A team at Johns Hopkins Medicine has identified the exact protein that a cancer-linked bacterial toxin latches onto in the gut, then used that discovery to build a molecular trap that reduced tumor formation in mice. The work, published in Nature in July 2026, centers on Bacteroides fragilis toxin, or BFT, a molecule produced by strains of a common intestinal bacterium. By running a genome-wide CRISPR screen, the researchers pinpointed claudin-4 as the host receptor BFT needs to reach colon cells and trigger the chain of damage that can lead to cancer. They then engineered a soluble version of claudin-4 that acts as a decoy, soaking up the toxin before it can bind to real tissue.
How a Gut Toxin’s Receptor Discovery Changes the Prevention Calculus
BFT works by binding to claudin-4 on the surface of colon cells and then cleaving E-cadherin, a protein that holds cells together. When E-cadherin is cut, the barrier between cells breaks down, allowing inflammation and abnormal cell growth. Identifying claudin-4 as the specific receptor was the missing piece: without knowing the lock, researchers could not design a precise key to block it. The CRISPR screen settled that question by systematically knocking out genes in human cells and observing which deletion made cells resistant to BFT.
The soluble claudin-4 decoy works like a sponge floating in the gut. Because it mimics the real receptor, BFT binds to the decoy instead of to the colon lining. In mouse experiments, this interception reduced the downstream damage that BFT normally causes, including disruption of cell junctions and inflammatory signaling. The approach is notable because it targets the toxin–receptor interaction itself rather than trying to eliminate the bacterium, which can also play harmless or even beneficial roles in the microbiome.
BFT is not the only bacterial product that damages colon DNA. Colibactin, produced by certain strains of Escherichia coli, causes harm through a different mechanism: it can crosslink DNA, creating lesions that can accumulate into mutations over time. The two toxins operate through separate pathways, BFT by breaking cell junctions and colibactin by directly damaging genetic material, but both are found in the same environment and sometimes in the same patient. That overlap raises an important question: could blocking one toxin change how the microbial community responds to the other, or how the host tissue tolerates ongoing damage?
Parallel Strategies for Colibactin and BFT Tested in Mice
The claudin-4 decoy is part of a broader wave of precision tools aimed at specific bacterial toxins rather than at entire microbial populations. In a separate line of research, scientists engineered probiotic bacteria to display ClbS on their surface. ClbS is the protein that colibactin-producing E. coli use to protect themselves from their own toxin. When these engineered bacteria were introduced into mice colonized with colibactin producers, the surface-displayed ClbS reduced genotoxic effects and tumor formation in the colon.
A third approach targets colibactin production at its source. A small-molecule inhibitor described in Nature Chemical Biology blocks the biosynthetic machinery that assembles colibactin inside the bacterium, preventing the toxin from ever being made. Each of these three strategies-the claudin-4 decoy for BFT, the ClbS-displaying bacteria for colibactin, and the small-molecule biosynthesis blocker-attacks the problem at a different step. No published study has yet tested them in combination or compared their efficacy head-to-head in the same animal model, leaving open basic questions about optimal dosing, timing, and potential synergy.
That gap matters because the gut is not a single-toxin environment. A person carrying toxigenic B. fragilis may also harbor colibactin-producing E. coli, along with many other microbes that modulate inflammation and epithelial repair. If sequestering BFT with a decoy alters the inflammatory state of the colon lining, it could change how vulnerable cells are to colibactin-induced DNA crosslinks, or vice versa. Co-colonization experiments in mice, measuring both cell-junction damage and DNA lesion counts simultaneously, would test whether these toxin pathways interact or remain largely independent. As of the current reports, no group has described such a combined study.
Gaps Between Mouse Models and Human Colon Cancer Risk
All three intervention strategies have been validated only in mouse models so far. The claudin-4 decoy reduced tumor-related outcomes in mice repeatedly exposed to BFT, and the ClbS-displaying engineered bacteria cut genotoxic markers in a similar setting for colibactin. But mouse colons differ from human colons in microbiome composition, immune response, and tissue architecture. The leap from blocking a toxin in a controlled animal experiment to preventing cancer in a person who carries the bacterium for decades is large and untested.
No human epidemiologic data directly link the BFT–claudin-4 interaction to specific patient tumors at this point. Observational studies have associated toxigenic B. fragilis with colorectal neoplasia, but they typically measure bacterial presence rather than toxin load, receptor expression, or downstream molecular fingerprints in tumors. Likewise, colibactin-producing E. coli have been tied to characteristic DNA damage patterns in human cancers, yet prospective data showing that neutralizing the toxin lowers cancer incidence are lacking.
Translating these mouse-based tools into prevention strategies for people will require several intermediate steps. First, researchers will need reliable biomarkers: assays that detect BFT and colibactin activity in stool, blood, or colon biopsies, and that can be followed over time. Second, they will have to define which patients are at sufficiently high risk to justify interventions, such as those with recurrent polyps, inflammatory bowel disease, or strong family histories of colorectal cancer who also carry the relevant toxigenic strains.
Safety questions loom as well. Claudin-4 is not just a passive docking station for BFT; it is part of the tight junction machinery that keeps the gut barrier intact. Any long-term therapy based on claudin-4 fragments or mimics must avoid interfering with normal junction function. Engineered bacteria that express ClbS or other protective proteins raise separate concerns about horizontal gene transfer, unintended immune responses, and ecological effects on the existing microbiome. Small-molecule inhibitors of colibactin biosynthesis must be specific enough not to disrupt other beneficial bacterial pathways.
Regulatory pathways for such interventions are also uncertain. A soluble claudin-4 decoy might be regulated like a biologic drug, while ClbS-expressing microbes would fall into the category of live biotherapeutic products. Demonstrating efficacy would likely require long, expensive trials with cancer or advanced precancerous lesions as endpoints, or carefully validated surrogate markers such as mutational signatures and high-grade dysplasia rates.
Despite these hurdles, the identification of claudin-4 as the BFT receptor and the parallel advances against colibactin mark an important conceptual shift. Rather than treating the microbiome as an undifferentiated target to be wiped out with antibiotics or broadly reshaped with diet, researchers are beginning to dissect specific molecular interactions that tip the balance toward cancer. If future work can map which combinations of toxins, hosts, and microbial communities pose the highest danger, precision toxin-blocking therapies could eventually join colonoscopy, lifestyle changes, and emerging chemopreventive drugs in the toolkit for reducing colorectal cancer risk.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
