Updated November 1, 2026: I rewrote this post to fix the Synlogic story, add two approved microbiome drugs and a 2026 colitis trial, and tighten the writing. The engineered bacteria here are research tools, not approved treatments.
Every drug you have ever swallowed was dead before it reached you
A pill dissolves. A molecule drifts through your bloodstream. It reaches its target, or it does not. After a few hours it is gone. That is how almost every drug works, and it is the opposite of a living therapeutic. You put a chemical in. A fixed thing happens. Then it ends.
A different kind of medicine is being built, and I find it one of the most exciting ideas in biology. These therapies are alive. Living therapeutics are bacteria engineered to read your body’s chemistry and act on what they find. When they sense inflammation, they make an anti-inflammatory molecule. When they reach the low-oxygen core of a tumor, they multiply and release a drug.
This is no longer only theory. Engineered bacteria have reached human trials. Researchers have shrunk tumors in mice and detected disease signals inside the body with real precision. The field is young and the regulatory path is unfinished. Even so, the core idea holds. You can program a microbe like software and send it into a patient to do a job.
What makes a bacterium a living therapeutic
A conventional drug floods the body with a molecule. The liver breaks it down, the kidneys flush it out, and side effects turn up far from the target. Bacteria work differently. They already live inside us, and your gut holds trillions of them. They sense their surroundings, respond to signals and carry out tasks with precision.
The idea behind living therapeutics is to redirect those abilities. Instead of a bacterium’s own programming, researchers write new instructions and install them as genetic circuits. The logic is simple: if this signal appears, make this response. I covered the microbial world these strains join in my post on the microbiome.
Most work starts with a safe, well-studied host. The favorite is Escherichia coli Nissle 1917, a probiotic people have taken for over a century. It is easy to engineer, its genome is mapped, and it settles into the gut. Researchers also use Lactococcus lactis and Lactobacillus species for gut work. For cancer, they turn to strains of Salmonella and Clostridium that seek out tumors on their own.
The tools are standard synthetic biology. CRISPR genome editing, circuits built from well-characterized parts, and promoters that switch genes on only under set conditions. The result is a bacterium carrying new medical instructions in its DNA, alongside everything it already knows how to do.
Living therapeutics for gut disease
The gut is where this field has gone furthest. Three reasons explain why. You can reach it by mouth, the simplest route in medicine. It already holds a dense bacterial community, so a new strain is not out of place. And inflammatory bowel disease is exactly the kind of problem these bacteria suit.
IBD affects more than six million people worldwide. Standard biologics like adalimumab, sold as Humira, suppress inflammation across the whole body. The trouble is that the disease sits in the gut while the drug circulates everywhere to reach it. That mismatch is what engineered strains aim to fix.
Researchers have built E. coli Nissle strains that detect the chemical marks of inflammation, such as nitric oxide. Where those signals appear, the bacteria switch on and make an anti-inflammatory molecule. Inflamed tissue gets the drug. Healthy tissue gets nothing.
A bacterium that senses bleeding and seals the wound
A 2025 study pushed this further. A team engineered E. coli with a blood-sensing circuit. It responded to bleeding at 100 parts per million, with a signal more than 100-fold above background. Once it sensed blood, the bacterium made two things. An adhesive built from barnacle cement proteins, and a protein that helps the gut lining repair itself.
The bacteria glued themselves to the wound and began patching it from the inside. In mouse models of colitis, the system cut bleeding, eased inflammation and strengthened the intestinal barrier. No single conventional drug does all of that at once. When I read it, it looked like a real turn in what these strains can do.
A separate group took aim at the same target as Humira. They engineered E. coli Nissle with a nitric oxide sensor tied to a system that releases anti-TNF nanobodies, small antibody fragments. The bacteria made the fragments only where the inflammation signal was present. Same target as the blockbuster drug, delivered to the spot that needs it.
Bacteria that hunt tumors
Cancer is harder, and more dramatic. Solid tumors build a strange microenvironment: low oxygen, low pH, high lactate, and an immune system held at bay. Those conditions repel normal tissue and most drugs. Some bacteria find them inviting, because they evolved to live in exactly such low-oxygen, nutrient-rich pockets. That quirk makes bacteria a logical weapon against tumors.
The idea is old. In the 1890s, the surgeon William Coley saw that some cancer patients who caught bacterial infections had their tumors shrink. He began injecting bacteria into tumors and recorded remissions his colleagues could not explain. Chemotherapy and radiation later pushed the approach aside. His core observation held up. Bacteria can home to tumors and stir an immune response against cancer.
Clostridium, a bacterium that hunts tumors
Clostridium sporogenes cannot survive where there is oxygen. That makes it a natural fit for cancer. Tumors carry large oxygen-starved cores, and Clostridium grows there while dying off in healthy, oxygenated tissue. The targeting is built into the organism.
Researchers at the University of Waterloo have engineered C. sporogenes for solid tumors. They faced two problems. The bacterium’s oxygen sensitivity meant it reached the core but died at the edges. The team borrowed a gene from a hardier relative to push it further in.
The second problem was control. They needed the killing function to switch on only inside the tumor, never in healthy tissue. Their answer was quorum sensing, the system bacteria use to take a headcount. Each cell releases a small signal molecule. When enough cells gather and the signal crosses a threshold, the whole population acts together.
The Waterloo team wired a quorum-sensing circuit so the therapy turns on only once the bacteria reach high density inside the tumor. The bacteria count themselves before they act. The work appeared in ACS Synthetic Biology in late 2025, with a fluorescent marker confirming the circuit fired right at the threshold.
E. coli against colorectal cancer
For tumors in the gut, E. coli Nissle offers another route. A 2024 study engineered it to find and treat colorectal cancer, releasing a tumor-killing payload once inside. A companion design added a self-destruct circuit. When the population hit a set density, the cells burst in unison and dumped their cargo at once.
A 2025 study from Tianjin University went further. Instead of one strain, the team built a consortium of engineered strains working together. It read three tumor signals at once: pH, low oxygen and high lactate. Depending on what it sensed, it released different payloads. One drained lactate to starve the tumor’s metabolism. Another freed a nanobody that blocks PD-L1, the shield tumors use to hide from the immune system. In mice, the system roused T cells and slowed tumor growth. That is close to biological decision-making. I would compare it with the mRNA cancer vaccines that aim at the same goal by a different road.
Building a kill switch
A bacterium that multiplies inside you is powerful, and it raises hard safety questions. What if it spreads past the target? What if the circuit misfires? What if the organism mutates its way out of control? These are real, and the field is working on them in the open.
The main tool is the kill switch. It forces the bacteria to die outside the place they belong. Two influential designs came from James Collins’s lab at MIT, named Deadman and Passcode. The Deadman switch needs a constant survival signal. Leave the target environment, lose the signal, and the cell flips to a lethal state. The Passcode switch demands several signals at once, like a combination lock. No single stray cue keeps the bacteria alive in the wrong place.
More recent designs point CRISPR at the problem. The circuit senses the absence of a specific molecule, then cuts the bacterium’s own genome to pieces. Because it attacks the DNA directly, it is harder to evolve around. I covered this gene-editing toolkit in artificial metabolism, a sister post on reprogramming the chemistry of cells.
Evolution is still the hardest part. A kill switch costs the bacteria fitness, so selection keeps favoring mutants that break it. Researchers answer with redundancy, several independent switches that must all fail at once. They also tie the therapeutic genes to the switch, so you cannot lose one without the other.
There is a second layer too. Some strains are engineered to need an artificial amino acid that exists nowhere in nature. Stop the supply and the bacteria cannot build their proteins, so they die within days. Clinical-grade strains now stack auxotrophy, kill switches and non-colonizing designs into several barriers at once.
From the lab to the clinic
Most living therapeutics are still in animals or early development. Two microbiome drugs have crossed the line, though neither is engineered. The FDA approved Rebyota in 2022 and Vowst in 2023, both made from donor stool to stop recurrent Clostridioides difficile infection. In Vowst’s trial, infection came back in 12% of treated patients against 40% on placebo. They prove regulators will approve a live bacterial product. They do not prove the engineered kind works in people.
For engineered strains, the clearest human test was phenylketonuria, or PKU. People with PKU cannot break down phenylalanine, an amino acid in most protein. It builds up and harms the brain, so patients follow one of the strictest diets in medicine.
Synlogic, a company founded on MIT research, engineered E. coli Nissle to eat phenylalanine in the gut before it reaches the blood. The drug, SYNB1934, carried enzymes that broke the amino acid down far faster than any natural strain. It won FDA Fast Track status, and Synlogic ran a global Phase 3 trial called Synpheny-3.
In February 2024 the company stopped the trial. An internal review found it unlikely to hit its main goal. Safety was not the issue. Synlogic then cut more than 90% of its staff. By early 2026 it had become a shell company, delisted from Nasdaq, with one employee and no research, hunting for a buyer.
I want to be plain about what that means. The earlier trials showed the approach was safe. The bacteria did not take up residence, and cleared from the gut within days of the last dose. What the failed Phase 3 did not show is that the treatment works. For now, safety is proven and benefit in patients is not.
The pipeline beyond PKU is still moving. In early 2026, Microbiotica reported that its oral drug MB-310 put 12 of 19 colitis patients into remission in an early trial. Other groups are chasing urea cycle disorders and chronic kidney disease, aiming the same gut-level trick at different toxins.
What has to go right
Several things are converging. The circuits are getting smarter, moving from single-input designs to logic gates and memory. The pairing of bacteria with immunotherapy is opening cancer work, turning cold tumors the immune system ignores into hot ones it attacks. And delivery is spreading past the gut toward injection straight into tumors and protective gels that carry bacteria to a target.
The FDA treats live biotherapeutic products as their own category, apart from drugs and gene therapies. It has issued early guidance on their safety. The framework is not finished. Synlogic’s trials, the failure included, handed regulators safety data and precedents to build on.
Why I keep watching this
A living drug is unsettling on its face. Medicine is meant to be controlled, and bacteria are not. They evolve, they respond, they multiply. Those are the very traits that make them useful, and the reason they need careful handling.
Every ordinary drug is inert. It does one thing, at one dose, for one stretch of time, whatever your body actually needs at that moment. A living therapeutic can sense whether inflammation is easing or getting worse and adjust as it goes. It can sit quiet in the gut until a flare starts, then switch on. That adaptive, local idea of medicine is more compelling to me than most of what the drug industry produces.
We are not there yet. The field is still working through safety, stability, manufacturing and the long gap between a mouse and a patient. The history of medicine is full of good ideas that died on contact with human biology. The Synlogic failure is a fresh reminder.
Even so, the early results are real. Engineered bacteria have cleared toxic metabolites in people and shrunk tumors in animals. They have found bleeding in the gut and sealed it, and made antibody fragments on cue. The living drug has stopped being a thought experiment. The open question is how far the programming can go, and whether the safety, manufacturing and regulation can keep up with it.
Your thoughts about living therapeutics
Will living therapeutics reach mainstream medicine this decade, or will manufacturing, safety and regulation hold them back? And does a replicating organism inside you give you pause, or does the precision make up for the strangeness? Tell me in the comments, especially if you work in synthetic biology or clinical development.
Bleeding Edge Biology recommends
Articles
- Developing a New Class of Engineered Live Bacterial Therapeutics to Treat Human Diseases
Mark R. Charbonneau, Vincent M. Isabella, Ning Li and Caroline B. Kurtz, Nature Communications, April 2020. The clearest published account of how the platform works end to end, written by the Synlogic team, from circuit design to manufacturing. - Engineered Bacteria for Disease Diagnosis and Treatment Using Synthetic Biology
Kai Jin, Yi Huang, Hailong Che and Yihan Wu, Microbial Biotechnology, January 2025. A current survey of biosensor design, microbial consortia and the cancer and gut-disease work. Strong on multi-strain systems. - “Deadman” and “Passcode” Microbial Kill Switches for Bacterial Containment
Clement T.Y. Chan, Joe H. Lee, Declan E. Taylor, Christopher Voigt and James J. Collins, Nature Chemical Biology, 2016. The paper that introduced the two kill-switch designs in this post, and it is candid about why they are hard to keep. - Construction and Functional Characterization of a Heterologous Quorum Sensing Circuit in Clostridium sporogenes
Sara Sadr, Bahram Zargar, Marc G. Aucoin and Brian Ingalls, ACS Synthetic Biology, December 2025. The Waterloo paper behind the tumor-hunting Clostridium work.
Talks
- Programming Bacteria to Detect Cancer (and Maybe Treat It)
Tal Danino, TED Fellow talk, 2015. Danino, a bioengineer at Columbia, is the best short entry point on quorum sensing and bacteria against cancer. - Hacking Bacteria to Fight Cancer
Tal Danino, TED-Ed animated lesson, 2019. Five minutes, from Coley to today’s engineered strains, no biology background needed. - How We’re Using AI to Discover New Antibiotics
Jim Collins, TED, 2020. Collins co-founded Synlogic and built the kill switches above. The talk is about antibiotics, but the logic of reprogrammable bacteria runs through it.
Documentaries
- The Invisible Extinction
Directed by Sarah Schenck and Eugene Jarecki, 2022. Two Rutgers microbiologists track the damage antibiotics and processed food do to the human microbiome. Good background on why the bacterial world inside us is worth protecting. - Hack Your Health: The Secrets of Your Gut
Directed by Flemming Lytved Hag, Netflix, 2024. An accessible primer on gut microbiome science through four test subjects.
Websites and resources
- Collins Lab at MIT
One of the founding labs of the field, behind Synlogic, the Deadman switch and CRISPR diagnostics. The best place to watch where the work is heading. - iGEM Foundation
The world’s largest synthetic biology competition, with open-source parts used by teams who often build their own engineered strains. - Live Biotherapeutic Products, Guidance for Industry
The FDA’s definition of live biotherapeutic products and the data it wants. Read it to see why moving from mouse to patient takes so long. - Synlogic Therapeutics
Synlogic took engineered bacteria further into human trials than anyone. Its PKU program ended when the Phase 3 stopped in 2024, and the company has since wound down. Its papers remain a record of what the field tried.
