Gut Homing and Tissue-Resident Memory: How T Cells Stay Put

A T cell does not choose which tissue to defend. The tissue chose it, during priming. A naive T cell has no tropism at all; it acquires one from the dendritic cell that activates it, and the instruction is written into which adhesion and chemokine receptors it puts on its surface. For the intestine that instruction is retinoic acid, and it produces one of the cleanest examples of tissue-specific immunity in mammalian biology — clean enough that it has yielded a gut-selective drug class.

Imprinting, entry, retention and residency. Open the interactive version to click any protein for its role and the matching validated reagent.

Imprinting is a swap, not an addition

Intestinal CD11c⁺ dendritic cells that have migrated from the lamina propria to the mesenteric lymph nodes express RALDH2 (ALDH1A2) and convert dietary vitamin A into retinoic acid. Signalling through RARα, together with TGF-β1, induces α4β7 and CCR9 on the CD4 and CD8 cells being primed.

The half that gets left out of most summaries is that the same signal reciprocally downregulates CCR7 and CD62L — the receptors a cell needs to enter peripheral lymph nodes. Gut imprinting is therefore a swap, not an addition. A cell cannot be optimised for the intestine and for the peripheral node network simultaneously, and the dendritic cell that sends it to the gut has taken away its access elsewhere. That trade-off is why vitamin A status affects mucosal immunity so specifically, and why it is a genuine variable in animal work rather than a footnote about diet.

Why anti-α4β7 is gut-selective and anti-α4 is not

α4β7 (LPAM-1) is a heterodimer: CD49d (α4) paired with integrin β7. Its ligand, MAdCAM-1, sits on intestinal post-capillary venules and essentially nowhere else. Block the pair and you block gut entry without touching trafficking anywhere else in the body.

Block α4 alone and you do not, because α4 has a second partner. Paired with β1 it forms VLA-4, which reads VCAM-1 — expressed widely, including on the blood–brain barrier. This is precisely the difference between vedolizumab and natalizumab, and the reason one carries a CNS risk profile the other does not. If you are dissecting a homing phenotype in a model, the anti-β7 versus anti-CD49d comparison is the experiment that separates "gut-specific" from "α4-dependent", and the two are not synonyms.

The chemokine arm follows the same logic. CCR9 reads CCL25 (TECK), made constitutively by small-intestinal epithelium. LFA-1 and ICAM-1 provide the firm-arrest step that every tissue shares, and CCR6–CCL20 plus CXCR3–CXCL9/CXCL10 layer inflammatory recruitment on top. CCR7 and CCL19 are the route the imprinted cell has given up.

Residency is the absence of an exit

Entry gets most of the attention, but the step that creates a resident cell is negative, and it is the most elegant part of the pathway.

KLF2 normally drives expression of S1PR1, and S1PR1 lets a cell follow the sphingosine-1-phosphate gradient out of tissue and back into lymph. It is the exit receptor. CD69, induced on activation in situ, binds S1PR1 and drives its internalisation; KLF2 falls; the exit receptor disappears from the surface and the cell is trapped. Nothing has been glued down. The cell simply cannot leave.

Everything else that defines a tissue-resident memory cell is built on that trapping. CD103 (αEβ7), induced by TGF-β, anchors the cell to epithelial E-cadherin. CD49a binds collagen IV and marks the most sessile subset. CD101 and PD-1 mark a restrained phenotype — and PD-1 here is worth reading carefully, because constitutive PD-1 on a resident cell is a barrier-tolerance mechanism, not necessarily exhaustion. A permanently armed cytotoxic cell living inside an epithelium needs a brake. Runx3 and Blimp-1 hold the transcriptional programme, and IL-15 keeps the cell alive.

The consequence nobody likes

TRM cells do not recirculate. That is the definition. It follows that blood immunophenotyping systematically misses the population that does most of the work in mucosal tissue, and that a peripheral readout can be flat while the tissue response is substantial — or vice versa.

This has a practical edge in vaccine and infection work. A parenteral immunisation that generates excellent circulating memory may generate very few resident cells at the mucosal surface where the pathogen actually arrives, and a blood-based correlate will not reveal the gap. If the question is mucosal protection, the tissue has to be sampled.

What the resident population actually does

Residents are the fastest responders in the tissue because they are already there and already armed. On re-encounter they produce granzyme B, IFN-γ, IL-17A and IL-22. IL-22 is the interesting one: it drives epithelial proliferation, antimicrobial peptide production and MUC2 — repair rather than killing. The same population is therefore protective in infection and pathogenic in colitis, depending on what is triggering it and how often. CD25⁺ regulatory cells share the niche and set the threshold at which any of it fires.

The same architecture, different tissues

Gut homing is the best-worked example, not a special case. The pattern generalises: a tissue-associated dendritic cell imprints a tissue-specific receptor pair, the tissue endothelium expresses the matching ligands, and the same CD69/S1PR1 switch converts arrivals into residents.

In skin, the imprinting signal is vitamin D rather than vitamin A, and the receptors are CCR10 and CCR4 rather than CCR9. In lung, recruitment leans on the CXCR3–CXCL9/CXCL10 axis, and residency is CD103-associated much as it is in gut epithelium. The retention step is the part that does not change: CD69 sequestering S1PR1 appears to be how residency works everywhere it has been looked for.

That conservation is why the pathway has a therapeutic mirror image. S1PR1 modulators — fingolimod and the newer selective agents — work by acting on the same receptor from the opposite direction, trapping lymphocytes in lymph nodes so they never reach tissue at all. One drug class blocks the address on the envelope; the other stops the letter leaving the sorting office.

Resident memory inside tumours

The reason this pathway has drifted into oncology is that tumour-infiltrating lymphocytes with a resident phenotype — CD103⁺, CD69⁺, often CD49a⁺ — are, across several solid tumour types, the subset whose abundance tracks with outcome. They are also the subset most likely to respond when a checkpoint is blocked.

The mechanistic reading is straightforward once residency is understood as trapping. A resident cell has already found its antigen, already stopped recirculating, and already sits in contact with the tissue it is responding to. Its constitutive PD-1 is not a marker of a spent cell so much as the brake that a permanently armed tissue-dwelling cell requires — which is exactly why releasing that brake can produce a response in a population that a blood-based assay had written off.

It also sharpens the caution above. If the informative population does not recirculate, then peripheral pharmacodynamic markers are measuring a different set of cells from the ones doing the work.

Diet, microbiota and a variable that is easy to ignore

Because the imprinting signal is a vitamin A metabolite, mucosal immunity in an animal model is directly sensitive to diet. RALDH2 activity in intestinal dendritic cells depends on retinol supply, and the enzyme itself is induced by microbial and dietary signals in the gut — so chow formulation and microbiota composition both feed into how strongly a gut-homing programme is written.

The practical point is not that this invalidates comparisons; it is that "we changed facility" or "we changed chow supplier" is a plausible explanation for a shifted mucosal phenotype, and worth excluding before more exotic ones. The same applies to TGF-β1, which is required both for imprinting and, later, for CD103 induction in the tissue — one cytokine acting at two separate stages of the same journey, which makes systemic TGF-β blockade a blunt tool here.

Blocking entry and blocking retention are different experiments

This map has an unusually deep functional-grade layer — eleven In Vivo targets — and the reason to care is that they interrogate different steps. An intervention that prevents cells arriving tells you nothing about cells already resident, and a phenotype that survives entry blockade is usually a resident-cell phenotype.

Two details worth checking in your own model

MAdCAM-1 is inducible, not just constitutive. Baseline expression is intestinal, but inflammation raises it and can extend it beyond the gut. A homing phenotype measured in inflamed tissue is not necessarily reporting the constitutive route.

CD69 is an activation marker and a residency marker. Because it is induced by activation, CD69 positivity alone does not establish residency — a recently activated recirculating cell is CD69⁺ too. CD103, CD49a and demonstrated failure to exit are what separate the two.

For research use only. Not for use in diagnostic or therapeutic procedures.