Electrophysiology readouts compared: what each measures and what it costs in effort

Four families of technique dominate functional electrophysiology on cultured human cells. They are frequently discussed as alternatives, which is misleading, because they measure different physical quantities. Choosing between them is not a matter of preference or budget in the first instance, it is a matter of which quantity answers your question.

The distinction that matters most: only patch clamp gives you the membrane potential in millivolts and the ionic current in picoamps. Everything else gives you a signal that correlates with electrical activity, in arbitrary or relative units, with a known distortion. That is often exactly what you want, and it is much cheaper. But it is a different measurement.

Multielectrode array

An MEA is a culture substrate with embedded electrodes. Cells grow on it, and each electrode records the extracellular voltage transient produced by nearby cells firing. There is no penetration and no dye, so the same culture can be recorded repeatedly over weeks.

What it is uniquely good at: population and network activity over time. Synchronised network bursting, connectivity, developmental maturation, chronic drug exposure. The 3D iPSC neural network work cited here followed activity development over roughly six and a half weeks on a single set of cultures, a study design that is only possible because the technique is non-destructive.

What it cannot see: subthreshold activity, membrane potential, and individual ionic currents. An MEA records the extracellular signature of firing. A cell that is depolarising but not firing is invisible.

The critical distinction inside MEA data is between spiking and network bursting. Isolated spikes on a handful of electrodes mean the culture is electrically alive. Synchronised network bursts across the array are what most pharmacology and network assays actually depend on, and they arrive substantially later, sometimes not at all without glial support. See MEA-ready, where this is the central issue.

Density variants. Conventional MEAs have tens of electrodes per well. CMOS-based high-density arrays have thousands, allowing sub-cellular spatial resolution and spike sorting into single units, at the cost of much larger data volumes and a heavier analysis burden. A high-density MEA study of a 3D iPSC brain model reported consistent spontaneous firing and network bursting parameters from seven-week-old cultures, the seven weeks being the honest part of the specification.

Cardiac MEA is a different assay on the same hardware. Cardiomyocytes are spontaneously active within days rather than weeks, and the readout is the field potential, whose duration is used as an in vitro correlate of the QT interval. It is one of the pillars of the CiPA framework, alongside ionic current measurement in heterologous expression systems and in silico integration. Open-source analysis tooling exists specifically for cardiomyocyte MEA data.

Effort: low per experiment, high in calendar time. Plating, coating and weeks of feeding are the cost, not the recording. Analysis is non-trivial: burst detection parameters are choices, and the same raw data yields different conclusions under different detection criteria. Insist on knowing the burst detection criterion behind any figure you are shown.

Patch clamp

A glass micropipette forms a high-resistance seal with the membrane, which is then ruptured or perforated to give electrical access to the cell interior. The membrane potential can be recorded (current clamp) or held at a commanded voltage while the resulting current is measured (voltage clamp).

What it is uniquely good at: everything quantitative about a single cell. Absolute resting membrane potential, action potential waveform and its parameters, specific ionic currents isolated pharmacologically or by voltage protocol, channel kinetics, and, with intracellular solution control, the effect of intracellular ions and second messengers.

What it costs: the cell. Whole-cell recording is terminal for that cell, and the dialysis of the cytoplasm by the pipette solution changes the cell’s internal composition over the recording. Manual patch clamp is also the most skill-dependent technique in the list by a considerable margin; throughput is a small number of cells per person per day and the experimenter is a variable in the result.

Automated patch clamp (APC) changes the throughput picture without changing the physics. Planar-array systems in 384-well format achieve gigaohm seal quality and support both voltage and current clamp, with intracellular solution exchange available on some platforms. Published work has run a 1,920-compound screen against NMDA receptors on such a system, and screened 36 arthropod venoms against Na_V_1.9 on a single 384-well plate.

Two caveats on APC that a buyer should hold onto. It is designed around small, homogeneous cells in suspension, which is why heterologous expression lines are its native application and why recording native cardiomyocytes required a different, fixed-well plate architecture to accommodate larger cells. And it needs a suspension, meaning your adherent culture must be dissociated, which is a perturbation and which post-mitotic neurons tolerate poorly.

Effort: very high for manual. High capital and consumable cost, plus method development, for automated. Data quality per cell is the highest available, and the cost per cell is correspondingly the highest.

Calcium imaging

A calcium-sensitive dye or genetically encoded calcium indicator (GCaMP and relatives) reports intracellular free calcium as a fluorescence change. In excitable cells, calcium entry follows depolarisation, so calcium transients serve as an activity proxy.

What it is uniquely good at: cheap, scalable, spatially resolved population activity in standard multiwell plates with standard imaging hardware or a plate reader. Every cell in the field is measured simultaneously and identified by position, which is something no electrode array gives you at the same cost.

The central limitation: calcium is a proxy, and a slow one. Indicator binding kinetics and intracellular calcium buffering low-pass filter the signal, so individual action potentials within a high-frequency burst are typically not resolvable. What you measure is the envelope of activity, not the spike train. For network-level questions this is often sufficient. For anything requiring spike timing it is not.

Second limitation: calcium is not only an electrical signal. Calcium release from intracellular stores, and calcium signalling in non-excitable cells, produce transients with no action potential behind them. Interpreting calcium as voltage requires that assumption to hold in your system.

The scalable calcium imaging work cited here is a useful illustration of what the technique buys and demands. It found that neuron-to-astrocyte ratio determined cellular distribution and surface adherence, that increasing neuron density decreased network spike frequency while increasing network spike amplitude, and that including GABAergic neurons increased network spike frequency while reducing amplitude. In other words, the culture composition is not background to the assay; it is the dominant variable in the readout. Any comparison across conditions with different cell compositions is measuring composition as much as pharmacology.

Effort: low to moderate. The main burdens are the loading step for dyes (or a reporter line for GECIs, which carries its own licensing and cost), phototoxicity management, and analysis pipelines for large image stacks.

Voltage indicators

Voltage-sensitive dyes and genetically encoded voltage indicators (GEVIs) report membrane potential optically. They sit between calcium imaging and patch clamp: optical and scalable like the former, measuring the right quantity like the latter.

Strengths: membrane potential directly, with sub-millisecond temporal resolution in principle, without impaling the cell. Subthreshold events are in scope, which they are not for MEA or calcium imaging. Modern engineered GEVIs are sensitive enough that spontaneous membrane voltage alterations have been resolved in non-excitable cell lines, using a construct reported as offering a 3.6-fold improvement in brightness change for fast voltage spikes over its predecessor.

Weaknesses, which are real: the signal is a small fractional fluorescence change confined to the membrane rather than the whole cell volume, so it is photon-starved. That drives fast cameras, bright illumination, phototoxicity and photobleaching, which in turn limits recording duration and makes chronic longitudinal recording hard. Voltage dyes can also be pharmacologically active and can perturb membrane properties. And the measurement is relative: without a calibration step you get ΔF/F, not millivolts.

Effort: high. This is the least turnkey of the four, most demanding of optics and analysis expertise, and the one where generation-to-generation improvements in the indicators matter most to whether it works for you.

Side-by-side

Choosing by question

What people get wrong

Treating MEA activity as a substitute for cellular electrophysiology. An MEA tells you a population is firing. It does not tell you the cells have mature membrane properties. A culture can produce spikes on an array while having a depolarised resting potential and an immature action potential waveform, and only patch clamp will show you that.

Treating calcium transients as spike trains. They are the low-pass-filtered envelope. Any analysis that counts calcium peaks as action potentials is counting something else.

Underestimating the maturation period. For iPSC-derived neurons the recording is quick and the wait is long: the cited studies work with cultures at roughly six to seven weeks. Budget for the incubator time, the medium and the risk of losing a plate, because that is where the money and the schedule actually go.

Ignoring culture composition. The calcium imaging work cited here found neuron-to-astrocyte ratio and GABAergic content dominating the network readout. The same is true on MEA. If your two conditions have different cell compositions, composition is your independent variable whether you intended it or not.

Assuming platform transferability. Seeding density, coating and expected timelines validated on one MEA platform are a starting point on another, not a transfer. Electrode geometry, well area and recording modality differ enough to matter. See the MEA platform compatibility matrix.

Buying the instrument before establishing the culture. The culture is the hard part and the recurring cost. An instrument with no reliable culture behind it records nothing.

Boundary cases

Sharp-electrode and perforated-patch recording sit inside the patch clamp family with different trade-offs: perforated patch avoids cytoplasmic dialysis at the cost of higher access resistance, which matters for long recordings where run-down would otherwise confound the result.

Impedance-based contractility measurement is not electrophysiology at all, it measures mechanical contraction via impedance change. It is frequently offered on the same instruments as cardiac MEA and answers a different question. Do not conflate the two endpoints.

3D constructs need 3D-capable arrays. A planar MEA measures only the surface it contacts. Dedicated 3D arrays record at multiple vertical positions, which is what allowed the cited work to resolve region-specific differences in synaptic composition within the depth of a hydrogel-based tissue.

Optogenetic stimulation is orthogonal to all four. It is a stimulation modality that combines with any of these readouts, and it introduces its own construct, licensing and illumination-crosstalk considerations, particularly with optical readouts, where excitation spectra must be separated.

The one-line summary

MEA measures population activity non-invasively over weeks; patch clamp measures membrane potential and ionic currents in absolute units at the cost of the cell; calcium imaging is a cheap, scalable, slow proxy; voltage indicators measure the right quantity optically but are photon-limited. Pick from the question, not the budget, then budget for the six to seven weeks of culture that most of these readouts sit on top of.