Your Th2 Phenotype Might Not Be a T Cell

If your knockout showed a type 2 phenotype and you attributed it to Th2 cells, there is a reasonable chance you were looking at something else. Innate lymphoid cells run T-bet, GATA3 and RORγt — not analogues of the Th1, Th2 and Th17 master regulators but the same genes, doing the same jobs, producing the same cytokines — in cells that carry no antigen receptor at all. IFN-γ from an ILC1 and IFN-γ from a Th1 are the same molecule. The only way to tell them apart is to remove one and look again.

Innate lymphoid cells — 37 clickable nodes across the epithelial alarmins, the shared ILC identity, and the ILC1, ILC2 and ILC3 modules with their tissue outputs. Open the interactive version to click any protein through to its ELISA kit or In Vivo antibody.

The same master regulators, minus the receptor

Evolutionarily the ILC module came first and the T cell bolted a rearranging receptor onto it, which is why the three lineages map so exactly onto the three T-helper subsets. What ILCs keep instead of antigen specificity is speed and position: they are already differentiated, so there is no priming delay, and they sit resident in tissue rather than recirculating. IL-7 keeps the whole family alive through CD127, and that dependency is one of the few clean functional separations between ILCs and NK cells.

The shared surface identity is short: CD127, Thy-1, and depending on subset NK1.1, CD25, ICOS and KLRG1. There is no lineage marker in the conventional sense, which is why every ILC in every paper is gated as lineage-negative CD127-positive. That gating convention matters more than it sounds — it means the definition of the cell type is operational, and two labs using different lineage cocktails are not always counting the same thing.

One consequence of that surface phenotype is easy to overlook and has bitten more than one study. CD25 is constitutive on ILC2, unlike on resting T cells, so ILCs sit permanently ready to consume IL-2 and compete for it with regulatory T cells. Low-dose IL-2 therapy, designed to expand Tregs, expands ILC2 alongside them — which means a type 2 signal appearing after IL-2 treatment is not necessarily off-target noise, it is a predictable second effect of the same dose.

With no receptor, the epithelium chooses the response

This is the conceptual inversion worth carrying away. In adaptive immunity the antigen selects the responding clone. Here there is no clone to select, so what decides which module fires is whichever alarmin arrives from the barrier. IL-33 is stored pre-made in the nucleus of healthy epithelium and released on necrosis, which means the signal is cell damage rather than infection — no transcription delay, no processing step. TSLP reports a different insult and sits upstream of the type 2 programme, which is why tezepelumab works in asthma regardless of eosinophil count. IL-25 comes specifically from tuft cells, the rare chemosensory epithelial cells that detect helminths, and signals through IL-17RB.

For the type 1 module the instruction is IL-12 with IL-18; for type 3 it is IL-23 with IL-1β. One assay caution is worth stating plainly because it is among the most commonly misread analytes in this area: soluble ST2 is a decoy receptor. A rising serum sST2 means less IL-33 reaching the cell, not more. Reading it as a marker of increased IL-33 signalling inverts the result.

Type 2 is a repair programme that also causes allergy

Amphiregulin is the node that reframes what ILC2 are for. It restores epithelial integrity after influenza — the cells are a tissue-repair system. The same cell in the same state drives goblet-cell metaplasia and smooth-muscle contraction through IL-13, which is asthma. Nothing switches between those two readings; the context does. IL-5 is constitutive rather than induced in tissue ILC2, which is why baseline eosinophil counts track ILC2 activity and why anti-IL-5 biologics work on a cell type that is not an eosinophil.

IL-9 is autocrine here, and that detail has clinical consequences: ILC2 make IL-9 and respond to it, so a transient alarmin becomes a self-sustaining response. It is a mechanistic reason why blocking the alarmin late so often disappoints. IL-4 arrives largely from basophils as an amplifying input rather than as an ILC2 product, and CRTH2 — the prostaglandin D2 receptor — is how mast cells join the circuit, which is the link between this map and the mast cell and IgE map. For where the eosinophils themselves go next, see the eosinophil and IL-5 map.

Type 3 has the mirror-image problem

IL-22 acts only on epithelium — no immune cell carries the receptor — which makes it the cleanest example in immunology of an immune cell talking exclusively to a barrier. It induces REG3γ, an antimicrobial peptide that keeps bacteria physically off the epithelial surface rather than killing them. IL-17A comes from the same cell and is straightforwardly inflammatory. That shared origin is the therapeutic problem: blocking IL-17 in inflammatory bowel disease removed the protection along with the pathology, and trials worsened rather than improved disease.

Two further nodes on this module are worth knowing. AhR means diet and microbiota feed directly into immunity — ligands from cruciferous vegetables and from tryptophan-metabolising bacteria maintain ILC3, and removing them empties the compartment. And ILC3 express MHC class II while carrying no co-stimulation, so a T cell that engages them is deleted rather than primed: an antigen-presenting cell whose function is tolerance. GM-CSF is the output that recruits and licenses myeloid cells, connecting to the myeloid reprogramming map, while the Th17 side of the same axis is drawn in the Th17 and IL-23 map.

ILC1, NK cells, and a decade of confusion

ILC1 and NK cells share NK1.1, which is why the two were conflated for years and why a single depletion experiment cannot separate them. The useful distinction is developmental and functional rather than phenotypic: NK cells are cytotoxic circulating cells, ILC1 are tissue-resident cytokine producers, and ILCs are IL-7-dependent in a way NK cells are not. Under IL-12 and IL-18 an ILC1 will produce IFN-γ and TNF with no recognition event whatsoever, which is the purest available demonstration that these cells read cytokines rather than antigens. The cytotoxic side of the family is drawn separately in the NK cell activation map.

Plasticity complicates this further. TGF-β pushes ILC1 towards ILC3 and back, and IL-10 from regulatory populations damps the whole family, so a cell's module is a state rather than a fixed identity. In practice that means a single timepoint can misassign a cell that was something else a week earlier.

Which cell actually made your cytokine?

Because the effector molecules are shared with T-helper subsets, the practical question is almost never "is this cytokine present" but "which compartment produced it". The reagents on this map answer that directly, and this is the unusual case where the functional-grade antibodies are the method rather than an accessory to it.

The experiment that settles it

Twelve of the thirty-seven nodes carry a functional-grade In Vivo antibody, and two of them are not merely targets but the standard tools of the field. Anti-Thy-1 in a T-cell-deficient host is the experiment that assigned function to ILCs in the first place, and anti-CD127 is how the compartment is both defined and removed. Also blockable: NK1.1, CD25, IL-2, IL-1β, IL-4, IFN-γ, TNF, GM-CSF, IL-10 and pan-TGF-β.

What is not blockable in this range, said plainly: there is no functional-grade anti-IL-33, anti-ST2, anti-TSLP, anti-IL-25, anti-IL-5, anti-IL-13, anti-IL-17A, anti-IL-22 or anti-IL-23. The alarmins that start this pathway and the type 2 and type 3 cytokines that finish it are measurable analytes here, not intervention points. That constraint points at the better experiment anyway. Rather than neutralising a cytokine and inferring which cell made it, delete the cell and measure what disappears — which is precisely the design the CD127 and Thy-1 reagents support, and precisely the design that reassigned so many Th2 and Th17 phenotypes to this family in the first place.

For research use only; not for use in diagnostic or therapeutic procedures. Explore the full library of interactive pathway diagrams.