Diagnosing a flare of inflammatory bowel disease usually means a colonoscopy or an endoscopy. Both work, but they are expensive, uncomfortable, and not the sort of thing anyone wants repeated every few weeks to see whether a treatment is holding. So a team at Caltech and Rice tried a different approach. They took a probiotic microbe that people already swallow, taught it to notice the chemical signature of inflammation, and gave it a way to announce what it found. The announcement comes through as an ultrasound image.
The work, published in Nature Communications in August 2025 by Marjorie Buss and colleagues, rests on a strain called E. coli Nissle. It has been used as a gut probiotic for a century and is comfortable living in the intestine for a while before washing out. That persistence is exactly what you want in a living sensor. The trick was getting a signal out of the body without cutting anything open.
Why light does not work down there
Bacteria have been engineered before to report on things happening in the gut. The usual reporters glow, either through fluorescence or bioluminescence, or they change color in a way you can spot after plating a stool sample. None of that is much good for a real-time readout. Light barely travels through tissue, so a bacterium glowing a few centimeters deep is invisible from outside. Plating feces tells you something happened, but not where and not quite when.
Ultrasound has the opposite problem set. It passes through the body cheaply and reaches deep, with resolution below a millimeter, and the machines already sit in most clinics. The catch is that bacteria do not naturally show up on it. To fix that, the researchers borrowed a structure from aquatic microbes called a gas vesicle. These are tiny protein shells filled with air. Because air scatters sound so differently from surrounding tissue, a cell packed with gas vesicles stands out as bright contrast on an ultrasound scan.
A switch wired to inflammation
The engineering had two halves that had to work together. First, the bacteria needed to sense the right thing. The team tuned genetic circuits to respond to thiosulfate and tetrathionate, two sulfur compounds that build up in an inflamed gut. When those markers are present, the circuit flips on. Second, flipping on had to mean something visible, so the sensor was connected to the genes that build gas vesicles. Detect inflammation, make gas vesicles, scatter sound.
Getting both halves to behave took a lot of tuning. A circuit that fires too easily gives false positives; one that fires weakly produces contrast too faint to see. The researchers reworked the signaling to be sensitive to the biomarkers while still producing strong, stable ultrasound contrast. They also modified the strain so it would colonize the gut only transiently, which matters for a diagnostic you might not want lingering indefinitely.
Then they tested it in mice. Using antibiotics to provoke gut inflammation, they fed the animals the engineered bacteria and imaged them. The inflamed guts lit up on ultrasound in a way healthy ones did not, without any surgery or injection. A living cell had reported on a disease state from deep inside the body, and a standard imaging machine picked it up.
What the study can't say yet
This is a mouse result, and mouse guts are not human guts. Antibiotic-induced inflammation is also a clean, controllable model, not the messy chronic disease that patients live with. Whether the same circuits fire reliably in a human intestine full of competing microbes, varied diets, and real Crohn's or colitis is an open question the paper does not answer.
There are practical hurdles too. The bacteria are genetically modified organisms meant to be swallowed, which brings a regulatory and safety review that glowing lab strains never had to face. The team built in transient colonization partly for that reason, but containment and long-term behavior will need careful study. And ultrasound contrast, while useful, is a coarser signal than a lab assay; reading it consistently across patients and operators is its own problem.
Still, the core idea travels well. The same platform could in principle be pointed at other gut biomarkers by swapping the sensing module, turning one probiotic chassis into a family of noninvasive tests. If it holds up, monitoring bowel disease might one day look less like a scope in a procedure room and more like a quick scan with a machine that is already down the hall.
Comments