Bacterial small molecule metabolites implicated in gastrointestinal cancer development

Abstract

Numerous associations have been identified between cancer and the composition and function of the human microbiome. As cancer remains the second leading global cause of mortality, investigating the carcinogenic contributions of microbiome members could advance our understanding of cancer risk and support potential therapeutic interventions. While fluctuations in bacterial species have been associated with cancer progression, studying their small molecule metabolites offers one avenue to establish support for causal relationships and the molecular mechanisms governing host–microorganism interactions. In this Review, we explore the expanding repertoire of small molecule metabolites and their mechanisms implicated in the risk of developing gastrointestinal cancers.

Table of contents blurb

In this Review, Crawford and Turocy examine diverse small molecule metabolites produced by the human microbiota, their role as potential risk factors for cancer development as well as novel mechanistic insights demonstrating their association with gastrointestinal cancer.

Introduction

Every individual hosts a community of archaea, fungi, viruses, and bacteria, known as the human microbiota, and their collection of genes, referred to as the human microbiome¹. Collectively, the human microbiota contributes in various ways to overall health and disease states, supporting its role as an organ.

As the ability to measure the composition and function of the microbiota has evolved through the use of advanced sequencing and spectrometry techniques, machine learning, and database development, the roles of specific microbiota members in the development of gastrointestinal cancer have gained increasing attention. Genetic studies indicate that inherited genetic factors play a minor contribution to cancer development, accounting for 5% to 45% risk of gastrointestinal cancer development^2,3. These findings support the notion that environmental factors, including dietary and microbial factors, contribute to the sporadic gastrointestinal cancer majority.

Among the microbiota contributors to gastrointestinal cancers, bacteria have emerged as a focus of extensive investigation. Research comparing microbial compositions between healthy individuals and patients with cancer has identified thousands of correlations between microbial dynamics and disease outcomes^4,5. Exploring these correlations is an entry point to establishing metabolic and mechanistic implications.

Key microbial-induced carcinogenic mechanisms include inflammatory stimulation and the production of carcinogenic metabolites⁶. Further, microbial members can contribute to oncogenesis by activating pro-carcinogenic pathways, spurring bacterial composition changes, impacting the tumour microenvironment, and participating in wide cellular and metabolic reprogramming^7–10. While this Review focuses on carcinogenic molecules, other metabolites serve antitumorigenic roles and many are being researched for their diagnostic, prevention, and therapeutic potential^11,12.

Helicobacter pylori, an inhabitant of the gastrointestinal microbiome, is a key player in cancer risk. H. pylori induces reactive oxygen species production and triggers inflammation, thus contributing to the development of various gastrointestinal cancers^13,14. Escherichia coli has been linked to colorectal cancer through its production of the genotoxin colibactin^15,16. Additionally, indolimines and small molecule toxins like tilimycin and tilivalline, represent expanding classes of bacterial metabolites with potential carcinogenic mechanisms^17–19.

Other members of the microbiota — viruses, fungi, archaea — have also been implicated in cancer progression. Human papillomaviruses (HPVs) and hepatitis B virus (HBV) are known risk factors for cervical cancer and hepatocellular carcinoma, respectively^20,21. HPV and HBV integrate into the host genome during chronic infections resulting in genomic changes associated with cancer development. Metabolic connections between fungi and cancer involve Candida albicans, a gastric, oral, and colorectal cancer risk factor. In addition to initiating diverse proinflammatory signaling, isolates of C. albicans generate carcinogenic nitrosamines, implicated in oncogene activation²². Further, shot-gun metagenomic analysis revealed that patients with colorectal cancer had enhanced levels of halophilic archaea and reduced methanogenic archaea compared to healthy controls, correlatively suggesting their potential role in colorectal cancer²³.

Beyond bacterial gastrointestinal cancer risk factors, the oropharyngeal and genitourinary tracts are home to microorganisms with implications for cancer risk (FIG. 1)^6,24–27. For example, the oral etiological agent Porphyromonas gingivalis induces inflammation and oral tissue destruction, contributing to its association with gingival and oral squamous cell carcinoma^28,29. Further, reduction in certain Lactobacillus species in the gynecologic microbiota has been associated with ovarian and cervical cancer due in part to their role in pH reduction, HPV infection protection, and immunomodulation³⁰.

Fig. 1. Microbiota members associated with cancer risk at different body sites.

a) Microbiota members from the oropharyngeal, gastrointestinal, and genitourinary regions are implicated in cancer development and progression. Specifically, Fusobacterium nucleatum, Porphyromonas gingivalis, Escherichia coli, Helicobacter pylori, and Lactobacillus spp. are recognized bacterial risk factors for cancer. F. nucleatum can contribute to cancer development by inhibiting immune cell activity and activating tumorigenic pathways through outer membrane adhesins, Fap2 and FadA²⁹. Further, F. nucleatum is known to translocate widely throughout the body, leading to its carcinogenic effects beyond the oral microbiome, including colorectal cancer¹⁸⁵. P. gingivalis is equipped to inhibit cancer cell apoptosis through the activation of host signalling pathways, the expression of microRNAs, the secretion of nucleoside diphosphate kinases, and the inhibition of tumour suppressor p53²⁹. P. gingivalis infection can activate various inflammatory pathways that induce the expression of pro-matrix metalloproteinase-9 (MMP-9)²⁹, which is activated by P. gingivalis-secreted gingipains, leading to increased cell invasion, migration, and metastatic growth²⁹. Enhanced P. gingivalis abundance has been observed in pancreatic, gingival, oesophageal, and colorectal cancer studies^{28,29,186–190}. Lactobacillus spp. help maintain reproduction health, but the role of each species varies¹⁹¹. Reduced numbers of Lactobacillus species have been observed in samples of patients with urinary, ovarian, gastric, colorectal, and breast cancer^192–195. Due to a decrease in lactic acid production, the pH shifts and creates a less hostile colonization niche for non-Lactobacillus species, enhancing susceptibility to infection. Lactobacilllus crispatus and Lactobacillus jensenii exhibit a reductive trend in cervical cancer while Lactobacillus iners enrichment is associated with a possible detrimental role^30,196. H. pylori and E. coli are well-known contributors to gastric and colorectal cancer, respectively. b) Metabolites with links to cancer development include reactive oxygen and nitrogen species like hydrogen peroxide (contributes to sperminal), (geno)toxins like tilimycin, tilivalline and colibactin, and indole-functionalized metabolites including ʟ-tryptophan and indolimines.

In this Review, we focus on microbial-derived small molecule metabolites with established and emerging mechanistic links to the development of gastrointestinal cancer, given the broad scope of this field. Important molecular families of metabolites such as bile acids, short-chain fatty acids, and oligosaccharides are not covered. The reader is also directed to available reviews discussing the connections between microorganisms and specific cancers, as well as the roles of particular strains in carcinogenesis^6,24,29–31 Specifically, we explore the carcinogenic roles of reactive oxygen and nitrogen species, toxins and genotoxins, and functionalized indole metabolites. We highlight their bacterial producers, biosynthetic production methods, and etiological mechanisms, aiming to provide mechanistic insight underlying metabolite–cancer associations and to further support the need for bacterial metabolic and mechanistic elucidation.

Reactive oxygen and nitrogen species

The diversely populated gastrointestinal tract is home to bacteria with established associations to cancer. Leveraging the ease of stool sample collection to represent the colonic microbiome, comparative analyses have uncovered a range of potential cancer risk factors⁴. For instance, reactive oxygen and nitrogen species produced by microbiota members can contribute to cancer development by causing DNA damage, promoting genomic instability, and altering cellular signalling pathways³².

Reactive oxygen species and virulence factors of Helicobacter pylori

Notably, H. pylori stands out as the first bacterial group 1 carcinogen recognized by the World Health Organization³³. Acknowledged as the strongest risk factor for gastric cancer, H. pylori exerts its influence through a variety of mechanisms and metabolites, imparting significant, long-term carcinogenic consequences on its host.

H. pylori utilizes small molecule reactive nitrogen species (RNS) and reactive oxygen species (ROS) to promote the progression of gastric cancer. RNS, such as nitric oxide (·NO) and peroxynitrite (ONOO⁻), are stimulated by H. pylori³⁴. ROS, such as superoxide (·O₂⁻), hydrogen peroxide (H₂O₂), and hydroxyl radicals (·OH), are natural (by)products of cellular processes³⁵. While RNS and ROS serve essential roles in host signalling and defence, uncontrolled production can have negative cellular consequences including DNA damage^35,36. In carcinogenesis, ROS contribute to various stages of cancer progression, with evidence pointing to both anti- and pro-tumorigenic effects, which are dependent on the context and tissue³⁷. The virulence factors employed by H. pylori in carcinogenesis, primarily vacuolating cytotoxin A (VacA) and cytotoxin-associated gene A (CagA) (FIG. 2), have been implicated in promoting ROS. Reviewed thoroughly elsewhere, VacA and CagA target host cellular pathways that contribute to carcinogenesis^38,39. Briefly, VacA binds to cell surface components before internalization, where it goes on to activate inflammatory responses^38–40. VacA induces vacuolation by allowing chloride ions to pass through VacA-formed membrane channels, leading to ATPase activity, proton pumping, and ammonium ion trapping to regulate the impacted cellular compartments. This leads to osmotic swelling and, ultimately, cell vacuolation⁴¹. The other major virulence factor, CagA, is secreted by a type IV secretion system⁴². After entering the cell, CagA alters various signalling pathways causing the disruption of tight-epithelial barrier function in Madin-Darby canine kidney (MDCK) cells when delivered long-term, activation of nuclear factor kappa B (NF-ĸB), and induction of the oncogene SHP2^43–45. Additionally, crosstalk between the virulence factors has been investigated, with a study reporting the accumulation of CagA in dysfunctional autophagosomes of gastric epithelial cells in the presence of VacA⁴⁶.

Fig. 2. Reactive oxygen species and virulence factors produced by persistent Helicobacter pylori infections are implicated in gastric cancer.

H. pylori infection can result in prolonged inflammatory response and DNA damage due to reactive oxygen species (ROS) production. The virulence factors VacA and CagA from H. pylori contribute to ROS production by increasing immune modulation and inflammation^38,39. a) VacA binds to cell-surface components and activates inflammatory responses once internalized, forming anion-selective membrane channels¹⁹⁷. Chloride ions (Cl⁻) pass through these channels increasing intralumenal chloride levels. To counteract this influx, vacuolar ATPase activity is upregulated, increasing proton (H⁺) pumping and reducing compartmental pH. Weak bases, like ammonia (NH₃), diffuse into these compartments and are trapped following protonation, leading to osmotic swelling and cell vacuolation⁴¹. b) CagA is delivered inside gastric epithelial cells by a type IV secretion system, where it is phosphorylated by host cell kinases, and interacts with host cell proteins to alter signaling pathways. Phosphorylated CagA binds and activates the oncoprotein SRC-homology 2 (SHP2), which induces the extracellular signal-regulated kinase (ERK) via RAS-dependent and -independent mechanisms⁴³. Additionally, CagA activates NF-ĸB, upregulating the expression of proinflammatory cytokines interleukin (IL)-8 and tumour necrosis factor (TNF)-α in gastric epithelial cells⁴⁴. Long-term delivery of CagA disrupts tight-epithelial barrier function, alters cell morphology, and prevents normal tight junction formation⁴⁵. c) In infected macrophages, ROS are produced by spermine oxidase. After being transported into the macrophage, ʟ-arginine is converted to ʟ-ornithine by arginase, generating urea as a byproduct⁵¹. ʟ-ornithine undergoes polyamine conversion to putrescine by a decarboxylase, metabolism by spermidine synthase to spermidine, and subsequent conversion to spermine by spermine synthase. Spermine oxidase (SMOX) reverts spermine back to spermidine, generating hydrogen peroxide (H₂O₂). Sperminal and spermindial could induce double-strand DNA breaks^{53–55,198,199}. d) In polymorphonuclear neutrophils (PMN), ROS are produced by nicotinamide adenine dinucleotide phosphate (NADPH) oxidase⁴⁹. Neutrophils engulf pathogens into phagosomes and use NADPH-dependent ROS within the phagosome to eliminate pathogens. During phagocytosis, the catalytic subunit of NADPH oxidase is activated and receives an electron from cytoplasmic NADPH, which is donated to oxygen either inside or outside the phagosome, forming superoxide (O₂⁻) that is converted to membrane-permeable H₂O₂. Hydrogen peroxide forms hypochlorous acid (HOCl), which can lead to toxic monochloramine (NH₂Cl) production⁴⁹. NH₂Cl is the result of the reaction between HOCl and ammonium, which is available through urea-to-ammonia conversion mediated by the urease activity of H. pylori⁴⁹. Additionally, hydrogen peroxide can react non-enzymatically with iron (Fe²⁺) or copper (Cu⁺) ions to generate hydroxyl radicals. e) H. pylori persists in the gastric environment and evades immune responses, driving long-term inflammation and carcinogenesis. Part a is adapted from ref.³⁹, CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Part d is adapted from ref.⁴⁹, Springer Nature Limited.

H. pylori strains expressing CagA substantially increased ROS production in the thoracic aorta of mice via CagA-containing exosome induction, compared to mice infected with H. pylori that did not express CagA⁴⁷. Further, ROS production was increased in gastric epithelial cells following transfection with the CagA gene⁴⁸. Alternatively, VacA increases ROS levels by promoting CagA activation. Treatment of gastric cancer cells with VacA resulted in elevated ROS levels and proliferation⁴⁸. Together, CagA and VacA are dually implicated in increasing the risk of gastric cancer through their individual mechanisms and their promotion of ROS.

In the early stages of cancer development, the accumulation of autophagosomes induced by H. pylori can lead to elevated ROS production and persistent oxidative stress. The increase in ROS primarily comes from the nicotinamide adenine dinucleotide phosphate (NADPH)-dependent reduction of oxygen to superoxide, or spermine oxidase acting within the spermine metabolic pathway to produce hydrogen peroxide (FIG. 2)⁴⁹. In polymorphonuclear neutrophils, ROS are produced by NADPH oxidase when pathogens are first engulfed into phagosomes in an attempt to eliminate pathogens⁵⁰. During phagocytosis, the catalytic oxidase subunit is activated and primed to receive an electron from cytoplasmic NADPH. The electron can be donated to oxygen inside or outside of the phagosome to form superoxide. Superoxide can be converted to hydrogen peroxide, which can be transformed into hydroxyl radicals or hypochlorous acid⁴⁹. Hypochlorous acid can be converted to toxic monochloramine through a reaction with ammonium, available via the activity of H. pylori urease. Additionally, hydrogen peroxide can react non-enzymatically with iron or copper ions to generate hydroxyl radicals. In a healthy stomach, hypochlorous acid and hydroxyl radicals eliminate pathogenic bacteria in phagocytes. However, H. pylori evades phagocyte–ROS eradication, leading to further elevated ROS levels and increased cytotoxicity. In the alternative spermine metabolic pathway that produces ROS, ʟ-arginine is transported into H. pylori-infected macrophages by the ʟ-arginine transporter SLC7A2⁵¹. In H. pylori-infected macrophages, ʟ-arginine is then converted to ʟ-ornithine by arginase, generating urea as a byproduct. ʟ-ornithine serves as a precursor for the synthesis of the polyamine putrescine by ornithine decarboxylase 1 (ODC1). Putrescine is subsequently metabolized by spermidine synthase to spermidine, and to spermine by spermine synthase. Reversion of spermine back to spermidine by spermine oxidase generates hydrogen peroxide, previously shown to participate in mitochondrial membrane depolarization^51,52. Recently, it was shown that products generated from peroxide oxidation of spermine, sperminal and spermindial, induce a double-strand break response (γH2AX) in a human cervical cancer cell model and cause DNA damage in vitro (FIG. 2), further supporting alternative indirect mechanisms for ROS-mediated DNA damage^53–55. ROS and virulence factor production collectively enable H. pylori to persist in the stomach, promoting chronic inflammation, DNA damage, and ultimately contributing to the development of gastric cancer (FIG. 2)⁵⁶.

Reactive oxygen species produced by the microbiota

Alongside H. pylori, the microbiome harbours other ROS-producing members that have been associated with gastrointestinal cancers. Select Enterococcus, Streptococcus, Escherichia, Salmonella, and Bifidobacterium species participate in ROS production, impacting immune response and disease progression⁵⁷. Enterococcus faecalis, for example, is linked to the induction of colorectal cancer through the generation of extracellular superoxide⁵⁸. In E. faecalis, superoxide is formed through the autoxidation of membrane-associated demethylmenaquinone. In the mildly acidic intestinal environment, superoxide converts to hydroxyl radical and hydrogen peroxide⁵⁸. Subsequently, hydrogen peroxide can infiltrate colonocytes, leading to DNA damage. This hypothesis is supported by both in vitro and in vivo studies, which used comet assays to demonstrate that peroxide generated from E. faecalis superoxide damages DNA in tissue culture cells and rat luminal colon cells⁵⁸. Furthermore, research has revealed that superoxide produced by E. faecalis plays a pivotal role in promoting chromosomal instability by inducing cyclooxygenase-2 (COX2) in macrophages⁵⁹. This process may initiate a bystander effect that exacerbates colorectal cancer, where COX2 byproducts induced by E. faecalis superoxide propagate genomic instability in neighbouring cells⁵⁹.

Compositional impact of reactive oxygen species

Beyond impacting the host, the production of ROS can disrupt the microbiota composition⁶⁰. Facultative anaerobes are upregulated by excess ROS production, since ROS-mediated products like nitrates and nitrogen or sulfur oxides can serve as electron acceptors, supporting bacterial growth through anaerobic respiration⁶¹. For example, in a mouse colitis model, Salmonella enterica serovar Typhimurium gained a growth advantage by producing tetrathionate, a respiratory electron acceptor, through a reaction between inflammatory ROS and endogenous sulfur⁶². Colitis is a chronic inflammatory condition in the colon and a known risk factor for colorectal cancer⁶³. The interplay between colitis-associated inflammation and bacterial mediators is acknowledged as a driver of colitis-associated colorectal cancer. Further, an increase in ROS during inflammation can create a microaerobic niche in which preferential E. coli colonization occurs^64,65. This microbial niche, partially driven by ROS, can create a feedback loop that promotes genotoxin-producing E. coli expansion^8,66. These studies demonstrate the impact of microbial ROS on the host and microbiota, highlighting their significance in relation to gastrointestinal cancer risk.

Toxins and genotoxins

Microbial toxins and genotoxins are often produced by gastrointestinal microorganisms to limit predation and generally compete in their polymicrobial and host-microbial environments⁶⁷. Toxins can contribute to carcinogenesis by promoting chronic inflammation, suppressing immune response, enhancing cell proliferation, inducing angiogenesis, and causing DNA damage^68–71.

Tilimycin and tilivalline

Klebsiella, a gram-negative gammaproteobacterial genus, has drawn interest due to its diverse pathogenic potential, causing infections ranging from pneumonia to septicemia⁷². While commonly found in the human gastrointestinal tract, specific strains of Klebsiella have evolved into opportunistic pathogens, prompting investigations into their role in conditions such as antibiotic-associated hemorrhagic colitis (AAHC)¹⁸. AAHC is characterized by severe colonic inflammation posterior to antibiotic treatment. Two small molecule toxins, tilivalline and tilimycin, produced by certain β-lactam-resistant Klebsiella species, have shed light on the mechanisms underlying AAHC and the potential implications of toxin-induced inflammation and cancer risk^17,18.

Initial investigations into Klebsiella oxytoca pathogenicity led to the identification of tilivalline, a pyrrolobenzodiazepine metabolite¹⁷. In human Hep2 cells, tilivalline was found to emulate phenotypic associations observed in AAHC patients and corresponding animal models. Tilivalline is synthesized through a nonribosomal peptide synthetase (NRPS) assembly line. While investigating the biosynthesis of this pathway, a second pyrrolobenzodiazepine, tilimycin, with even stronger activity was identified⁷³. Biosynthesis of tilimycin and tilivalline starts with the loading of 3-hydroxanthranilic acid and ʟ-proline onto the NRPS modules. Modular processing results in the reductive release of N-acylprolinal, which spontaneously undergoes ring closure to form tilimycin, forming an electrophilic motif upon dehydration. Tilimycin can undergo further spontaneous conversion to tilivalline via this dehydration reaction and an electrophilic aromatic substitution reaction using free indole as the nucleophilic partner (FIG. 3a)⁷³.

Fig. 3. Klebsiella oxytoca produces enterotoxins tilimycin and tilivalline.

a) A nonribosomal peptide synthetase (NRPS) pathway composed of the three proteins NpsA, ThdA and NpsB, releases ʟ-N-(3-hydroxyanthraniloyl) prolinal, which undergoes spontaneous cyclization to tilimycin diastereomers. Tilivalline is the product of nonenzymatic condensation of the tilimycin dehydration product with indole. Letters in circles indicate adenylation (A), thiolation (T), condensation (C) and thioester-reductase (Re) domains of NRPS proteins. b) Tilivalline binds and stabilizes microtubules, leading to increased polymerization of tubulin into microtubules within asters, which are a structural component of the cell made of microtubules constructed immediately preceding mitosis. Tilivalline ultimately leads to downstream mitotic arrest. c) Tilimycin, a known DNA alkylator, permeates colonic crypts and causes colonic somatic mutations following prolonged exposure.

Both tilivalline and tilimycin individually induce intestinal epithelial cell apoptosis, a hallmark phenotype of AAHC^17,18. However, they function through different molecular targets⁷⁴. Tilivalline was found to target microtubules, a discovery made when observing disrupted spindle morphologies and micronucleation in lung and colon cancer cell lines⁷⁴. By binding to microtubules, tilivalline inhibits depolymerization, leading to cell cycle arrest (FIG. 3b). Conversely, tilimycin functions as a genotoxin and binds to the minor groove of double-stranded DNA, triggering DNA damage responses⁷⁴. The genotoxic phenotypes of tilimycin were confirmed in mice infected with tilimycin-producing bacteria, which showed heightened intestinal epithelial genome instability and lesion burden in the ceca of infected mice⁷⁴. DNA adduct analysis revealed its ability to alkylate the N2 position of the guanine nucleobase⁷⁵. Research further established the role of tilimycin as an enteric mutagen that permeates murine colorectal crypt lumen and increases somatic stem cell mutation frequency, potentially impacting epithelial resilience (FIG. 3c)⁷⁶. The accumulation of somatic mutations over time, followed by clonal expansion, may contribute to the development of cancer over the long term.

Colibactin

Colibactin is a small molecule genotoxin encoded by a 54 kilobase NRPS–polyketide synthase (PKS) hybrid genomic island (FIG. 4a)⁷⁷. While the island was first discovered in E. coli, it has since been found in Klebsiella pneumoniae, Enterobacter aerogenes, Citrobacter koseri, and Frischella perrara^77–79. The colibactin gene cluster contains 19 genes that encode three PKSs (ClbC, ClbI, and ClbO), three NRPSs (ClbH, ClbJ, and ClbN), two hybrid PKS–NRPS systems (ClbB and ClbK), and eleven tailoring and accessory proteins⁷⁷. The pathway requires posttranslational activation of its carrier protein domains by the phosphopantetheinyl transferase ClbA. Amino acid, acyl-CoA, and malonyl-CoA building blocks are selected and loaded onto the enzymatic modules for the assembly line production of precolibactin. Amidase ClbL promotes heterodimerization to generate the proposed asymmetric structure precolibactin 1491^80,81.

Fig. 4. Colibactin is a microbial risk factor for colorectal cancer.

a) The colibactin biosynthetic gene cluster encodes for the production of predicted mature precolibactin 1491. The colibactin gene cluster contains 19 genes that encode three polyketide synthases (PKSs), three nonribosomal peptide synthetases (NRPSs), two hybrid PKS/NRPS systems, and 11 tailoring and accessory proteins. Amino acid, acyl- and malonyl-CoA building blocks are selected and loaded onto the enzymatic modules to produce precolibactin. Amidase ClbL promotes heterodimerization to generate a proposed asymmetric structure precolibactin 1491. b) Precolibactins are transported into the periplasm before being cleaved by peptidase ClbP to release a dual-warhead metabolite, which can undergo further cyclizations. ClbP peptidase activation can be blocked by inhibitors featuring a boronic acid motif. Colibactin(s) traffic to host cell nuclei by an unknown mechanism, indicated by ‘?’. c) Colibactin electrophilic spirocyclopropane moieties can cross-link DNA. Additional electrophilic sites of colibactin have the potential for additional unexplored reactions such as addition at the C4 of the lactam ring or addition at the central dicarbonyl spacer separating the two warhead motifs. Crosslinking leads to double-stranded DNA breaks, which contribute to colorectal cancer risk through increased cell proliferation, inflammation, and senescence. Part b and c are adapted from ref.²⁰⁰, Springer Nature Limited.

In this model, precolibactin 1491 is transported by ClbM from the bacterial cytosol into the periplasm and undergoes two-fold cleavage by ClbP, a membrane-bound peptidase with an active site facing the periplasmic space^82–85. These cleavage events release two N-acyl-D-Asparagine prodrug motifs, facilitating cyclization cascades to establish two genotoxic ‘warheads’^10,80,86. The mature genotoxin is a heterodimer containing two electrophilic spirocyclopropane moieties. A two-carbon linker, derived in part from an α-amino-malonyl polyketide extender unit, separates two central thiazole rings. Under aerobic cell culture conditions, the linker undergoes oxidation and hydrolysis, to a detectable diketone product. Intriguingly, it was found that the diketone spacer also serves as an electrophilic motif and can undergo hydrolytic degradation⁸⁷. While synthesizing stable mimetics of colibactin that attenuated the reactivity of the linker, it was found that colibactin could further cyclize to form a β-hydroxy-γ-lactam (FIG. 4b)⁸⁸. Additional cyclizations of (pre)colibactin have also been observed, including 14 degradation products isolated using a colibactin-overproducing strain, E. coli-50, as well as a macrocyclic structure that was found to induce double-strand DNA breaks in vitro and in cell culture^89–91. For example, macrocyclic colibactin-645 induced DNA breaks through a copper-mediated oxidative mechanism, where it was proposed to bind to copper in the intestinal lumen and be reduced and ‘activated’ by oxygen coordination in host cells⁹⁰. It is worth noting that these metabolites were initially identified in a ClbQ thioesterase mutant, whereas clbQ is required for cellular genotoxic effects. Further colibactin–DNA structural studies are needed to support the extent of cyclization in DNA–colibactin adducts and crosslinks. Regardless, evidence supports that at least one of the electrophilic spiorcyclopropane moieties alkylates the N3 position of a bis-adenine DNA crosslink (FIG. 4c)^80,92,93. The ring-opened form is subject to peroxidation and generation of an additional electrophilic site at the C4 position of the lactam ring, identified during ClbS drug resistance mechanism investigations⁹⁴. In the absence of a suitable nucleophile, exocyclic double-bond isomerization was observed (FIG. 4c)⁹². Given the number of reactive sites in colibactin, it is likely that other cross-linking reactions could occur (for example, colibactin–histone adducts).

Mechanistic studies of the colibactin metabolites are largely in agreement with cellular studies on the colibactin pathway. Exposure of mammalian epithelial cells to colibactin-producing bacteria stimulates cellular enlargement (megalocytosis), the formation of double-strand DNA breaks (γH2AX phosphorylation), cell cycle arrest, and genomic instability, which is exacerbated under inflammatory conditions^77,95,96. The colibactin pathway also activates the Fanconi anaemia pathway DNA cross-link repair machinery, as demonstrated by the recruitment of Fanconi anaemia repair protein, FANCD2, upon bacterial infection of mammalian cells⁹⁷. Adherent invasive E. coli (AIEC) strains carrying the colibactin pathway (for example, E. coli NC101) have been shown to initiate colorectal cancer in multiple mouse models^8,16,66. Clinical studies complement these findings by revealing an increased abundance of colibactin-positive (clb⁺; commonly referred to as pks⁺, but we refer to it here as clb⁺ to avoid confusion with the broader PKS field) E. coli in patients with familial adenomatous polyposis (68%), colorectal cancer (67%), and inflammatory bowel disease (40%) in comparison to healthy individuals (22%)^{16,66,98–100}. Prolonged exposure to clb⁺ E. coli was found to promote the emergence of two distinct mutational signatures in human organoids. The analysis of a patients with colorectal cancer reported that 2.4% of colorectal cancer driver mutations matched the colibactin motif, appearing half as frequently as the most recognized colorectal cancer gene mutation, Apc (HGNC:583)¹⁰¹. Another study reported colibactin-induced DNA damage signatures in 10% of genomes from individuals with colorectal cancer¹⁰². These mutational signatures parallel those observed in cancer patients and were consistent with the colibactin structures that crosslink DNA¹⁰². A study investigated the contribution of clb⁺ E. coli mutations using whole-genome sequencing on a matched clinical data set, comparing the genomes of healthy individuals to distant normal crypts, normal crypts that were adjacent to the tumour, and cancer genomes from the same individual¹⁰³. They found that patients with colorectal cancer had an increased incidence of clb⁺ mutational signatures compared to healthy individuals. Interestingly, they found that the mutational signatures were highly enriched in both tumour and matching distant and adjacent normal samples compared to healthy controls, supporting the role of these signatures in early tumorigenesis, as well as the role of clb⁺ E. coli as a driver of human colorectal carcinogenesis¹⁰³.

Although structural and mechanistic studies of colibactin establish its role as a significant risk factor for colorectal cancer, studies to suppress colibactin carcinogenic activity are ongoing. For example, several boronic acid small molecules were found to block the genotoxic effects of colibactin, including characteristic adduct formation, FANCD2 ubiquitination, and cell cycle arrest, by ClbP inhibition in mammalian cell cultures (FIG. 4b)¹⁰⁴. These inhibitors were designed to mimic the structure of precolibactin, containing a ClbP-recognizable ᴅ-asparagine side chain and a lipid group, where they form a covalent bond with the catalytic serine in the peptidase active site. Moreover, questions about colibactin dependency on bacterial cell-to-human cell contact⁷⁷, precise DNA–adduct structures, additional electrophilic motif functionality, and the relevance of metabolite oxidation remain unresolved.

Cytolethal distending toxin and typhoid toxin

Other pathogens serve as genotoxin producers with distinct metabolic outputs and mechanisms of action. Bacterial genotoxins like tilimycin and colibactin target DNA, leading to genomic instability, cell senescence if unrepaired⁹⁵, and an increased risk of health issues, including AAHC risk for tilimycin due to prolonged cellular malfunction¹⁰⁵. Beyond tilimycin and colibactin, two protein genotoxins have been well studied: cytolethal distending toxin (CDT) and typhoid toxin. While the scope of this review focuses on small molecule metabolites, CDT and typhoid toxin regulate cancer-associated processes and potentially impact metabolite distribution. Briefly, CDT comprises three subunits — CdtB, the catalytically active subunit, and CdtA–CdtC, the binding complex — and induces G₂/M cell cycle arrest and cell enlargement in mammalian cell models, as well as causing double-strand DNA breaks, classifying it as a protein-based genotoxin^106,107. In contrast to all other CDTs, typhoid toxin, found in Salmonella enterica serotypes, lacks CdtA and CdtC, utilizing pertussis and subtilase-like toxin subunits for binding instead¹⁰⁸. Closely associated with gallbladder cancer, typhoid toxin shares CdtB-dependent activities with CDT while employing distinct toxin delivery and bacterial internalization mechanisms¹⁰⁹. The impact that these toxins may have on metabolite presence is yet to be fully understood. By comparing a rat model inoculated with recombinant CdtB versus vehicle control, researchers found ileal gene expression and microbiome differences¹¹⁰. For example, Lactobacillus spp. and Desulfovibrio spp. levels varied between the cohorts, suggesting potential changes in lactic acid or sulfur species abundance, respectively. Further, plasma samples from patient were used to find typhoid carrier biomarkers, identifying glutaric and caproic acid, as well as three known metabolites, as metabolic biomarkers with unexplored toxin or mechanistic implications¹¹¹.

Protein toxins

Bacterial toxins, Shiga toxin, cycle inhibiting factor (Cif), enterotoxigenic Bacteroides fragilis toxin (BFT), and several species-specific Clostridium toxins, target non-DNA components. Shiga toxin, a multi-subunit protein complex produced by certain E. coli and Shigella spp., damages vascular epithelium and can lead to haemolytic-uremic syndrome^112–114. Mechanistically, Shiga toxin binds to glycosphingolipid globotriaosylceramide (Gb3) via its B subunit, forming membrane tubules for bacterial uptake, where subsequent cytosolic A subunit cleavage in the ribosomal subunit blocks protein synthesis^115,116. Gb3 is highly expressed in many types of cancers, with studies demonstrating its upregulation in colorectal cancer, compared to its restricted expression levels in normal tissues^116,117. Shiga toxin has not been implicated in cancer risk, however, specific binding to Gb3-expressing cancer cells highlights the potential of Shiga toxin as a cancer therapy when coupled to a therapeutic compound¹¹⁶. Further, butyrate upregulates Gb3, suggesting metabolic interplay between Shiga toxin-producers and butyrate-producers within the microbiota¹¹⁸. Cif, an effector of a type III secretion system first identified in enteropathogenic E. coli, induces stress fibre formation by rearranging the actin cytoskeleton, which mediates cell cycle arrest^119,120. By inhibiting cullin-ring ubiquitin ligase activity, Cif leads to the accumulation of p21 and p27 regulator proteins in the cell, altering metabolites and metabolic pathways that have yet to be explored¹²¹.

Enterotoxigenic B. fragilis (ETBF) exacerbates cancer risk through its biofilm and toxin production. ETFB biofilms induce chronic inflammation and tissue injury contributing to carcinogenesis^31,122. Biofilms present in the colon have also been shown to influence the cancer metabolome, upregulating polyamines, including N¹, N¹²-diacetylspermine, in colon cancer tissue samples¹²³. By disrupting polyamine metabolism, ETBF biofilms may indirectly influence microbial metabolites such as spermidine modulation of colibactin production¹²⁴. Additionally, the 20-kilodalton BFT toxin produced by ETFB instructs myeloid-derived suppressor cell recruitment into tumours, activates mitogen-activated protein kinases (MAPK) and signal transducer and activator of transcription 3 (STAT3) pathways, increases spermine oxidase expression, disrupts tight junctions, and promotes cell proliferation^31,125. Further, BFT cleaves E-cadherin, thus activating β-catenin and leading to loss of membrane association and increasing colon permeability, which drive the early progression of colorectal cancer^24,126. Known to activate chemokine secretion by polarizing proinflammatory interleukin (IL)-17 and IL-23 production, BFT may participate in the interplay with available bacterial or tumour metabolites such as lactic acid, a small molecule secreted by tumour cells that activates the proinflammatory IL-17–IL-23 pathway^31,127. Unlike the link of polyamines to biofilms, small molecule metabolites impacted by BFT remain unclear and represent a promising area for future research.

Clostridium species produce diverse toxins. For example, Clostridium septicum produces four toxins, and its α-toxin has been strongly correlated with patients with colorectal cancer compared to healthy controls¹²⁸. The toxin leads to cell death via cell membrane disruption and mitochondrial dysregulation and can downregulate immune response by inducing neutrophil apoptosis^128,129. Recombinant α-toxin was shown to induce pores in MDCK cells, resulting in increased charged molecule transport (potassium chloride) as well as potential wider metabolic transport effects¹³⁰. Toxin production by Clostridium perfringens varies among its strains, with some producing up to six distinct toxins. The toxins work through the activation of various intracellular pathways followed by downstream cell death¹³¹. Clostridioides difficile secretes toxin A and B, which inactivate key actin cytoskeleton regulators through glucosylation⁶⁸. The disruptive effects of these toxins increase intestinal barrier permeability and stimulate inflammatory cytokines⁶⁸. A study treating Apc^min/+ mice (mice that develop multiple intestinal adenomas) with patient-derived toxic C. difficile strains found that colorectal cancer tumorigenic phenotypes were dependent on toxin B (TcdB)¹³². TcdB was linked to the activation of Wnt signalling and altered immune responses, as well as an increase in ROS. Although many connections between metabolites and C. difficile infections have been identified, their mechanistic and carcinogenic roles remain largely unexplored¹³³.

Reviewed thoroughly elsewhere, these protein toxins serve as potential bacterial risk factors for cancer^68–71. While studies continue to probe the mechanistic roles of protein toxins, interactions between toxins and bacterial metabolites remain understudied. A deeper understanding of the interplay between toxins and metabolites could provide new insights into cancer development or prevention strategies.

Indole-functionalized metabolites

Indoles

Indoles are a class of heterocyclic organic compounds that feature an indole moiety derived from tryptophan metabolism. This aromatic group provides indoles with a range of biological activities in various physiological processes¹². Under typical physiological conditions, a majority of available tryptophan is metabolized to ʟ-kynurenine in the liver and extrahepatic tissues by tryptophan 2,3-dioxygenase (TDO) and indoleamine 2,3-dioxygenase (IDO1), respectively¹². The remaining tryptophan fraction is metabolized by the microbiome. Many bacterially derived indoles and ʟ-kynurenine serve as ligands for the aryl hydrocarbon receptor (AhR), a multifunctional ligand-activated transcription factor. Through AhR activation, indolic compounds can drive both pro- and anti-inflammatory effects. For example, indoles can promote anti-inflammatory IL-10 expression and reduce proinflammatory NF-ĸB and IL-8 in select immune cells¹³⁴. Indoles, such as indole-3-carboxaldehyde and indole-3-carbinol, can induce IL-22^135,136. Although the production of IL-22 can enhance gut barrier function and prevent colitis-induced microbial dysbiosis, its production during late stages of cancer can promote tumour progression^137,138. Together, indolic compounds can regulate inflammation through cytokine production and immune cell modulation.

Disruption of tryptophan metabolism can impact cancer risk. Specifically, IDO1 is overexpressed in various cancers, driven in part by enhanced proinflammatory COX2 expression and interferon-gamma (IFNγ) signalling¹². This augmented expression of IDO1 is often triggered by sustained proinflammatory conditions perpetuated by microorganisms such as E. coli, E. faecalis, or F. nucleatum. Consequently, this dysregulation depletes tryptophan levels, resulting in elevated ʟ-kynurenine and constitutive activation of AhR. The impact of enhanced AhR activation is still being explored, but studies have shown that it plays a role in cancer migration and invasivity¹³⁹. In primary colorectal cancer cell lines, AhR was identified as the downstream effector responsible for increased expression of the migratory mediator MEMO1¹⁴⁰. Further, it induces MMP-9 through the c-Jun pathway to enhance invasiveness in a gastric cancer cell line¹⁴¹. Together, this series of metabolic events establish an immunotolerant environment that is conducive to tumour growth (FIG. 5a). Supporting constitutive AhR activation by increased kynurenine availability in a tumorigenic microenvironment, a faecal metabolite comparison between patients with colorectal cancer and healthy control subjects recently revealed an elevated kynurenine-to-tryptophan ratio (41% increase) accompanied by a correspondingly diminished indole-to-tryptophan ratio (45% decrease) in colorectal cancer patient samples samples^9,12,142. Using various AhR reporter assays, indolic compounds, including indole, tryptamine, indole-3-acetaldehyde (I3A), indole-3-acetate (I3AA), and indole-3-aldehyde (IAld), have been identified as ligands of the AhR receptor^137,143. For example, a study using a luciferase gene expression system in rat hepatoma-derived reporter cells showed that tryptamine induced the activation of AhR in a concentration-dependent manner (28% of positive control levels at 80 micromolar stimulation)¹⁴⁴.

Fig. 5. Microbial tryptophan-indole metabolism influences colonic tumorigenesis.

a) Microorganisms can convert tryptophan into other indoles via different enzymatic reactions. The enzymes involved in tryptophan transformations are indicated in the coloured legend. b) Tryptophan conversion is dysregulated in tumorigenic environments. The overexpression of indoleamine 2,3-dioxygenase (IDO1) and the depletion of tryptophan by microbial metabolism contribute to increased ʟ-kynurenine levels and constitutive aryl hydrocarbon receptor (AhR) activation. Overproduction of interleukin (IL)-17 and IL-22, as well as the activation of regulatory T cells (T_reg), drive immune tolerance and hyperproliferation. c) Morganella morganii produces a family of genotoxins, termed indolimines. The functional imine group forms through a proposed condensation of primary amines and the aldehyde of indole-3-aldehyde (IAld). DNA damage induced by indolimines has been shown to exacerbate colonic tumour burden. IAM, indole-3-acetamide; I3A, indole-3-acetaldehyde; I3P, indole-3-pyruvate; ILA, indole-3-lactic acid; I3AA, indole-3-acetic acid; IA, indoleacrylic acid; IPA, 3-Indolepropionic acid.

The role of indoles extends beyond their tumorigenic AhR associations. Certain indole-related small molecules, including indole-3-lactate (ILA), exhibit anti-cancer effects. ILA boasts anti-inflammatory properties through its modulation of immune responses and the reduction of pro-inflammatory factors (for example, IL-8 and NF-ĸB)¹². In one example, ILA produced by Bifidobacterium infantis was shown to curtail TNF-α and lipopolysaccharide-induced IL-8 levels in gut epithelial cells by activating AhR and the Nrf2 pathways¹⁴⁵. Moreover, the anti-inflammatory effects of ILA extend to Lactobacillus gallinarum, an intestinal ILA producer. L. gallinarum inhibited colorectal cancer cell viability and tumorigenesis in an Apc^Min/+ mouse model, and in an ILA-dependent manner, the strain triggered apoptosis in colorectal cancer cell staining assays¹⁴⁶. These findings illustrate the potential of harnessing specific bacterial strains and their metabolomic profiles as potential living therapeutic agents.

Indolimines

Indolic compounds can also serve as reactive precursors in the formation of other indole-functionalized metabolites, such as in the conversion of tilimycin to tilivalline described above. A study screening diverse microbiota members for genotoxic activities in vitro led to the identification of a new set of conjugates termed the ‘indolimines’ (ref¹⁹). Over 100 unique bacterial isolates from patients with inflammatory bowel disease were evaluated. In the primary screen, linearized plasmid DNA was co-incubated with individual bacterial isolates, assessing the DNA damage produced by each isolate. From there, the 24 most damaging isolates underwent a secondary cell culture screening, where a substantial fraction of this population (75%) harboured cellular DNA damage activity distinct from that of colibactin, which was used as a positive control¹⁹. Among the identified strains was Morganella morganii, a Gram-negative bacterium shown through meta-analyses to be enriched in faecal samples from patients with inflammatory bowel disease and colorectal cancer^147–149. Comparative metabolomics and bioactivity-guided characterization techniques led to the identification of electrophilic indolimines that were imine conjugates of primary amines and IAld (FIG. 5b). The characterized family encompasses three active genotoxins: indolimine-214, indolimine-200, and indolimine-248, generated by conjugation with isoamylamine, isobutylamine, and phenethylamine, respectively. The protein coded by the aat gene, a member of the aspartate aminotransferase (AAT) family of pyridoxal-dependent decarboxylases, was established as an essential enzyme for indolimine synthesis in M. morganii. The DNA-damaging activity of the indolimines was established through both cell-free and cell-based assays. Additionally, in a gnotobiotic mouse model subjected to azoxymethane and dextran sulfate sodium, M. morganii alone or in a mock community increased intestinal permeability and tumour burden in an indolimine pathway-dependent manner.

A recent study further examined the signalling role of indolimines¹⁵⁰. Using mouse and human reporter cell lines, the study found that indolimine-200, indolimine-214, and indolimine-248 act as direct AhR ligands. Indolimines, especially indolimine-248, were found to increase CYP1A1 protein (also known as aryl hydrocarbon hydroxylase) expression and promote AhR-dependent CYP1A1-mediated carcinogen benzo(a)pyrene metabolization. Previous studies illustrated the role of AhR in priming the IL-6 promoter, suggesting AhR-mediated IL-6 promoter repression may explain the increased IL-6 expression in the presence of inflammatory signals, such as IL-1β¹⁵¹. By treating human colorectal adenocarcinoma (Caco2) cells with IL-1β and each indolimine, the authors of the study demonstrated combinatorial IL-6 activation, supporting indolimine-mediated AhR ligand activity¹⁵⁰. Contrasting downstream signalling roles for AhR activation may be both cell type- and ligand-dependent. For example, studies have shown that AhR activation attenuates colitis in mouse models and can beneficially reshape mouse small intestinal epithelial programming to promote intestinal resilience^152–154. Alternatively, AhR activation may lead to enhanced tumour outgrowth with comparative analyses often finding a high level of AhR expression versus healthy tissue^155,156. Therefore, metabolites acting as AhR activators may represent a double-edged sword in cancer, taking on tumour preventative or promotive roles depending on their microenvironment.

Bacterial translocation and metabolite distribution

Translocation enables microbiota members (and their metabolites) to reach beyond their original point of colonization¹⁵⁷. For example, the oropharyngeal microbiome is rich in colonization diversity, and many of its members have been investigated for their association to head and neck cancers¹⁵⁸. However, the oral inhabitant F. nucleatum, comprised of four subspecies (nucleatum, animalis, vincentii, and polymorphum) that have been proposed as independent species¹⁵⁹, is recognized as an emerging pathogen¹⁶⁰. Studies support that F. nucleatum present in the gut, typically rare in a healthy gut but enriched in the gut of subjects with colorectal cancer, originates from the oral cavity. Researchers found identical strains of F. nucleatum in patients’ saliva and colorectal cancer tissue samples, indicating oral–gut translocation¹⁶¹. This translocation likely occurs through hematogenous transfer or by descending the digestive tract. A study comparing oral gavage versus intravenous inoculation of F. nucleatum in an orthotopic mouse model found that blood-borne F. nucleatum colonized more effectively, suggesting a preference for the circulatory colorectal cancer colonization route¹⁶². F. nucleatum can translocate far beyond the oral microbiome, and it has been observed enriched in gastric, colorectal, prostate, and urinary tract cancers^{27,30,160,163,164}. Further mechanistic studies are needed to investigate preferred translocation routes of F. nucleatum and associated metabolite mobility.

The pathogenic profile of F. nucleatum is thought to proceed mainly through two outer membrane exposed adhesins Fap2 and FadA^29,165. Specifically, localization of F. nucleatum to colorectal cancer tumours is mediated through the binding of Fap2 to highly tumour-expressed ᴅ-galactose and N-acetyl-ᴅ-galactosamine (Gal–GalNAc)¹⁶⁶. This interaction not only creates an enrichment of F. nucleatum in colorectal cancer but also leads to the inhibition of immune cell activity in cell culture via Fap2 inhibition of the TIGIT immune receptor on natural killer cells¹⁶⁷. Additionally, F. nucleatum activates β-catenin signalling following its attachment to E-cadherin via the fusobacterial adhesin FadA²⁹. Further, F. nucleatum activates tumorigenic p38 and TLR4/MYD88 in epithelial cells and mice, respectively^165,168,169. It was demonstrated that F. nucleatum contributes to colorectal cancer metastasis in multiple mouse models through the upregulation of caspase activation and recruitment domains 3 (CARD3) and downstream autophagy activation¹⁷⁰. Recently, a study identified a distinct clade (FnaC2) within the known colorectal cancer-associated F. nucleatum animalis subspecies (proposed F. animalis¹⁵⁹), which dominates the colorectal cancer niche¹⁷¹. By comparing FnaC1 and FnaC2 genomes, the authors observed that FnaC2 harboured genetic features associated with gastrointestinal cancer colonization, including fap2. Compared to FnaC1, human colorectal cancer cell lines cultured with FnaC2 had higher cell invasion levels, and Apc^+/− mouse models treated with FnaC2 had increased intestinal adenoma numbers. In human cohorts, 16S ribosomal RNA gene sequencing revealed that only FnaC2 was enriched in tumour tissues, a trend similarly observed in stool samples from patients with colorectal cancer. Further, FnaC2 had distinct metabolic capabilities, including possession of 1,2-propanediol and ethanolamine metabolic operons and increased glutathione metabolism, which may contribute to driving carcinogenic conditions.

Studies have begun to explore possible metabolic roles of F. nucleatum in cancer. For example, increased lactic acid and decreased propionic acid levels were observed in samples from patients with colorectal cancer as a result of F. nucleatum-mediated microbial composition changes, highlighting the need for further mechanistic investigation¹⁷². Further, formate produced by F. nucleatum was shown to induce AhR nuclear translocation activity in vitro. Enhanced Th17 cell infiltration as well as tumour incidence and size were observed in vivo. Cancer stemness was increased in vivo with F. nucleatum or formate treatment, suggesting a biological role for F. nucleatum-produced formate¹⁷³. The exact formate-derived AhR activation mechanisms require further investigation.

Conclusions

While many correlations between host microbial composition and disease states exist, the distinction between whether these correlative links contribute to cancer development or are consequences of tumour presence remains enigmatic. Moreover, the microbiome is subject to an individual’s genetics, diet, lifestyle, age, and antibiotic regimen, features that also correlate with cancer risk¹⁷⁴. Mendelian randomization addresses causal relationships between modifiable exposures and disease outcomes by using inherent genetic variation and can serve as supporting causal evidence¹⁷⁵. For example, the use of mendelian randomization analysis on human cohort collections revealed that fatty acids, such as omega-3 and docosahexaenoic acid, were potentially causal in colorectal cancer development; however, follow up studies would be needed to probe the mechanistic relationship¹⁷⁶. Mechanistic studies implementing both epidemiological and experimental methods are required to test and confirm causal relationships.

There are still many questions surrounding the role of metabolites in cancer risk. For example, further research is required to understand how certain metabolites leverage their often dual pro- and anti-cancer activity profiles and if their mechanistic roles act independently or in combination with other factors. Developing therapeutics to address interindividual variation across the microbiota and its metabolic profile requires unexplored multivariable consideration. To explore this, metabolic modelling (for example, BacArena) computationally leverages metabolomic and metagenomic datasets to predict interactions and activities within complex systems¹⁷⁷. A personalized approach would benefit from theoretical modeling-driven hypotheses complemented with biological studies using complex consortia. Another study found that, in patients across seven cancer types, tumours and their adjacent tissues harboured distinct microbial compositions, further suggesting the importance of evaluating the tumour microenvironment as an additional cancer-contributable microbiota and metabolic layer¹⁷⁸.

By characterizing bacterial small molecules, researchers can establish mechanistic connections between microbial metabolic pathways and host phenotypic responses, providing a pipeline for therapeutic intervention. For example, cancer cells adapt to a higher level of ROS, so anticancer drugs with the ability to overwhelm the adaptation through further ROS elevation are being designed³⁵. Additionally, the urea cycle is upregulated during colorectal cancer, promoting a pro-tumoral phenotype in mice¹⁷⁹. This activation is triggered by impaired microbial ureolytic capacity, particularly be reduced Bifidobacterium spp., supporting the use of Bifidobacterium spp. supplements to treat urea-mediated colorectal carcinogenesis. Pathogenic microbiota members, recognized to mechanistically promote cancer, may serve as targets for elimination. In one example, a narrow-spectrum phage therapy using E. coli-targeting CRISPR-Cas machinery selectively reduced E. coli loads in mouse and minipig models¹⁸⁰.

Alternatively, bacteria can be used to produce anti-cancer effects through the production of active metabolites, such as the anti-colorectal cancer activity of Lactobacillus gallinarum via ILA production. Similarly, ILA produced by Lactobacillus reuteri was shown to suppress colorectal tumorigenesis in Apc^Min/+ mice¹⁸¹. Atorvastatin treatment resulted in increased microbial tryptophan levels and subsequent increased L. reuteri abundance, highlighting the importance of studying microbiota variability impacted by individual drug regimens. Further, by blocking the Gb3 receptor in a pretreated human colorectal cancer cell line, total Lactobacillus casei metabolites neutralized Shiga toxin-induced cell cytotoxicity¹⁸².

Engineered strains and combinatorial-chemotherapy strategies can additionally be considered. An engineered Salmonella enterica Typhimurium strain was shown to selectively inhibit pancreatic tumour growth and increase immune cell infiltration through the secretion of ClyA, a pore-forming toxin, into the tumour microenvironment¹⁸³. A dual-function E. coli BL21 strain was engineered to contain the chemotherapeutic drug doxorubicin conjugated to its surface and to overexpress glucose dehydrogenase. A combination of doxorubicin and glucose dehydrogenase overexpression leads to an increase in NADPH production and its subsequent conversion to toxic ROS within the tumour, while selective tumour colonization helps reduce the unwanted side effects produced by doxorubicin alone¹⁸⁴. These examples illustrate the potential of combining microbiota metabolite production with an expanse of diagnostic and therapeutic applications. Thus, by exploring the mechanistic complexity of the human microbiota and its metabolites, therapeutic avenues to address gastrointestinal cancer risk may be uncovered.