Two Asthma Drugs, Opposite Eosinophil Counts, Both Working

Two asthma biologics have opposite effects on the blood eosinophil count, and both of them work. Mepolizumab empties the eosinophil compartment; dupilumab can transiently raise it. If that seems contradictory, the map below is the resolution: the type 2 cytokines are not interchangeable, and they divide the labour cleanly. IL-5 owns the eosinophil — production, survival, almost nothing else. IL-4 and IL-13 own the tissue. Anti-IL-5 therefore empties the blood and leaves the mucus; anti-IL-4Rα does the reverse and raises the blood count precisely because it blocks the cells' exit into tissue rather than their manufacture. Click any protein for the matching Assay Genie ELISA kit or In Vivo antibody.

The type 2 granulocyte axis end to end: epithelial alarmins, the ILC2 output, the division of labour between the cytokines, the eosinophil's own surface, and the granule proteins that report what actually happened. Open the interactive version to click any protein for its role and the matching validated reagent.

It starts with damage, not with an allergen

IL-33, TSLP and IL-25 are alarmins: epithelial proteins released in response to protease activity, viruses, pollutants or simple mechanical damage. No antigen recognition takes place. They act on ST2, TSLPR and IL-17RB on ILC2 cells, which then produce the full type 2 cytokine set with no T-cell receptor involved anywhere in the process. This is why "allergic" inflammation happens in people with no demonstrable sensitisation, and why viral exacerbations look immunologically identical to allergen exacerbations.

It is also why the newest and broadest drug in this space acts at the very top. Tezepelumab targets TSLP, and because it intervenes before the cascade branches it reduces eosinophils, IgE, FeNO and exacerbations together — including in patients whose blood eosinophil count was never elevated and who were therefore ineligible for everything else. Every other drug on this map is a single-branch drug. One practical caveat on assays: soluble ST2 acts as a decoy receptor, so a rising serum sST2 does not mean rising IL-33 signalling — it may mean the opposite.

The division of labour, and why it explains the clinic

GATA3 licenses the whole type 2 locus, but its products do different jobs. IL-5 acts on IL-5Rα to drive eosinophil production in the marrow and survival in the tissue. IL-4 and IL-13 share IL-4Rα and own the tissue phenotype: IgE class-switching, MUC5AC and goblet-cell metaplasia, smooth-muscle change, and periostin as the serum marker of that remodelling arm.

So the two drug classes are not competing versions of the same idea. Mepolizumab (anti-IL-5) and benralizumab (anti-IL-5Rα) remove the eosinophil and leave mucus and remodelling largely intact. Dupilumab (anti-IL-4Rα) attacks the tissue arm and leaves eosinophil production running, which is why blood counts can rise on treatment while the patient improves. Benralizumab has an extra wrinkle worth knowing: it is afucosylated, so it depletes by NK-mediated ADCC rather than by blockade alone, which makes its depletion faster and more complete than the mechanism name suggests.

IL-3 and GM-CSF matter because they share the βc receptor chain with IL-5 and partially substitute for it, which is one honest explanation for why anti-IL-5 does not help everyone with high eosinophils. IL-9 is the least-exploited branch. IL-10 is the resolution arm, restraining cytokine production rather than tissue damage.

Recruitment needs two signals, and only one is a chemokine

The eotaxins — CCL11, CCL24 and CCL26 — all signal through CCR3. That single fact explains two decades of results: blocking one eotaxin achieves very little because the others cover for it, and CCR3 looked like the obvious drug target it never became. Species and chronicity matter here too. CCL11 dominates acute recruitment, while CCL26 becomes the dominant human eotaxin in chronic disease and is IL-13-inducible — so if you are measuring one eotaxin in a chronic cohort, CCL11 is probably the wrong choice.

Before an eosinophil can follow any gradient it has to stop moving. CD62L handles the initial tether, CD49d (the VLA-4 α chain) provides firm arrest against VCAM-1 on IL-4-activated endothelium, and integrin β7 directs the gut-tropic pool that matters in eosinophilic gastrointestinal disease. Note the dependency: VCAM-1 is induced by IL-4, so the tissue cytokine controls the arrest surface. That is the coupling that makes the whole response type-2-specific rather than generically inflammatory. Low surface CD62L, incidentally, means the cell has already been activated and shed it — not that it lacks the receptor.

The granule proteins are the damage, and they are the assay you are missing

A blood eosinophil count tells you cells exist. It does not tell you whether they degranulated, and degranulation is the pathology. EPX generates epithelium-damaging oxidants; ECP and EDN are cytotoxic ribonucleases; major basic protein is the most abundant and the most toxic to epithelium. Galectin-10 forms Charcot-Leyden crystals, which were dismissed as a histological curiosity for a century and are now understood to be an inflammatory stimulus in their own right.

These are measurable, and measuring them changes conclusions. An intervention that blocks arrest should reduce granule protein in tissue while blood eosinophils stay high — a result a cell count would score as a failure. Siglec-8 is the one true off-switch on the map: engaging it induces eosinophil apoptosis, which is the mechanism lirentelimab was built on. The lipid receptors CysLT1 (montelukast's target) and CRTH2 sit on the eosinophil itself, not only on smooth muscle, and GATA1 is the lineage factor — a GATA1 promoter deletion is how the eosinophil-null mouse was made, and it remains the cleanest genetic tool in the field.

An assay checklist

The In Vivo angle — stated plainly

Seven of the 37 nodes carry a functional-grade In Vivo antibody, and they are concentrated in the trafficking and shared-cytokine layer rather than on the IL-5 axis: IL-4 (clone 11B11), GM-CSF, CD62L, CD49d, integrin β7, pan-TGF-β and IL-10. That set is enough to stop an eosinophil arresting and entering tissue, in mouse and in human, which given the arrest bottleneck described above is arguably the most interesting thing you can block here.

The IL-5 axis itself is covered by ELISA and by research-grade mepolizumab and benralizumab biosimilars rather than by functional-grade antibodies, and it is worth saying that directly rather than implying otherwise. The natural configuration for this map is therefore: block the trafficking, measure the cytokines and the granule proteins. Pair a CD49d block with EPX and ECP ELISAs and you have a clean test of whether arrest, rather than production, is what your model depends on.

The one-line version

Alarmins start it, ILC2 cells run it without any antigen, IL-5 owns the cell and IL-4/IL-13 own the tissue. Which arm you block determines which readout improves — and a blood eosinophil count is the one measurement that can move in the wrong direction while the patient gets better.

For the mast-cell side of the same type 2 response, see mast cells, IgE and type 2 immunity; for the IL-17 counterpart programme, see the Th17 / IL-23 axis; and for the adhesion cascade in general, see the leukocyte adhesion cascade.

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