Gut microbiota-derived indole-3-propionic acid promotes lymph node metastasis in gastric cancer via the aryl-hydrocarbon receptor signaling pathway

Abstract

Background

Gut microbiota (GM) regulates the tumor microenvironment through microbial metabolites. Indole 3-propionic acid (3-IPA) is one such metabolite that regulates gastrointestinal barrier function. In this study, we investigated the effects of 3-IPA on the progression of lymph node metastasis of gastric cancer (GC) and the molecular mechanisms that underlie them.

Methods

The microbial metabolites were identified using a fecal metabolomic assay in GC patients. Lymphangiogenesis was evaluated using tube formation and wound healing assays in vitro. The expression of aryl hydrocarbon receptor (AHR), CYP1A1, and vascular endothelial growth factor receptor 3 (VEGFR3) were assayed using quantitative real-time PCR (qRT-PCR) and western blot (WB) analyses. Matrigel plug and popliteal lymph node metastasis model were employed to validate the influence on lymphangiogenesis and lymph node metastasis in vivo.

Results

Fecal metabolomic and microbiome profiling was drastically different between GC patients with lymph node metastasis (GC-LM) and those without metastasis. The GC-LM group showed high 3-IPA expression in the feces; 3-IPA had no significant effect on GC cells; Human lymphatic endothelial cells showed greater tube formation and promoted migration after 3-IPA administration. Also, upregulation of AHR, CYP1A1, and VEGFR3 was observed. Moreover, administration of the AHR inhibitor suppressed tube formation and lymph node metastasis both in vitro and in vivo.

Conclusions

Our findings suggest that gut microbiota-derived 3-IPA functions as a lymph node metastasis promoter through the AHR/CYP1A1–VEGFR3 axis in GC. 3-IPA could serve as a prognostic biomarker and conceivably a therapeutic target for GC lymph node metastasis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40170-026-00438-1.

Introduction

Gastric cancer (GC) is one of the most prevalent malignant tumors of the digestive system worldwide. Lymph node metastasis is the primary route of spread, which can occur early in the progression of the disease, with up to 70% of advanced gastric cancers exhibiting lymph node involvement [1, 2]. Research has shown that lymph node metastasis is an independent predictor of poor prognosis in patients with gastric cancer. Furthermore, such metastasis can facilitate subsequent hematogenous spread [3]. These findings highlight the importance of lymph node metastasis as a critical determinant of patient outcomes in gastric cancer [4, 5].

The gut microbiota (GM), frequently described as the human second genome, plays a multifaceted role in the regulation of intestinal functions, nutrition, and metabolism, and is intricately associated with the initiation and progression of cancer. The association between GM and the development of colorectal cancer has been the subject of extensive characterization [6–8]. Nevertheless, research investigating the role of GM in the regulation of tumorigenesis in distant organs remains relatively scarce. In the context of gastric cancer, the gastric microbiota, especially Helicobacter pylori, has been shown to promote angiogenesis and lymphangiogenesis [9, 10]. The potential of GM to influence lymphatic metastasis in gastric cancer, as well as the underlying mechanisms, has not been thoroughly investigated. Although GM does not directly interact with gastric tissue during the progression of gastric cancer, it can indirectly modulate the tumor immune microenvironment through its metabolic byproducts and derivatives, thereby impacting cancer progression [11, 12]. This study breaks new ground by shifting the focus to the impact of gut microbiota and its metabolites on lymphangiogenesis in gastric cancer, rather than concentrating solely on the role of gastric microbiota in this process.

The present study employed non-targeted metabolomics to investigate the differences in small molecule metabolites in intestinal faeces of gastric cancer patients with and without lymph node metastasis. The findings suggest that the intestinal flora metabolite indole 3-propionic acid (3-IPA) may be associated with gastric cancer lymphatic metastasis. Subsequent experiments demonstrated that IPA does not exert a direct effect on gastric cancer cells; rather, it participates in gastric cancer metastasis by inducing lymphangiogenesis in human lymphatic endothelial cells. A review of the literature and molecular docking studies revealed that IPA may promote lymphatic metastasis of gastric cancer by binding to AHR and activating the AHR-CYP1A1–VEGFR3 signalling pathway, which induces lymphangiogenesis in lymphatic endothelial cells. Furthermore, in vivo and in vitro experiments demonstrated that the process of lymphangiogenesis was markedly diminished following the administration of the AHR inhibitor CH223191. The present study sought to elucidate the intrinsic mechanism of GM-GC linkage, thereby offering a novel perspective on the regulation of GM progression in distant organs.

Methods

Patients

The clinical protocols recruiting patients were approved by the Ethics Committee of Shandong Provincial Hospital (NSFC: NO. 2020–442). All experiments and therapeutic procedures adhered to relevant guidelines. We enrolled patients diagnosed with GC by pathology at Shandong Provincial Hospital between August 2020 and November 2021 in the study. All the patients volunteered to participate in the study and the informed consent forms were signed. All samples were harvested preoperatively. However, patients with distant metastases detected by imaging evaluation before surgery, received preoperative neoadjuvant radiotherapy, diagnosed with inflammatory bowel diseases were excluded from the experiment. The 8th edition of the American Joint Committee on Cancer tumor/node/metastasis (T/N/M) system was used for lymph node staging and histological classifications after GC resection. Data on age, gender, cell differentiation, size and site of tumor, and TNM stage were collected for each patient. An adequate amount of stool samples was collected and stored properly at −80 °C until analysis.

Microbial metabolism analysis

For fecal metabolomics, approximately 100 mg of fecal content was collected from each patient before surgery and ground at −80 °C, and the homogenate was resuspended in prechilled 80% methanol and 0.1% formic acid by vortexing. The samples were incubated on ice for 5 min and then centrifuged at 15,000 × g at 4 °C for 20 min. The supernatant was diluted to a final concentration of 53% methanol using LC-MS grade water. The samples were subsequently transferred to fresh Eppendorf tubes and centrifuged at 15,000 × g at 4 °C for 20 min. Quality control (QC) samples were prepared by pooling aliquots of all fecal samples and were processed using the same procedure as that used for the experimental samples. Finally, the samples were injected into the LC-MS/MS system for analysis [13]. LC-MS/MS analyses were performed using a Vanquish UHPLC system (Thermo Fisher Scientific, Germany) coupled with an Orbitrap Q Exactive TMHF-X mass spectrometer (Thermo Fisher Scientific, Germany) at Novogene Co., Ltd. (Beijing, China). Chromatographic separation was performed using a Hypesil Gold column (100 × 2.1 mm, 1.9 μm) and a 17-min linear gradient at a flow rate of 0.2 mL/min. The eluents for the positive polarity mode were 0.1% formic acid in water (eluent A) and methanol (eluent B). The eluents for the negative polarity mode were 5 mM ammonium acetate at pH 9.0 (eluent A) and methanol (eluent B). The solvent gradient was set as follows: 2% B, 1.5 min; 2–100% B, 12.0 min; 100% B, 14.0 min; 100–2% B, 14.1 min; 2% B, 17 min. The Q Exactive TMHF-X mass spectrometer was operated in positive/negative polarity mode with a spray voltage of 3.2 kV, capillary temperature of 320 °C, sheath gas flow rate of 40 arb, and auxiliary gas flow rate of 10 arb [14]. Metabolic analyses were performed using Compound Discoverer 3.1 software (Thermo Fisher Scientific, Germany). Metabolites with fold change > 2 and p-value < 0.05 were identified as differential metabolites. Based on the found differential metabolites, pathways enrichment was studied using the KEGG database.

Gut microbiome analysis

Fecal microbial genomic DNA was extracted using the CTAB/SDS method, and DNA concentration and purity were monitored on 1% agarose gels. The bacterial 16S rRNA gene V3–V4 region was amplified using region-specific primers with sample-specific barcodes. All PCR mixtures contained 15 μL Phusion® High-Fidelity PCR Master Mix (New England Biolabs), 0.2 μM of each primer, and 10 ng of template DNA. PCR cycling conditions were 98 °C for 1 min, followed by 30 cycles of 98 °C for 10 s, 50 °C for 30 s, and 72 °C for 30 s, with a final extension at 72 °C for 5 min. PCR products were purified using the Qiagen Gel Extraction Kit. Sequencing libraries were generated using the NEBNext® Ultra™ II DNA Library Prep Kit (E7645) following the manufacturer’s recommendations. Library quality was evaluated using a Qubit 2.0 Fluorometer and an Agilent Bioanalyzer 2100, and libraries were sequenced on an Illumina NovaSeq platform to generate 250-bp paired-end reads.

Paired-end reads were assigned to samples based on unique barcodes and merged using FLASH (v1.2.11) to generate raw tags. Quality filtering was performed using fastp (v0.20.0) to obtain high-quality clean tags. Chimera sequences were detected and removed using Vsearch (v2.15.0) against the SILVA reference database to generate effective tags. Amplicon sequence variants (ASVs) were inferred using the DADA2 (default) module in QIIME2 (QIIME2-202006), and ASVs with abundance < 5 were filtered out. Taxonomic assignment and phylogenetic analyses were performed in QIIME2 using the SILVA database. For downstream diversity analyses, sequencing depth was normalized by rarefying each sample to the minimum number of reads across all samples, and alpha and beta diversity metrics were calculated based on the normalized ASV table.

Molecular docking

Chemical structures of the drugs and 3D protein structures were obtained from the PubChem and AlphaFold databases, respectively. Ligands and proteins were preprocessed, including hydrogen addition and charge assignment, using AutoDockTools. Grid boxes were configured to fully encompass the anticipated active sites. Molecular docking was executed via AutoDock 4.2.6 using 100 iterations, with default settings for search space, docking algorithm, and population size. A binding energy threshold of < −5 kcal/mol was used as a screening criterion to indicate potential binding, together with pose convergence and reasonable binding geometry. Finally, binding energies, hydrogen bond formations, and ligand-protein interactions were evaluated and visualized using PyMOL 3.0.0.

Cell culture

MKN-45 cells were donated by Key Laboratory for Experimental Teratology of the Ministry of Education, Department of Pathology, School of Basic Medical Sciences, Shandong University. AGS cells were purchased from ATCC. Both the cell lines were cultured in RPMI-1640, containing 10% heat-inactivated fetal bovine serum (FBS), 1% penicillin/streptomycin, in humidified 5% CO₂/95% atmosphere at 37 °C. Human lymphatic endothelial cells (HLEC) were obtained from ScienCell Carlsbad, CA, USA）and cultured in ECM, containing 5% FBS, 1% penicillin/streptomycin.

Viral transfection

Lentivirus with eGFP-Tag and luciferase were obtained from Genomeditech, Shanghai, China. MKN-45 cells were transfected with complete RPMI-1640 medium containing lentivirus and 0.1% polybrene for 24-48 h, following the selection using puromycin for a week.

Administration of AHR inhibitor CH223191

The AHR inhibitor CH223191 (MCE, Monmouth Junction, NJ, USA) was dissolved in 20% dimethyl sulfoxide (DMSO; Solarbio, Beijing, China) and applied to HLECs at a concentration of 10 µM for 24 h to inhibit AHR. The cells in the NC group were treated with 20% DMSO.

Tube formation assays

Tube formation assays were performed using growth factor–reduced Matrigel (product# 354,248; Corning, NY, USA). Ninety-six–well plates were pre-chilled on ice, and each well was coated with 50 μL Matrigel and allowed to polymerize for 60 min at 37 °C. HLECs) were harvested at 70–80% confluence, resuspended in complete endothelial medium, and seeded onto polymerized Matrigel at 2 × 10⁴ cells/well (equivalent to 100 μL of a 2 × 10⁵ cells/mL suspension). Cells were treated with vehicle DMSO or 3-IPA at the indicated concentrations immediately after seeding. Where indicated, cells were pretreated with CH223191 for 24 h before seeding and maintained in the corresponding treatment condition during the assay.

Plates were incubated at 37 °C with 5% CO₂ and imaged at a fixed time point (6 h) using an inverted microscope under identical acquisition settings (objective 4× or 10×, consistent exposure and focus). For each well, 5 non-overlapping fields were captured at predefined positions (center and four quadrants). Each condition was plated in triplicate wells, and experiments were repeated independently at least three times.

Tube formation was quantified in a blinded manner using ImageJ (NIH) with the Angiogenesis Analyzer plugin. All images were processed using an identical workflow (background subtraction followed by a fixed threshold for binarization) applied uniformly across all groups. The primary endpoint was the number of meshes (reported as “Mesh number” by Angiogenesis Analyzer), which was used as a surrogate for “tube number” (capillary-like network formation). For each well, five non-overlapping fields were analyzed and averaged to obtain one value per well. Each condition was plated in triplicate wells, and experiments were repeated independently at least three times.

Wound-healing assays

The migration capacity of cell lines was determined using a wound-healing assay. Suspension of cells (2 × 10⁶) was seeded into each well of a 6-well plate and incubated at 37 °C, 5% CO₂ for 12 h to allow the cells to adhere. After decanting the medium, the monolayer was scratched using a 200 µl pipette tip and then washed with PBS twice. The fresh medium containing 1% FBS, followed by the addition of DMSO or 3-IPA at the indicated concentrations, was added to the well and cultured in humidified 5% CO₂/95% atmosphere at 37 °C. The scratches were photographed at 0, 24, and 48 h using a microscope with a CCD color camera attached to the microscope, and wound closure was calculated using the Image J software.

Transwell cell invasion assays

Transwell assays were carried out to measure cell migration and invasion. 100 μL of the diluted Matrigel matrix (Corning, NY, USA) was incubated at 37 °C for 1 h to form gel in the upper chambers. 200 μL of cell suspension with a density of 4 × 10⁵ cells/mL was added into inserts and cultured with serum-free RPMI-1640 medium. The lower chambers were coated with 600 μL medium containing 10% FBS. Perforated cells in the lower chamber were fixed with 4% paraformaldehyde for 20 min, stained with 0.1% crystal violet for 30 min after routine culture for 24 h. The slides were observed and captured randomly using a microscope with a camera attached to, and the number of invaded cells was calculated using the Image J.

Trans-endothelial cell migration assays

Trans-endothelial cell migration assays were employed to evaluate the ability of GC cells to migrate through endothelial cell barriers. 100 μL of endothelial cell suspension (1 × 10⁵/mL) was placed in each Transwell upper chamber. The non-adherent HLEC were removed and 600 μL of complete RPMI-1640 medium was added to each lower chamber the next day. Then 200 μL of MKN-45 cells with eGFP-Tag (3 × 10⁵/mL) were shifted into the upper chamber for 12-24 h. After the migration, the whole process of fixing and staining was under protection from lights. Observation and quantification of cells with fluorescence by EVOS M7000 (Thermo Fisher Scientific, Waltham, MA, USA).

Western blot analysis of AHR, CYP1A1, and VEGFR3 levels

Cells were washed with PBS twice and then lysed in RIPA lysis buffer supplemented with phenylmethanesulfonylfluoride (PMSF; Solarbio, Beijing, China) in a ratio of 1:100 for 25 min on ice. Protein concentrations in the lysate were assayed using a BCA Protein Assay Kit (Solarbio, Beijing, China). Proteins were resolved using 10% SDS-PAGE and transferred to a polyvinylidene fluoride membrane. The PVDF membrane was blocked for 1 h using 5% skimmed milk in TBS containing 0.05% Tween® 20 and then incubated with primary antibodies (mouse-anti-AHR, 67,785–1-Ig; rabbit-anti-CYP1A1, 13,241–1-AP at 1:5000 dilution; mouse-anti-β-actin（ACTB）, 60,004–1-Ig at 1:1000 dilution and obtained from Proteintech, Wuhan, China; rabbit-anti-VEGFR3, ab27278 at 1:1000 dilution and obtained from Abcam, Cambridge, UK) at 4 °C for 16 h. After washing three times with TBSt (10 min each), the membrane was treated by secondary antibodies (SA00001-1, SA00001-2 at 1:5000 dilution and obtained from Proteintech, Wuhan, China) for 1 h at room temperature and subsequently washed again with TBSt. In the end, the membranes were incubated with ECL regent (Beyotime, Shanghai, China) for target protein detection. Further gray value analysis of the blots was conducted using the Image J software. At last, data are presented as the expression ratio of the target protein amount relative to that of β-actin, taking the mean expression in the NC group as normalization.

qRT-PCR analysis of AHR, CYP1A1, and VEGFR3 mRNA levels

Total RNA from cell lines was extracted using the Total RNA Extraction Reagent (Vazyme, Nanjing, China), and the RNA underwent qualification and concentration determination using NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA). 1 μg total RNAs were reverse transcribed to produce cDNA, according to the manual instructions (Vazyme, Nanjing, China). mRNA expression levels were analyzed using the QuantStudio™ 1 System (Thermo Fisher Scientific, Waltham, MA, USA). primers (Sangon Biotech, Shanghai, China), and 10 μL of SYBR qPCR Premix (Vazyme, Nanjing, China). Quantitative real-time PCR was performed using a 20 μL reaction system consist of cDNA, primers (Sangon Biotech, Shanghai, China) and ChamQ Universal SYBR qPCR Master Mix (Vazyme, Nanjing, China) following the steps of pre-denaturation at 95 °C for 30s, then 40 cycles of denaturation at 95 °C for 15 s, and annealing and extension at 60 °C for 30 s, followed by the pre-set melt curve. At last, the expression of target mRNA was normalized to that of β-actin using the 2^-ΔΔCt method.

Animals

All animals were purchased from Viton Lihua (Beijing, China) and raised in the animal center of Shandong Provencial Hospital. Mice were kept under SPF conditions for at least a week before experiment procedures were taken. All experimental procedures were approved by the Institutional Animal Care and Use Committee of Shandong Provencial Hospital.

Matrigel plug assays

On day 0, 7–8-week-old male C57BL/6 mice (20–25 g) were shaved on the abdomen and injected subcutaneously on both abdominal flanks with 400 μL Matrigel per site. For the negative control group, Matrigel was diluted 1:1 in PBS. For the positive control group, Matrigel was diluted 1:1 in PBS containing VEGF-A (500 ng/mL), bFGF (500 ng/mL), and S1P (2 μM). Mice were randomly assigned to treatment groups.

During the experiment, mice received intraperitoneal injections of 3-IPA (10 mg/kg/day) and/or CH223191 (10 mg/kg/day) once daily, with both compounds dissolved in 0.9% normal saline and administered at an injection volume of 10 mL/kg. Vehicle controls received the same volume of normal saline on the same schedule.

On day 28, mice were euthanized and Matrigel plugs were harvested. Lymphatic vessels within plugs were assessed by immunofluorescence staining for CD31 and LYVE-1. Quantification was performed using ImageJ; for each plug, multiple non-overlapping fields were analyzed and averaged to generate one value per plug. Statistical analyses were performed using GraphPad Prism, with the plug treated as the experimental unit.

Popliteal lymph node metastasis model

5–6 weeks old BALB/c-nude mice (18–20 g) were randomly allocated to three groups (n = 5 per group). On day 0, 5 × 10⁵ MKN-45 cells stably expressing luciferase were inoculated into the footpads.

Mice were treated by intraperitoneal injection once daily with 3-IPA (10 mg/kg/day) and/or CH223191 (10 mg/kg/day) from day 0 to day 28. Both compounds were dissolved in 0.9% normal saline and administered at an injection volume of 10 mL/kg. Vehicle controls received the same volume of normal saline on the same schedule.

On day 28, mice were injected intraperitoneally with D-luciferin potassium salt (Beyotime, Shanghai, China) according to the manufacturer’s instructions. Bioluminescence signals from the primary footpad tumors and popliteal lymph nodes were acquired using an IVIS Imaging System and quantified as total flux (photons/sec) with regions of interest (ROIs) placed over the tumor and popliteal lymph node. Mice were euthanized immediately after imaging. Primary tumors and popliteal lymph nodes were collected, fixed, paraffin-embedded, sectioned at 4 μm, and evaluated by immunohistochemistry.

Immunohistochemistry (IHC) and immunofluorescence (IF)

IHC staining was carried out using the IHC Kit (PV-9000, ORIGENE, Beijing, China). Paraffin-embedded slides were incubated with primary antibodies (mouse-anti-AHR, 67,785–1-Ig, at 1:200 dilution, obtained from Proteintech, Wuhan, China; rabbit-anti-LYVE-1, ab218535, at 1:2000 dilution; rabbit-anti-Luciferase, ab185924, at 1:500 dilution, and both obtained from Abcam, Cambridge, UK). After washing three times with PBS, slides were treated with secondary antibodies, then stained with DAB and hematoxylin.

For IF staining, slides were treated with primary antibodies (mouse-anti-CD31, sc-376764, at 1:100 dilution, obtained from Santa cruz biotechnology, Dallas, Texas, USA; rabbit-anti-LYVE-1, ab218535, at 1:5000 dilution, obtained from Abcam, Cambridge, UK), washed and incubated with secondary antibodies (coraLite594-conjugated goat-anti-mouse IgG(H+L), SA00013-3, at 1:200 dilution; coraLite488-conjugated goat-anti-rabbit IgG(H+L), SA00013-2, at 1:500 dilution; both obtained from Proteintech, Wuhan, China), stained with DAPI reagent, then observed and captured with EVOS M7000 (Thermo Fisher Scientific, Waltham, MA, USA).

Statistical analysis

All quantitative data are expressed as the mean ± SD of three replicates. Statistically significant differences among all groups were determined using the Fisher’s exact test or Student’s t-test. Differences among multiple groups were analyzed by ANOVA. Statistical analyses were performed using SPSS v.26.0 (IBM Corp., Armonk, NY, USA), and p < 0.05 was considered statistically significant.

Results

Fecal metabolomic and gut microbiota profile differs between GC-LM and GC with no LM (GC-nLM)

Fecal samples from patients who met the inclusion criteria were collected before surgery. Table 1 shows lymph node metastasis status, including the AJCC TNM stage and univariate analyses for multiple factors such as age, gender, and histological grade. LC-MS/MS analyses were performed to quantify 2099 metabolic features. Based on the partial least squares discrimination analysis (PLS-DA) of these quantitative data, two distinct metabolomic profiles corresponding to the GC-LM and GC-nLM samples, respectively, were revealed (Fig. 1A). A “Y-scrambling” statistical validation was applied to test the predictive ability of PLS-DA (Fig. S1A, Supporting Information). Moreover, 74 microbial metabolic features were upregulated (ratio_GC-LM/GC-nLM > 2, p < 0.05) and 26 metabolic features were downregulated (ratio_GC-LM/GC-nLM < 0.5, p < 0.05) (Fig. 1B). The microbial metabolic features upregulated in the GC-LM samples were highly enriched in amino acids and lipid metabolism. Notably, the metabolomic analysis identified significant changes in 3-IPA abundance between the GC-LM and GC-nLM samples (Fig. 1C). To investigate whether these metabolomic alterations were associated with corresponding changes in the gut microbial community, we next analyzed the fecal microbiome profiles of the same patient cohort. The T-test was employed to find out the significantly different species at each taxonomic level, and GC-LM samples were significantly abundant in Escherichia-Shigella, Coprococcus, and Marvinbryantia at the genus level (p < 0.05, Fig. 1D). We also conducted Pearson’s correlation analysis on the relationship between the gut microbiota and its metabolic features in the fecal samples of patients. Because 16S rRNA sequencing provides taxonomic rather than functional resolution, these correlations indicate an association between microbial community shifts and fecal 3-IPA abundance, but do not establish a causal microbial tryptophan/indole metabolic pathway responsible for 3-IPA production. Strong correlations were noticed between the especially focused gut microbiota and metabolic features (p < 0.05, Fig. 1E). The results were visualized in chord diagram as well (Fig. S1D, Supporting Information).

Table 1.

Basic information and lymph node metastasis status in enrolled patients (n = 15)

Factors	Lymph node metastasis(n = 9)	Non-lymph node metastasis(n = 6)	p value*
Age(year)
＜65	6	5
≥65	3	1	ns
Gender
Male	8	6
Female	1	0	ns
Cell differentiation
Poor differentiation	4	3
Moderate differentiation	5	3	ns
Tumor size
＜1 cm	4	4
≥1 cm	5	2	ns
Site of tumor
Cardia	1	1
Body	3	2
Antrum	5	3	ns
Lymphatic metastasis
N0	0	6
N1	7	0
N2	2	0	＜0.01
Depth of cancer invasion
T1	5	4
T2	4	2	ns
Distal metastasis
Positive	0	0
Negative	9	6	ns

^*Significant thresholds set both to p < 0.05 and p < 0.01, ns—no significance, Fisher’s exact test

Fig. 1.

Fecal metabolomic and gut microbiota profile differs between lymph node metastasis in GC (GC-LM) and GC with no lymph node metastasis (GC-nLM). (A) Partial least squares discrimination analysis (PLS-DA) shows significant difference between GC-LM and GC-nLM in both positive and negative metabolites; (B) Volcano plot comparing the relative levels of the metabolites between GC-LM and GC-nLM. Significant thresholds set to p < 0.05, Student’s t-test; (C) Relative abundance of indole 3-propionic acid (3-IPA) is up-regulated in GC-LM group, *p < 0.05; (D) Abundance of Escherichia-Shigella, Coprococcus, and marvinbryantia are enriched at the genus level in GC-LVI group, p < 0.05; (E) the heatmap of the correlations between the concerned gut microbiome and metabolites at the genus level. Red and blue represent positive and negative correlations, respectively. The darker the color, the closer the relationship, *p < 0.05

The administration of 3-IPA has no significant effect on GC cells

Aiming at identifying 3-IPA modulating the migration and invasion ability of GC in vitro, colony formation assays, wound-healing assays, and transwell assays were applied in both AGS and MKN-45 cell lines. Interestingly, there were no significant difference between the 3-IPA (10 μM) and NC group (Fig. S2 A-F), indicating 3-IPA most likely promoted LM via other ways instead of facilitating migration or invasion capability of GC cells.

The administration of 3-IPA promotes endothelial tube formation in vitro

To assess the role of 3-IPA in lymphangiogenesis, we first performed CCK-8 assays in HLECs and observed no significant difference in cell proliferation between groups (Fig. 2A). We next evaluated the effect of 3-IPA on endothelial cell migration using a wound-healing assay. Compared with the control group, HLECs treated with 3-IPA exhibited significantly enhanced wound closure (p < 0.05, Fig. 2B). Consistently, 3-IPA also promoted tube formation in HLECs compared with the control group. Quantitative analysis further showed a dose-dependent increase in tube formation after 3-IPA treatment (p < 0.01, Fig. 2C).

Fig. 2.

3-IPA promotes lymphangiogenesis in vitro and is associated with activation of AhR signaling and increased CYP1A1/VEGFR3-related responses. (A) CCK-8 analysis of human lymphatic endothelial cells (HLECs) treated with increasing concentrations of 3-IPA. (B) Wound-healing assay of HLECs treated with 3-IPA. Representative images were captured at 0, 24, and 48 h. Scale bar: 200 μm. (C) Tube formation assay of HLECs treated with a concentration gradient of 3-IPA. Representative images and quantification of total tube numbers are shown. Scale bar: 100 μm. (D) Pathway/network analysis indicating the association of 3-IPA with lymphangiogenesis-related signaling, including AHR, CYP1A1, and VEGF-related pathways. (E) Molecular docking models of 3-IPA with AhR and the predicted CYP1A1/VEGFR3-related signaling context. (F) qRT-PCR analysis of AHR, CYP1A1, and VEGFR3 mRNA expression in HLECs after 3-IPA treatment. Transcript levels were normalized to β-actin. (G) Western blot analysis of AHR, CYP1A1, and VEGFR3 protein expression in HLECs treated with 3-IPA. β-actin was used as the loading control.Error bars represent the mean ± SD from three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001

AHR, CYP1A1, and VEGFR3 are significantly overexpressed after the application of 3-IPA

Ingenuity pathway analysis (IPA) was used to identify signaling pathways potentially linking 3-IPA to lymphangiogenesis and lymph node metastasis. IPA highlighted AHR signaling and associated downstream nodes, including CYP1A1 and VEGFR3, as candidate mediators relevant to the observed phenotype (Fig. 2D). Molecular docking further suggested potential interactions between 3-IPA and AHR-related signaling components (Fig. 2E).

We then experimentally examined the expression of these candidate mediators in HLECs. Compared with the control group, 3-IPA treatment significantly increased the mRNA levels of AHR (p < 0.001), CYP1A1 (p < 0.01), and VEGFR3 (p < 0.001) (Fig. 2F). Consistent with the qRT-PCR results, western blot analysis further showed that 3-IPA increased the protein expression of AHR, CYP1A1, and VEGFR3 (p < 0.01, Fig. 2G). Together, these findings indicate that 3-IPA is associated with activation of AhR signaling, accompanied by induction of CYP1A1 and VEGFR3-related lymphangiogenic responses in HLECs.

3-IPA facilitates lymphangiogenesis via the AHR/CYP1A1–VEGFR3 pathway in vitro

To examine the role of the AHR/CYP1A1–VEGFR3 pathway in the ability of 3-IPA to improve lymph node metastasis in GC, the AHR inhibitor (AHRi) CH223191 was administered to HLEC cultured in a medium enriched with 3-IPA. Proliferation assays were conducted and no difference between groups was found within the working concentration range (p > 0.05, Fig. 3A). Notably, CH223191 weakened the ability of endothelial cells to migrate (p < 0.05, Fig. 3B), or to form tubes (p < 0.05, Fig. 3C) compared to the 3-IPA administration group. The trans-endothelial migration of GC cells was inhibited in the CH223191 group as well (p < 0.01, Fig. 3D). Additionally, qPCR also showed that the AHR inhibitor significantly downregulated the mRNA levels of AHR/CYP1A1 and VEGFR3 compared with the group treated only with 3-IPA (p < 0.05, Fig. 3E). Similarly, protein expression analyses revealed that CH223191 significantly decreased the levels of AHR, CYP1A1, and VEGFR3 proteins compared to the 3-IPA group (p < 0.05; Fig. 3F). Together, these data demonstrated that the suppression of the AHR/CYP1A1 pathway counteracts the effect of 3-IPA on lymphangiogenesis.

Fig. 3.

3-IPA promotes lymphangiogenesis in vitro, and these effects are attenuated by AHR inhibition. (A) CCK-8 analysis showing that the AhR inhibitor CH223191 does not significantly affect HLEC proliferation in the presence of 3-IPA. (B) Wound-healing assay showing that CH223191 attenuates 3-IPA-induced migration of HLECs. Representative images were captured at 0, 24, and 48 h. Scale bar: 200 μm. (C) Tube formation assay showing that CH223191 reduces 3-IPA-induced capillary-like network formation in HLECs. Representative images and quantification of total tube numbers are shown. Scale bar: 100 μm. (D) Trans-endothelial migration assay showing that CH223191 reduces the ability of eGFP-tagged MKN-45 cells to migrate through the HLEC monolayer. Representative images and quantification of migrated cell numbers are shown. Scale bar: 125 μm. (E) qRT-PCR analysis showing that CH223191 decreases the mRNA expression of AHR, CYP1A1, and VEGFR3 in HLECs in the presence of 3-IPA. Transcript levels were normalized to β-actin. (F) Western blot analysis showing that CH223191 decreases the protein expression of AHR, CYP1A1, and VEGFR3 in HLECs treated with 3-IPA. β-actin was used as the loading control. Error bars represent the mean ± SD from three independent experiments. *p < 0.05, **p < 0.01, ***p < 0.001. AHRi, AhR inhibitor

3-IPA facilitates lymphangiogenesis and lymph node metastasis through AHR signaling pathway in vivo

To evaluate the effect of 3-IPA and AhR inhibition on lymphangiogenesis in vivo, we performed a Matrigel plug assay. Matrigel diluted 1:1 with PBS served as the negative control, whereas Matrigel supplemented with a lymphangiogenic VFS cocktail served as the positive control (Fig. 4A). Histological examination by H&E staining and immunohistochemical analysis revealed increased vascular infiltration in both the PBS and VFS subgroups after 3-IPA administration, accompanied by higher AhR expression, compared with the corresponding control groups (Fig. 4B, C). To further assess lymphangiogenesis, newly formed CD31⁺/LYVE-1⁺ lymphatic vessels beneath the implanted plugs were identified and quantified. Notably, the pro-lymphangiogenic effect of 3-IPA was significantly attenuated by co-administration of the AhR inhibitor (p < 0.01, Fig. 4D).

Fig. 4.

3-IPA facilitates lymphangiogenesis and lymph node metastasis through AhR signaling in vivo. (A) Schematic diagram of the Matrigel plug assay. (B) Representative H&E staining images of Matrigel plugs. Scale bars: upper, 100 μm; lower, 20 μm. (C) Representative immunohistochemical (IHC-P) staining images of AhR in Matrigel plugs. Scale bars: upper, 100 μm; lower, 20 μm. (D) Representative immunofluorescence images of CD31 and LYVE-1 in Matrigel plugs. Scale bar: 275 μm. (E) Schematic diagram of the popliteal lymph node metastasis model. (F, G) IVIS imaging of the popliteal lymph node metastasis model (n = 5) and quantification of bioluminescence in (F) foot-pad tumors and (G) popliteal lymph nodes. (H) Gross images of sample harvest after completion of the experiment. (I) Gross images of foot-pad tumors. (J, K) Representative IHC-P staining images of AHR and LYVE-1 in foot-pad tumors. Scale bars: upper, 100 μm; lower, 20 μm. (L) Gross images of popliteal lymph nodes. (M) Representative IHC-P staining images of luciferase in popliteal lymph nodes and quantification of metastatic cells. Scale bars: upper, 100 μm; lower, 20 μm. (N) Representative IHC staining images of CD31, D2-40, and AhR in clinical gastric cancer specimens from the GC-NLNM and GC-LNM groups. Scale bar: 100 μm.Error bars represent the mean ± SD from three independent experiments. *p < 0.05, **p < 0.01

To determine whether 3-IPA promotes lymph node metastasis in vivo, we next established a popliteal lymph node metastasis model. Luciferase-labeled gastric cancer cells were inoculated into the footpads of immunodeficient mice, and metastatic dissemination was assessed by IVIS imaging and immunohistochemical staining of popliteal lymph node sections using an anti-luciferase antibody (Fig. 4E). At week 4, no significant difference in primary footpad tumor burden was observed among the three groups (p > 0.05, Fig. 4F, I). In contrast, bioluminescence signals in the popliteal lymph nodes were markedly increased in the 3-IPA-treated group, particularly when the primary footpad tumors were shielded during imaging (p < 0.05, Fig. 4G). Immediately after IVIS imaging, footpad tumors and popliteal lymph nodes were harvested for gross and histological analyses (Fig. 4H).

As shown in Fig. 4I, primary tumor size did not differ significantly among groups. However, AhR expression was increased in the 3-IPA group and reduced after AhR inhibitor treatment, accompanied by parallel changes in LYVE-1⁺ lymphatic vessels (Fig. 4J, K). In addition, popliteal lymph nodes from 3-IPA-treated mice were visibly enlarged compared with those from the other groups (Fig. 4L). Immunohistochemical staining for luciferase further showed a markedly higher number of metastatic tumor cells in popliteal lymph nodes from the 3-IPA group, whereas this effect was attenuated by AhR inhibition (p < 0.01, Fig. 4M).

To further assess the clinical relevance of these findings, we performed immunohistochemical staining of human gastric cancer specimens with and without lymph node metastasis. Compared with the GC-NLNM group, the GC-LNM group exhibited increased staining for CD31, D2-40, and AhR, supporting enhanced angiogenic/lymphangiogenic activity and AhR signaling in metastatic clinical samples (Fig. 4N). Collectively, these in vivo and clinical data indicate that 3-IPA promotes lymphangiogenesis and facilitates lymph node metastasis in an AhR-dependent manner.

Discussion

Gastric cancer is recognized as one of the most prevalent malignant tumors within the digestive tract, with lymphatic metastasis and recurrence serving as significant determinants of patient prognosis. Consequently, the prevention and reduction of lymph node metastasis in gastric cancer have emerged as critical areas of contemporary research. The advancement of sequencing technologies, including microbiomics and metabolomics, has increasingly highlighted the role of gut microbiota in the onset and progression of digestive tract tumors, particularly in relation to colon cancer proliferation and metastasis. Notably, since the gut microbiota does not directly interact with gastric cancer tissues, its influence on gastric cancer may be easily underestimated. Our clinical observations indicate that there are discernible differences in gut microbiota composition and fecal metabolomics between early gastric cancer patients with lymph node metastasis and those without. This suggests that the gut microbiota may exert its effects on gastric cancer not through direct contact, but rather via its metabolic byproducts. However, current research on the impact of gut microbiota metabolites on lymphatic and angiogenic processes in gastric cancer remains limited, and the underlying mechanisms are yet to be elucidated. We note that fecal metabolite levels reflect microbial metabolic output and do not directly measure gastric tissue metabolism; therefore, we integrated fecal profiling with functional validation to define how 3-IPA mechanistically impacts lymphatic endothelial cells and lymphatic metastasis.

In this study, we confirmed phenotypic differences in gut microbiota composition and fecal metabolomics between early gastric cancer patients with lymph node metastasis and those without. Our 16S analysis identified genera enriched in the GC-LM group (e.g., Escherichia–Shigella and Coprococcus) [15], and correlation analyses showed significant associations between these taxa and fecal 3-IPA levels. However, we note that genus-level 16S profiling does not identify the specific 3-IPA–producing strains nor establish a causal metabolic link. Therefore, we interpret these findings as microbiota–metabolite associations that may reflect broader community restructuring and shifts in tryptophan/indole metabolic capacity, rather than direct evidence that these genera are the sole producers of 3-IPA. Identifying microbial producers and causal pathways will require functional microbiota manipulation and strain-level/functional profiling in future studies.

Indole-3-propionic acid (3-IPA) is a gut microbiota–derived tryptophan metabolite, and its elevation in the metastatic cohort is consistent with altered microbial tryptophan/indole metabolism. Mechanistically, 3-IPA represents a downstream output of microbial tryptophan/indole metabolism; thus, elevated fecal 3-IPA together with community shifts in indole-associated taxa is consistent with altered microbial tryptophan metabolic capacity, although functional confirmation requires pathway-level measurements (e.g., metagenomic profiling of tryptophan/indole genes and targeted metabolomics of tryptophan and key indole intermediates).

Functionally, we found that 3-IPA did not measurably alter gastric cancer cell proliferation, migration, or invasion, suggesting that its pro-metastatic effect is not mediated by tumor cell–intrinsic changes. Instead, 3-IPA primarily acted on lymphatic endothelial cells, promoting migration and tube formation. We further show that 3-IPA engages AhR signaling in HLECs, accompanied by induction of the canonical AHR target CYP1A1 and upregulation of VEGFR3-associated lymphangiogenic responses; importantly, these effects are attenuated by the AHR antagonist CH223191. Together, our data support a non–tumor-cell-autonomous mechanism whereby a microbiota-derived metabolite promotes lymphatic niche remodeling and facilitates lymphatic dissemination in gastric cancer.

3-IPA, as a member of the indole family, exhibits neuroprotective properties and has potential applications in the treatment of Alzheimer’s disease [16–19]. Recent research has indicated that 3-IPA is associated with the regulation of gastrointestinal barrier function, the alleviation of intestinal inflammation, and the mitigation of liver steatosis [20–23]. Furthermore, accumulating evidence suggests that IPA/3-IPA can modulate host antitumor responses in a context-dependent manner, including reported enhancement of immunotherapy responsiveness via immune-cell–intrinsic programs in certain models [24–26]; a recent study reported that microbial-derived IPA can potentiate anti–PD-1 immunotherapy across multiple tumor models by enhancing CD8⁺ T-cell stemness/progenitor exhausted programs. Importantly, the biological compartment and endpoint examined in that work differ from ours. In the present study, 3-IPA did not measurably alter gastric cancer cell proliferation, migration, or invasion; instead, it acted primarily on lymphatic endothelial cells to activate the AHR/CYP1A1–VEGFR3 axis, thereby promoting lymphangiogenesis and facilitating lymph-node metastasis, which could be attenuated by the AHR antagonist CH223191. We therefore propose a dual, compartment-specific model in which IPA/3-IPA may enhance antitumor immunity via T-cell–intrinsic programs in certain settings, while concurrently remodeling the lymphatic/stromal niche through AHR signaling to favor lymphatic dissemination in gastric cancer. The net consequence is likely to depend on tumor type, baseline microenvironmental state, and exposure dynamics. Collectively, these findings support a context-dependent, compartment-specific framework for IPA/3-IPA, in which tissue-protective and/or immune-enhancing actions reported in other settings may coexist with pro-metastatic lymphatic remodeling in gastric cancer. Previous studies have demonstrated that indole derived from bacteria activates the aryl hydrocarbon receptor (AHR), thereby facilitating immune suppression and tumor proliferation [27]. Among the primary agonists of AHR, 3-IPA has been extensively documented for its anti-inflammatory properties, particularly in conditions such as osteoarthritis, gastrointestinal inflammation, neuritis, and vasculitis [28–31]. While the immunotherapy-related effects of IPA have been increasingly recognized, its impact on the lymphatic niche and lymphatic dissemination has remained less explored, which is directly addressed by our findings in gastric cancer. Previous research conducted by Wiggins [32] et al. has demonstrated that the activation of AHR reduces the response of vascular and lymphatic endothelial cells to inflammatory stimuli, thereby inducing a state of quiescence in these cells. This quiescence is characterized by an increase in the proportion of cells in the G0/G1 phase and a concomitant decrease in the number of cells in the S phase, ultimately limiting the proliferation of both vascular and lymphatic endothelial cells. Importantly, the findings of this study indicate that AHR activation in endothelial cells primarily serves to inhibit excessive proliferation without compromising the functional characteristics of these cells. Furthermore, the activation of AHR induced by 3-IPA did not significantly impact the proliferation of HLECs; however, it did enhance their migratory capacity and ability to form tubular structures by upregulating the expression of VEGFR3 in HLECs. Research indicates that the activation of AHR enhances the transcriptional activity of signal transducer and activator of transcription 1 (STAT1) while simultaneously downregulating the expression of VEGF in a murine model of lower limb ischemia. Yang [33] et al. have demonstrated that in thyroid cancer, ten-eleven translocation 3 (TET3) interacts with AHR, thereby activating the hypoxia-inducible factor 1-alpha (HIF-1α)/VEGF signaling pathway, which facilitates lymphangiogenesis in breast cancer cells. In this study, we present novel findings that 3-IPA activates AHR in HLECs, leading to the upregulation of VEGFR3 and the promotion of lymphangiogenesis. These findings offer valuable insights into the mechanisms of lymph node metastasis in gastric cancer, particularly from the perspective of gut microbiota.

At present, targeted therapies directed at the gut microbiota have emerged as a focal point in cancer treatment, including fecal microbiota transplantation (FMT), despite the potential limitations in its applicability. Recently, Li et al. [32] introduced the targeted-bacterium-depleted (TBD) model, which aims to deplete specific pathogenic bacteria, thereby serving as a strategy for both disease prevention and treatment [34].

Conclusions

In conclusion, our findings indicate that the gut microbiota-derived metabolite 3-IPA stimulates the AHR/CYP1A1–VEGFR3 pathway in human lymphatic endothelial cells, promoting the formation of lymphatic vessels and subsequently affecting lymph node metastasis in gastric cancer. This implies that the gut microbiota’s impact on tumors may not solely be through direct binding to membrane receptors but could also involve influencing the tumor microenvironment through its metabolic byproducts. Moreover, the precise modulation of the gut microbiota and its metabolites may exert a meaningful influence on cancer treatment and prevention.

Limitation

Several limitations should be acknowledged. First, although fecal 3-IPA was elevated in the metastatic cohort and 16S profiling suggested an association between microbial community alterations and 3-IPA abundance, we did not directly manipulate the microbiota (e.g., gnotobiotic/pseudo-germ-free models or fecal microbiota transplantation) to establish an upstream causal link between specific microbial communities and 3-IPA production. In addition, because we used 16S rRNA sequencing, our analysis is limited to genus-level taxonomic resolution and cannot assign 3-IPA production to specific strains or quantify tryptophan/indole pathway gene abundance or metabolic flux. Metagenomic functional profiling and targeted metabolomics of tryptophan and key indole intermediates will therefore be required to validate producer taxa and pathway-level metabolic capacity.

Second, we did not directly quantify circulating or gastric tissue 3-IPA levels following administration; thus, the exact systemic exposure and gastric-site concentrations achieved under our dosing regimen remain to be determined in future targeted LC–MS/MS pharmacokinetic studies. Accordingly, we do not claim a precise equivalence between the in vitro concentrations used and in vivo tissue exposures.

Third, although our data support AHR as a key mediator of 3-IPA activity, we did not perform genetic or pharmacologic perturbation of downstream candidates (e.g., CYP1A1 or VEGFR3) to establish their necessity for the lymphangiogenic and metastatic phenotypes. Therefore, CYP1A1 and VEGFR3 should be interpreted as candidate downstream mediators consistent with AhR pathway engagement, and future studies using CYP1A1/VEGFR3 inhibition or knockdown will be required to determine causal contribution. In addition, although molecular docking and AhR blockade support receptor involvement, the precise structural basis of 3-IPA–AhR interaction and the broader downstream transcriptional network following AhR activation in lymphatic endothelial cells warrant further investigation.

Finally, our study was not designed to directly evaluate immune checkpoint blockade efficacy or immune-cell–intrinsic mechanisms; therefore, our conclusions specifically pertain to lymphangiogenesis and lymph node metastasis driven by the lymphatic endothelial compartment. In addition, the clinical cohort was relatively small (n = 15; LN+ n = 9, LN− n = 6), which limits statistical power and restricts broad clinical generalizability. Accordingly, the patient-derived multi-omics results should be interpreted as exploratory, discovery-level evidence. Larger, independent cohorts—ideally from multiple centers and with comprehensive clinical covariates—will be required to validate fecal 3-IPA as a biomarker for lymph node metastasis in gastric cancer and to further assess potential confounding factors.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

GC

Gastric cancer

LVI

Lymphovascular invasion

GM

Gut microbiota

3-IPA

Indole-3-propionic acid

AHR

Aryl hydrocarbon receptor

CYP1A1

Cytochrome P450 1A1

HLEC

Human lymphatic endothelial cells

VEGFR3

Vascular endothelial growth factor receptor 3

Author contributions

Conceptualization, F.T. and C.J.; methodology, F.T.; software, Q.P.,Q.W. and Z.S.; validation, M.W., Y.W. and M.B.; formal analysis, M.W.; investigation, Y.W.,Z.Z.; resources, M.W.,G.C. and S.Y.; data curation, M.W.; writing—original draft preparation, M.W.; writing—review and editing, M.W. and Y.W.; visualization, Y.W.; supervision, F.T.,L.L.and C.J.; project administration, C.J.; funding acquisition, F.T., C.J. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China under Grant 81900524; the Natural Science Foundation of Shandong Province under Grant ZR2020MH252, ZR2020MH205; the China Postdoctoral Science Foundation under Grant 2020M672102; the Science and Technology Development Program of Jinan under Grant 202134027.

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Decla-ration of Helsinki, and approved by the Ethics Committee of Shandong Provincial Hospital (NSFC: NO. 2020–442, 18 March 2020). Informed consent was obtained from all subjects involved in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no conflict of interest.