Quiet Breakout: The Tuberculosis Drug Named Muddy

In a compost heap behind a university in Durban, in 2010, an undergraduate named Lilli Holst peeled back the skin of a rotten eggplant and scraped some of the slime into a tube. The slime contained, among other things, a previously unknown virus. Holst was taking a teaching lab that asked students to find viruses that eat bacteria — bacteriophages — in whatever soil they could get their hands on. She named her find Muddy, packed it up, and shipped it to a freezer at the University of Pittsburgh that already held a few thousand other student samples.

On 31 March 2026, the United States Patent and Trademark Office issued patent number 12,589,127 to the University of Pittsburgh. Claim 20 of that patent is the legally protected formula for a pharmaceutical composition consisting of exactly five named bacteriophages, one of which is Muddy_HRMN0157-2 — a clinical-grade descendant of Lilli Holst’s eggplant. The other four have names like AdephagiaΔ41Δ43 and FionnbharthΔ45Δ47. The little Greek deltas mean the genes between those numbers have been deleted on purpose. The patent claims this cocktail as a treatment for the deadliest infectious disease in the world.

Tuberculosis killed 1.23 million people in 2024, according to the World Health Organization. That is more than any other infectious agent. Of the 10.7 million new cases that year, around 390,000 were resistant to the standard first-line drugs. The current cure for drug-sensitive TB still takes six months of three to four antibiotics; the cure for the resistant version takes longer, costs more, and fails more often. The pharmaceutical industry’s pipeline against the disease is thin. No truly new mechanism of action has been approved for TB in a generation other than the diarylquinoline bedaquiline.

Pittsburgh’s patent claims a different mechanism altogether. Instead of a small molecule that jams an enzyme, it claims a living virus, or rather five of them, that infect and kill Mycobacterium tuberculosis by hijacking it. The cocktail can be administered intravenously or, per claim 11 of the earlier sibling patent issued two weeks before, as an aerosol. A TB patient could inhale a swarm of trained viruses directly into the lung where the bacteria live.

What is genuinely strange about this is where the active ingredients came from.

The freezer that ate the pharmacopoeia

Since 2008, the Howard Hughes Medical Institute and the University of Pittsburgh have run an undergraduate program called SEA-PHAGES, in which first-year biology students dig up dirt — from gardens, from compost heaps, from the cracks in their dorm sidewalks — and use that dirt to hunt for bacteriophages. The hunters are required to name their phages whatever they like. According to HHMI, more than 50,000 students at roughly 175 institutions have now participated. Graham Hatfull, who runs the program from his lab at Pitt, told the trade journal Drug Discovery World in 2025 that the program has accumulated over 50,000 student researchers and thousands of named, sequenced phages.

For most of that program’s history, the phages were academic curiosities. Mycobacteriophage genome papers got modest citation counts. The biggest commercial application of mycobacteriophages was a diagnostic — phages that light up when they detect live, drug-resistant TB in a sputum sample.

Then in 2018 a teenager named Isabelle Carnell-Holdaway, recovering from a double lung transplant at Great Ormond Street Hospital in London, came down with a runaway Mycobacterium abscessus infection — TB’s nontuberculous cousin. She was being moved to palliative care when her doctors emailed Hatfull on the off chance he had a phage for the strain. Hatfull’s lab had one. It was Lilli Holst’s. They engineered it into a more aggressive killer, paired it with two others, and infused a billion phages twice daily into the girl’s bloodstream and dabbed more on her skin lesions. Within months she was eating again, baking, posting on social media. STAT News and NPR ran the story. The original case was published in Nature Medicine in May 2019.

The provisional patent for the TB cocktail was filed nine days later, on 21 May 2019.

What changed in March

A search of US grants going back three decades shows almost no patents specifically claiming phage cocktails for Mycobacterium tuberculosis. The earliest entries from the 1990s are for diagnostic phages, not therapeutics. A single TB-therapy patent issued to the same Pittsburgh group in mid-2024. Then nothing for eighteen months. Then, in March 2026, two more in fifteen days. The most recent, US 12,589,127, lists Graham Hatfull, Rebekah Marie Dedrick and Carlos Andrés Guerrero as inventors, and the University of Pittsburgh as the sole assignee. According to the patent itself, the underlying research was funded by NIH grant GM116884; the government holds march-in rights.

The patents are not theoretical. The infectious-disease community has now treated roughly 48 patients with engineered mycobacteriophages on a compassionate-use basis, Hatfull told Drug Discovery World. A 2023 paper in Clinical Infectious Diseases documented 20 of those cases. Most were against M. abscessus, not full TB. Of the 20, 11 had favourable or partially favourable responses, 5 were inconclusive, and 4 did not respond. Eight patients developed neutralising antibodies against at least one of the phages, although in some of them the therapy still worked. There are now active phage trials at Vancouver Coastal Health and another Phase 1 starting July 2026, both still focused on M. abscessus. None yet on M. tuberculosis itself.

That is what makes the new patents interesting: they push past the cousin disease. The cocktail in claim 20 was specifically assembled to kill diverse clinical strains of M. tuberculosis itself, according to the Hatfull lab’s 2021 mBio paper. A patent does not equal a drug. It is, however, the way American universities tell the market they are ready to license. With government rights attached and an NIH grant number in the file, the next step is some combination of a non-profit licensee, a small TB-focused biotech, and a Phase 1 trial in a country with enough drug-resistant TB to enroll quickly.

Why this is worth watching

The economics here are upside-down compared with normal pharma. Big pharmaceutical companies have abandoned the TB pipeline because the patients live in low-income countries that cannot pay $80,000 a year for a new drug. The active ingredient in 12,589,127 was, in effect, discovered by undergraduates working for tuition credit. The R&D cost of the discovery phase was a federal teaching grant. The IP is owned by a non-profit university. The cocktail itself is a manufactured biological — closer in production to a vaccine than to a small molecule — meaning it could in principle be made cheaply in a fermentation tank by a non-profit manufacturer. If it works, it routes around the part of the pharmaceutical industry that has refused to make new TB drugs.

For an R&D director or a VC, the watchable signal is whether Pittsburgh now spins out a phage-therapy company specifically for mycobacterial disease, the way Yale spun out Felix Biotechnology for general phage work and the way Adaptive Phage Therapeutics emerged from Patrick Secor’s lab. The Pittsburgh portfolio — three TB therapy patents since June 2024, an NTM patent issued in November 2025, and now the cocktail of record — is large enough to anchor a company. The freezer is full enough to feed it.

For a journalist or a clinician, the watchable signal is the first compassionate-use TB case. Hatfull has done it 48 times for cousin diseases. Once he does it once for full pulmonary TB, the question is no longer whether the cocktail kills the bacterium — it is whether the patient’s immune system lets it. That is the antibody-neutralisation question raised in the 2023 paper, and the next year of trial data should answer it.

The phages are ready. The patent is granted. Muddy is in the freezer, sixteen years out of a Durban compost heap and a year of legal drafting away from being the active ingredient in something that an industrial fermenter in Pittsburgh, or Mumbai, or Cape Town can grow at scale to kill the single bacterium that has killed more humans than any other pathogen in history.

Method note. Patent counts and titles came from the USPTO weekly grant feed indexed by Signalnet; we read the full claims and description of US 12,576,118 and 12,589,127 directly. The five-phage cocktail composition is from claim 20 of US 12,589,127; the route-of-administration claim is from claim 11 of US 12,576,118. Inventor and assignee details were cross-checked against the public USPTO record via Google Patents. Tuberculosis epidemiology numbers are the WHO Global TB Report 2024. The 20-patient compassionate-use outcomes are from Dedrick et al., Clinical Infectious Diseases 76(1), January 2023. Trial details for NCT07228702 came from ClinicalTrials.gov. The Lilli Holst / Muddy phage origin story has been reported in STAT News, NPR, Nature Medicine (May 2019), and Pitt Magazine; we relied on those sources for biographical detail. SEA-PHAGES participation numbers are from HHMI’s program page and Hatfull’s 2025 Drug Discovery World interview. The Signalnet Research Bot wrote this brief; a human reviewed it before publication.