Organ-on-Chip Perfusion Platform

From one perfused model to a dataset

A microphysiological model stops being a demonstration the day it has to produce numbers, and numbers mean replicates. At that point the setup starts limiting the experiment. One pump drives one sample, so six samples means six pumps, six sets of tubing and six things that can go wrong overnight.

The medium arithmetic surprises most people. At 10 µL/min a single sample gets through 14.4 mL of medium in a day, so six samples over two weeks is around 1.2 L on a single pass, and the expensive part of that volume, the growth factors, was never consumed. What the tissue itself secretes leaves with it, diluted below what the cells or the assay can pick up. The options differ in how they handle that.

Several syringe pumps. One pump, one line, so the shelf grows with the experiment. A syringe carries a fixed volume, so a run of days means refills, and each refill stops the flow and opens a way in for air. Nothing returns to the reservoir: every millilitre passes once and goes to waste.

A multichannel peristaltic pump. The honest baseline for recirculation, since it draws from a bottle for as long as the run lasts. Rollers squeeze the tubing to move the liquid, so the flow reaches the sample as a train of pulses and the wall shear stress on a barrier rises and falls with every roller that passes. The pulse rate is tied to the flow rate, so the low rates a barrier model tolerates are where the pulsation is worst. The tubing is a consumable under mechanical fatigue as well, so the rate on day ten sits off the one you set on day one.

A rocker or a gravity plate. Many samples to analyse, nothing to connect, and the reason many groups never buy a pump. The flow is a by-product of tilt: bidirectional, rising and falling through each rock, with the tilt angle and the interval as the only handles. The shear swings, and there is no value you can set it to or report.

A circuit you build yourself. Published protocols describe how, down to the milled fluidic board, the return valves and the control software. It works, and in parts it is cheap. What it costs is the months to build and characterise it, and the fact that the record of what it did is whatever your own script happened to write.

A platform built around the circuit. Air pressure moves the medium, so no moving part pushes it and the flow arrives smooth. One circuit carries a single recirculating medium to six samples in parallel. You set the wall shear stress your model calls for, and the platform holds it there, measuring what flows and correcting itself as the run proceeds. The sequence, the cleaning steps and the run record come with the instrument.

Perfuse six samples from one circulating medium.

Six lines run in parallel from a single circuit, driven together, so the perfusion is uniform across all six. What sits at the end of each one is yours to decide: six separate microfluidic chips, or six inputs on a single multi-well plate. A study that used to take four runs on two devices can then be laid out in one, and every sample in it saw the same perfusion. The medium circulates, which is what makes the volume manageable at that count: your tissues share one batch, and what they release stays in circulation with them. Where a model needs its own separate medium, the platform can be configured with several independent lines instead, and we size that with you.

Keep what your tissue secretes.

The medium-to-cell ratio in a perfused model sits far from the ratio in the body, and a single-pass circuit makes it worse: autocrine and paracrine factors are carried away as fast as they appear, too dilute to act on the cells and often too dilute for the assay you wanted to run. A closed circuit changes that. The medium returns to the reservoir and comes back, unidirectional at the sample, so the flow never reverses on the cells. That direction is part of the cue: an endothelium reads which way the flow goes as well as how hard it pushes, and a flow that swings back and forth aligns cells differently from one that holds its direction. Conditioned medium accumulates instead of being pushed to waste, and the same batch keeps going round for days or weeks. Recirculation also decides what medium you can afford to work in. A defined medium carrying growth factors, or patient serum, is out of reach when every millilitre passes once; it becomes workable over a two-week culture when the same batch circulates. For a vascular or barrier model held for weeks, the tissue then sits in a signalling environment it built for itself.

Put your barrier under the shear stress you chose.

Wall shear stress is what an endothelium or an epithelial barrier actually reads, so it is the number the experiment turns on. You set it on the platform, and holding it is the harder half, since flow drifts as tubing warms and as a reservoir level moves. The platform measures what is actually flowing and corrects itself, so the shear stays at the value you chose from the first hour of a culture to the last. No moving part pushes the medium either, so the flow arrives smooth and free of the pulsing a roller or a plunger adds.

Load a small volume or a large one.

The reservoir’s volume has no bearing on the shear stress your cells feel. A 2 mL aliquot of a compound solution reaches your samples at the same shear stress as a litre of culture medium, and swapping one reservoir for another changes nothing in the settings. That matters at both ends of the same study: expensive reagents belong in small aliquots close to the samples, while a long perfusion run gets through volume. It also means a dose or a washout can be a reservoir change, with the same shear stress before it and after.

Keep a multi-week loop clean.

Air pressure moves the medium, so no moving part pushes it and no pump head sits in the liquid. In a closed loop that counts for more than in a single-pass line, because whatever gets into the medium comes back round with it, again and again, for as long as the run lasts. The platform runs a cleaning step at the end of a study, so what circulated for two weeks is cleared out and the next experiment starts on a clean circuit.

Start it from the front panel, and follow it as it runs.

A touchscreen on the platform sets the perfusion and builds the sequence, so a run starts with no computer to connect first and no script to maintain. Build the sequence once, store it on the platform, recall it for the next study, or load a pre-defined one. An initialization step brings the platform up before anything reaches your samples, and a cleaning step clears the circuit at the end so the next study starts clean. While the run proceeds, the data is plotted as it is acquired, so a glance tells you the perfusion is still doing what you asked. A trigger signal lines the run up with a microscope, so an acquisition lands at a point in the sequence you chose.

Reproducible perfusion across a study.

Perfused models have an inter-laboratory reproducibility problem, and part of it is instrumental rather than biological. In a review of the published literature, 55% of the articles disagreed with each other on how a given biomarker responds to flow, and only 3% characterised the magnitude of the flow they applied at all. A shear stress that is set, held and recorded takes that variable out of the argument. The protocol expertise stays yours; what changes is that the perfusion your samples received is a number you can put in a methods section and another lab can reproduce.

A record that goes in a qualification file.

As a run proceeds, the platform logs its full history, every value it held and every step it ran, with a timestamp. That log is a 21 CFR Part 11-compliant record, meeting the FDA rule for electronic records and audit trails. For a safety group or a facility running assays for other people, the shear stress your models received becomes documented, auditable data. That weighs more now than it did: as regulators open the door to data from non-animal methods, a model is judged on whether it is fit for purpose, and the perfusion history is part of that file.