One Cytokine, Two Opposite Clinical Effects

Low-dose IL-2 expands regulatory T cells. High-dose IL-2 expands the cells they exist to restrain. Same cytokine, same receptor family, opposite clinical intent — and the explanation is a receptor subunit that performs no signalling whatsoever. CD25 raises the receptor's affinity for IL-2 roughly a hundredfold and contributes nothing to the intracellular signal. It is a capture subunit. A Treg wins the competition for scarce IL-2 simply by carrying more of a component that does nothing on its own. Raise the dose and that advantage evaporates, because the intermediate-affinity receptor on memory CD8 and NK cells is engaged too.

Regulatory T-cell biology and the IL-2 paradox — 38 clickable nodes from the receptor chains through the FoxP3 programme, the two suppression mechanisms, the TNF receptor family and the loss of the lineage. Open the interactive version to click any protein through to its ELISA kit, biosimilar or In Vivo antibody.

The paradox is a subunit that does not signal

The IL-2 receptor exists in two functional forms, and almost everything clinically interesting about IL-2 follows from the difference between them. CD122 paired with the common gamma chain CD132 binds IL-2 with intermediate affinity and is carried constitutively by memory CD8 and NK cells. Add CD25 and affinity rises about a hundredfold — but CD25 has essentially no cytoplasmic tail and transduces nothing at all.

So when IL-2 is scarce, the cells carrying the high-affinity receptor take almost all of it, and those are the Tregs. Saturate the system and the intermediate-affinity receptor is occupied too, and the same molecule now expands cytotoxic lymphocytes. That is the whole of the dose paradox. It is also why IL-2 muteins engineered away from CD122 binding are pursued in autoimmunity while the opposite bias is pursued in oncology — the two programmes are built on the same observation read in opposite directions.

There is a practical corollary that is easy to miss. At the IL-2 node the clone is the experiment. The functional-grade anti-mouse IL-2 stocked here is JES6-1A12, which occludes the CD122-binding face and therefore biases the IL-2 complex towards CD25-high cells rather than simply neutralising the cytokine. Other clones do the reverse. Two papers can both report "anti-IL-2" and be running opposite manipulations, and the methods section is the only place that distinction appears.

Suppression by contact is removal, not inhibition

CTLA-4 is routinely described as an inhibitory receptor. That description is misleading in a way that changes experimental design. Its dominant mechanism is trans-endocytosis: it physically strips CD80 and CD86 off the antigen-presenting cell and degrades them intracellularly. The Treg is not signalling into the APC — it is removing the costimulatory ligands that CD28 would otherwise use.

This matters because ligand removal and competition-with-CD28 predict different results when CD80 and CD86 are blocked separately — and both are separately blockable in functional grade, so the distinction is testable rather than rhetorical. CD80 and CD86 are frequently treated as interchangeable. They are not: their turnover rates on the APC surface differ, and so does the consequence of removing each one.

CD28 itself carries a similar trap. It is required to generate Tregs in the thymus — CD28-deficient mice are Treg-deficient — and it is opposed by them in the periphery. One molecule, two opposite roles, and which one you observe depends entirely on when you interrupt it. Alongside the contact machinery runs a second, contact-independent route: CD39 hydrolyses extracellular ATP to AMP and CD73 converts that to adenosine, an enzymatic suppression module that needs no cell contact and no cytokine.

The TNF receptor family tunes it in both directions

This band of the pathway is where published results most often appear to conflict, and almost none of it is genuine disagreement. TNFR2 signalling expands and stabilises Tregs, which means anti-TNF therapy removes a Treg-supporting signal at the same time as an inflammatory one — a fact that reframes what "TNF blockade" means in an autoimmune setting.

OX40 is constitutive on Tregs, unlike on effectors, and its agonism opposes FoxP3. That is the reverse of its effect on conventional T cells, so an OX40 agonist runs two mechanisms simultaneously and the net result depends on the ratio of the two populations present. GITR is the most misread node on the map: agonism expands Tregs and renders responder cells resistant to them, so reported net effects run in both directions depending on which cell type dominates the model. 4-1BB agonism has the same structure of problem, and connects directly to the crosslinking question covered in the agonist-antibody map in this library.

None of these is a discrepancy to be averaged away. They are consequences of the same receptor sitting on both sides of the interaction under study, which is the design problem any experiment here has to solve before it can produce an interpretable number.

The lineage is not fixed

The FOXP3 locus is held open by STAT5 binding its CNS2 enhancer, and that STAT5 signal comes from IL-2. Withdraw IL-2 and CNS2 remethylates. IL-6 opposes the same signal, which is the molecular basis for Treg conversion into an IL-17 producer: IL-23 cannot initiate the conversion but sustains it once started, and the cell begins making IL-17A. Under IL-12 or in a strongly inflamed environment the same cells convert instead towards IFN-γ production, and IFN-γ-producing Tregs are recoverable from human autoimmune lesions.

The metabolic layer explains why one drug behaves unexpectedly here. Effector T cells depend on mTOR-driven glycolysis; Tregs run on oxidative metabolism and are relatively mTOR-independent. That is precisely why rapamycin enriches for Tregs rather than suppressing them along with everything else — an effect that looks paradoxical until the metabolic difference is on the table. JAK3 signalling, bound constitutively to the gamma chain, is the shared route every common-gamma cytokine takes, which is also why JAK inhibitors hit Treg maintenance despite being described as selective.

What you can actually block, and what you cannot

27 of the 38 nodes on this map carry a functional-grade antibody — the deepest such layer of any pathway in this library, and the reason it is worth drawing in full. The distribution is as useful as the count.

The gaps, stated plainly. CD122 and CD132 are research grade only — clones 5H4 and 3E12 — and that is a real limitation rather than a cosmetic one, because the intermediate-affinity receptor is half of the paradox this map is about. CD39, CD73, GITR and IL-23 are biosimilar only; no functional-grade equivalent exists for any of them. And FoxP3, STAT5, JAK3 and mTOR are intracellular, so they are measurement nodes rather than intervention nodes. That last point is also why every Treg sorting strategy in routine use relies on the CD25-high, CD127-low surface surrogate instead of the transcription factor that actually defines the cell.

Why the two suppression routes need separating

A Treg suppresses by several mechanisms at once, and most published experiments do not distinguish between them. That is usually defensible, but it stops being defensible the moment the result is used to predict what an intervention will do — because the routes have different dependencies and fail under different conditions.

Contact-dependent suppression through CTLA-4 requires the Treg and the antigen-presenting cell to be in the same place, and it scales with the density of CD80 and CD86 on that APC. The soluble route does not: IL-10 and pan-TGF-β act at a distance and continue working when contact is prevented. The enzymatic route through CD39 and CD73 is different again — it depends on extracellular ATP being available, which is itself a function of how much cell stress is present in the tissue. In a quiet tissue that arm contributes almost nothing; in an inflamed one it can dominate.

The consequence for design is straightforward. If contact and soluble mechanisms are both blockable in the same animal — and here they are — then the informative experiment blocks them independently rather than removing the cells and attributing the whole effect to "Tregs". LAG-3 deserves particular care: its role in suppression is genuinely contested, and LAG-3-deficient Tregs suppress normally in many systems. Having it in functional grade is what allows that to be settled in a given model rather than assumed from a review.

One distinction worth getting right before you start

More published CD25 results are depletion results than the wording usually admits. Clone PC61 depletes Tregs rather than merely blocking IL-2 capture, and an experiment that removes the cells answers a different question from one that starves them. The same split is available at CTLA-4, where three functional-grade clones are stocked including the depleting-versus-blocking pair — which makes this one of the few axes where the ambiguity can be resolved directly rather than argued about.

A second, smaller point: Tregs are anergic to TCR stimulation alone and need IL-2 supplied in trans, so an anti-CD3 protocol calibrated on effector cells will read as a non-response in Tregs. It is not a non-response. It is the wrong protocol.

And a third that costs people time. ICOS marks the most suppressive Treg population and its high expressers are IL-10-biased, while CD62L separates central from effector Tregs — the CD62L-high population recirculates through lymph nodes and carries systemic tolerance, the CD62L-low population is tissue-directed. Both are available in functional grade, which means the phenotype you gate on can also be the thing you interrupt. Sorting a population and then blocking something else entirely is the most common way a clean Treg experiment becomes uninterpretable.

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