Building an MEA-based neurotoxicity assay

The request is usually some version of “we need to know whether these compounds are neurotoxic”, and the honest first answer is that a microelectrode array will not tell you that. It will tell you whether a compound changes the electrical behaviour of a human neuronal network that you grew, at concentrations you chose, at a timepoint you picked. That is a genuinely useful thing to know and it is not the same claim.

The reason to be precise about it is commercial. An MEA neurotoxicity assay has a long, unforgiving setup — the cells have to mature for weeks before they are worth recording from, and there is no way to shorten that. Buying the wrong configuration costs you the money and the weeks, and you find out at the end.

This page is the route from the request to a purchase order. The short version: a 24-well pilot with co-cultured human iPSC neurons and astrocytes is roughly $3,000 to $3,500 of MODELLED materials, takes six to nine weeks from order to first dosing, and the binding constraint is maturation, not procurement.

What the assay actually measures, and what has been run through it

Two published EPA screening campaigns define the shape of this assay better than any vendor document. One screened 86 compounds during neural network formation on microelectrode arrays, reporting effects on developing cortical networks. A subsequent paper evaluated chemical effects on network formation in cortical neurons grown on MEAs across a larger compound set. Together they establish two things worth knowing before you design anything.

There are two distinct assay designs, and they answer different questions.

The commercial difference is large and rarely stated. In an acute design the network is a reusable instrument and the plate serves several compounds through wash-out. In a network formation design every well is committed to one treatment for the entire multi-week run, so the plate count scales directly with the compound count and the cell requirement scales with it. A network formation screen of twenty compounds is not the same purchase as an acute screen of twenty compounds; it is several times the purchase.

Decide which one you are building before you cost anything. If the request came from a developmental neurotoxicity concern, it is the second. If it came from a seizure liability or pharmacology concern, it is usually the first.

The three-way instrument decision

This is the decision that determines the shape of everything else, and it should be made in week zero.

No MEA vendor publishes instrument or plate list pricing. That is a deliberate commercial structure rather than an oversight, and its practical effect is that a first-time buyer cannot sanity-check a quote. The only public transaction anchors we have found are federal procurement records and one European order confirmation, all set out in what MEA plates and consumables actually cost.

For the outsourcing route, one published fee schedule gives a shape: the Boston Children’s Hospital IDDRC Human Neuron Unit publishes MEA recording at $49.00 internal / $64.00 external academic / $84.00 external industry per recording, plus a per-new-project planning fee of $290.00 / $377.00 / $493.00. That is one facility’s published rates for its own customer classes, not a market price, and “per recording” needs defining before you can build a total from it. Use it to sanity-check a commercial quote, not as a budget.

Choosing the plate format

The plate format decision is made under a constraint most people meet only after they have bought: neurons have a density floor below which they do not network at all.

Axion’s published best-practice density sweep for the Maestro system covers 5,000 to 80,000 cells per well and reports little to no activity at or below 20,000 cells per well, with robust spiking at 40,000 and above. That is not a gradual falloff. Under-seeding an MEA does not give you a weaker signal; it gives you no signal, and the whole plate is wasted.

The decision rule: choose the format by electrodes per well needed for your endpoint, then accept the well count that comes with it. A network burst synchrony metric computed from eight electrodes is not the same measurement as one computed from sixteen, and no amount of well count compensates. Spike rate is more forgiving than synchrony.

MEA plate manufacturers publish well volume, electrode count and recording area but not a growth area per well, and you should not derive one from the plate drawing — the wells taper to a raised bottom above the array and are not flat-bottomed cylinders. Seeding is done by droplet spotting over the array, and vendors publish cells per well directly. The full explanation is in vials to wells.

Cells: neurons, and the astrocyte decision

The published prices

The BrainXell rows carry a caveat we repeat rather than hide: those figures come from a same-day storefront capture and the site returned HTTP 403 to automated re-fetch, so they are recorded rather than live-verified and must be re-confirmed before you rely on them.

They still make the point. Price per million falls 46 percent from a 1M vial to a 5M vial at one supplier. Before concluding that any vendor is expensive, check you are comparing the same pack size. See iPSC neuron price per million cells.

Whether to co-culture with astrocytes

This is a scientific decision with a direct commercial consequence, and the published data points the same way on both.

Vendor titration data reports that human iPSC glutamatergic neurons co-cultured with astrocytes at a 10:1 neuron-to-astrocyte ratio reach a useful network at 50,000 neurons per 24-well MEA well, against 100,000 for the monoculture. Co-culture accelerates network maturation, so a lower neuron density works.

Halving the neuron requirement is a real saving. Against it: astrocytes are usually a different SKU, frequently from a different vendor, which adds a second order, a second lead time and a second lot to control. A published MEA characterisation of human iPSC-derived neurons in co-culture with primary human astrocytes reports spontaneous firing at 11 days in vitro with no network activity at that point, and synchrony developing over roughly six weeks — which is the timeline you should plan against rather than the one on any marketing page.

Our position: co-culture for anything where network synchrony is the endpoint, monoculture for spike-rate endpoints where you want one supply chain. Then hold the ratio and the density fixed across the whole campaign, because changing either mid-programme silently changes your baseline.

MODELLED cost: two pilots

These are models. Every unit price is real and sourced with a date. Every quantity is our assumption. They are not quotations, and no supplier has quoted against them.

Pilot A — 24-well co-culture, acute design

Eight compounds at two concentrations plus vehicle and positive control, in triplicate across two plates.

Cell arithmetic, using the published co-culture density and a 15 percent droplet-spotting overage:

• Wells: 24 × 2 plates = 48 wells
• Neurons: 48 × 50,000 = 2,400,000, × 1.15 = 2,760,000
• Astrocytes: 48 × 5,000 = 240,000, × 1.15 = 276,000

The plate line carries the biggest uncertainty and we are stating it plainly: we do not have a published or derived price for a 24-well MEA plate. The $488.80 figure is derived from a federal award for 48-well plates and is used here only as an order-of-magnitude anchor. Get a real quote for the plate before you believe this total.

Pilot B — 48-well co-culture, network formation design

The same eight compounds, but dosed from plating and recorded across the maturation window, so every well is committed for the whole run. One plate, using the published 48-well co-culture density.

• Neurons: 48 × 120,000 = 5,760,000, × 1.15 = 6,624,000
• Astrocytes: 48 × 20,000 = 960,000, × 1.15 = 1,104,000

Compare the two subtotals. The network formation design costs roughly 59 percent more in materials for the same eight compounds, before counting the extra weeks of medium and the extra recordings. That gap is the price of asking a developmental question instead of a pharmacological one, and it is worth knowing before you promise either.

The timeline, and the part that cannot be compressed

The single most important scheduling fact: the vendor’s “network signal from day 12” and the published “synchrony over roughly six weeks” are both true, and they describe different things. Day 12 is when something is detectably firing. Six weeks is when a network is stable enough that a compound effect is not confounded by continued maturation. Which one you need depends on your endpoint, and promising a date based on the first when your endpoint requires the second is the most common way this assay slips.

If the deadline is under eight weeks and you do not already have matured cultures on an array, the honest answer is that the assay does not fit. See sourcing under a grant deadline for what can be compressed and what cannot.

What breaks

Under-seeding to save money. Below the published density floor you do not get a weaker network; you get no network. The saving is the entire plate.

Buying neurons and astrocytes from different vendors without checking the media. Two products with two recommended media and one shared well is a problem you should solve on paper before it becomes a problem in an incubator. Ask both suppliers what medium the co-culture runs in, and whether they have published a co-culture protocol together.

No positive control. Without a compound that reliably moves your endpoint in your hands, a null result is uninterpretable. Agree the assay quality metric — Z-prime, on your control pair — before dosing, not after.

Ignoring the plate-edge artefact. Multi-week cultures in outer wells evaporate more than interior wells. Over a six-week maturation this is a systematic position effect that looks like a compound effect. Randomise treatment position across the plate, or leave the perimeter as buffer, and state which in the method.

Lot boundaries across conditions. Reserve the entire neuron and astrocyte quantity from one lot each and get it in writing. If a single lot cannot cover it, split the boundary across replicates rather than across compounds.

Forgetting the plate is single-use. A repeat run is a new plate at a new lead time, and MEA plates are frequently made to order with a two-to-eight-week lead. Buy the repeat plate with the first one.

Assuming your instrument takes the plate. Electrode count, well count and plate geometry are platform-specific. Check the MEA platform compatibility matrix before ordering anything, including which neuron products have published protocols on which recording platform, and where the honest answer is “unpublished” rather than “incompatible”.

The specification to send

Cells

• Neuronal subtype: cortical glutamatergic / GABAergic / mixed cortical / spinal motor
• Genotype: normal / isogenic control / named disease mutation; reporter requirement none / GFP / GCaMP
• Quantity: [total], stated as total at freeze or viable post-thaw
• Pack size: quote at 1 × 10⁶, 2.5 × 10⁶ and 5 × 10⁶ with price per million at each
• Lot: entire quantity from one lot, reserved, stated on the order confirmation
• Astrocytes: [quantity], species (human iPSC-derived / primary human / rat), and whether the supplier publishes a co-culture protocol with the neuron product
• Post-thaw viability specification floor, method (trypan blue or AO/PI), timepoint
• Media, supplement and coating: included / excluded, quoted as separate lines if excluded

Plate and platform

• MEA platform and instrument model
• Plate catalogue number, well count and electrodes per well
• Number of plates, including at least one spare for a repeat run
• Confirmed stock position and lead time in weeks, in writing

Assay design to state

• Design: acute activity or network formation
• Seeding: cells per well, neuron-to-astrocyte ratio, droplet volume and concentration
• Maturation target: days in vitro at first dosing, and the network metric that defines “mature” for this study
• Endpoints: mean firing rate / active electrodes / network burst frequency / synchrony index, named explicitly
• Positive control compound and expected direction of effect

Commercial

• Use rights: internal research, fee-for-service, or commercial derivative
• Certificate of analysis and QC panel included at this price
• Remedy for a lot failing the viability specification, and the claim window in days
• Repeat-order price for the same specification in three months

If you are outsourcing, add: raw per-well traces supplied as a deliverable, the recording schedule in days in vitro, who owns the data, and what happens if the culture fails to network.

Where we fit

We put that specification to named suppliers, normalise the responses onto one unit and one counting convention, and check lot availability and plate stock before you raise a purchase order — because on this assay a plate stock-out does not cost you a plate, it costs you the maturation window.

We do not manufacture and we do not resell any platform, which is why we can tell you when the right answer is to co-culture and halve the neuron order, to drop from 96-well to 24-well because your endpoint needs the electrodes, or to send the study out rather than buy an instrument you cannot commission before the deadline.