An optical MEA pixel that separates the spike from the motion

What the work claims

Leronni and Moreddu claim that a membrane-based optical microelectrode array can measure cardiac electrical excitation and mechanical contraction simultaneously and separately, in the same unit cell, label-free with respect to the cells. This is a primary experimental result with a supporting physics model, not a position piece. The device sits on a silicon chip with a centrally suspended silicon nitride membrane that separates an upper biological chamber from a lower optical chamber. Through-membrane nanoholes 100 nm in diameter are electrodeposited with gold to form vertical 3D contact electrodes, and 1 micrometer nanopillars are added on the bases to roughen the surface for cell adhesion. The readout electrode is patterned as concentric squares: a central contact electrode that the cell touches, surrounded by a frame electrode on the lower side that never contacts the cell and senses only through the fluid.¹

The claim rests on a subtraction, not a new dye. In drug-free medium, a spontaneously beating hiPSC-derived cardiomyocyte deforms the membrane over the contact electrode each beat, so the contact trace carries a composite electromechanical signal while the non-contact frame reports predominantly membrane displacement. After 5 micromolar blebbistatin suppresses actomyosin contraction, the non-contact trace collapses to its noise level while the contact trace keeps periodic events, which the authors assign to action-potential-associated charge redistribution. Four signal classes result: composite, electrical-dominated, mechanical-dominated, and noise, all from the same fluorophore bath.¹

How it works

The transduction mechanism is electrokinetic, and the paper does the unusually honest work of modeling it rather than hand-waving it. Cationic rhodamine 6G at 0.2 mg per mL in ethylene glycol fills the lower chamber. The action potential is imposed as a surface charge on the contact electrode through a specific membrane capacitance of 0.01 F per square meter, and the beat is imposed as a 2 micrometer downward membrane displacement lagging depolarization by 20 ms and outlasting it by 100 ms. A time-dependent Poisson-Nernst-Planck model solved in COMSOL tracks the fluorophore and chloride counter-ion fluxes under diffusion, electromigration, and the advection induced by membrane motion, with fluorophore diffusivity 1.61e-11 square meters per second scaled by the water-to-ethylene-glycol viscosity ratio.¹

The model reproduces the measured phenomenology. In the purely electrical case, depolarization flips the electrode charge and the fluorophore concentration at the contact electrode swings from minus 63 percent to plus 27 percent of bulk, confined to a nanometric enrichment layer at the surface. Add membrane motion and the concentration overshoots to plus 62 percent, relaxes slowly, and, because the mechanical contribution does not recover within one beat, drifts upward cycle over cycle to plus 106 percent by the fifth beat, while the electrical case stays stationary. At the uncharged non-contact electrode the electrical case produces nothing and the electromechanical case plateaus near 70 percent, identifying the frame as a purely mechanical reporter. Predicted peak-to-peak ranges of 0.90, 1.26 to 1.58, and 0.77 relative units scale with the measured values of 0.15, 0.26, and 0.18 by factors of 6, 4.8 to 6.1, and 4.3 respectively, consistent with one common transduction mechanism across all three channel types.¹

The experiments were run on human induced pluripotent stem cell cardiomyocytes from a commercial source, recorded after 13 days in vitro when synchronous beating was stable, using a resonant-scanning confocal microscope at 30 frames per second with 561 nm excitation and 570 to 620 nm collection. Quantified amplitude statistics come from 48 detected events across 12 electrodes on 3 devices. Cell-free controls show drift and low-amplitude fluctuation but no periodicity, which excludes the optical system, the perfusion chamber, and the dye bath as sources of the rhythmic signal.¹

Where a skeptic should push

The most load-bearing assumption is that blebbistatin cleanly deletes the mechanical channel without touching the electrical one. Blebbistatin is a myosin II inhibitor widely used to uncouple motion from excitation, and the residual periodic contact signal after treatment is interpreted as purely electrical. But the waveform does change after drug application, the authors expect it to, and the separation therefore leans on a pharmacological intervention rather than on waveform decomposition. A reviewer should also note the small statistics: mean waveforms from 48 events over 12 electrodes on 3 devices establish that the discrimination exists, but they do not establish yield, longevity, or cross-lab reproducibility. The device has not been benchmarked head-to-head against patch clamp, conventional MEAs, voltage imaging, or contractility assays, and the authors say so in their limitations.¹

The second push is on absolute quantitation. The method discriminates signal components but does not recover membrane voltage or contractile force; it is a comparative readout, not a calibrated instrument. The model-to-measurement amplitude scaling factors of roughly 4 to 6 are accepted as consistency evidence, but they also tell you the model is not quantitatively fitted to the data. And the 30 frames per second confocal acquisition is two to three orders of magnitude slower than the kHz sampling an electrical MEA runs at, so the action-potential waveform here is an envelope, not a resolved spike shape. That is fine for rhythm and contractility screening; it is not fine for spike sorting. Finally, the label-free claim needs a qualifier: the cells see no dye, but the measurement consumes a fluorophore dispersion in a separate optical chamber, refreshed through microfluidics, so the system trades cellular labeling for a fluidic and optical overhead of its own.¹

What contactless optical pixels change for arrays

For microelectrode array hardware, the interesting move is architectural: the pixel is electrically floating. The contact electrode communicates with the cells but its signal leaves the chip as light, through a fluorophore layer interrogated confocally from below. No per-pixel amplifier, no per-pixel ADC, no multiplexed metal routing to the cell-facing side; the density limit moves from the CMOS wiring pitch to the optical resolution of the readout. Anyone who has fought the trade between electrode pitch and channel count in a wired high-density array should recognize what that removes. The concentric contact and non-contact pair is equally suggestive: because the two regions of interest see the same illumination, dye environment, and focal conditions, common-mode optical drift cancels in the comparison, which is the same differential principle as a wired instrument's CMRR, achieved here by geometry rather than by matched transistors.¹

The non-obvious implication runs in the other direction, and it is a caution rather than an opportunity. This paper demonstrates, with a physics model and a drug control, that an electrode in intimate contact with a beating cell records a composite in which contraction-driven membrane displacement contaminates the electrical signal, and that the contamination accumulates cycle over cycle because the mechanical perturbation does not relax within one beat. Every conventional MEA plated with spontaneously contracting tissue, cardiac organoids included, has this component sitting in its traces. In extracellular voltage it masquerades as slow baseline wander, drift between beats, or unstable spike amplitude, and it is usually treated as a motion artifact to be filtered out. The deeper point is that the artifact is not noise; it is a real, separable, biologically meaningful signal that a standard electrode cannot decompose because the electrical and mechanical channels share one conductor and one amplifier. The paired-pixel architecture is one answer; adaptive subtraction using a motion reference channel is the cheaper answer wired arrays could adopt tomorrow.

The genuine opportunity is in the acquisition chain economics for high-throughput cardiotoxicity and organoid screening, where the bottleneck is parallel, artifact-free readout of electromechanical function rather than sub-millisecond waveform fidelity. The genuine threat is bandwidth: a 30 fps scanned confocal readout cannot resolve fast electrophysiology, so this architecture competes for the screening niche, not the spike-recording niche, and anyone projecting it as a replacement for wired MEAs has the physics wrong. There is also an instrumentation hygiene lesson the array field should absorb: if your recording interface couples mechanically to living tissue, its motion response is part of the transfer function, and an honest datasheet for such a device would specify it, the way this paper's model does.¹

The bottom line

Established, in vitro: a membrane-based optical microelectrode array with paired contact and non-contact electrodes can report cardiac electrical and mechanical activity as separable optical channels; blebbistatin at 5 micromolar removes the mechanical component from the non-contact channel while the contact channel retains periodic events; cell-free controls show no periodic signal; and a Poisson-Nernst-Planck model with imposed action potential and displacement reproduces the four signal classes with measured-to-predicted amplitude scaling consistent across channels. Not established: absolute recovery of membrane voltage or force, pharmacology-free signal separation, head-to-head performance against patch clamp or electrical MEAs, throughput or yield beyond 12 electrodes on 3 devices, or anything faster than a 30 fps envelope. What would confirm the platform is a waveform-shape or timing-based separation that works without blebbistatin and a quantitative calibration against patch clamp. What would break it is the discovery that residual motion or dye-photophysics contaminates the electrical-dominated channel when contraction is suppressed by other means, because then the subtraction at the heart of the method has no clean reference.