Cardiac or skeletal muscle: choosing an actuator substrate

The choice between cardiac and skeletal muscle is not a performance question. It is a control question, and it is decided before you know anything about the forces.

Cardiac muscle contracts spontaneously, rhythmically and forever. You do not switch it on; you switch it off, or more accurately you modulate a rhythm that was already running. Skeletal muscle does nothing at all until commanded, then does exactly what you commanded and stops.

If your device wants a pump, a swimmer or anything else whose natural state is periodic motion, cardiac gives you the oscillator for free. If your device wants a gripper, a walker, a valve or anything that must hold still and then act, skeletal is the only sensible substrate — a cardiac gripper that will not stop gripping is not a gripper.

Everything else follows from that.

Head to head

Why cardiac dominated first, and then didn’t

The earliest and most visually striking biohybrid devices are almost all cardiac, and the reason is the spontaneity. Feinberg and colleagues’ 2007 muscular thin films — neonatal rat ventricular cardiomyocytes on micropatterned PDMS, released from a thermally sensitive substrate — performed gripping, pumping, walking and swimming at centimetre scale, generating specific forces as high as 4 mN/mm². That is a single paper covering four device classes, and it works because the tissue supplies its own drive.

That lineage continues directly. Nawroth and colleagues built the “medusoid”, a freely swimming jellyfish from dissociated rat tissue and silicone, designed by quantitatively matching stroke kinematics and animal–fluid interactions. Park and colleagues built a phototactically guided soft-robotic ray. Williams and colleagues put cardiomyocytes on a PDMS filament and got a self-propelled low-Reynolds-number swimmer at 5–10 µm/s, and a two-tailed version at 81 µm/s.

Then Lee and colleagues did something more interesting than any of them. Their 2022 biohybrid fish transferred two regulatory features of the heart into a swimming machine: mechanoelectrical signalling, so that each contraction fires automatically in response to the stretch of an antagonistic partner muscle, plus an engineered pacing node to entrain the loop. The result is self-sustained body-caudal-fin swimming with closed-loop control implemented in the tissue itself rather than in electronics.

That is the strongest argument for cardiac substrate that exists, and it is worth stating plainly because it is easy to miss: cardiac muscle lets you put the controller inside the actuator. No skeletal construct does that.

The reason the field nonetheless drifted toward skeletal is equally simple. You cannot tell a cardiac actuator to stop. For a swimmer that is a feature. For a machine that must do something on command, it is disqualifying.

Why skeletal took over for controlled machines

Skeletal muscle’s quiescence is the whole product. Raman and colleagues built modular light-controlled bioactuators from channelrhodopsin-expressing C2C12 producing up to 300 µN (0.56 kPa) on optical command — non-invasive, spatially addressable, and silent until addressed. Morimoto and colleagues built a jointed robot from an antagonistic pair of skeletal tissues achieving roughly 90° of joint rotation by selectively contracting one side or the other, and demonstrated pick-and-place manipulation with it. Neither device is expressible in cardiac tissue.

Skeletal also trains, and this is the second decisive advantage. Chronic pacing is the largest documented lever on force in the entire field: one week of intermittent 1 Hz stimulation raised tetanic specific force in human myobundles to 19.3 ± 0.63 mN/mm² from 9.1 ± 0.38 mN/mm² unstimulated. Chen and colleagues pushed C2C12 to 7.05 mN tetanic (8.51 mN/mm²) with a purpose-built self-training protocol, against a field where most constructs sit under 1 mN. Guix and colleagues went further and built the training into the mechanics — a 3D-printed serpentine spring skeleton whose restoring force cyclically stimulates the muscle as it spontaneously contracts, improving force output with no external stimulus at all, and reaching 800 µm/s (3 body lengths per second) at 5 Hz.

Cardiac tissue offers no comparable improvement curve. It arrives close to what it will ever do.

The lifetime problem, which nobody solves well

This is where both substrates disappoint, and where published numbers are thinnest.

Morimoto and colleagues state the skeletal problem precisely: it has been difficult to achieve both large and long-term actuation, because skeletal muscle tissue spontaneously shrinks over the course of culture. Their antagonistic-pair design is not primarily a control innovation, it is a durability one — the two tissues hold each other in tension, preventing the shrinkage that kills single-tissue constructs, and that is what bought them roughly a week of useful life.

One week. That is a headline result in a high-profile venue, and it is the honest ceiling to plan against for an untethered skeletal device.

Cardiac constructs generally run longer in published demonstrations, and the swimming work is built around sustained autonomous operation. But “longer” here is still measured against a 37 °C incubator, a media supply and a sterile environment. Neither substrate gives you a device you can put in a drawer.

The only published route out of that constraint is not mammalian at all: Akiyama and colleagues’ insect dorsal vessel tissue contracts spontaneously at room temperature with no CO₂ incubator and no stimulation, at 20 µN for a whole tissue. That is three orders of magnitude below the skeletal constructs above, and there is no commercial supply whatsoever — every published use is a dissection. It is the right answer only when ambient autonomy outranks force, which is rare but not never. The operating envelope reference sets out that trade in full.

The supply position, which favours cardiac

Here the commercial market inverts the research picture. Cardiac is the better-supplied substrate, because the pharmaceutical industry needs human cardiomyocytes for cardiotoxicity screening and has funded a real catalogue into existence.

Note the asymmetry hiding in the first two rows. Cardiac has the better human supply; skeletal has the better cheap supply. A group iterating on device geometry wants many cheap replicates, and C2C12 at $577 a vial with indefinite expansion beats any cardiomyocyte product on that axis by a wide margin. A group that needs human-relevant contractile tissue tomorrow is better served by the cardiomyocyte catalogue.

And the same limitation applies on both sides: nobody publishes a force acceptance criterion. You can buy cells of either lineage, and you can buy a finished tissue of either lineage. You cannot buy either specified to a force.

Choosing, in one pass

Choose cardiac when the device motion is inherently periodic — pumping, swimming, ciliary transport; when you want the oscillator supplied by the biology rather than by electronics; when human relevance matters and you want it from a catalogue today; or when you are attempting tissue-embedded closed-loop control of the kind Lee and colleagues demonstrated.

Choose skeletal when the device must act on command and be still otherwise; when you need antagonistic pairs, joints or graded force; when you intend to train the tissue toward higher force with chronic stimulation; or when you need many cheap iterations, which is the situation at the start of essentially every project.

Choose insect dorsal vessel tissue when ambient, incubator-free, self-powered operation is a hard requirement and 20 µN is enough — accepting that you will be dissecting insects, because there is no supplier.

Choose neither yet when you have not written down the force, the stroke, the duty cycle and the required lifetime. That specification decides the substrate far more reliably than any table can, and writing it is cheaper than a failed six-week culture.

How we can help

We source cardiac and skeletal cell products against the same written specification and return the quotes on a cost-per-construct basis, including the casting device and the contract culture capacity where you do not have it in-house. Where a requirement is stated as a force, we will tell you which published work supports it on each substrate and which does not — and we will say plainly that no supplier currently contracts to a force acceptance criterion on either, because knowing that before the purchase order is worth more than a favourable quote after it.