Scientists Cracked How a Gut Bacterium’s Toxin Quietly Drives Colorectal Cancer

Up to 20% of healthy adults carry enterotoxigenic Bacteroides fragilis in their gut without knowing it. For most of them, nothing obvious happens. But a new study published in Nature suggests that the toxin this bacterium secretes, known as BFT, may be quietly dismantling the colon's protective barrier in a way that sets the stage for cancer. Researchers have now identified the molecular accomplice that makes this possible: a tight-junction protein called claudin-4.

How BFT Uses Claudin-4 to Destroy E-Cadherin

BFT is a metalloprotease, meaning it cuts proteins the way a molecular scissors would. Its preferred target is E-cadherin, a well-studied tumor suppressor protein that holds epithelial cells together and keeps them behaving normally. When BFT cleaves E-cadherin's outer domain, the epithelial barrier falls apart, inflammation spikes, and cells start proliferating in ways that can seed tumors. What had never been clear was how BFT accomplishes this: previous biochemical assays could not show BFT cleaving E-cadherin directly, which implied a host cell factor was doing something essential.

The new work identifies that factor as claudin-4, a protein already associated with epithelial malignancies and often overexpressed in cancers of the colon, breast, and ovary. A genome-wide CRISPR knockout screen in HT29 colon carcinoma cells pointed unmistakably to claudin-4 as the top hit for BFT host factors. Cells lacking claudin-4 showed 20-fold resistance to BFT-mediated E-cadherin cleavage and a total absence of stable BFT binding at the cell surface. When claudin-4 was re-introduced into normally resistant 293T cells, BFT could suddenly bind and cleave E-cadherin robustly. A single amino acid substitution, T45N, within claudin-4's first extracellular sequence was enough to wipe out stable BFT binding entirely.

The structural logic, modeled using AlphaFold 3, is that claudin-4 acts as a positioning device: it recruits BFT close enough to the membrane to reach the E-cadherin cleavage site, which is otherwise sterically hidden. Without claudin-4 bridging the gap, BFT simply cannot get to its target.

Colorectal Cancer Risk and the Bacterial Toxin Problem

Colorectal cancer is the second-leading cause of cancer death in the United States, and the proportion of cases appearing in people under 50 has been rising for decades. Bacterial toxins have been increasingly recognized as contributors to tumor development, triggering inflammation, disrupting cell-cycle regulation, and remodeling the gut environment in ways that favor malignancy. The colibactin story offers a parallel: childhood exposure to that gut bacterial toxin has been linked to early-onset colorectal cancer through DNA damage imprinted on colon cells years before a tumor appears.

BFT operates differently, working through protein cleavage rather than DNA damage, but the implication is similar: a common commensal organism may be running a slow carcinogenic program in a large fraction of the population. The double knockout of both claudin-4 and the closely related claudin-3 in HT29 cells produced approximately 500-fold resistance to BFT, suggesting the toxin has a limited backup receptor option. Claudin-3 ranked fourth in the CRISPR screen, and the researchers believe it can partially compensate for claudin-4 absence, functioning as a lower-affinity receptor when claudin-4 is gone.

The CRISPR Screen and Cell Biology Behind the Discovery

The research team performed the genome-wide CRISPR knockout screen by treating HT29 cells with BFT, then using surface E-cadherin staining and cell sorting to pull out the rare cells that resisted the toxin. The screen's hits clustered heavily around plasma membrane proteins or factors involved in localizing proteins to the membrane, which already narrowed the field toward claudin-family candidates.

To confirm claudin-4's role in barrier function rather than just E-cadherin cleavage, the researchers measured transepithelial electrical resistance in T84 colon monolayers, a standard test of tight-junction integrity. Claudin-4-knockout monolayers showed substantially reduced BFT-induced resistance loss. A catalytically inactive version of BFT did not disrupt the monolayer at all, confirming that enzymatic activity, not just binding, is required for barrier damage. Co-immunoprecipitation and direct pulldown experiments with a soluble claudin-4 analogue called CLN4sol confirmed a physical 1:1 complex between claudin-4 and BFT.

Open Questions About BFT's Access to E-Cadherin

Several important mechanistic gaps remain. BFT is secreted from the apical surface of bacteria in the gut lumen, but E-cadherin sits on the basolateral side of epithelial cells, behind tight junctions. How the toxin crosses or bypasses those junctions to reach its target is not yet understood. It is also unclear whether BFT preferentially cleaves E-cadherin molecules embedded in junctions or those outside them. The AlphaFold 3 ternary model of claudin-4, BFT, and E-cadherin was generated with low confidence, meaning the fine structural details of the binding interface remain unresolved. The critical role of residue T45 in claudin-4 is evident from the mutagenesis data but is not explained by the current model.

The CRISPR screening strategy itself has a built-in limitation: it cannot identify genes that are essential for cell survival, so other potential BFT receptors or co-receptors may have been missed.

Soluble Claudin-4 as a Potential Therapeutic Decoy

The most immediately translatable finding may be the therapeutic proof of concept. Because claudin-4 physically docks with BFT before cleavage occurs, a soluble form of the protein could act as a decoy, intercepting the toxin before it reaches the cell surface. The researchers tested this idea with CLN4sol, a soluble claudin-4 analogue lacking the transmembrane domain. In HT29 cells, CLN4sol blocked BFT-mediated E-cadherin degradation. In a mouse caecal injection model, co-injecting CLN4sol with BFT significantly reduced epithelial cell shedding and tissue swelling.

This decoy strategy has appeal precisely because it targets a host-toxin interaction rather than the toxin's catalytic site directly, which could reduce the risk of interfering with the structurally similar mammalian ADAM metalloproteases that perform normal tissue remodeling. Whether a claudin-4-based therapy could be developed into a clinical intervention for people with confirmed ETBF colonization remains a long way off, but the mechanistic clarity provided here gives that goal a real molecular foundation to build on.

Evidence rating

EMERGING — 78/100

Caveats:

• This is primarily a mechanistic cell and mouse study — human clinical evidence linking claudin-4/BFT interaction to colorectal cancer outcomes has not yet been established.
• Mouse caecal injection results show promise for a soluble claudin-4 blocker, but no human trials of this therapeutic approach exist.
• Funding source was not disclosed by the authors, so independent verification of the absence of conflicts cannot be confirmed.

Mechanistic study combining genome-wide CRISPR knockout screening in human colon carcinoma cell lines with biochemical binding assays and a mouse caecal injection model. Establishes a detailed causal chain from claudin-4 receptor binding to BFT-mediated E-cadherin cleavage and epithelial barrier disruption, but human clinical data are absent and the findings have not yet been independently replicated.