to visualize the metabolic differences between OC II-III and OC IV (Figure S2).

Identification of microbiome-derived metabolite changes between OC stage II-III and OC stage IV

The same analysis using the MiMeDB database identified 16 metabolites potentially linked to the microbiome, including those associated with bacterial species, Eukaryota/Fungi, and Archaea. The corresponding phyla for the OC II-III vs. OC IV comparison are provided in Table 3.

Table 3 Metabolites potentially derived from the microbiota identified in the comparison between ovarian Cancer stage II-III and stage IV

The 16 differentiated potential microbiota-derived metabolites between OC II-III and OC IV showed significant changes. Specifically, 1-methylhistidine (a.), 3-hydroxyanthranilic acid (b.), 4-pyridoxic acid (c.), biliverdin (e.), butyryl-L-carnitine (f.), hydroxypropionic acid (g.), indole (h.), lysophosphatidylinositol 18:1 (LPI 18:1) (i.), mevalonic acid (j.), N-acetyl-L-phenylalanine (k.), and nudifloramide (m.) were elevated in OC IV ascites samples compared to OC II-III. On the other hand, benzyl alcohol (d.), naringenin (l.), o-cresol (n.), octadecanedioic acid (o.), and phenol (p.) were significantly reduced in the OC IV group relative to OC II-III. These changes suggest distinct metabolic shifts between ovarian cancer stages II-III and IV, potentially reflecting the progression of the disease. The corresponding metabolite alterations are visualized in Fig. 3C.

Fig. 3

Significant Metabolite Changes Between OC II-III and OC IV Groups. (A) Volcano Plot: Red dots represent metabolites upregulated in OC IV, while blue dots represent those upregulated in OC II-III. Thresholds: fold change ≥ 1.2 and raw p-value ≤ 0.05. (B) Circular Heatmap: Displays intensity variations of 16 significant microbiota-derived metabolites across 7 OC II-III and 3 OC IV biological replicates. Analysis was performed using t-tests (raw p-value ≤ 0.05, fold change ≥ 1.2), with clustering based on Ward’s hierarchical method and Euclidean distance. (C) Raincloud Plots: Show significantly different metabolites between OC II-III and OC IV groups: (a) 1-methylhistidine, (b) 3-hydroxyanthranilic acid, (c) 4-pyridoxic acid, (d) benzyl alcohol, (e) biliverdin, (f) butyryl-L-carnitine, (g) hydroxypropionic acid, (h) indole, (i) LPI 18:1, (j) mevalonic acid, (k) N-acetyl-L-phenylalanine, (l) naringenin, (m) nudifloramide, (n) o-cresol, (o) octadecanedioic acid, and (p) phenol. Unpaired t-tests were applied for comparisons (*p <.05, **p <.01, ***p <.001, ****p <.0001). Dots represent individual data points; the boxplot’s central line indicates the median, with quartiles represented by box edges, and half-violins illustrating data distributions. (D) Sankey Plot: Depicts microbial phyla and their respective superkingdoms associated with altered metabolites, emphasizing the dominant role of bacteria and their interactions with Eukaryota/Fungi and Archaea in the ascites of advanced ovarian cancer. Abbreviations: LPC, lysophosphatidylcholine; MG, monoacylglycerol; Phe-Phe, phenylalanine-phenylalanine; Phe-Leu, phenylalanine-leucine; CerP, ceramide phosphate; SM, sphingomyelin; LPI, lysophosphatidylinositol; PC, phosphatidylcholine; PPA, phenylpropionic acid; PE, phosphatidylethanolamine

The broad metabolic variations in potential microbiota-derived metabolites across the 7 OC II-III and 3 OC IV ascites samples, highlighting distinct shifts between early and advanced stages of ovarian cancer (Fig. 3B). The microbial sources and categories of the metabolites that exhibited alterations between the OC II-III and OC IV groups, including their associated phyla and respective superkingdoms (Fig. 3D). Bacteria emerged as the primary source of these metabolites, followed by Eukaryota/Fungi, with Archaea contributing the least. These findings suggest that the microbiota’s involvement in ovarian cancer progression may stem from a diverse array of microbial species, each potentially playing a specific role in influencing the tumor microenvironment and disease progression.

Correlations between potentially microbiota-derived metabolites and cytokines/chemokines in Ascites

Ovarian Cancer vs. Gastrointestinal Cancer

To investigate potential interactions between microbiota-derived metabolites and cytokines, a correlation analysis was performed between individual metabolites from the OC vs. GI comparison and cytokine levels (Fig. 4A). Due to the small sample size of ascites samples, the raw p-value was used for this comparison instead of the FDR-adjusted p-value. The FDR-adjusted p-value information is included in Supplementary Table S7. The results revealed several significant correlations amongst metabolites and cytokines/chemokines. Cytokine IL-23 showed a positive correlation with D-glucurono-6,3-lactone and a negative correlation with trimethylamine N-oxide. IL-18 was negatively correlated with caffeine. IL-10 demonstrated significant positive correlations with glucosamine, D-tagatose, trimethylamine N-oxide, caffeine, LPC 22:4, and LPC 20:1, while showing negative correlations with benzamide and thymol. TNF-α positively correlated with D-glucurono-6,3-lactone. IFN-γ was negatively correlated with levulinic acid and trimethylamine N-oxide, while IFN-α2 showed a positive correlation with D-glucurono-6,3-lactone. Chemokine MCP-1 was negatively correlated with D-tagatose.

Fig. 4

Heatmap of Correlations Between Potential Microbiota-derived Metabolites and Flow Cytometry Cytokines/Chemokines. (A) OC vs. GI, (B) OC II-III vs. OC IV, the heatmap illustrates the correlations between microbiota-derived metabolites and cytokines measured by flow cytometry. A single asterisk (*) indicates a significant correlation (p < 0.05), while a double asterisk (**) represents a highly significant correlation (p < 0.01). Correlations were calculated using Pearson method for normally distributed data. Abbreviations: LPI, Lysophosphatidylinositol

Ovarian Cancer stage II-III vs. stages IV

A similar correlation analysis was performed between microbiota-derived metabolites from the OC II-III vs. OC IV comparison and cytokine levels (Fig. 4B). Given the small sample size, raw p-values were used instead of FDR-adjusted p-values. Supplementary Table S8 presents the FDR-adjusted p-values. The analysis identified significant correlations. Chemokine IL-8 showed negative correlations with 3-hydroxyanthranilic acid and nudifloramide. MCP-1 was positively correlated with benzyl alcohol, naringenin, o-cresol, octadecanedioic acid, and phenol, while negatively correlated with 1-methylhistidine, 4-pyridoxic acid, and mevalonic acid. Cytokine IL-6 demonstrated negative correlations with butyryl-L-carnitine, indole, nudifloramide and N-acetyl-L-phenylalanine. IL-1β exhibited positive correlations with 1-methylhistidine, 4-pyridoxic acid, LPI 18:1 and mevalonic acid, and negative correlations with benzyl alcohol, naringenin, o-cresol, octadecanedioic acid, and phenol.

Discussion

Malignant ascites forms due to a combination of increased fluid production and reduced lymphatic absorption [25, 26]. Normally, the peritoneum absorbs excess fluid through lymphatic channels, but in malignancy, the tumor’s growth and invasion disrupt these processes. Tumor-induced neovascularization [7], driven by vascular endothelial growth factor (VEGF), increases vascular permeability, leading to fluid leakage into the abdominal cavity [8, 27]. Matrix metalloproteinases (MMPs) also degrade tissue barriers, further promoting fluid accumulation [28, 29]. In addition, hormonal changes activate the renin-angiotensin-aldosterone system, causing sodium and fluid retention [6].

Emerging research suggests the microbiota may influence ascites development. Disruption of gut bacteria has been linked to inflammation, altered immune responses, and cancer progression [30,31,32]. In malignancies, an imbalanced gut microbiota can impair the intestinal barrier, increasing permeability and allowing bacterial translocation [33, 34]. This worsens inflammation and fluid retention, creating a cycle that perpetuates ascites formation. The role of gut microbiota in modulating immune and metabolic pathways presents a potential area for further exploration in managing malignant ascites [35].

LC-MS based metabolomic profiles of OC stages II-III, IV and GI Ascites samples

Metabolic analysis of ascites from the OC II-III, OC IV, and GI groups revealed that OC IV shares more similarities with the GI group, while OC II-III is distinctly different from both. The GI group included patients with various cancer origins and histological subtypes; notably, two patients had received neoadjuvant therapy, and one patient was male. Despite this heterogeneity, we included all available GI cases to provide a general overview. Our aim was to capture the broader metabolic landscape of advanced disease, as OC at stage IV exhibits features increasingly similar to those observed in GI cancers with peritoneal metastasis. This pattern aligns with the progression of ovarian cancer, where advanced stages involve cancer cells detaching from the primary tumor, surviving in the peritoneal fluid, and spreading to organs such as the liver, lungs, spleen, intestines, and lymph nodes [36,37,38].

One metabolite, phenylalanylphenylalanine, a dipeptide of two phenylalanine molecules, was found at higher levels in OC IV. Elevated phenylalanine levels and altered phenylalanine-to-tyrosine ratios have been associated with inflammatory conditions, including cancer [39]. Phenylalanine metabolism also influences T-cell function, regulate T-cell proliferation and activation and affecting the following immune response [39]. This suggests that higher levels of phenylalanylphenylalanine could be linked to more advanced cancer stages. Similarly, SM 36:3;O2, a sphingomyelin, was significantly elevated in OC IV. Increased sphingomyelin levels have been linked to cancer development, with altered sphingomyelin metabolism observed in metastatic tumor cells and various cancers [40,41,42]. These changes in sphingomyelin metabolism, including higher synthesis and reduced breakdown, contribute to disrupted lipid balance, promoting cancer growth, particularly in ovarian and breast cancers [43, 44]. The higher levels of sphingomyelin in ascites may, therefore, reflect more aggressive tumor behavior and advanced disease.

LC-MS analysis also detected a wide range of metabolites in the ascites samples, many of which are not directly related to human metabolism. Several of the significantly different compounds were identified as exogenous substances, including drugs, food-derived compounds, plant oil-derived, and anesthetics. These findings highlight the complexity of ascites as a mixture of both endogenous and exogenous compounds. The presence of these exogenous compounds can be explained by several factors. Medications such as chemotherapeutics, anesthetics, and pain management drugs are commonly used in ovarian cancer treatment, and residual traces can accumulate in bodily fluids [45]. Additionally, diet and environmental exposure significantly influence an individual’s metabolic profile [46, 47]. Plant-derived compounds or food additives can enter the bloodstream and appear in ascites, particularly in patients undergoing systemic changes due to disease or treatment [48]. This suggests that ascites is influenced by both internal metabolic processes and external factors, complicating the interpretation of LC-MS data.

Potential microbiota-derived metabolomic features in malignant Ascites samples

Through MiMeDB analysis, we identified 17 microbiota-derived metabolites in the OC vs. GI comparison and 16 in the OC II-III vs. OC IV comparison. Both sets revealed a predominance of bacterial-origin metabolites in malignant ascites, consistent with the human gut microbiota profile [49]. While the proportions of Archaea and Eukaryota/Fungi-derived metabolites showed slight differences between the two sets, the findings align with studies highlighting the role of gut microbiota in carcinogenesis, immune surveillance, and responses to immunotherapy [50,51,52].

In OC patients, continuous immune checkpoint blockade (ICB) therapy with poly (ADP-ribose) polymerase inhibitors (PARPi) has demonstrated efficacy in prolonging progression-free and overall survival [53,54,55,56]. For GI cancers, ICB strategies vary by tumor origin. Immunotherapy is now standard in first-line treatment for advanced colorectal cancer with high microsatellite instability [57,58,59,60]. Additionally, in advanced gastric cancer, combining immunotherapy with chemotherapy or with HER2-targeted therapy has shown significant and lasting survival benefits in HER2-positive patients [61,62,63,64].

Despite advances in immunotherapy, a significant proportion of patients exhibit primary or acquired resistance to treatment [65, 66]. Additionally, immunotherapy-related adverse reactions pose a clinical challenge, particularly with the expanded use of combination therapies and multi-agent immunotherapy [67,68,69,70,71,72,73,74]. To address these challenges, studies have identified host-associated genomic and molecular biomarkers predictive of immunotherapy response [75,76,77]. Emerging evidence also implicates the gut microbiome, particularly specific microbial taxa, in modulating immune checkpoint blockade (ICB) efficacy [78].

Research suggests that gut microbiome composition may be both predictive and prognostic of therapeutic response to ICB, highlighting its potential as a biomarker [78]. These insights have driven the development of microbiome-targeted strategies aimed at enhancing treatment efficacy and minimizing adverse effects by modulating the patient’s gut microbiota [79, 80].

The microbiota-derived metabolites identified in ascites contribute significantly to the tumor microenvironment. This study revealed distinct microbiota profiles across ovarian cancer stages and gastrointestinal cancers, highlighting the importance of larger sample sizes and advanced tools like 16 S rRNA sequencing to enhance our understanding. Further research is needed to explore how these microbial profiles correlate with immunotherapy side effects, tumor reduction efficacy, and clinical outcomes, providing insight into their role in ascites formation and cancer progression.

Emerging interventions, such as fecal microbiota transplants (FMT), prebiotics, probiotics, antibiotics, and dietary modifications, show promise in modulating the gut microbiome [81,82,83]. Characterizing microbiota and its systemic effects will be key to identifying actionable targets for future therapeutic interventions and clinical assessment.

Microbiota-Derived metabolomic profiles in Ascites from OC and GI

Several bacterial-derived metabolites identified in the OC vs. GI comparison are associated with immune-metabolic pathways and may affect immune responses in ovarian and gastrointestinal cancers. Lysophosphatidylcholines (LPCs), known pro-inflammatory lipids [84], are influenced by microbial taxa such as Bacteroidetes and Firmicutes [85]. Bacteroidetes contribute to lipid absorption and metabolism, while firmicutes play a role in immune modulation through lipid pathways [86,87,88]. In cancer, LPCs contribute to inflammation, tumor growth, and immune evasion [89]. Dysbiosis between these taxa may alter LPC levels and functions, impacting cancer progression, immune response, and gut health. The presence of LPCs (LPC 18:1, LPC 20:1, LPC 20:2, LPC 22:1, LPC 22:4) and sphingolipids (sphinganine) indicates significant involvement in membrane turnover and lipid signaling, both commonly disrupted in cancer [90,91,92].

Additionally, metabolites such as 3-methylindole and trimethylamine N-oxide suggest shifts in gut microbiota metabolism that may impact cancer progression or immune function [93, 94]. Detoxification metabolites, like D-glucurono-6,3-lactone, reflect an active response to cellular stress, arising from cancer or external treatments such as chemotherapy [95]. This aligns with the clinical profiles of the eight GI patients, two of whom underwent neoadjuvant chemotherapy.

Differential potential microbiota-derived metabolites in ascites: OC II-III vs. OC IV

In the comparison between OC stages II-III and IV, several metabolites reflect metabolic changes, immune modulation, and inflammatory responses typical of the tumor microenvironment, particularly in advanced stages. Metabolites such as mevalonic acid, butyryl-L-carnitine, and LPI 18:1 suggest shifts in lipid metabolism, likely driven by heightened energy demands, membrane synthesis, and signaling activity in ovarian cancer cells [96, 97]. Additionally, 3-hydroxyanthranilic acid, indole, and naringenin indicate the presence of immune-modulating and inflammatory metabolites, potentially promoting immune evasion and supporting a pro-inflammatory environment in advanced cancer [98,99,100,101].

LPI 18:1, a bioactive lipid involved in cell signaling, is linked to tumor growth, migration, and immune suppression, and its levels may be influenced by gut microbiota, especially in cases of dysbiosis [96, 97]. Indole, a microbial byproduct of tryptophan degradation, can affect immune responses and inflammation, potentially aiding immune evasion in cancer [100]. These findings underscore the interaction between microbial metabolites and the tumor microenvironment, highlighting the gut-tumor axis’s role in cancer progression and suggesting therapeutic strategies that target the microbiota to improve outcomes.

Naringenin, known for its anti-inflammatory, antioxidant, and anticancer effects [101,102,103], was elevated in OC II-III compared to OC IV. It modulates inflammation by suppressing cytokine production and enhancing cytokine degradation [104], while also regulating cell growth, apoptosis, and metastasis [105, 106]. Its higher levels in OC II-III suggest naringenin may help regulate immune responses and inhibit cancer progression in early stages, with reduced activity as the disease advances.

Interactions between microbiota-derived metabolites and cytokines/chemokines

To investigate how microbiota-derived metabolites interact with the immune landscape in cancer progression, we conducted correlation analyses between these metabolites and cytokines in OC vs. GI and OC stage II-III vs. IV comparisons. The anti-inflammatory cytokine IL-10 showed multiple positive correlations with metabolites such as glucosamine, D-tagatose, TMAO, caffeine, LPC 22:4, and LPC 20:1, suggesting these metabolites may contribute to immune suppression in the tumor microenvironment. This immune tolerance could facilitate tumor evasion from immune surveillance. Conversely, IL-10 was negatively correlated with benzamide and thymol, illustrating how different metabolites exert opposing influences on cytokine regulation. Notably, IL-10 in ascites has been associated with both the migration of ovarian cancer cells [107] and, in some cases, longer survival in patients receiving cell-free and concentrated ascites reinfusion therapy (CART) [108].

MCP-1 chemokine that recruits monocytes and macrophages to the tumor site, was positively correlated with metabolites like benzyl alcohol, naringenin, o-cresol, octadecanedioic acid, and phenol, potentially promoting immune cell recruitment and inflammation. In contrast, 1-methylhistidine, 4-pyridoxic acid, and mevalonic acid were negatively correlated with MCP-1 chemokine, potentially reducing immune cell recruitment. This dual influence suggests a complex delicate balance of pro- and anti-inflammatory signals in advanced ovarian cancer, where shifts in metabolite levels may influence MCP-1 bioactivity. Additionally, MCP-1’s positive association with infertility in endometriosis patients highlights its role in inflammatory immune reactions within the peritoneal cavity [109,110,111].

Together, these findings illustrate an intricate network of interactions between metabolites and cytokines/chemokines that likely impact cancer progression, immune evasion, and patient outcomes. Observed stage-dependent differences underscore the influence of metabolic shifts on immune responses within the tumor microenvironment. The balance of pro- and anti-inflammatory metabolites, in particular, may play a critical role in shaping this environment.

However, the small sample size in this study may increase the risk of type I statistical errors, potentially leading to false-positive findings. This is particularly relevant for the GI group, which included only 8 samples and exhibited substantial heterogeneity in cancer origin and histological subtype. Additionally, two patients had received neoadjuvant therapy, and one patient was male. We also did not include BMI as a parameter, and we would like to clarify this decision. In patients with malignant ascites, body weight does not accurately reflect true body mass due to the accumulation of large fluid volumes. This can distort BMI calculations and introduce further confounding into the analysis. While we initially considered including BMI, we ultimately excluded it to avoid misinterpretation. Future approaches that estimate or approximate dry body weight may offer more meaningful insights.

Other clinical variables—such as prior chemotherapy, antibiotic use, diet, medications, and infection status—were not controlled for in this exploratory study but are important potential confounders that should be considered in future research.

This study represents an initial step toward characterizing the metabolic landscape of ascites in cancer. Validation in larger, more homogeneous cohorts or through multi-center collaborations is strongly warranted. Expanding the GI group in particular would allow for meaningful stratification by cancer origin and histological subtype, thereby improving the precision and interpretability of the results.

Additionally, the detection and interpretation of bacterial metabolites could be enhanced by integrating LC-MS with targeted microbial profiling methods such as 16 S rRNA sequencing. This combined approach would enable more accurate characterization of the microbiome’s contribution to the ascitic metabolome and provide deeper insights into its role in cancer progression.

We summarized our findings, especially the microbiota-derived metabolomic differentiation in ascites samples from OC II-III and OC IV (Fig. 5).

Fig. 5

Graphical abstract to present microbiota-derived metabolomic differentiation in ascites samples from OC II-III and OC IV. Red and blue arrow means increase and decrease in OC IV ascites compare to OC II-III

Conclusion

This study utilized novel trapped ion mobility spectrometry time of flight (timsTOF) mass spectrometry (MS) to profile metabolites in malignant ascites samples, comparing gastrointestinal cancers and different stages ovarian cancers. Both reversed-phase and hydrophilic interaction liquid chromatography were employed to ensure comprehensive separation of non-polar and polar metabolites. Key findings revealed distinct potential microbiota-derived metabolic changes in OC versus GI cancers and across ovarian cancer stages, particularly involving lipid metabolism, with significant alterations in sphingolipid and phospholipid pathways. Additionally, potential microbiota-derived metabolites were identified and correlated with cytokine and chemokine levels, indicating a possible interaction between microbiota and the immune response in ovarian and gastrointestinal cancers. These findings show the presence of microbiota-derived metabolites in malignant ascites and contribute as first step to a better understanding of the interplay between microbiota and intestinal malignancies. Future studies upon malignant ascites shall follow, assessing the exact impact of individual microbial metabolites upon the tumor microenvironment to improve patient care and support the success of novel cancer therapies.