A single Mycobacterium tuberculosis bacterium is small enough to ride an exhaled droplet all the way down into the deepest pocket of the lung, where a single layer of alveolar cells is all that separates air from blood.

What happens next is one of the oldest and least understood clashes in medicine. In some people, the immune system walls off the invader and the infection goes silent for decades. In others, the bacteria multiply, the tissue dies, and a cavity opens in the lung. More than a million people a year fall on the wrong side of that divide.

In a recent study published in Science Advances, researchers at The Francis Crick Institute in London, England, teaming up with Swiss biotech company AlveoliX, have developed a lung-on-chip model constructed entirely from a single human donor’s induced pluripotent stem cells (iPSCs). They call the device the iLoC — short for iPSC-derived lung-on-chip.

It houses four cell types: alveolar type I and type II epithelial cells, vascular endothelial cells, and macrophages. They are all genetically identical, arranged to recreate the architecture of the alveolar-capillary membrane, and seated in a microfluidic chip that rhythmically stretches to simulate breathing.

“From the moment a person breathes in the bacteria to when they manifest clinical symptoms, it takes months, if not years,” said Jakson Luk, PhD, the study’s lead author and a postdoctoral fellow at the Crick. “It’s a very long black box.”

Building a Lung From One Person’s Cells

Getting four cell types to cohabit on a microfluidic chip for weeks proved to be the central engineering challenge.

“You want to have the epithelial-endothelial connection, the barrier that mimics the blood and the alveolar space,” said Maximiliano Gutierrez, PhD, the study’s senior author and principal group leader at the Crick. Getting the right cells in the right proportions — the flat type I alveolar cells that exchange oxygen and carbon dioxide, along with scattered type II cells that secrete surfactant — was essential.

The team spent months searching for a single culture medium that would keep epithelial, endothelial, and immune cells alive simultaneously. Some cells sat on the device for 22 days. “The main challenge is finding the environment where all of those cells are going to be happy,” Gutierrez said, “and actually not only being happy but also reflecting the function that they have in a human body.”

The first time it worked — with all four cell types maintaining their identity together on the chip — the lab opened champagne.

Single-cell RNA sequencing of roughly 27,000 cells confirmed the effort had paid off: The chip’s cellular profiles mapped closely to those in the Human Lung Cell Atlas. But the analysis also revealed something unanticipated. Endothelial cells on the vascular side were reshaping macrophage behavior on the airway side — driving immune subpopulations toward phagocytosis and antigen presentation. Without endothelial cells, macrophages drifted into a quiet, transitional state. The lung, it appeared, relies on its blood vessels to arm its immune sentinels.

Then the machine began to breathe. Unlike systems that stretch cells in two dimensions, AlveoliX’s device uses negative pressure, mimicking the diaphragm’s pull, to create three-dimensional curvature. The chip’s data showed that this mechanical stretch substantially altered immune pathways in all three cell types.

But every infection experiment was conducted under static conditions. Luk said biosafety level 3 constraints forced the decision and that breathing infection experiments are now underway.

An Infection Unlike the Petri Dish

When the team infected the device with M tuberculosis strain H37Rv, the infection looked strikingly different from conventional cell culture, in which bacteria often overwhelm host cells within days. Here, both macrophages and epithelial cells took up the pathogen, but most were not permissive to replication, meaning the bacteria entered but did not multiply. Growth was slow.

“This is actually a step closer to what is actually happening in humans,” Luk said. “We are getting one step closer to what we want to model inside the lab.”

At certain sites, macrophages began to swarm into large clusters, with dead cells at the center, live macrophages pressing in from the edges, and bacteria replicating within the necrotic core. Luk compared it to a rugby scrum: “Where’s the ball? Everyone just goes there and tries to stop the wave.”

These structures bore an early resemblance to granulomas — not mature but heading in that direction.

One notable absence: The chip did not observe M tuberculosis cord formation, which is when virulent bacteria grow in rope-like structures. A 2023 lung-on-chip study showed cord formation drive immune evasion and antibiotic resistance. Luk framed this as a future variable worth exploring across different donor genetics.

An Outside View

Darrell N. Kotton, MD, founding director of the Center for Regenerative Medicine of Boston University and Boston Medical Center in Boston and a leading figure in iPSC lung research who was not involved in the study, called the work “an advance” and not a minor one.

“It’s extremely difficult to have four different cell types together,” Kotton said. “The lung might be made of 40 cell types. So we have a way to go. We all know that.” But getting from single cell-type cultures to four interacting lineages on a chip with physiologic breathing, then applying that to a disease model and extracting biological insights, represents real progress.

On the question of whether iPSC-derived cells truly behave like their counterparts in the body, Kotton argued that the criticism misses the point. “Even primary cells don’t look like primary cells once taken out of the body and put in these chips,” he said. “This is just the tricky business of doing lung disease modeling research.” From what he’s seen, iPSC-derived cells are “better than existing models, even those based on primary cells.”

The Promise in the Platform

Because every cell on the chip derives from a single iPSC line, future versions could be built from donors carrying mutations linked to tuberculosis (TB) susceptibility so researchers can figure out why one person clears infection while another can’t.

“The power of iPSCs is that now we can get cells from individuals with mutations that predispose them to any disease, including TB, build a lung-on-a-chip with them, and try to mimic that in vitro,” Gutierrez said.

The platform is already expanding. Gutierrez’s lab is using the chip to study COVID, lung cancer, and chronic obstructive pulmonary disease. The same architecture will answer different questions.

But Kotton sees two futures for platforms like this on different clocks. Testing drugs, identifying therapeutic targets, modeling host defense — “that’s all in the near future,” he said.

The personalized medicine vision is another matter.

With only four of the lung’s 40 cell types and no full immune system, predicting individual patient responses remains “just theoretical at the moment and probably further off,” Kotton said.

No drug has yet been tested on the device, and breathing experiments with live bacteria are still underway. But for a disease that still kills by the millions, even a narrow window into the black box is worth having.

One study co-author reported being employed by and holding minor equity in AlveoliX. All other authors declared having no competing interests.