Alexis Cheshire

Dirt Beat Cancer! A consortium of two bacteria isolated from tumors eliminates entire tumors in mice bearing human colorectal, ovarian, and pancreatic cancers, and does so in animals that lack functional T cells, B cells, or both. The organisms are Proteus mirabilis found in most common dirt, labeled A-gyo by the investigators, and the photosynthetic bacterium Rhodopseudomonas palustris, labeled UN-gyo. Injected intravenously in a starting ratio of roughly 3:97, they colonize hypoxic tumor tissue. Inside the tumor the ratio inverts to approximately 99:1 as A-gyo expands. Within 24 hours the treated tumors darken from selective thrombosis and vascular collapse driven by local cytokine release, fibrin deposition, and platelet aggregation. The same bacteria proliferate, form biofilms, and secrete cytolysins. A-gyo also elongates from short rods into long filaments when exposed to tumor metabolites such as fumarate and lactate, increasing its ability to penetrate and lyse cancer cells. Iron is depleted from the local environment through upregulated bacterial siderophore systems. These events produce widespread necrosis. The paper records 100 percent complete responses after a two-dose intravenous schedule in BALB/c-nu/nu, SCID, and NOD-SCID mice carrying Colon26 tumors, and complete disappearance of human HT29, SKOV3, and BxPC3 xenografts in nude mice. In an orthotopic pancreatic model, pancreas weight and the marker CA19-9 returned to levels measured in non-tumor-bearing animals. The work was led by Eijiro Miyako at the Japan Advanced Institute of Science and Technology, with collaborators from Daiichi Sankyo and the University of Tsukuba. The principal report appeared in Nature Biomedical Engineering in August 2. An earlier 2023 paper in Advanced Science had already shown that the same consortium, recovered from colon-cancer tissue, produced complete regressions and long-term survival in immunocompetent syngeneic models of colorectal cancer, sarcoma, metastatic lung cancer, and drug-resistant triple-negative breast cancer, accompanied by immune-cell infiltration. The 2025 study demonstrates that the same outcome occurs when adaptive immunity is absent, indicating that the dominant mechanism in the later experiments is direct vascular disruption and oncolysis rather than immune activation. Safety observations in the 2025 paper include the absence of cytokine-release syndrome and the absence of systemic organ toxicity at the double-dose regimen that produced complete responses. A single high dose above 7.8 × 10^9 CFU ml⁻¹ produced lethal cytokine release that could be reduced by dexamethasone or by splitting the dose. The bacteria remained sensitive to imipenem, which cleared them from treated animals. Body weight stayed within normal limits, and organ histology at 150 days showed no residual damage in the longest-surviving cohorts. The investigators note that certain artificial ratios outside the 3:97 window killed mice within 48 hours, so the composition of the inoculum is critical. Bacterial cancer therapy is not new. William Coley documented regressions after deliberate bacterial infections in the 1890s. Modern engineered strains of Salmonella and Listeria have entered clinical trials. The present consortium uses unmodified environmental and tumor-resident isolates whose mutual regulation appears to limit pathogenicity outside the tumor while amplifying it inside. Miyako has stated that the group aims to form a startup and open a Phase 1 trial within six years. Whether the same vascular-collapse mechanism operates in human tumors that differ in vessel architecture and metabolic profile is unknown. But very promising. Those limits are stated in the discussion of the Nature Biomedical Engineering paper and in subsequent interviews. The primary reports are available at nature.com/articles/s41551-0… and onlinelibrary.wiley.com/doi/….