The Frog's Pharmacy

A Frog, a Bacterium, and 400 Million Views

In December 2025, a research team at the Japan Advanced Institute of Science and Technology published a paper in Gut Microbes with findings that sounded almost too cinematic to be real: a bacterium isolated from the intestines of a Japanese tree frog had eliminated 100% of cancerous tumors in mice after a single intravenous injection.¹ When the researchers re-introduced cancer into the cured animals, the tumors couldn't grow back.

The paper would have been a quietly impressive piece of preclinical oncology research. Instead, it became a social media phenomenon. The headline traveled faster than most actual drugs: "FROG GUT MICROBE ELIMINATES 100% OF CANCEROUS TUMORS AFTER SINGLE DOSE IN MICE." One viral post from epidemiologist Nicolas Hulscher racked up millions of views. Comments ranged from the hopeful ("This is the cure") to the conspiratorial ("Big Pharma will bury this").²

I understand the excitement. A single dose. Complete elimination. No detectable toxicity. Those are the words every cancer patient wants to hear. But I have been covering medicine long enough to know that mice are not people, single studies are not proof, and the bacterium in question has a history that the viral posts conveniently left out.

Bacteria vs. Cancer Is a 130-Year-Old Story

Using bacteria to fight cancer is not new. It is, in fact, one of the oldest ideas in oncology. In 1891, a New York bone surgeon named William B. Coley injected live Streptococcus bacteria directly into a patient's inoperable tumor. The patient developed a raging infection. The tumor dissolved. Coley spent the next 40 years refining his approach, eventually switching to heat-killed bacterial mixtures known as "Coley's toxins," and treated over 1,000 patients with various cancers.³

The idea was elegant and, as we now understand, scientifically sound: the bacterial invasion triggered the patient's immune system, and that immune response attacked the tumor along with the bacteria. Coley's work is now considered the foundation of modern cancer immunotherapy. The Cancer Research Institute, founded by his daughter in 1953, exists because of it.⁴

But here is the part that matters for evaluating the Ewingella americana study: despite 130 years of work, no live bacterial therapy has become a standard cancer treatment. BCG, a weakened tuberculosis bacterium used to treat bladder cancer, is the closest success story, and it was approved in 1990 for a very specific, non-metastatic indication.⁵ The National Institutes of Health lists over 50 completed clinical trials of live bacteria for cancer treatment, using species like Salmonella, Listeria, and Clostridium. None has reached routine clinical use.⁶

This is not because the science is wrong. Bacteria genuinely can target tumors. Solid tumors contain hypoxic, low-oxygen regions that are essentially invisible to the immune system but hospitable to certain anaerobic and facultatively anaerobic bacteria. The bacteria colonize the tumor, secrete toxins, and trigger immune infiltration. The problem is translation: what works in a genetically uniform mouse with a single implanted tumor does not reliably work in a human with a complex, heterogeneous, immune-adapted cancer.⁷

45 Strains, 9 Candidates, 1 Standout

The study, led by Professor Eijiro Miyako at JAIST, took an unusual approach to the sourcing problem. Rather than engineering a known pathogen like Salmonella, the team screened gut bacteria from amphibians and reptiles, organisms whose immune systems operate under very different evolutionary pressures than mammals. They isolated 45 bacterial strains from the intestines of species including the Japanese tree frog (Dryophytes japonicus), geckos, and turtles.¹

Nine of those 45 strains showed some antitumor activity when tested against cancer cells in vitro. But one stood dramatically apart: Ewingella americana, a gram-negative, facultatively anaerobic bacterium found in the gut of the Japanese tree frog. Where other strains showed modest cytotoxic effects, E. americana displayed potent, selective cancer cell killing in culture.¹

The name may sound obscure, but Ewingella americana is not unknown to medicine. It was first identified in 1983 and named after the clinical bacteriologist William H. Ewing. It belongs to the Enterobacteriaceae family, the same family as E. coli, Salmonella, and Klebsiella. It has been found in a variety of environments, including mushrooms, water, and, as it turns out, the digestive tracts of Japanese tree frogs.⁸

The proposed mechanism is a dual-action system. First, E. americana selectively accumulates in the hypoxic tumor microenvironment, a behavior common to many facultative anaerobes. The tumor's low-oxygen core provides a growth advantage that healthy tissues do not. Within 24 hours of intravenous injection, bacterial counts inside the tumors increased approximately 3,000-fold, while the bacteria were rapidly cleared from normal organs.¹

Second, the bacterial colonization triggers a robust immune response. The tumor becomes flooded with immune cells, particularly neutrophils, T cells, and B cells, along with a cascade of inflammatory signaling molecules. The combination of direct bacterial cytotoxicity and immune-mediated attack appears to produce the observed tumor destruction.¹

One Mouse Model, One Study, Zero Humans

Let me be specific about what makes these results look impressive and what makes them less impressive than the headlines suggest.

What looks impressive: The 100% complete response rate is genuinely unusual, even in mouse studies. Most bacterial cancer therapies achieve partial responses or work in a fraction of treated animals. The comparison to anti-PD-L1, a class of drug that has transformed human oncology, and to doxorubicin, a foundational chemotherapy agent, provides a meaningful benchmark. The rechallenge experiment, where cured mice resisted re-implantation of the same cancer, suggests that E. americana induced lasting immunological memory, not just a temporary clearance.¹

What looks less impressive: The CT26 model is the most commonly used syngeneic colorectal cancer model in immunotherapy research, and it is notoriously immunogenic. This means the mouse immune system already has a relatively easy time recognizing CT26 cells as foreign. Many therapies that produce spectacular results in CT26 fail when tested against less immunogenic models like B16 melanoma or 4T1 breast cancer.⁹ The study tested a single tumor type, in a single mouse strain, with a single dosing schedule. That is the absolute minimum for a first-in-concept demonstration.

The safety data in mice is clean. But the safety question in humans is not theoretical, and this brings us to the part of the story that the viral posts uniformly ignored.

The Bacterium That Causes Sepsis in Cancer Patients

This is the detail that changes the conversation. Ewingella americana is not just a benign frog gut commensal. It is a recognized, if rare, human pathogen. A 2024 narrative review in PMC cataloguing all known human E. americana infections found cases of bacteremia, peritonitis, respiratory tract infections, and urinary tract infections.¹⁰ Most occurred in immunocompromised patients. Several occurred in patients who were immunocompromised because they had cancer and were undergoing chemotherapy.

In 2025, a case report in Cureus documented a cancer patient who developed E. americana sepsis during a red blood cell transfusion.¹¹ A 2020 report in Frontiers in Pediatrics described E. americana meningitis in a term newborn, demonstrating that the organism can cause serious infection even outside the immunocompromised population.¹² A case study in the Indian Journal of Health Sciences described a multidrug-resistant E. americana strain causing catheter-associated urinary tract infection.¹³

The irony is pointed: the patient population most likely to receive an experimental cancer therapy, people with advanced malignancies whose immune systems are already compromised by disease and treatment, is the exact population in which E. americana has demonstrated the ability to cause life-threatening infection.

None of this means E. americana cannot eventually be developed as a cancer therapy. It means that the safety barrier is not trivial. Any clinical development program would need to address whether a bacterium with documented pathogenic potential can be administered intravenously to immunocompromised cancer patients without unacceptable infectious risk. That is a question that will take years of careful clinical work to answer.

The Long History of Bacteria That Cured Cancer in Mice

The bacterial cancer therapy field is littered with spectacular preclinical results that collapsed on contact with human biology. Understanding this history is essential for evaluating the Ewingella americana data.

Salmonella typhimurium VNP20009 was perhaps the most prominent example. An attenuated Salmonella strain engineered for tumor targeting, it produced dramatic tumor regression in multiple mouse models throughout the late 1990s and early 2000s. A Phase 1 clinical trial in 24 patients with metastatic melanoma and renal cell carcinoma, published in 2002, found that the bacteria colonized tumors in only 3 of 24 patients, and no objective tumor responses were observed. The maximum tolerated dose in humans was far below the effective dose in mice.¹⁴

Clostridium novyi-NT, developed at Johns Hopkins, showed potent antitumor activity in rodent and canine models. A Phase 1 human trial initiated in 2014 treated a single patient with leiomyosarcoma, who showed tumor response but also developed significant toxicity including sepsis-like symptoms. The program has proceeded slowly, with no Phase 2 results published to date.¹⁵

More recently, a 2025 study in Nature Biomedical Engineering described a tumor-resident bacterial consortium of Proteus mirabilis and Rhodopseudomonas palustris that achieved complete tumor remission in mice via selective intratumoral thrombosis. The results were striking. But the gap between mouse data and human therapy remains, as always, vast.¹⁶

The pattern is consistent: elegant mechanism, spectacular mouse data, then a collision with the complexity of human tumors, human immune systems, and human pharmacology. This does not mean bacterial cancer therapy will never work. It means that spectacular preclinical results are the beginning of the conversation, not the end.

Interesting Biology, Premature Headlines