T. Ching*, J. L. Teo*, W. Wangsrikhun, X. Song, C. S. Chen
*Co-first author
bioRxiv, doi: https://doi.org/10.64898/2026.07.22.740104
No pumps, no tubing. Just a dish on a stir plate.
Watch the dye dropped from a pipette get pulled into the channel right away. Shear stress in human arteries ranges from about 10 to 70 dyn/cm². PROPEL covers up to 60 dyn/cm², from venous to high arterial levels. Reaching 60 dyn/cm² takes about 2 mL of medium flowing through each channel every second.
Blood vessels reach almost every tissue in the body, and many major diseases start in them or spread through them. Heart attack, stroke, diabetic complications, cancer growth and many inflammatory diseases all involve the vasculature. Cardiovascular disease alone causes about 1 in 3 deaths worldwide.
Vessels also decide where drugs go. Nearly every drug taken by mouth or injection travels through the bloodstream and has to cross the endothelial lining to reach its target. In the brain, that lining is tight enough to block most drugs. A model that gets the endothelium wrong can mislead both disease research and drug development.
Endothelial cells line every one of these vessels, and they sit under flowing blood that drags along their surface and creates shear stress. Shear stress controls how these cells look and behave: their shape, alignment, gene expression, barrier function and inflammatory state. Without it, you are studying a different cell.
Yet most vascular biology research still uses static culture. Endothelial cells grow in flasks and well plates with no flow at all.
Researchers know flow matters. The problem is the tools. Pump-based perfusion needs tubing, fittings and bubble traps, and every added channel is one more line that can leak or get contaminated. Reaching arterial shear in channels large enough to collect useful numbers of cells takes mL per second of recirculating flow, which is hard to sustain with syringe or peristaltic pumps. Orbital shakers are simpler, but the shear they make is uneven and changes direction across the well. So most labs test one or two flow conditions at a time, if they use flow at all.
I learned this the hard way. Every time I set up a perfusion experiment, I spent more time on plumbing than on biology. Each channel needed its own fittings, tubing, pump line and reservoir, so a dozen channels meant dozens of connections, each one a chance for a leak, a bubble or contamination. Scaling up was a nightmare.
So I designed PROPEL around one question: what if you could remove the fittings and tubing altogether?
And instead of one pump channel per chip, could a single input, like one rotating magnet from an ordinary stir plate, drive flow through many channels at once?
At the core of PROPEL is a small impeller driven by a standard magnetic stir plate. As it spins, the blades push fluid outward past their tips. This leaves a region of negative pressure under the impeller, while the surrounding medium stays at atmospheric pressure. That pressure difference is what draws medium through the channels, with no pump or tubing needed.
Using this pressure difference is simple. Each channel's outlet sits inside the negative pressure region under the impeller, and its inlet sits outside at atmospheric pressure. Medium is pulled from the inlet, through the channel, and out the outlet. Because the negative pressure region surrounds the whole impeller, many channels can be arranged around it. One rotation drives flow through all of them at once, up to 20 channels in a single dish.
PROPEL fits entirely inside a standard 10 cm Petri dish. The dish sits on a magnetic stirrer, which spins a magnet embedded in the impeller. Beneath the impeller is the channel array, with each channel's outlet placed under the impeller and its inlet outside it. A top bracket holds the impeller in place above the array, and the dish's own lid closes the system. As the impeller turns, medium is drawn in through every inlet, flows over the cells growing in the channels, and exits under the impeller. There is no tubing, fitting, or external pump, and the whole dish goes into the incubator like any other culture.
With PROPEL, we exposed six types of human endothelial cells, taken from different blood vessels, to four shear levels at once: static, low, intermediate, and high. Each cell type had its own threshold for when it began to align, elongate, and polarize in the direction of flow. RNA sequencing showed a shared core response that shifted steadily with shear, along with responses unique to each cell type. Together, the results suggest that endothelial cells from different vessels differ in both when they respond to flow and how. This may help explain why some vascular beds are more prone to disease than others.
Static culture would have missed these differences.
Endothelial cells in PROPEL align in the direction of flow over 3 days at 15 dyn/cm².
The full study is available as a preprint on bioRxiv: [link].