A team at Johns Hopkins University has solved a 17-year-old puzzle about how a common gut bacterium drives colon cancer, identifying the specific receptor the bacterial toxin uses to invade cells. The discovery, published in Nature in April 2026, reveals that the toxin from Bacteroides fragilis must first bind to a protein called claudin-4 before it can damage the colon's protective lining and trigger inflammation that leads to tumors.

Since a landmark 2009 study, researchers knew that Bacteroides fragilis, found in up to 20% of healthy individuals, secretes a toxin called BFT that damages the colon and can lead to colorectal cancer. But the exact mechanism remained elusive for nearly two decades. The new research not only identifies how the toxin works, but also demonstrates a potential strategy to block it using a molecular decoy that successfully protected mice from BFT-induced damage.

What Makes This Discovery Different From Previous Research?

The breakthrough came through an innovative approach using CRISPR gene-editing technology. Maxwell White, an MD/PhD candidate in the lab of senior researcher Cynthia Sears, led a genome-wide CRISPR screen in collaboration with Harvard Medical School researchers. By systematically knocking out genes in colon epithelial cells, the team identified claudin-4 as the critical link between the toxin and cell damage.

What surprised the researchers was the identity of the receptor itself. Scientists had long expected the receptor to be a signaling protein, such as a G-coupled protein receptor, but claudin-4 is neither. In fact, the team could not identify any other toxin in the scientific literature that functions this way, with most proteases going straight to their targets rather than binding a separate receptor first.

"We've made several attempts over time to identify the receptor, so this is an exciting moment. Understanding how bacterial toxins work can open doors to new approaches for detection and therapy for associated diseases, including diarrhea, colorectal cancer and bloodstream infections," said Cynthia Sears, Bloomberg-Kimmel Professor of Cancer Immunotherapy and professor of medicine at Johns Hopkins.

Cynthia Sears, Bloomberg-Kimmel Professor of Cancer Immunotherapy at Johns Hopkins University

The researchers confirmed their findings through multiple experimental approaches. Structural biologists at the Molecular Biology Institute of Barcelona used biophysical analysis to demonstrate that BFT and claudin-4 form a tight, one-to-one complex in a test tube, providing the first physical evidence of the binding interaction.

How Does This Lead to New Cancer Prevention Strategies?

The research team moved beyond laboratory findings to test their approach in living systems using mouse models. They created a decoy version of claudin-4, a soluble protein that displays claudin-4 sequences but cannot be damaged by the toxin. When introduced, the decoy successfully competed with the actual receptor on colon cells, preventing BFT from binding and causing damage.

This decoy strategy represents a promising foundation for future therapies. The team is now exploring which molecular approaches might be most successful to block the toxin, with potential options including small molecules or other biologics that could have better pharmacological properties for human use.

• The Decoy Strategy: A soluble protein displaying claudin-4 sequences acts as a molecular trap, preventing the bacterial toxin from binding to actual colon cells and causing damage.
• Small Molecule Development: Researchers are exploring synthetic compounds that could block the claudin-4 receptor more effectively than protein-based decoys for clinical applications.
• Biologic Therapies: Alternative biological approaches are being investigated to provide better pharmacological properties and potentially longer-lasting protection against BFT toxin.

The discovery also highlights the limitations of current artificial intelligence tools in structural biology. While AI modeling tools such as AlphaFold have revolutionized protein structure prediction, they were not able to fully resolve the exact experimental structure of the interaction between BFT and claudin-4. This suggests that even as AI advances, some biological puzzles still require traditional experimental validation and structural biology approaches.

"It took a while to get the assay working and validate the approach, but once we were able to do the screen, claudin-4 was a clear, resounding top hit. That was an exciting moment," explained Maxwell White, MD/PhD candidate in the Sears lab.

Maxwell White, MD/PhD Candidate, Johns Hopkins University

The research was supported in part by the National Institutes of Health and involved collaboration between multiple institutions, including Harvard Medical School and the Molecular Biology Institute of Barcelona. This multi-institutional approach allowed the team to combine expertise in genomics, structural biology, and in vivo disease modeling to solve a problem that had resisted solution for nearly two decades.

For patients at risk of colorectal cancer, this discovery offers hope for new preventive approaches. Colorectal cancer remains a leading cause of cancer-related deaths worldwide, and understanding the bacterial mechanisms that drive tumor formation could lead to targeted interventions that reduce risk without requiring broad-spectrum antibiotics that might disrupt the beneficial microbiome.