Buying intestinal tissue models: planar epithelium, crypts, organoids and perfused tubules

The short answer: the gut is the one tissue category where you must choose a geometry before you choose a supplier. A sealed organoid sphere, a flat monolayer on a transwell, and a perfused tubule are three different physical objects made from the same donor stem cells, and they support mutually exclusive assays. Pick the shape first. The supplier list follows from it.

The second thing to know: regional identity is a purchasable specification here, and it is not in most people’s RFQ. One supplier states it supplies six regions — duodenum through descending colon, each with its own biology. Duodenum and descending colon differ in transporter expression, enzyme complement and immune tone. “Intestinal model” is not a specification.

The third thing: the default comparator is a 1970s colon carcinoma line, and every supplier in this category is selling against it. Caco-2 is the incumbent, and the honest way to read vendor material here is as a case against Caco-2 — which may well be correct, and is still a sales argument.

What is actually available

The geometry decision, made concrete

This is the part of gut sourcing that determines everything downstream, and the clearest statement of it comes from Altis, which is explicitly selling against organoids:

Organoids grow in three dimensions as closed spheres with uneven shapes and inconsistent surface areas. That makes robust, repeatable assays hard to build and hides the apical surface inside the sphere.

That is a vendor argument, and it is also physically true. The apical surface of an intestinal organoid faces the lumen, and the lumen is on the inside of the sphere. To dose the apical surface of a standard enteroid you must microinject it, invert its polarity, or fragment it — all of which are real techniques and none of which are high-throughput.

The three shapes and what each one is for:

Flat monolayer on a transwell. Both surfaces are open. You dose one side and sample the other, which is exactly what absorption, secretion, efflux and barrier integrity measurements require. Altis states RepliGut Planar runs in 96, 24 and 12-well transwell plates, which is the format most laboratories already own. This is the right geometry for permeability, transporter and efflux work, and for anything where you need TEER as an inline control.

Sealed organoid sphere. Self-organising, crypt–villus architecture, multiple differentiated lineages, and the closest thing to native tissue architecture available in a dish. It is the right geometry for developmental biology, lineage and stem-cell questions, disease modelling from patient material, and screening where the readout is imaged or lysate-based rather than transported. It is the wrong geometry for a transport assay unless you invert or inject it. See organoids versus engineered tissue for the full distinction.

Perfused tubule. MIMETAS grows adult-stem-cell-derived colon organoids as tubules in the OrganoPlate, retaining apical and basal access while adding flow. This is a genuine third thing: the architecture is tubular rather than spherical or flat, and the tissue experiences shear. It is the right geometry when flow is part of your biological question and the wrong one when you need a plate-reader-simple barrier assay.

How to choose in one question: does anything have to cross the tissue and be measured on the other side? If yes, you need open apical and basal compartments — planar or perfused tubule. If no, the sphere is available to you and brings better architecture.

The Caco-2 comparison, recorded as a supplier claim

Altis publishes a direct comparison against Caco-2. It is reproduced here as the supplier’s published positioning, not as an independent benchmark — we have not tested any row of it. It is worth reading because it names the axes on which any intestinal model should be interrogated.

How to use this table honestly. Every row is a legitimate question to put to any supplier, including Altis, with a request for the underlying data. Three of them are load-bearing:

• The esterase profile. Caco-2’s non-human esterase activity is a well-known problem for ester prodrugs specifically. If your compound is an ester prodrug, this is not a marginal consideration — it can invert your absorption conclusion. If it is not, the row matters much less to you.
• CYP3A4. Low CYP3A4 in Caco-2 is the standard reason intestinal first-pass metabolism is under-predicted. If your question involves gut-wall metabolism rather than passive permeability, a model with functional CYP3A4 is answering a different and more relevant question.
• Time to barrier. Three weeks versus roughly one is a real operational difference — three weeks of incubator occupancy, feeding and contamination exposure per assay plate. Over a screening campaign that compounds into a schedule difference, not just a cost one.

And the row that cuts the other way: Caco-2 is the incumbent because of the comparability of two decades of published data. A regulatory or internal-decision context where the historical Caco-2 dataset is the reference frame is a context where switching models costs you comparability. That is not a scientific argument for Caco-2 — it is a real institutional one, and it is why MIMETAS catalogues a perfused Colon Caco-2 product alongside its organoid-derived one.

Regional identity: the specification most people omit

Altis states RepliGut supports six regions, duodenum through descending colon, each with its own biology. This is unusual and it is worth using.

The intestine is not a uniform tube. Bile acid handling, active transporter expression, CYP3A4 gradient, immune cell density, mucus thickness and microbial load all vary along its length. Consequences that show up in real programmes:

• Oral absorption is dominated by the proximal small intestine — duodenum and jejunum. Modelling it in a colonic model is a category error, and Caco-2 is colonic in origin.
• Colonic delivery, IBD and microbiome-adjacent work belongs in ileum or colon.
• Bile-acid and lipid handling is regionally specific, concentrated in the distal ileum.
• Regional toxicity — the classic pattern of GI injury from a compound hitting one segment and sparing another — is invisible in a single-region model, and it is precisely what a multi-region panel is for.

Practical rule. If you are running a GI toxicity screen and your compound produced a signal in vivo in one segment, buying the matching region is the single highest-value specification decision in this category. If you are running absorption, name the proximal small-intestinal region rather than accepting a colonic default.

Time to a working barrier is a scheduling number

Altis states 7 to 8 days from thaw to a working polarised epithelium. MatTek’s EpiIntestinal and MIMETAS’s OrganoReady models ship ready to use, moving the growth window entirely onto the supplier.

Three different procurement models sit behind those statements, and the right one depends on how often you run the assay:

You grow it, from a kit. Time cost is 7–8 days of bench and incubator time per run, plus the technical competence to do it reproducibly. Advantage: you control timing entirely, and you can start when you want rather than when a production slot exists. This suits a laboratory running the assay continuously.

It arrives finished. No growth window, no growth risk, and the supplier absorbs the variability. Advantage: the schedule is a delivery date. Disadvantage: you inherit the production calendar and the shelf life, which is the constraint described in production calendars and shelf life. This suits occasional or campaign-based use.

You buy the study. Altis, MatTek and MIMETAS all run services arms — permeability, transporter and efflux, metabolism, drug–drug interaction, inflammation screening and toxicity. If you need the answer rather than the capability, and you need it once, the deliverable is data and the comparison is against a CRO rather than against a plate price.

The break-even is genuinely a run-rate question. One or two studies a year is a services purchase. Weekly assays are a kit purchase. The awkward middle — a handful of campaigns a year — is where buying finished tissue makes most sense, and it is also where the production calendar hurts most.

Barrier control: TEER is the inline acceptance test

Altis states transepithelial resistance confirms a tight, intact barrier before and after each incubation. That phrasing is the correct practice and worth adopting regardless of supplier.

Before the incubation, TEER establishes that the monolayer was intact when dosing began. Without it, a high apparent permeability is ambiguous — it may be transport, or it may be a hole.

After the incubation, TEER establishes that the barrier was still intact at the end, which separates genuine transport from compound-induced barrier damage. A compound that is cytotoxic to the epithelium will look highly permeable if you only measure the receiver compartment.

Both readings are needed and the second is the one most often skipped. For any permeability number you report, the paired pre- and post-TEER values belong in the record next to it. This is the same release-and-acceptance logic described for airway tissue, where the supplier publishes an explicit Ω·cm² release floor.

Mucus, immune cells and what the base model leaves out

The plain epithelial monolayer is a simplification in three specific directions, and each has a purchasable answer.

Mucus. Altis states RepliGut Planar has a mucus layer where Caco-2 does not, and separately catalogues intestinal mucus as a product. Mucus matters for particle and nanoparticle delivery, mucoadhesive formulations, bacterial adhesion, and the effective concentration reaching the apical membrane. A mucus layer can lower apparent permeability substantially and correctly.

Immune compartment. Altis catalogues RepliGut Immune Co-Culture and an InflammaScreen service. For IBD, cytokine-driven barrier disruption, or any immune-mediated endpoint, the epithelium-only model contains none of the cells that produce the phenomenon.

Flow. Static transwell culture has no shear. MIMETAS and CN Bio both address this by perfusion. Whether flow changes your answer depends on the endpoint — it matters more for endothelial and mechanotransduction biology than for passive permeability — and the decision framework is in organ-on-chip versus transwell versus static culture.

What is still missing across the whole category: the microbiome. Anaerobic bacterial co-culture against a living human epithelium is an active research area, not a catalogue item, and any supplier implying otherwise should be asked precisely what organisms, at what oxygen tension, for how long.

What a gut model purchase does not include

• The transwell plates, unless bundled — confirm whether 96, 24 or 12-well plates are supplied or assumed.
• The TEER instrument and electrodes, and a chamber geometry matched to your insert size.
• Analytical chemistry. A permeability study is an LC-MS/MS study with a tissue model attached. The bioanalysis is usually the larger line item and is frequently forgotten in the model budget.
• Reference compounds. Permeability controls, efflux substrates and inhibitors for P-gp and BCRP.
• The instrument, for the perfused route. A MIMETAS OrganoPlate or CN Bio PhysioMimix plate is a consumable for a platform you must already own or acquire.
• The cells, for the seed-it-yourself route. MIMETAS’s OrganoReady Collagen and CN Bio’s plates arrive without cells, and for gut work the cells are not a trivial line.
• Regional panels. Six regions is six products, not one.

Specification checklist for an intestinal model quote

• Geometry, decided first: planar monolayer, sealed organoid, or perfused tubule.
• Anatomical region — duodenum, jejunum, ileum, ascending/transverse/descending colon — named explicitly, not “intestinal”.
• Cell source: primary adult donor stem cells, iPSC-derived, or an immortalised line, and if primary, how many donors and whether donor identity is held constant across the study.
• Differentiated lineages present: enterocytes, goblet, enteroendocrine, Paneth — asked as a list, with the characterisation data.
• Mucus layer: present or absent, and if present, whether it is characterised.
• Transporter function: P-gp, BCRP, and the data demonstrating activity, not just expression.
• Metabolic competence: CYP3A4 and UGT activity, measured with a named substrate.
• Time from thaw or delivery to a usable barrier, in days.
• TEER acceptance range, and the instrument and chamber it was measured on.
• Plate format: 96, 24 or 12-well transwell, matched to your existing hardware.
• Immune co-culture, if the endpoint is inflammatory.
• Whether you are buying a kit, finished tissue, or a study — and for a study, who owns the data and the report format.
• Production lead time and, for finished tissue, the order-by date against your start date.

Where to go next

• Organoids versus engineered tissue — the geometry distinction in full, and why it decides which assays are possible.
• Engineered tissue map — where gut sits among the three availability clusters.
• Airway and respiratory models — the other barrier-tissue category, with published TEER release criteria.
• Organ-on-chip vs transwell vs static culture — whether perfusion is worth the lock-in for your endpoint.
• 3D tissue model suppliers compared — gut alongside skin, liver and airway.

How we can help

The expensive mistake in this category is running a proximal-absorption question on a colonic model because the colonic model was the one with a catalogue number, then discovering the regional mismatch after the bioanalysis is paid for. The second is buying an epithelium-only model for an immune-mediated endpoint.

We fix the geometry decision first — planar, sphere or tubule — then match the anatomical region to the actual biological question, ask each supplier for transporter and CYP activity data rather than expression data, check the plate format against the TEER hardware you already own, and put the kit route, the finished-tissue route and the buy-the-study route side by side on total cost including bioanalysis, which is the comparison that usually decides it.

Tell us the compound class, the segment of interest, the endpoint and whether you need the capability or just the answer.