Studying intestinal barrier function

The intestinal barrier cannot be captured by a single measurement.

It is an integrated biological system formed by epithelial cells and their junctional complexes, mucus, immune and stromal compartments, microbial communities, vascular and lymphatic interfaces, and the signalling networks that connect them. Barrier function emerges from the behaviour of these components together.^¹,²

This creates a methodological challenge.

A study may measure transepithelial electrical resistance, molecular flux, tight-junction proteins, gene expression, circulating biomarkers, microbial composition or histological integrity and describe the result as a measure of “barrier function”.

These observations are related.

They are not equivalent.

The central methodological question is therefore not simply whether a study has measured the intestinal barrier.

It is:

Which dimension of barrier biology was measured, in which model, and what does that measurement actually allow us to conclude?

That distinction becomes increasingly important as barrier research moves from controlled experimental systems towards human disease.

No single model is the intestinal barrier

Experimental models exist because biological systems must be simplified before individual mechanisms can be investigated.

The advantage of simplification is control.

Its cost is context.

A cell-culture model can isolate epithelial transport from much of the surrounding biology. An organoid can reproduce important features of epithelial organisation while excluding many systemic influences. An animal model can preserve interactions among the intestine, microbiota, immunity, metabolism and other organs, but introduces questions of species translation. Human studies provide direct clinical relevance while sacrificing much of the experimental control available in laboratory systems.^¹–³

None of these approaches is inherently sufficient.

They answer different questions.

The methodological strength of barrier science comes from understanding those differences rather than treating results from different models as interchangeable.

Cell monolayers isolate epithelial function

Epithelial monolayers have been central to the study of intestinal transport and tight-junction biology.

Models such as Caco-2 cells can form polarised monolayers with apical and basolateral compartments and develop measurable transepithelial resistance. This makes it possible to manipulate nutrients, microbial products, cytokines or candidate therapeutics under highly controlled conditions and examine their effects on epithelial permeability.^³

Two common functional approaches are transepithelial electrical resistance (TEER) and molecular flux.

TEER measures electrical resistance across an epithelial monolayer. Changes in resistance can indicate alterations in ionic conductance across the epithelium, particularly through paracellular pathways.

Flux assays ask a different question. Labelled molecules of defined size are placed on one side of the monolayer and their passage to the other side is measured.

The distinction matters because the paracellular pathway is not a single homogeneous route.

Changes affecting ion-selective pores may alter electrical resistance without producing equivalent changes in the passage of larger molecules. Conversely, alterations in the leak pathway may influence macromolecular flux differently from ionic conductance.^⁴,⁵

TEER and molecular flux therefore provide complementary rather than interchangeable information.

Reductionism is both the strength and limitation of cell models

The experimental precision of a monolayer is also its major limitation.

An epithelial cell line does not reproduce the full intestinal environment.

It lacks much of the cellular diversity of the native epithelium. It does not reproduce the normal organisation of mucus, immune cells, vasculature, enteric nerves and stromal cells. Its interaction with the microbiota is necessarily artificial, and immortalised cell lines may differ substantially from primary human epithelial cells.^³

A result obtained in a cell monolayer may therefore establish that a biological mechanism can occur under defined experimental conditions.

It does not establish that the same mechanism dominates in an intact human intestine.

This distinction between mechanistic possibility and physiological importance is fundamental to translational interpretation.

Organoids restore epithelial complexity

Intestinal organoids changed what could be studied experimentally by allowing primary intestinal epithelial tissue to be maintained in three-dimensional culture.

The landmark demonstration that individual Lgr5-positive intestinal stem cells could generate crypt–villus-like structures provided the foundation for much of modern intestinal organoid biology.^⁶

Human intestinal organoids can preserve multiple epithelial lineages and important characteristics of the tissue from which they originate. They can therefore model epithelial physiology and disease in a way that conventional immortalised monolayers cannot. Recent reviews have highlighted their use in studying absorption, secretion, barrier function, inflammatory signalling and host–microbiome communication.^⁷,⁸

This is a major advance.

But an organoid is not a miniature complete intestine.

Conventional epithelial organoids generally lack the immune, stromal, vascular and neural compartments of native tissue. They do not automatically reproduce normal microbiota, luminal flow, peristaltic forces, oxygen gradients or systemic signalling.^⁷

Their biological relevance therefore depends on the question being asked.

Co-culture restores relationships

The limitations of epithelial-only organoids have driven development of increasingly sophisticated co-culture systems.

Organoids can now be combined experimentally with microbes, immune cells, stromal cells and neural components. A 2025 review by Al-Qadami and colleagues maps these approaches and their methodological trade-offs in detail.^⁹

This development matters because intestinal barrier function is relational.

Epithelial cells receive signals from immune cells.

Microorganisms and microbial metabolites influence epithelial and immune pathways.

Stromal cells help define the tissue environment.

Neural and endocrine signals participate in regulation.

A co-culture therefore asks a different question from an epithelial monoculture: not only what can this cell do?, but what happens when these biological compartments communicate?

Yet increasing model complexity does not remove uncertainty.

It changes it.

Every added cell type or microbial component introduces questions of source, ratio, timing, viability, culture conditions and reproducibility. Organoid–microbe systems must also confront the profound difference between oxygen requirements at the epithelial interface and those of many obligate anaerobes.^⁹

A more complex model can be more physiologically informative without becoming equivalent to the organism.

Barrier-on-chip models add physical context

Microfluidic intestinal systems add another dimension: physical physiology.

The intestinal epithelium in vivo exists under fluid flow, mechanical deformation, regional oxygen gradients and continual luminal exposure. Conventional static culture reproduces little of this environment.

Barrier-on-chip and intestine-on-chip systems can introduce flow, mechanical forces, separated tissue compartments and controlled gradients. Some platforms combine epithelium or organoids with endothelium, microbes or immune cells. Modern systems can also incorporate biosensors, advanced imaging and multi-omics measurements.^¹⁰

A 2024 review of microfluidic intestinal-barrier biosystems highlights how oxygen gradients, peristaltic-like forces, bioflow and multicellular architecture can be deliberately engineered rather than ignored.^¹⁰

These systems are powerful precisely because they allow selected dimensions of physiology to be reconstructed.

But the same principle applies:

greater engineering sophistication does not eliminate the need to define what the model represents.

An intestine-on-chip is not evidence merely because it looks more like an intestine.

Its validity depends on whether the biological feature relevant to the research question has been reproduced and measured appropriately.

Ex vivo tissue preserves native architecture

Ex vivo approaches provide another level of evidence.

Intestinal biopsies or excised tissue retain native epithelial organisation and important subepithelial structures that are lost in cell culture.

Ussing chambers have long been used to investigate transepithelial ion transport and molecular permeability across intact intestinal tissue. Electrical parameters can be recorded while molecular probes are used to assess passage across the tissue.^¹¹

This gives ex vivo systems an important advantage: permeability can be examined in native tissue architecture.

But excision also changes the system.

Blood flow is lost.

Neural and endocrine inputs are altered.

Systemic immune signalling is removed.

Tissue viability becomes time-dependent.

Preserving tissue architecture does not preserve the intact organism.

Again, the method determines the inference.

Animal models restore system-level biology

Animal models allow barrier regulation to be studied within an integrated organism.

Diet, microbiota, immunity, metabolism, endocrine signalling and behaviour remain connected. Experimental exposures can be controlled, tissue can be collected at defined time points, and mechanisms can be manipulated genetically or pharmacologically.

This has been directly relevant to my own research.

In wild-type and zonulin-transgenic mice, we investigated the prophylactic effects of bovine colostrum on intestinal microbiota and behaviour.^¹² The experimental design allowed microbial and behavioural outcomes to be examined in animals differing in zonulin-related barrier regulation. The work was conducted within the Mucosal Immunology and Biology Research Center at Massachusetts General Hospital and included the broader mechanistic context developed in collaboration with Alessio Fasano and colleagues.^¹²

In subsequent work, we examined intestinal transcriptomic responses in the same model.^¹³ That study included Takumi Konno and colleagues at MGB and identified treatment-associated transcriptional changes involving tight-junction and inflammatory signalling, with sex-specific patterns in the response.^¹³

These observations provide a level of mechanistic resolution that would be difficult to obtain in humans.

They also demonstrate why measurement level must remain visible in the conclusion.

A transcriptomic change demonstrates altered gene expression.

It does not itself demonstrate altered protein localisation.

It does not itself demonstrate changed paracellular flux.

And it does not itself demonstrate restoration of intestinal permeability.

Those are different biological claims.

Animal models do not remove the translation problem

An intact animal restores systemic interactions, but it introduces another inferential boundary.

A mouse is not a human.

Species differ in intestinal anatomy, immune responses, metabolism, microbial ecology, diet and environmental exposure. Genetically modified animals may deliberately isolate or amplify a pathway in ways that do not reproduce the distribution or magnitude of that pathway in humans.

The appropriate question is therefore not whether an animal model perfectly reproduces a human condition.

It rarely does.

The question is whether the model is appropriate for the mechanism being investigated and whether its findings converge with evidence at other experimental levels.

Animal studies are strongest when they are mechanistically informative and their translational limits remain explicit.

Human permeability tests move closer to physiology

Human studies provide direct relevance to human physiology, but intestinal barrier function remains unusually difficult to measure.

Functional permeability testing commonly relies on administration of probe molecules followed by measurement of their appearance in urine or blood. Sugar probes such as lactulose, mannitol, rhamnose and sucralose have been widely used, often in combinations designed to reduce the influence of non-mucosal factors.^¹⁴,¹⁵

The principle is elegant.

The interpretation is not always simple.

Probe size and physicochemical properties matter.

Absorption site matters.

Gastric emptying and intestinal transit matter.

Renal clearance matters.

Timing and completeness of urine collection matter.

Background dietary exposure may matter.

And the time window of sample collection can influence which intestinal region contributes most strongly to the result.^¹⁴,¹⁵

Recent methodological reviews continue to emphasise these issues rather than identifying a single universal permeability test.^¹⁵,¹⁶

A functional permeability assay is therefore valuable precisely because it measures molecular passage.

But it does not directly identify the molecular mechanism producing that passage.

Regional permeability matters

The gastrointestinal tract is not one homogeneous barrier.

The stomach, small intestine and colon differ in epithelial architecture, luminal environment, microbial density and transport function.

A permeability result therefore has an anatomical dimension.

The probe must reach and be absorbed within the region being investigated, and probes intended to assess colonic permeability must survive the microbial and chemical environment sufficiently to remain interpretable. Recent systematic methodological work has emphasised the importance of probe selection and collection timing for regional discrimination.^¹⁶

This matters when broad claims are made about “gut permeability”.

A result predominantly reflecting the small intestine should not automatically be interpreted as evidence about the entire gastrointestinal tract.

A biopsy describes the sampled tissue.

A functional probe describes the route and time window it can interrogate.

A blood biomarker may have still less anatomical specificity.

The anatomical scale of the conclusion should match the anatomical scale of the evidence.

Biomarkers do not all measure the same biology

Because direct functional permeability testing is demanding, substantial attention has shifted towards circulating and faecal biomarkers.

This is an area in which the literature has advanced rapidly.

Candidate blood-based markers include intestinal fatty acid-binding protein (I-FABP), citrulline, zonulin-related measurements, lipopolysaccharide-binding protein (LBP) and soluble CD14. But these markers do not represent one biological process. They may reflect epithelial injury, epithelial mass, microbial translocation, immune response or aspects of tight-junction regulation.^¹⁷

A major 2026 review by Wouda and colleagues makes this point explicitly: no single blood biomarker captures intestinal barrier function in its entirety, and complementary markers may be more informative when their biological meanings are kept distinct.^¹⁷

This is a critical methodological correction.

Calling several measurements “gut-barrier biomarkers” does not make them interchangeable.

I-FABP is not zonulin.

Zonulin is not LBP.

LBP is not a direct permeability probe.

And none should automatically be treated as equivalent to measured transepithelial molecular flux.

Zonulin illustrates the measurement problem particularly clearly

Zonulin-related measurements have become widely used in human studies because they offer an apparently simple route to investigating a complex biological process.

But the relationship between a reported zonulin concentration, the zonulin regulatory pathway and functional intestinal permeability is not one-to-one.

This was central to my systematic review and meta-analysis of zonulin-dependent intestinal permeability in children diagnosed with mental disorders.^¹⁸

The available studies suggested altered zonulin-related measurements in some populations, but methodological heterogeneity limited the strength of inference. Differences in biomarker concentrations could not simply be translated into evidence that functional intestinal permeability was altered, still less into evidence of a causal pathway from permeability to psychiatric phenotype.^¹⁸

The lesson extends beyond zonulin.

A biomarker becomes useful when we understand what biological process it reflects — not when we give it the name of the system we hope to measure.

Tissue, protein and transcriptomic measurements answer different questions

Barrier studies frequently combine functional assays with tissue morphology, immunohistochemistry, protein quantification and gene-expression analysis.

This can substantially strengthen a study.

But only if the measurements remain conceptually distinct.

Histology can reveal structural damage or tissue organisation.

Immunofluorescence can show where a junction-associated protein is located.

Protein assays can quantify abundance.

Transcriptomics can identify changes in gene expression and pathway activity.

Functional permeability assays determine whether movement across the barrier has changed under the conditions being tested.

These measurements may converge.

They may also diverge.

A tight-junction transcript may increase without a corresponding increase in protein.

Protein abundance may change without correct junctional localisation.

A molecular response may compensate for injury rather than cause it.

Functional permeability may change without dramatic histological disruption.

The absence of equivalence is not experimental noise to be removed.

It is biological information.

Microbiome measurements introduce another inferential layer

The microbiota is deeply integrated with barrier biology, but microbiome research adds a separate methodological domain.

16S rRNA sequencing provides information primarily about taxonomic composition.

Shotgun metagenomics provides greater taxonomic resolution and information about microbial genetic potential.

Metatranscriptomics asks which microbial genes are being expressed.

Metabolomics examines chemical products present within a biological sample.

These methods do not answer the same question.

Taxonomic composition does not directly demonstrate microbial function.

Genetic potential does not establish expression.

Expression does not establish that a metabolite reaches the epithelial surface at a biologically meaningful concentration.

And an association between microbial composition and permeability does not establish causal direction.

This distinction was relevant in our bovine-colostrum mouse study, where intestinal microbiota and behavioural outcomes were examined within a model of altered zonulin biology.^¹²

Observing changes across several systems can reveal biological relationships.

It does not automatically establish which system drove the others.

Human observational research magnifies the causality problem

The difficulty becomes greater in human observational studies.

Suppose a cross-sectional study reports:

• altered microbiota;
• higher inflammatory markers;
• increased concentrations of a barrier-associated biomarker;
• and greater symptom severity.

Several causal structures remain possible.

Microbial changes could influence barrier regulation.

Inflammation could alter both microbiota and barrier function.

Diet could influence all three.

Medication could alter microbial composition and epithelial physiology simultaneously.

Disease activity could produce the entire pattern.

Stress, age, adiposity, infection and other exposures could confound several variables at once.

Or some measured variables may simply correlate without lying on the same causal pathway.

Cross-sectional association cannot resolve these possibilities.

Longitudinal studies add temporality.

Intervention studies can strengthen causal inference.

Mechanistic experiments can test individual links.

But no study design rescues a poorly specified measurement.

Convergence is stronger than a single marker

Because no single method captures the complete intestinal barrier, stronger inference often comes from convergence across levels.

A molecular pathway identified in vitro becomes more compelling when reproduced in organoids or native tissue.

A junctional response in an animal model becomes more informative when accompanied by functional permeability measurements.

A human biomarker association becomes stronger when it aligns with direct functional testing and a plausible mechanism.

A microbiome association becomes more informative when microbial function or metabolites are measured rather than inferred solely from taxonomy.

This is not an argument for measuring everything.

It is an argument for choosing complementary measurements that interrogate the same biological question from different directions.

Convergence should increase explanatory resolution.

It should not be used to blur methodological differences.

New technologies do not abolish old inferential problems

Barrier science is becoming technologically sophisticated.

Human organoids, co-culture systems, microfluidic chips, biosensors, advanced imaging, spatial methods and multi-omics can now generate levels of biological detail that were impossible a decade ago.^{⁷–¹⁰,¹⁶}

This is enormously valuable.

But technological resolution and inferential validity are not the same thing.

A multi-omic dataset can still be cross-sectional.

A sophisticated chip can still model only selected compartments.

An organoid can still lack immune or microbial context.

A biomarker panel can still contain markers of biologically different processes.

More data do not automatically produce a stronger conclusion.

The central question remains unchanged:

What was actually measured, and how far does that observation allow us to go?

Model, measurement, mechanism and meaning

A useful way to read intestinal-barrier research is to keep four levels separate.

Model

Where was the observation made?

Was it an immortalised epithelial cell line, a primary-cell system, an organoid, a co-culture, an intestine-on-chip platform, excised tissue, an animal model or a human population?

Each preserves some aspects of intestinal biology and removes others.

Measurement

What was actually quantified?

Electrical resistance, molecular flux, tissue morphology, junctional localisation, protein abundance, gene expression, microbial composition, microbial metabolites, circulating biomarkers and urinary probe recovery are not alternative names for the same variable.

They measure different things.

Mechanism

What biological process does the evidence directly support?

Does the experiment demonstrate altered paracellular transport?

Does it identify a signalling pathway capable of modifying junctional organisation?

Does it show epithelial injury?

Does it demonstrate microbial translocation?

Or does it identify an association from which a mechanism is being inferred?

Meaning

How far can the observation legitimately travel beyond the model and measurement in which it was generated?

A mechanistic effect in a cell system may justify further investigation in intact tissue.

An animal experiment may demonstrate biological plausibility and causal mechanisms under controlled conditions.

A human observational study may demonstrate an association in the population studied.

A functional permeability test may demonstrate altered molecular passage under defined test conditions.

None automatically supplies the evidence required by the next level.

This separation between model, measurement, mechanism and meaning is not semantic.

It determines the strength of the scientific conclusion.

The problem of translation

Translation is sometimes described as movement from bench to bedside.

For intestinal-barrier research, that image is too simple.

Evidence does not move along one straight line from cell culture to animal to human.

Different models answer different questions, and findings may move backwards as well as forwards between them.

A human association may generate a mechanistic hypothesis that is then tested experimentally.

An unexpected response in an animal model may lead investigators back to epithelial signalling.

An organoid derived from a patient may allow a clinical phenotype to be examined under experimentally controlled conditions.

Translation is therefore better understood as an iterative process of convergence and challenge.

Evidence at one level informs another.

It should not be used to replace it.

This becomes especially important when the biological narrative is compelling.

The microbiota can influence epithelial and immune signalling.

The intestinal barrier can regulate exposure of the host to luminal material.

Systemic immune and metabolic signals can communicate with distant organs, including the brain.

Each of these propositions has experimental support.

But combining them into a complete causal sequence requires evidence for the connections between them.

A plausible pathway is a hypothesis until those connections have been demonstrated.

Negative and discordant findings matter

Convergence is scientifically valuable, but so is disagreement.

If gene expression changes while functional permeability does not, the discrepancy may reveal compensation, timing or a difference between molecular regulation and physiological outcome.

If a circulating biomarker changes without corresponding alteration in a direct permeability test, that may indicate that the biomarker reflects another component of barrier biology.

If an effect observed in a cell model disappears in an organoid or intact animal, the additional biological context may be modifying the mechanism.

If an animal finding does not translate into humans, that does not necessarily mean the animal experiment was wrong.

It may reveal where the biology differs.

Discordant findings should therefore not automatically be treated as failed replication or inconvenient noise.

They can identify the boundaries of a mechanism.

That is particularly important in a field where several related but non-equivalent measurements are frequently grouped under broad terms such as barrier function, gut integrity or leaky gut.

Precision sometimes means allowing the data not to converge.

A living barrier requires dynamic methods

The intestinal barrier is not static.

Epithelial cells are continually renewed.

Junctional complexes are remodelled.

Immune activity changes.

Microbial communities and their metabolic output vary with diet, medication, infection, environment and time.

Permeability itself can change physiologically.

A single measurement therefore represents a biological state within a temporal process.

This has consequences for experimental design.

Repeated measurements can reveal trajectories that a single endpoint cannot.

Longitudinal human studies can distinguish persistent differences from transient fluctuations.

Time-course experiments can show whether a molecular response precedes, accompanies or follows a functional change.

Continuous electrical monitoring can identify disruption and recovery that would be invisible in a single endpoint measurement.

Temporal resolution is therefore another dimension of barrier measurement.

The question is not only what was measured.

It is also when.

The future is multidimensional — but interpretation must remain disciplined

The next generation of intestinal-barrier research will increasingly combine experimental platforms and data types.

Organoid systems can incorporate immune, microbial, stromal and neural components.^⁹

Microfluidic models can add flow, mechanical forces and controlled gradients.^¹⁰

Continuous impedance measurements can capture barrier dynamics rather than single endpoints.^⁸

Human studies can combine functional permeability tests with biomarkers, microbial profiling, metabolomics and longitudinal phenotyping.^{¹⁶,¹⁷,¹⁹}

These developments make it possible to ask better questions.

They also create a new risk.

As datasets become richer, it becomes easier to construct convincing biological narratives from correlated measurements.

A multidimensional dataset does not abolish the distinction between association and mechanism.

Nor does a machine-learning model or multi-omic signature automatically identify causal direction.

The sophistication of the analytical method should therefore increase, rather than reduce, the discipline applied to biological interpretation.

Studying the barrier means respecting its complexity

There may never be one measurement that can be called the measure of intestinal-barrier function.

That may be the wrong objective.

The intestinal barrier is not one variable.

It is a living interface whose behaviour emerges from epithelial structure, selective transport, junctional regulation, immune activity, microbial interaction, tissue physiology and systemic context.

Different methods allow different parts of that system to become visible.

The scientific task is to choose the method that matches the question, understand the assumptions built into it, and resist extending the conclusion beyond what the experiment has actually demonstrated.

That is why methodology belongs at the centre of barrier science rather than at its margins.

The method does not merely produce the result. It defines what the result has earned the right to mean.