Organ-on-chip experiments usually depend on incompatible, closed peripheral equipment, which makes models hard to combine or move between labs. The authors build and test STARTER, a stand-alone platform in which pumps, reservoirs, in-line sensors and three organ-on-chips mount as ISO 22916-compliant building blocks (following the openly published TOP Design Rules) on a microplate-sized fluidic circuit board; every channel runs to a single swappable routing block, so replacing that one part reconfigures the whole fluidic circuit without tubing. They demonstrate, with measurements, parallel and

STARTER fits pumps, reservoirs, sensors and three organ-on-chips onto one microplate-sized board. Every channel ends at a single swappable routing block — change that block and the same parts run a different experiment.

Three identical, isolated experiments.

Board, channels and module shapes come from the open STARTER design files (CC BY 4.0). Routing-block channels are reconstructed to reproduce the circuits of Fig. 2a, 2c, 4a and 4b; the flow animation is schematic and sped up.

Same board, same modules — only the routing block differs.

Watch configuration 2 fill

The authors filmed the board from above as the pumps started on the combined routing block. Two kinds of path fill in turn: red loops that each stay with their own chip, then a green stream that threads all three chips one after another. Pick a moment to jump there.

One plug standard for everything

Organ-on-chips are small devices in which living cells or tissue are fed by a slow, controlled flow of medium, so that they behave more like a real organ. They need pumps, reservoirs and sensors around them — and today most of that equipment comes from closed systems whose connectors fit only themselves. Combining a chip from one lab with a pump from another usually means custom tubing and adapters.

STARTER follows a shared rulebook instead. The international standard ISO 22916 fixes where fluid ports may sit (on a 1.5 mm grid) and how big a module may be (in steps of 15 mm). The openly published TOP Design Rules narrow this to a concrete set of footprints and port layouts — you can design with them in the next section. Any module that obeys them — a microfluidic building block — can be screwed onto the board.

The board itself has the footprint of a standard microplate, so it fits ordinary microtiter-plate holders. Inside it run 400 µm square channels; soft silicone O-rings in small pockets seal every port. Pumps, reservoirs and sensors mount on top; the routing block and the organ-on-chips mount underneath, above imaging windows. Below is the real board, redrawn from its design files.

Tap a module footprint to trace its channels (its dimensioned drawing opens below), or tap a channel. Dashed outlines sit on top of the board; dotted outlines are mounted underneath. The grid toggles show the ISO rules per module: sizes in 15 mm steps, ports on a 1.5 mm grid measured from each footprint corner.

Modules come from different makers and materials: in-house PDMS chips, 3D-printed intestinal explant chips adapted by TNO, commercial reservoirs on a milled base, a commercial thermal flow-sensor chip and an optical pH spot turned into 15 × 15 mm blocks. They fit because their footprints and ports follow the same rules.

Every block, to the millimetre

The paper's supplementary drawings (Fig. S1) give the board and each block their exact dimensions. Pick a part: the drawing shows the published measurements, and the rings mark the O-ring pockets in the board under that block, taken from the design files — so you can see each port land on one.

Measurements: Fig. S1 of the paper's supplementary information, redrawn (mm). Pocket positions: STARTER design files; pocket size Ø 2.75 mm from the Methods. Ø 0.40 mm holes are drawn larger so they stay visible.

Design a chip that plugs in

How does a chip made in one lab fit a board made in another? Through a short rulebook. ISO 22916 sets the outer limits: ports on a 1.5 mm grid, sizes in steps of 15 mm. That still leaves a lot of freedom, so the TOP Design Rules — openly published drawings that the paper builds on — pick a handful of footprints and fix exactly where each port goes. Every chip, reservoir and sensor slot on STARTER's board is one of these layouts.

Pick a layout below, then try to break it: drag a port, resize the footprint, or cut an opening through the chip. The rule check reacts as you go, and the dashed orange rings mark where STARTER's own board ports sit.

Drag a port or the opening, or tap it and use the arrow keys (one step = one 1.5 mm grid step). Port names follow the TDR drawings: a letter and a number counting grid steps from the top-left corner.

Why keep-clear areas?

On STARTER, every port seals against a soft silicone O-ring recessed into the board. The TDRs reserve a 6 × 6 mm square around each port on the face that meets the board, and a 1 mm rim on the face the clamp presses on: both must stay free of obstructions and chip features so the connection stays leak-free. Squares of neighbouring ports simply merge into one bar.

The same rules let STARTER take chips of different make — in-house PDMS chips and 3D-printed gut explant chips adapted by TNO — through the same clamps.

Footprints, port positions, keep-clear squares and clamping rim redrawn from the TOP Design Rules: E. Safai, Translational Organ-on-chip Platform (TOP) Design Rules (TDRs), 4TU.ResearchData, 2024, doi:10.4121/2558bd4c-d7ad-4e17-bc54-8c335b4c1c01 (CC BY 4.0) — reference 33 of the paper. STARTER's board ports and slots come from its open design files. Channels drawn inside the chip and the Ø 8 mm opening are illustrations.

Be the routing block

Each port under the routing block already leads, through the board, to one specific module port: a pump inlet, a reservoir chamber, one end of a chip channel. The routing block's only job is to join those ports in pairs. It is a single milled plastic layer sealed with tape, so two channels can never cross — they would simply merge. The authors' collaborators even wrote an automatic router for this. Here you decide what to connect; a built-in router finds a channel path that keeps clear of every other port and channel.

Tap a port, then a second port to lay a channel. Tap a channel to remove it.

Every module has two ports: medium enters one and leaves the other. Pumps only push one way (in → out). A working loop needs a pump, a reservoir to draw medium from, and a chip channel to perfuse. The router keeps each channel centre 1.45 mm from other ports and 0.75 mm from other channels — the same rules used to reconstruct the circuits in the 3D model above.

The pump block, on the bench

Six commercial peristaltic pumps — rollers that squeeze a soft tube to push liquid forward in small pulses — sit in one 105 × 30 mm block with a custom driver board and an onboard Arduino that takes commands over Bluetooth Low Energy. Each pump is rated for 0.18 to 180 µL/min. The team put a commercial thermal flow sensor, rebuilt as a 15 × 15 mm block, in line to measure what the pumps really deliver.

Drive frequency → flow

The stepper motors are driven at a set frequency. Turn it and read the in-line flow sensor.

Flow-sensor trace: illustrative, shape after Fig. 3a (pulses grow and quicken with frequency).

Points: mean ± spread over 5 min per frequency, read from Fig. 3b (approximate). The relation is almost linear; the sensor's range limited the test to 0.5–120 µL/min.

Two pumps, one colour

Run a blue and a yellow pump into the same channel. Their flow ratio sets the shade — a concentration gradient made by the pump block alone.

Photographs: the five measured ratios, Paul et al. 2026, Fig. 3c (CC BY 4.0), cropped. Colours sampled from these photos; in-between shades are interpolated.

A sensor in the stream

The board has three 15 × 15 mm sensor slots that can be placed anywhere in a circuit. For pH, a commercial optical sensor spot sits inside a channel and an optical fibre looks down on it from exactly above; the spot's glow changes with acidity. The spots cover pH 6.0 to 8.0 — the range that matters for living cells. After a two-point calibration, the team recirculated each test solution for 30 minutes at 27 µL/min.

Feed the sensor

Send a solution past the sensor spot and watch the reading settle.

Readings read from Fig. 3d (approximate). The sensor was most accurate near pH 7.0, deviated by at most 0.2 pH and showed no drift over 3 h. Spot colour is illustrative.

Three days of watching

The same pH spots then monitored medium on two platforms during a 3-day culture of blood-vessel cells.

After a short settling period at the start, both platforms held pH 7.1 for three days — no significant change, which is what cells need.

Redrawn from Fig. 4a iii (approximate). Blue: commercial reservoirs; red: 3D-printed reservoirs.

Prime with a dummy, then swap in the living chip

Cells need days to settle into a chip, and a freshly assembled platform has to be sterilised and filled before anything living touches it. Because every block on STARTER has its own screws, any one of them can come off without disturbing the others, so the two jobs can run side by side. The paper's supplementary protocol (Fig. S3) does it in three steps, with a stand-in dummy chip holding each chip slot until the real one is ready.

Redrawn from Fig. S3 of the supplementary information. The loop is schematic: on the real board every path runs through the routing block. Cells in the chips are drawn, not imaged.

Why no bubble? The soft silicone O-rings are hydrophobic, so a port keeps a hanging drop when a chip comes off. A chip that arrives already filled meets that drop liquid-to-liquid, so swapping traps no air (section 4.1 of the paper). The port cross-section is a schematic, not to scale.

Grow the chips off the board

Sterilise and prime STARTER

Integrate the chips

Vessel-on-chip

Endothelial cells were seeded at 2 × 10⁶ cells per mL, left still for 4 h at 37 °C, then rocked (10°, in 1-hour intervals) with fresh medium twice a day until they formed a monolayer. Only then were the chips connected to STARTER, for 3 days at 1 µL/min, and imaged daily from the moment of connection.

Gut explant

The explant chips were primed on the platform itself, the day before: flushed with 20% Biofilm solution, then PBS, then filled with Williams E medium with 1% BSA overnight at 33 µL/min in the incubator. Next day, fresh pig colon tissue, mounted within 4 h of collection, was clicked into the chips, the apical and basolateral media were added, and perfusion ran at 33 µL/min for 24 h.

Does the biology notice?

A platform is only useful if cells and tissue do as well on it as in the usual setups. The team ran two kinds of biology on STARTER: a human blood-vessel model grown inside the chips (in vitro) and freshly collected pig gut tissue (ex vivo).

Illustrative monolayer: cell density and day-3 coverage follow Fig. 4a iv (read from the plot); cell shapes are drawn, not imaged.

Human umbilical vein endothelial cells (HUVECs, the cells that line blood vessels) grew in the two narrow middle channels of each chip under a steady 1 µL/min flow for 3 days. Controls sat on a rocking platform that sloshes medium back and forth. Two STARTER platforms ran side by side: one with commercial reservoirs, one with 3D-printed ones.

Real image after 3 days: VE-cadherin cell borders (white), nuclei (blue). Paul et al. 2026, Fig. 4a ii (CC BY 4.0), cropped.

With commercial reservoirs, cell numbers and coverage matched the controls, and cell–cell junction protein (VE-cadherin), cell shape and alignment with the flow showed no significant differences. With 3D-printed reservoirs, cells declined — the authors suspect substances leaching from the printed resin under the daily fluorescence imaging.

Pig colon tissue from a slaughterhouse was punched into discs and clamped in intestinal explant barrier chips (IEBCs, from TNO, adapted to the 30 × 30 mm footprint). The tissue separates two compartments: the apical side (the gut's inner surface) and the basolateral side (towards the blood). STARTER perfused each side with its own loop at 33 µL/min for 24 hours, three chips at once, and the results were compared with IEBCs in their usual setup.

Is the tissue still alive?

LDH is an enzyme that leaks out of damaged cells. Less LDH in the medium means healthier tissue.

Values read from Fig. 4b ii (approximate). ns: not significant at every time point.

Is the barrier still tight?

Atenolol is a drug that crosses the gut wall only through the gaps between cells. If little reaches the basolateral side, the barrier is intact.

Values read from Fig. 4b iii (approximate).

Same biology, a much smaller footprint

The gut-explant results were compared with explant chips run in their usual setup. The authors photographed both workflows on the same bench: two STARTER platforms carrying six explant chips, next to the usual setup for eight. Pick a side to see what each one takes.

STARTER column: from the paper (microplate-sized board, tube-less connections between modules, a 12 V power adaptor as the only peripheral). Usual-setup column: what the photo shows.

An open starter kit

Every design file — the board, the pump block with its driver electronics and firmware, the reservoirs — is published openly (GitHub: TOP-OoC/Starter-Kit), alongside the TOP Design Rules, an explainer of the ISO standard and an automatic routing-block designer (the MMFT Routing Block Channel Router, from collaborators at the Technical University of Munich). The idea is a shared, pre-competitive starting point: a chip developed for one standardized platform can move to others, including future high-throughput systems that follow the same rules.

What comes next, according to the authors

Other labs Running STARTER outside the founding labs is the key next step to validate its translatability, design choices and robustness.

Easier connections Simpler mechanical connections would make set-up faster in time- and space-limited workflows.

Fewer cables As sensors multiply, integrating electronics and read-out on the platform would avoid a web of cables.

Valves in the routing block Would let the circuit be re-programmed actively, instead of swapping the block by hand.

STARTER: a stand-alone reconfigurable and translational organ-on-chip platform based on modularity and open design principles. Aniruddha Paul, Eric R. Safai, Laura E. de Heus, Anke R. Vollertsen, Kevin Weijgertse, Bjorn de Wagenaar, Hossein E. Amirabadi, Evita van de Steeg, Mathieu Odijk, Andries D. van der Meer, Joshua Loessberg-Zahl. Lab on a Chip 2026, 26, 604–617. doi:10.1039/d5lc00756a · Open access under CC BY 4.0.

Figures and photographs from the paper (Fig. 1a, 1b, 1c, 2b, 2d, 3c, 4a ii, 4b ii) and its supplementary information (Fig. S6 bench photo; Supplementary Video 1, trimmed to the platform and re-encoded): © the authors, CC BY 4.0; cropped. Frames and outlines on the video and on Fig. S6 are added here.

Board outline, holes, port pockets and all 60 channel paths were extracted from the STARTER design files, and the reservoir and pump-base meshes used directly (github.com/TOP-OoC/Starter-Kit, CC BY 4.0); re-rendered here.

Pump-block photo: TOP-OoC/Starter-Kit repository README (CC BY 4.0).

Block dimensions (section 01) and the priming and chip-integration protocol (section 06): redrawn from Figs. S1 and S3 of the paper's supplementary information. Experiment-specific preparation steps from the Methods (sections 6.8 and 6.10).

Chip-design section: footprints, port positions, keep-clear squares and clamping rim redrawn from E. Safai, TOP Design Rules (TDRs), 4TU.ResearchData 2024, doi:10.4121/2558bd4c-d7ad-4e17-bc54-8c335b4c1c01 (CC BY 4.0).

Routing-block channel layouts in the 3D model and the puzzle are a reconstruction that reproduces the published circuits, not the authors' routing-block files. Plot values marked "approximate" were read from the published figures. The 3D scene, flow animation, cell mosaic and sensor-spot colours are illustrations.