Buying corneal and ocular tissue: four validated models and what separates them

The short answer: four commercially available corneal models are named in the current OECD eye irritation guideline, and each one has a different pass mark. EpiOcular passes a chemical as unclassified above 60 percent mean tissue viability. SkinEthic HCE passes above 60 percent for liquids but 50 percent for solids. LabCyte CORNEA-MODEL24 uses 40 percent. MCTT HCE uses 35 percent for liquids and 60 percent for solids.

Those numbers are not interchangeable, and a result quoted without naming the model and the protocol is not interpretable.

The second thing to know, and the guideline is blunt about it: these methods cannot tell you a chemical is an irritant. They identify chemicals that do not require classification. A low viability result yields no prediction on its own — the guideline says so in those terms — because the methods cannot resolve UN GHS Category 1 from Category 2, and because RhCE methods show a high false-positive rate. If you are buying corneal tissue expecting a positive answer, you are buying the wrong assay.

The third thing: the supplier is required to prove each batch, and you are entitled to that evidence. TG 492 states data demonstrating compliance with all production release criteria should be provided by the tissue developer to the test method user so it can go in the test report. Ask for it by name.

The four models named in the guideline

Note the first row carefully. EpiOcular is built from primary human epidermal keratinocytes — skin cells, not corneal cells. That is what the guideline states. It is a cornea-like epithelium: the guideline’s own term is “reconstructed human cornea-like epithelium”, and the requirement is that the construct is similar to the in vivo corneal epithelium in three-dimensional structure, non-keratinised, with at least three layers of viable epithelial cells. Two of the four models use genuine corneal cells, one uses an immortalised corneal line, and the reference method uses keratinocytes.

For hazard identification this is validated and accepted, because the endpoint is cytotoxicity from chemical penetration and the model was proven against that endpoint. For ophthalmic drug delivery, corneal wound healing, dry eye or corneal infection research, cell provenance matters far more, and that is a different product — MatTek catalogues EpiCorneal separately for exactly those applications, distinct from EpiOcular.

The procurement consequence: never buy “the eye model”. Buy either the regulatory hazard identification model or the corneal biology research model. They are different products from the same supplier and they are not substitutes.

The prediction models, exactly as the guideline states them

This is Table 4 of TG 492, reproduced verbatim in substance. The right-hand column is the one people misread.

Three things follow that change how you should plan a programme.

The cut-off is model-specific and protocol-specific. The same chemical returning 45 percent viability is “no prediction” on EpiOcular and “No Category — no further testing required” on LabCyte CORNEA-MODEL24. Choosing the model is choosing the pass mark. That is not gaming the system — each cut-off was set by that method’s own validation — but it does mean a model switch mid-programme invalidates comparability.

Below the line is not a positive result. The guideline states that if mean viability falls at or below the cut-off, no prediction can be made from this result in isolation, and further information is required for classification. Reporting such a result as “irritant” overstates what the method delivers.

Borderline results have a defined procedure. The guideline states that a single test of at least two tissue replicates suffices when the result is unequivocal, but for non-concordant replicates or mean viability at 60±5% (VRM1, VRM2 liquids, MCTT solids), 50±5% (VRM2 solids), 40±5% (LabCyte) or 35±5% (MCTT liquids), a second test should be considered — and a third if the first two disagree. Budget for that. A borderline chemical is a three-run chemical, and if your quote assumed one run per compound your cost model is wrong for exactly the compounds you care most about.

QC batch release: the numbers your supplier must hit

TG 492 publishes the acceptance ranges each tissue developer must meet to release a batch. This is the strongest published quality specification in any tissue category on this site, and it is enforceable because it is in a guideline rather than a brochure.

How to use this in procurement. The guideline states the RhCE construct should only be used if the developer demonstrates that each batch meets defined production release criteria, of which viability and barrier function are the most relevant, and that compliance data should be provided to the test method user for inclusion in the test report.

So the batch certificate is not a courtesy document. It is a guideline requirement, and the specific question to ask is: what was the measured ET50 or IC50 for this batch, and where does it sit inside the published acceptance range? A batch at the very edge of the range is compliant, and it is also information you want before a borderline chemical lands on it.

The barrier requirement exists for a specific reason the guideline spells out: the construct must resist rapid penetration of cytotoxic benchmark substances, and its containment properties must prevent passage of the test chemical around the edge of the viable tissue. Leakage around the insert edge produces a false positive that looks exactly like cytotoxicity.

On morphology, the guideline requires histology showing cornea-like structure with at least three layers of viable epithelial cells and a non-keratinised surface, and states this has been established by the developer and does not need re-demonstration by the user per batch.

Exposure protocols differ per model — and per physical form

The liquid and solid protocols are separate methods, not settings. From the guideline:

• Liquids. Tissues are incubated for 1 minute (LabCyte CORNEA-MODEL24), 10 minutes (MCTT HCE), or 30 minutes (EpiOcular and SkinEthic HCE) under each method’s standard conditions. For the two VRMs and MCTT HCE, exposure is at 37±2 °C, 5±1% CO₂, ≥95% relative humidity; the LabCyte liquid protocol is 1 minute (±5 seconds) at room temperature.
• Solids. Substantially longer — 6 hours (±0.25 h) for EpiOcular, 4 hours (±0.1 h) for SkinEthic HCE, and up to 24 hours in the LabCyte solids protocol.
• Surface moistening. For the two VRMs and MCTT HCE, the construct surface is moistened with calcium- and magnesium-free DPBS before application, to mimic the wet conditions of the human eye.
• Dose. A sufficient amount to uniformly cover the epithelial surface while avoiding an infinite dose — the guideline defines an infinite dose as an amount exceeding what is needed to completely and uniformly cover the surface.
• Replicates. At least two tissue replicates per test chemical and per control substance in each run.
• Positive control response, which the guideline uses to prove the tissue can respond: methyl acetate (VRM1, VRM2, MCTT HCE), ethanol (LabCyte liquids) or lauric acid (LabCyte solids), with required mean viability < 50% for VRM1, ≤ 30% (liquids) or ≤ 20% (solids) for VRM2, ≤ 35% for MCTT HCE and ≤ 40% for LabCyte.

MatTek’s own published protocol for the EpiOcular EIT matches this: N=2 tissues per condition, liquids at a 30-minute topical exposure to 50 µL, solids at a 6-hour topical exposure to 50 mg of test material per tissue.

The tissue-count consequence. Two replicates per chemical, plus negative and positive controls per run, plus a probable second run on borderline chemicals. A screen of twenty compounds is not forty tissues. Work the arithmetic against kit size before requesting a quote — as in every other living-tissue category, the kit is the purchasable unit, not the tissue.

What these models explicitly cannot do

The guideline is unusually direct, and this section exists because the limitation is routinely oversold.

They cannot replace the in vivo test on their own. TG 492 states the methods cannot be used alone to replace the Draize eye test across the full range of responses, and recommends testing strategies such as those in TG 467 and TG 492B.

They cannot distinguish Category 1 from Category 2. Serious eye damage and reversible eye irritation are not separable by these methods. That differentiation requires another tier — the guideline names TG 437, 438, 460, 491, or as a last option TG 405.

They are designed for one direction of inference. They sit as an initial step in a Bottom-Up approach — where a chemical is expected not to require classification — or as one of the last steps in a Top-Down approach. Used outside that logic, a result is not meaningful.

They have a high false-positive rate, which the guideline states plainly as one reason a below-cut-off result yields no prediction.

If your programme needs a definitive irritancy category, the corneal model is one tier of a strategy and you should budget for the tier above it from the start.

Beyond hazard identification: corneal models as research tissue

Regulatory eye irritation is where the published specifications live, but it is not the whole category. MatTek catalogues EpiCorneal as a distinct 3D corneal model positioned for ophthalmic drug delivery, wound healing and tissue regeneration, disease modelling such as dry eye, and corneal infection — a non-animal alternative for questions that TG 492 does not address.

Three things change when you move from hazard identification to research use.

Cell provenance becomes central. For a delivery or wound-healing study, whether the construct is built from corneal epithelial cells or from keratinocytes is no longer a validated technicality — it determines the barrier chemistry, the transporter complement and the healing response you are measuring.

There is no published acceptance criterion. The ET50 and IC50 ranges above exist because a guideline demanded them. Outside the guideline you are back to the general condition of this market: ask what is measured at release, in what units, with what range, and expect that the answer may be qualitative.

Longer culture and repeat dosing become the question. Hazard identification is an acute exposure measured in minutes or hours. Delivery and disease modelling need days. Ask for the maintenance protocol and the validated culture duration, as you would for airway tissue, where suppliers publish months.

What a corneal model purchase does not include

• The MTT, WST-8 or WST-1 reagent and the extraction solvent, which differ by method.
• Control substances. Methyl acetate, ethanol, lauric acid, SDS or Triton X-100 as the method requires, plus the negative control.
• The HPLC/UPLC route for coloured or MTT-interfering chemicals, which the guideline accommodates and which is a separate analytical capability.
• The second and third runs for borderline results.
• The higher tiers of the testing strategy — TG 437, 438, 460 or 491 — when a result falls below the cut-off.
• Study execution, if you do not run it yourself. Contract laboratories offer the EIT as a service, and for a handful of chemicals a year that is usually the cheaper route.

Specification checklist for a corneal tissue quote

• Which of the four test methods — this fixes the cut-off, the protocol and the dye.
• Liquids or solids protocol, since exposure time and, for two models, the cut-off both change.
• Batch ET50 or IC50 measured value, and its position within the published acceptance range.
• Written confirmation that production release criteria were met, supplied for the test report as the guideline requires.
• Tissue count arithmetic: chemicals × 2 replicates + controls per run + contingency for borderline repeats.
• Kit size and format, and whether partial kits are available.
• Delivery date against the production calendar, and the shelf life on arrival — see production calendars and shelf life.
• Whether the chemical is coloured or MTT-interfering, which changes the readout route.
• For research rather than regulatory use: the actual cell source, the validated culture duration, and what is measured at batch release.
• Whether you are buying tissue or a study, and if a study, whether the report is written to the guideline’s reporting requirements.

Where to go next

• Engineered tissue map — where ocular sits among the availability clusters.
• OECD TG 439 and RhE model requirements — the same regulatory logic applied to skin, where the acceptance criteria are also published.
• Oral, gingival and mucosal models — the adjacent non-keratinised epithelia, mostly outside the guideline system.
• Airway and respiratory models — published TEER release criteria and much longer culture windows.
• Buying reconstructed human epidermis — the most mature validated-model category.

How we can help

The expensive mistake here is a screen designed around one tissue per chemical, which collides with the two-replicate minimum, the control burden per run and the repeat-testing rule for borderline results — a collision that surfaces after the kit has been bought. The second is reporting a below-cut-off result as a positive, which the guideline does not support and a reviewer will catch.

We size the tissue count against the real protocol before anything is ordered, confirm which method’s cut-off applies to the physical form of your chemicals, get the batch ET50 or IC50 and the release documentation requested up front rather than after the run, and compare running it in-house against buying it as a contract study, which for small compound sets is usually the cheaper answer.

Tell us the number of chemicals, whether they are liquids or solids, whether any are coloured, and whether you need a regulatory submission or a research answer.