Indole-3-lactic acid suppresses colorectal cancer via metabolic reprogramming

The content of ILA decreases in the intestine of CRC patients

To investigate the gut microbiota in CRC patients, we performed metagenomic shotgun sequencing on a total of 132 fecal samples (83 CRC patients and 49 healthy controls) (Figure 1(a)). The clinical information of all patients is given in Table S1. Following quality control filtering of the sequences (Table S2), high-quality reads were aligned to the Integrated Gene Catalog of Gut Microbes to obtain species abundance data for the samples. The alpha diversity of the gut microbiota was compared between the two groups. No significant differences were found at the species level (Figure S1A), but a decrease was detected in the CRC group at the gene level (Figure 1(c)). In this study, we analyzed gut microbiota diversity in individuals with CRC compared with healthy controls via principal coordinate analysis (PCoA) techniques. The results showed a significant difference in microbial communities between the two groups (Figure 1(c) and Figure S1B). Additionally, linear discriminant analysis effect size (LEfSe) was used to identify variations in bacterial communities between the two groups, with a focus on assessing differences at the genus and species levels (Figure 1(d)). In conjunction with the linear discriminant analysis (LDA) score and taxonomic cladogram derived from the LEfSe analysis, our findings indicate that Bacteroides, Proteobacteria, Enterobacterales, and Escherichia were significantly more abundant in the CRC group, whereas Clostridia, Firmicutes, Lachnospiraceae, Bifidobacterium, and Ruminococcaceae were notably enriched in the healthy control group (Figure 1(d) and Figure S1C). In order to elucidate the microbial metabolic processes, present in CRC patients and healthy control subjects, we annotated metagenome-analyzed microbial genes within the Gut Metabolic Modules database (GMMs) (Table S3). Our analysis revealed notable alterations in Trp degradation-related metabolism (Figure 1(e)), which are in agreement with findings reported in previous studies.^24,25 Furthermore, our analysis suggested that alterations in various bacterial taxa may impact Trp metabolism, with Bacteroides being enriched in CRC patients and Clostridia and Bifidobacterium being enriched in the healthy control cohort.

Trp metabolites in CRC and healthy control feces were analyzed using targeted Trp metabolomics, which revealed 26 downstream Trp molecules, including indole derivatives and L-Kynurenine. The orthogonal partial least squares discriminant analysis (OPLS-DA) reveals notable differences in the distribution of Trp metabolites between the CRC and healthy control groups (Figure S1D). Compared with those in the healthy control group, 14 metabolites were upregulated, whereas 12 metabolites were downregulated in the CRC group (Figure 1(f) and Table S4). Among the variables under observation, fecal ILA levels were significantly lower in the CRC group than in the healthy control group (Figure 1(g)). In contrast, the fecal indoxyl sulfate (IS) levels exhibited the most pronounced increase in the CRC group, while no statistically significant differences were observed (Figure S1E). Subsequently, an analysis was conducted to assess the relationships between the fecal ILA concentration and various clinical parameters. The results indicated that the level of ILA was not related to sex, age, or tumor location (Figure S1F-H). However, compared with the healthy control group and early stage patients, the level of ILA in the feces of late stage CRC patients was significantly reduced (Figure 1(h)). Additionally, patients with poorly differentiated CRC exhibited a lower ILA content than those with medium to high differentiation, although this difference was not statistically significant (Figure S1E). Furthermore, the relationship between the ILA concentration and the tumor proliferation indicator Ki67 revealed a negative correlation (p = 0.002, R² =-0.459, Figure 1(i)). Our above results revealed a significant decrease in fecal ILA levels in late-stage CRC patients compared to healthy individuals. These clinical findings suggest that ILA may be a valuable metabolite for prevention and may contribute to the progression of CRC.

ILA treatment inhibits the progression of CRC in vivo

Considering the decreased content of ILA in the feces of CRC patients, particularly those in advanced stages, it is hypothesized that ILA may mitigate the progression of CRC. To investigate this hypothesis, we conducted an in vivo study utilizing the Azoxymethane (AOM)/Dextran Sodium Sulfate (DSS) induced tumor model, which serves as a model for inflammation-driven colorectal adenocarcinoma (Figure 2(a)). Our findings indicated that the exogenous administration of ILA resulted in a significant reduction in tumor number, size, and load (Figure 2(b,c)). Furthermore, no notable differences in mortality rates were found between the control group and the experimental group of mice during tumor progression (Figure S2A). No significant difference was observed in the efficacy of ILA supplementation via oral administration compared with that via intraperitoneal injection. Hematoxylin and Eosin (H&E) examination revealed a notable reduction in tumor-induced damage to villous structure with ILA supplementation (Figure 2(d)). Immunohistochemical analysis revealed lower positive rates of Ki67 and PCNA in ILA-supplemented tumors than in control tumors, suggesting decreased tumor cell proliferation (Figure 2(d)). Subsequently, an evaluation of innate and adaptive immune cells within colon tumors was conducted. Compared with the control group mice, ILA supplementation did not significantly change spleen weight, the proportion of tumor infiltrating macrophages, and the proportion of bone marrow-derived suppressor cells (MDSCs) (Figure 2(e) and Figure S2B-2C). However, there was an observed increase in the proportion of tumor infiltrating CD8⁺ T cells, consistent with prior research findings (Figure 2(e) and Figure S2D).⁸ These findings suggest that ILA may elicit antitumor effects by bolstering antitumor immunity or directly targeting tumor activity.

Subsequently, we assessed the anti-tumor efficacy of ILA supplementation in mice bearing unilateral tumors (MC38 cells) with intact immune function (Figure 2(f)). Consistent with previous research findings, the administration of ILA via peritumoral and intraperitoneal injection led to a notable decrease in tumor size, and no impact on the spleen weight or the body weight of the mice (Figure 2(g–i) and Figure S2E-S2F). To assess tumor progression, the mice were euthanized on the 24th day, and samples of MC38 tumor tissue samples were obtained. Immunohistochemical analysis revealed reduced expression levels of Ki67 and PCNA in the ILA-treated group (Figure 2(j)). Following ILA supplementation, there was an observed increase in the infiltration of CD8⁺T cells into the tumor, although this difference was not found to be statistically significant (Figure 2(k)). To elucidate the role of the immune system in the anti-tumor mechanisms of ILA, we administered anti-CD8 antibodies to inhibit mouse CD8⁺ T cells. Notably, despite this intervention, ILA still effectively reduced tumor volume and weight (Figure S3A-S3C). Additionally, our experiments with nude mice bearing HCT116 tumors demonstrated that supplementation with ILA led to a reduction in tumor volume (Figure S3D-3 G), suggesting that ILA can elicit antitumor effects through mechanisms independent of CD8⁺T cells. In summary, our results showed that ILA supplementation inhibits CRC progression in vivo. The efficacy of ILA in reducing tumor volume has been demonstrated in several mouse models, suggesting potential mechanisms beyond CD8⁺ T cell involvement.

ILA treatment inhibits the viability of CRC cells

To investigate the potential direct effects of ILA on tumor cells, functional analyses were conducted in vitro using two CRC cell lines, MC38 and HCT116. Coculturing ILA with the CRC cell lines at various concentrations and durations revealed that ILA inhibited the proliferation of CRC cells, as evidenced by CCK8 and EdU tests after 48 hours, in a concentration-dependent manner (Figure 3(a,b)). Additionally, colony formation experiments demonstrated that ILA can inhibit colony formation in CRC cell lines (Figure 3(c)). By using transwell assays, ILA inhibited the migration of MC38 and HCT116 cells (Figure 3(d)). Subsequently, the proportion of apoptotic cells was increased upon ILA treatment (Figure 3(e)). These results showed that ILA directly inhibited proliferation, colony formation, and migration of CRC cells in a concentration-dependent manner, indicating its potential as a therapeutic agent to inhibit cell viability and progression.

ILA treatment modulates tumor cell glycolysis by inhibiting HK2 expression

Next, to clarify how ILA inhibits the activity of CRC cells, we performed RNA-seq analysis on the CRC cell line HCT116 cocultured with ILA. After ILA stimulation, 744 genes were downregulated, whereas 650 genes were upregulated in HCT116 cells (Table S5). We conducted gene enrichment analysis and found that the genes whose expression changed were enriched mainly in the glycolytic pathway (Figure 4(a,b)). The Warburg effect in tumor cells is related to impaired mitochondrial oxidative phosphorylation, abnormal expression of glucose metabolism enzymes, oncogene activation, TSG inactivation, and changes in the tumor microenvironment.^26,27 Several key glycolytic genes HK2, PFKL, PKM2, and LDHA were downregulated after ILA stimulation (Figure 4(c,d) and Figure S4A). Similar results were found in HCT116 cells using RT-qPCR (Figure 4(e)). Furthermore, Western blotting (WB) was used to detect the changes in glycolytic protein levels and subcutaneous tumor tissue after supplementation with ILA (Figure 4(f)). We found that HK2 showed the most significant changes at the mRNA and protein levels. Next, we investigated whether ILA can regulate glycolysis and oxidative phosphorylation of HCT116 cells in vitro. We found that after treatment with ILA, the lactate production, glucose consumption, and ATP production were reduced (Figure 4(g)). A Seahorse XFe24 extracellular flux analyzer was used to measure glucose metabolism in HCT116 cells. The results indicated that ILA significantly reduced the glycolytic level but had no effect on oxidative phosphorylation (Figure 4(h) and Figure S4B).

To validate that ILA inhibits glycolysis by suppressing the expression of HK2, we transiently increased the expression of HK2 in HCT116 cells and assessed the levels of HK2 post-enhancement using WB (Figure 4(i)). Furthermore, we conducted biological functional assays to evaluate the impact of ILA on cell activity subsequent to HK2 overexpression. The results demonstrated that the upregulation of HK2 counteracted the impact of ILA (Figure 4(j-m)). Additionally, we administered 2-deoxyglycose (2-DG) to ILA-supplemented HCT116 cells to assess changes in cell viability following glycolysis inhibition. Our results indicated that ILA did not exert additional inhibitory effects on cell viability under these conditions (Figure S4C-F). These above results show that ILA inhibits CRC cell glycolysis by suppressing HK2 expression, resulting in reduced lactate production and glucose consumption. This highlights the potential of ILA to inhibit CRC progression by targeting tumor cell metabolism.

ILA reduces HK2 expression independent of AHR

As an indole derivative, ILA functions as an exogenous ligand of the aryl hydrocarbon receptor (AHR). Previous research has demonstrated the ability of ILA to activate AHR and elicit its effects, leading to speculation regarding the potential antitumor effects of ILA mediated through AHR^{23,24,28–30}(Figure 5(a)). We conducted protein-level validation and observed a significant increase in the expression of AHR and CYP1B1 in HCT116 cells following stimulation with ILA (Figure 5(b)). CYP1B1, a member of the cytochrome P450 family, was identified as a marker for AHR activation.^31,32 After ILA stimulation, there was a significant increase in the nuclear displacement of AHR (Figure 5(b) and Figure S5A). To investigate the potential mechanism by which ILA inhibits tumor cell activity through inhibiting glycolysis via AHR, we utilized the AHR receptor inhibitor CH223191.³³ Surprisingly, our findings indicated that AHR inhibitors failed to rescue the inhibitory effect of ILA on cell activity (Figure 5(c-f)). Additionally, the classic ligand 6-Formylindolo[3,2-b] carbazole (FICZ) of the AHR receptor was used as a positive control, revealing its inability to inhibit tumor cell activity. AHR knockdown cell lines were generated and subsequently subjected to ILA stimulation (Figure 5(i)). Upon AHR knockdown, ILA maintained its ability to suppress the proliferation and migration of HCT116 cells, as well as its ability to induce apoptosis (Figure S5B-5E). The stable shAHR cell line was utilized in xenograft model experiments, revealing that in vivo AHR knockdown did not negate the inhibitory influence of ILA on tumor growth (Figure 5(j,k)). Furthermore, our investigation demonstrated that the suppression of tumor cell glycolysis by ILA was not affected by AHR receptor inhibitors (Figure 5(g,h) and Figure S5F-5 G). Our results further demonstrated that ILA reduces HK2 expression independently of the AHR pathway, as evidenced by experiments showing that AHR inhibitors and knockdown did not reverse the inhibitory effects of ILA on CRC cell activity and glycolysis. These results suggest that the modulation of tumor metabolism by ILA is not mediated by the AHR, highlighting a new mechanism in CRC progression.

ILA inhibits the viability of CRC cells through the p-STAT3-HK2 axis

Phosphorylation of STAT3 is necessary for its translocation from the cytoplasm to the nucleus for its functional activity.^34,35 Our investigation revealed that the expression level of STAT3 remained constant in tumor cells treated with ILA, while its phosphorylation level decreased significantly (Figure 6(a)). The phosphorylation of STAT3 predominantly takes place within the Src homology 2 (SH2, residues 584–688) domain,³⁶ with tyrosine 705 (Tyr705) being the most frequently modified residue.^37,38 Subsequently, the phosphorylated Tyr705 (pTyr705) interacts with the SH2 domain of the opposing STAT3 subunit, thereby facilitating STAT3 dimerization and culminating in the phosphorylation process.^35,39,40 Through molecular docking simulations, we demonstrated that ILA exhibits a strong binding affinity to the SH2 domain, forming a robust π-cation interaction with the Arg609 residue (Figure 6(b)). In contrast, the binding of pTyr705 to the SH2 domain is mediated solely by hydrogen bonds (Figure S6A). ILA competes for the SH2 docking pocket, thereby inhibiting the binding of pTyr705 to SH2 and consequently suppressing the phosphorylation process of STAT3 (Figure S6B). To further elucidate this mechanism, we utilized the online database JASPAR（http://jaspar.genereg.net/) to identify a consensus binding site for STAT3 in the HK2 promoter region, with a high score of 11 (Figure 6(c)).⁴¹ Then chromatin immunoprecipitation quantitative polymerase chain reaction (ChIP-qPCR) analysis revealed a marked decrease in the enrichment of STAT3 at the HK2 promoter following ILA stimulation (Figure 6(d)). Furthermore, ectopic expression of STAT3 effectively counteracted the decrease in tumor cell proliferation and migration capacity, and the increase in apoptosis triggered by ILA (Figure 6(e-i)). Notably, the upregulation of STAT3 negated the suppressive effect of ILA on HK2 expression and glycolysis levels (Figure 6(j,k)).

Subsequently, the expression of HK2 and STAT3 in CRC were examined. By utilizing IHC and WB assays revealed that the expression of HK2 and STAT3 was notably upregulated in tumor tissue compared with adjacent tissues (Figure 6(m) and Figure S6C). Furthermore, analysis of the Gene Expression Omnibus (GEO) dataset (GSE24550, Table S6) corroborated the increased expression levels of HK2 and STAT3 in tumor tissues, and there was a positive correlation between their expression levels (Figure 6(m)). The same results were also obtained in the TCGA database (Figure S6D-E, Table S7). Our analysis delved deeper into the role of ILA in the context under study. The results of the WB experiment revealed a significant decrease in the expression of HK2 in the high-ILA group compared to that in the low-ILA group, as well as a notable reduction in the amount of p-STAT3 in the tumor tissue of the high-ILA group (Figure 6(n)). The quantification of HK2 protein levels in CRC tumor tissue and the subsequent analysis of its association with ILA content revealed a significant negative correlation (Figure S6F). These findings show that ILA inhibits CRC cell viability through the p-STAT3-HK2 axis by directly targeting STAT3 phosphorylation, resulting in decreased p-STAT3 enrichment at the HK2 promoter.

In conclusion, our study revealed a decrease in ILA levels in the intestines of CRC patients, which was significantly associated with tumor progression. Our findings suggest that ILA exerts a direct inhibitory effect on tumor cells by limiting their proliferation, migration, and antiapoptotic capabilities. Furthermore, our results indicate that ILA acts independently of AHR by directly inhibiting the activation of STAT3, leading to a reduction in HK2 expression and ultimately inhibiting the glycolytic pathway in tumor cells. This study provides new insights into the role of ILA in the inhibition of CRC and establishes a new theoretical basis for future research.

Limitations of the study

This study, based on a small single-center clinical sample, may have biases due to sample limitations. To accurately assess ILA distribution in colorectal cancer patients, future research should involve more diverse medical centers and populations. Our current understanding of ILA’s role in glycolysis regulation is limited, despite some progress. We aim to explore additional molecular mechanisms and analyze ILA’s impact on glycolysis from various angles to enhance the theoretical framework. While ILA shows promise in tumor inhibition, it is not yet ready for clinical use. We have not conducted clinical trials, so its effectiveness and safety in humans remain unverified.

Resource availability

Human samples

To investigate alterations in the gut microbiota and metabolites in CRC patients, fecal samples were obtained from Nanjing Drum Tower Hospital, and control samples were collected from the physical examination center of the same hospital. Inclusion criteria for the colorectal cancer cohort are as follows: (1) Participants must be aged between 18 and 80 years; (2) A confirmed diagnosis of colorectal cancer via pathological assessment, with plans for radical surgical intervention; (3) Absence of antibiotic, microbial preparation, or corticosteroid usage within the 3 months preceding specimen collection; (4) A body mass index (BMI) within the normal range of 18.5 to 25; (5) Provision of informed consent by all participants. Exclusion criteria include (1) A history of major abdominal surgeries, with the exception of cholecystectomy and appendectomy; (2) Presence of complete intestinal obstruction; (3) Diagnosis of chronic diarrhea, inflammatory bowel disease, or other conditions known to affect gut microbiota; (4) Undergoing neoadjuvant chemoradiotherapy; (5) Severe dysfunction of the heart, brain, liver, lungs, kidneys, or other vital organs; (6) Belonging to ethnic minorities with distinct dietary practices; (7) Demonstrating poor compliance or inability to cooperate. The average age of the CRC patients was 83, whereas that of the control donors was 49. Additional clinical and pathological details of the patients are presented in Table S1.

Fecal sample collection and DNA extraction

Fecal samples were collected from all participants for metagenomic sequencing after confirming that they had not undergone antibiotic treatment for a minimum of 1 month prior to sample collection. Additionally, individuals refrained from consuming probiotic-rich foods, such as yogurt, in the week leading up to sample collection. Each sample was promptly frozen at −80℃ or temporarily stored in personal −20℃ freezers before being transported to the laboratory within a 24-hour timeframe. DNA extraction procedures were conducted in accordance with established protocols.

Metagenomic sequencing

The samples were sequenced on the NovaSeq 6000 platform with paired-end reads, an insert size of 350 bp, and a read length of 100 bp. Adaptor sequences and low-quality reads were removed from the raw data, and the remaining reads were filtered to exclude human host DNA via the human genome reference (hg18) as outlined in a previous study.⁵³ A total of 1,403.5 Gb of high-quality paired-end reads were obtained for the 132 samples, averaging 10.6 Gb per sample after the removal of human DNA reads. The obtained sequences were deposited in the NCBI database (https://dataview.ncbi.nlm.nih.gov/object/PRJNA1157583?reviewer=94q0mr5gq6ocslojevd04j420b and https://dataview.ncbi.nlm.nih.gov/object/PRJNA1163867?reviewer=he36gb400guur438mj1o1j6jf7). We utilized the Integrated Gene Catalog of Gut Microbes (IGC) as the reference database for the gene set of human gut microbiota. Sequence alignment results exhibiting a similarity exceeding 90% were selected as the final alignment outcomes. Subsequently, we normalized the length information of each gene in the background database to obtain their relative abundances. For gene alignment, we utilized the Short Oligonucleotide Alignment Program (SOAP).

Subsequently, we annotated the genes obtained using the Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation information corresponding to the IGC background database. These genes were then processed through computational algorithms to derive the relative abundance data of gene functions present in each sample. Further annotation of genes and KEGG Orthology (KO) terms was conducted utilizing several widely used databases, including Carbohydrate-Active enZyme (CAZy), Gut Metabolic Modules (GMMs), and Comprehensive Antibiotic Resistance Database (CARD), to ascertain the relative abundance of functional units at various levels. According to the official documentation of MetaPhlan3.0, we selected the recommended parameters for species annotation analysis and utilized the algorithm provided by the software to compute relative abundance information at the phylum, genus, and species levels. The MetaPhlan algorithm identifies unique clade-specific marker genes from approximately 17,000 reference genomes and subsequently compares them with pre-extracted marker genes using Bowtie2 to classify species and determine their relative abundance.

Trp metabolism

The participants’ fecal samples were treated with an extract solution (1:10, w/v; methanol/acetonitrile/water = 2:2:1, precooled at −40°C, containing 0.1% formic acid and an isotopically labeled internal standard mixture) and homogenized. Following centrifugation, the resulting supernatant (400 μl) was evaporated under N2, and then reconstituted with 100 μl of 0.1% formic acid. Subsequent centrifugation was performed to prepare samples for analysis. The target compounds were chromatographically separated using UHPLC with an ExionLC System (SCIEX) equipped with a Waters Acquity UPLC HSS T3 (100 × 2.1 mm, 1.8 μm) liquid chromatography column. The analysis utilized a SCIEX 6500 QTRAP+ triple quadrupole mass spectrometer with an IonDrive Turbo V electrospray ionization (ESI) interface, with specific parameter settings including curtain gas at 40 psi, IonSpray voltage at ±4500 V, temperature at 500℃, ion source gas 1 at 30 psi, and ion source gas 2 at 30 psi. Mass spectrum data acquisition and quantitative analysis of target compounds were conducted using SCIEX Analyst Work Station Software and SCIEX MultiQuant software.⁵⁴ Trp metabolism analysis was performed with the assistance of Shanghai Biotree Biotech Co.

Mice

Six-week-old female wild-type C57BL/6J (GemPharmatech, No. N000013) mice and four-week-old female BALB/c nude (GemPharmatech, No. D000521) mice were procured from the Model Animal Research Center of Nanjing University. The mice were housed in a controlled environment with independent ventilation systems, ensuring specific pathogen-free conditions, within the Animal Facility at Nanjing University. The researchers involved in the mouse experiment possessed valid animal breeding licenses.

Cell culture

The CRC cell lines HCT116 (ATCC, CCL-247) and MC38 (Kerafast, CVCL_B288) were procured from the Type Culture Collection of the Chinese Academy of Sciences in Shanghai, China. These cell lines were grown at 37℃ with 5% CO₂ in RPMI 1640 (Gibco, 11875500BT) supplemented with 10% heat-inactivated FBS (Celligent, CG0430B), and 1% antibiotic-antimycotic (Gibco 15,140–122). HCT116 and MC38 cells were seeded onto different types of plates and treated under different conditions for further experiments when the cell density reached 80%. Dissolve the necessary pharmacological agents, including ILA (MCE HY-113099), CH223191 (MCE HY-12684), FICZ (MCE HY-12451), and 2-DG (MCE HY-13966), in DMSO. The experimental protocol ensures that the concentration of DMSO in the experimental group is equivalent to that in the control group. For the assessment of AHR functionality, cells were pre-treated with CH223191 (10 μM) for 4 hours prior to the administration of ILA (250 μM) and FICZ (100 nM). The experimental operators were blinded to the treatment conditions, with drug preparation conducted by separate personnel. The operators processed the cells based on their assigned numerical identifiers. All in vitro cell experiments involved a 12-hour starvation period using serum-free culture medium before cell stimulation.

In vivo animal studies

Each mouse was meticulously tagged with an ear tag for unique identification. Subsequently, by leveraging a random number table, the mice were randomly assigned to different experimental groups in strict accordance with the sequence of the randomly generated numbers. During the experimental process, a single – blind design was adopted, where only the experimental operators were kept unaware of the group assignments. Other dedicated personnel were solely responsible for drug preparation. The operators administered treatments to the mice strictly based on the assigned numbers.

In the xenograft model of CRC, 4-week-old female C57BL/6J (GemPharmatech, No. N000013) mice were subcutaneously injected with 1 × 10⁶ MC38 cells, while 4-week-old female nude mice were subcutaneously injected with 5 × 10⁶ HCT116 cells. The animals were randomized to treatment or control groups when tumors were palpable, at approximately 80 mm³. ILA (400ug/20 g/d) was administered subcutaneously (s.c.) or intraperitoneally (i.p.) around the tumor every day,^8,22,55 while the control group received PBS subcutaneously. Tumor volume was measured and recorded throughout the experiment. The mice were euthanized using CO₂ inhalation when the tumors reached a final size of 1 cm³, and the tumors were collected and weighed for further analysis. To induce colitis-associated CRC in C57BL/6J mice at 8 weeks of age, intraperitoneal injections of azoxymethane (AOM) at a dosage of 10 mg/kg were administered on day 1 (n = 7 per group). Following a 14-day interval, 2% dextran sulfate sodium (DSS) was added to the drinking water for 7 consecutive days, with three cycles of DSS treatment utilized. The mice were euthanized on day 120 after tumorigenesis induction. The administration of ILA commenced 10 days postinjection of AOM, either through oral ingestion by dissolving ILA in drinking water (p.o.) or intraperitoneally (i.p.) by dissolving ILA in PBS. Colon tissues, tumor tissues, feces, serum, spleens, and mesenteric lymph nodes (mLNs) were collected on day 120. Tumor number, tumor size, and total tumor mass (tumor load) were measured for each mouse. After they were macroscopically assessed, the colons were used for histopathological analysis.

Quantitative real-time PCR (qRT-PCR) and RNA-sequencing (RNA-seq)

TRIzol reagent (Vazyme, R411–01-AA) was utilized for the extraction of total RNA from the cell samples. Cytoplasmic and nuclear RNA were isolated using a Cytoplasmic & Nuclear RNA Purification Kit (Norgen Biotek, Thorold, ON, Canada) following the manufacturer’s protocol. The extracted RNA was then reverse transcribed into cDNA using HiScript® III RT SuperMix (+gDNA wiper) (Vazyme, R223–01) for qRT-PCR analysis. The ABI StepOne Plus instrument with PowerUp™ SYBR® Green (Applied Biosystems, USA) was employed for the qRT-PCR analyses, with Actin serving as the internal control for normalization. A list of the qRT-PCR primers can be found in Table 1. For the RNA-seq experiments, isolated RNA samples were submitted to the BGI Genomics (Guangdong, China) for sequencing. The raw tag data were generated using the sequencing-by-synthesis methodology, and the institute subsequently processed the transcriptomic data. The raw data have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE271802 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE271802 secure token: efkricwqfnmbpen).

Western blotting (WB)

Tumor tissue was homogenized with three steel balls in lysis buffer (Solarbio, R0010) containing 1% phosphatase inhibitor (MCE, HY-K0023) and 1% protease inhibitor (MCE HY-K0010) in 200 µl of buffer per 20 mg of tissue. Proteins were extracted by centrifugation at 14,000 g for 20 minutes. Cell samples were treated with 200 µl of lysis buffer per 1 × 10⁶ cells and centrifuged at 14,000 g for 20 minutes. Protein concentration was measured using a BCA kit (Beyotime, P0010). A total of 25 µg of protein per lane was separated on 8% and 12% polyacrylamide gels and transferred to polyvinylidene difluoride membranes. Membranes were blocked with 5% bovine serum albumin (Biofroxx 9048-46-8) in Tris-buffered saline with 0.1% Tween 20 (Biosharp 9005-64-5), then incubated with specific antibodies at 4℃ overnight. The next day, secondary antibodies were applied and visualized using an imaging system. Details of all relevant antibodies are provided in Table 1.

Plasmid construction and short hairpin RNA (shRNA) transfection

The human HK2 sequence and STAT3 sequence were synthesized by Corues Biotechnology (Nanjing, China) and subsequently subcloned into a pcDNA3.1 vector sourced from Addgene (Cambridge, MA) (Table S8). shRNAs targeting the AHR were procured from Corues Biotechnology, specifically shAHR no.1-TRCN0000021258, shAHR no.2-TRCN0000245285, and no.3-TRCN0000245286. Transfection of shAHR into cells was carried out using Lipofectamine 3000（Thermo Fisher, L3000001）, and plasmid transfection was also conducted using the same reagent. Infections were performed in accordance with the manufacturer’s instructions. Details of all relevant shRNA information are provided in Table 1.

Cell growth and proliferation assays

The cell growth rate was assessed via CCK‐8 cell proliferation assays. Seven independent replicates of MC38 and HCT116 cells treated with various conditions were incubated in 96-well plates. Subsequently, 10% CCK-8 solution (Dojindo Laboratories, CK04) was added to each well at specified time points, and the samples were incubated for an additional 2 hours. The absorbance of each well was measured at 450 nm. 5-Ethynyl-2’-deoxyuridine (EdU) assays were performed following the manufacturer’s instructions via the EdU Cell Proliferation Assay Kit (KeyGEN, KGA9602–100). Cell colony formation was assessed via a colony formation assay, in which 1,000 cells were seeded in 6-well plates containing complete medium and incubated at 37°C in a 5% CO₂ environment for 2 weeks. Following fixation with methanol for 30 minutes, cells were stained with 0.2% crystal violet (Solarbio, G1062) solution, and colonies consisting of more than 50 cells were quantified.

Transwell assay

Cell migration assays were conducted with millicell cell culture inserts (24-well plates; 8 µm pore size; Merck KGaA, Darmstadt, Germany). MC38 and HCT116 cells were introduced into the upper chamber, while 1640 medium (600 µl) supplemented with 20% fetal bovine serum was added to the lower chamber. Following 24–48 hours of incubation, cells were fixed for 20 minutes with 4% paraformaldehyde, stained for 30 minutes with crystal violet (Solarbio, G1062), and subsequently imaged.

Flow cytometry analysis

Apoptosis was assessed through flow cytometric analysis via an Annexin V FITC/PI double-stain assay according to the manufacturer’s instructions (Vazyme， A211–01). Flow cytometric analyses were performed via a CytoFlex (Beckman), and data interpretation was carried out via FlowJo software (Treestar, Inc., San Carlos). Tumor tissues were enzymatically digested with collagenase solution and subsequently filtered through a 200-mesh sieve to obtain a single-cell suspension. Details of all relevant antibodies are provided in Table 1.

Metabolite measurements

The concentrations of glucose and lactate in the medium were quantified separately using the Glucose Assay Kit (Beyotime, S0201S) and the Lactate Colorimetric Assay Kit (KeyGEN, KGA7401–48), respectively, following the protocols provided by the manufacturers at the 24-hour time point. Intracellular ATP levels were measured using the Enhanced ATP Assay Kit (Beyotime, S2007). All metabolite measurements were conducted independently, with each assay performed a minimum of three times.

Seahorse metabolic analysis

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured with an Agilent Seahorse XFe96. The measurements were conducted employing the Seahorse XF Glycolysis Stress Test Kit (Agilent 103,020–100) and the Seahorse XFp Cell Mito Stress Test Kit (Agilent 103,010–100). Cells were subjected to a 12-hour starvation period in the absence of FBS, followed by treatment with 250 μM ILA for an additional 24 hours in a complete medium. Prior to each assay, the XFe96 cartridge was hydrated overnight in water within a no-CO₂ incubator maintained at 37℃. On the day of measurement, the water was replaced with pre-heated Seahorse XF Calibrant Solution (Agilent 100,840–000) and incubated for 90 minutes. The cell medium was subsequently removed and replaced with 180 μL per well of Agilent Seahorse XF RPMI Medium at pH 7.4 (Agilent 103,681–100), and the cells were incubated for 1 hour in a no-CO₂ incubator at 37℃. Following this preparation, the medication was added according to the instructions provided in the kit.

Immunohistochemistry (IHC)

Immunohistochemistry (IHC) was conducted according to established protocols,⁵⁶ involving rehydration of tumor tissue slides, antigen retrieval, blocking, and incubation with primary antibodies. Subsequently, secondary antibodies were applied and the slides were stained using an IHC kit (Beyotime, P0603), with DAB peroxidase substrate. Details of all relevant antibodies’ information are provided in Table 1.

Chromatin immunoprecipitation (ChIP) assay

Chromatin immunoprecipitation (ChIP) assay was performed following the instructions provided by the manufacturer via a ChIP Assay Kit (BersinBio, Bes5001). HCT116 cells were treated with 1% formaldehyde for 10 minutes at room temperature to cross-link proteins and DNA. The cross-linking reaction was stopped by the addition of a stop-fix solution containing glycine (125 mm), aprotinin (1 μg/ml), pepstatin A (1 μg/ml), and Pefabloc (1 mg/ml) for 5 minutes at room temperature. The samples were then centrifuged at 2,500 × g and 4°C for 2 minutes, washed with ice-cold PBS, and subjected to chromatin immunoprecipitation using IgG (BersinBio) and anti-STAT3 (Cell Signaling Technology, 9139) antibodies. The association of STAT3 with HK2 promoter was measured by ChIP-PCR with the following primers: forward, 5’-GCACTGGTGGAGAATGGCAAG-3’, and reverse, 5’-CTGTACACATTGCAGACAGCACC-3’.

Molecular docking

To obtain the three-dimensional structure of ILA, access PubChem (https://pubchem.ncbi.nlm.nih.gov) and export the structure using PyMol. The phosphorylated dimer crystal structure of STAT3 is available under PDB ID: 4e68. According to the literature, phosphorylated Tyr705 (pTyr705) binds to the SH2 domain.³⁹ Therefore, PyMol was utilized to isolate and export the SH2 domain for subsequent protein-ligand docking studies. According to literature reports, the pTyr705 binding pocket position was utilized as the substrate docking site.³⁹ Indole lactate and SH2 were subjected to 100 docking simulations using AutoDock, with the lowest energy conformation being selected as the output structure. The interaction forces between ILA and SH2 were subsequently analyzed using PyMol.

Data source

RNA-sequencing expression profiles (level 3) and associated clinical data for colorectal cancer (CRC) were obtained from The Cancer Genome Atlas (TCGA) dataset (https://portal.gdc.cancer.gov). Additionally, the GSE24550 dataset was retrieved from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo). The raw data were acquired in MINiML file format. RNA sequencing data acquisition address: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE271802.

Quantification and statistical analysis

The data presented in this study are displayed as the mean ± SD, unless otherwise specified, and are elaborated upon in the corresponding section of the methodology and figure legend. An unpaired, two-tailed t-test was used to compare two sample sets that adhered to a normal distribution and homogeneity of variance. When sample sets deviated from normality and variance, or when normality and variance could not be assessed, the Mann–Whitney U-test was used. The Mann–Whitney U-test was also used for pairwise comparisons involving three or more samples. The Pearson correlation test was used to analyze correlations between two continuous variables. The Benjamini–Hochberg (BH) method was used to correct for multiple testing in the analysis of metagenomic sequencing, metabolomics, and RNA-seq data. Specific statistical methods are described in the corresponding figure legends. The p values were calculated using either SPSS 26.0 or GraphPad Prism 9.0, with significance levels denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ns indicating no significance. Details of the bioinformatics analyses are provided in the aforementioned sections.