One Integrin Chain, Two Partners, Brain or Gut

A chemokine gradient does not move a leukocyte anywhere. It only tells one that has already stopped which way to go. That sentence is the whole map. Trafficking is a four-step cascade and the steps are not interchangeable: tether and roll on selectins, then inside-out integrin activation triggered by a chemokine receptor, then firm arrest, and only then diapedesis. Block arrest and no gradient in the world gets the cell into tissue — which is exactly how natalizumab and vedolizumab work, and exactly why one of them carries a PML risk and the other does not. Every cell in every other map in this library had to get where it is somehow, and this is how. Click any protein for the matching Assay Genie ELISA kit or In Vivo antibody.

The adhesion cascade in order: the chemokine layer and its convergences, the receptors that trigger inside-out signalling, rolling on selectins, firm arrest through the integrin–ligand pairs that decide the destination, and the residency programme that keeps the cell there. Open the interactive version to click any protein for its role and the matching validated reagent.

The order of the steps is the mechanism

Selectins catch a cell moving at arterial shear and convert it into a slow roll. P-selectin appears within seconds from platelet and endothelial granules; E-selectin requires transcription and so appears hours later, which is why the two report different phases of the same injury. Both, and L-selectin on the leukocyte itself, bind PSGL-1 — and only when it is correctly glycosylated and sulfated. These are low-affinity, high-off-rate bonds, and they survive shear precisely because they break and reform. CD44 provides a secondary rolling route through hyaluronan.

A rolling cell then encounters chemokine displayed on the endothelial surface, and its receptor switches on RAP1 and talin-1 within milliseconds. Talin binds the integrin β tail and physically extends the ectodomain from a bent, low-affinity conformation into an upright, high-affinity one. That is inside-out signalling, and it is mechanical rather than transcriptional. Only then does firm arrest occur.

Miss this ordering and the experiment stops making sense. A selectin blocker and a chemokine-receptor antagonist both reduce recruitment, and they look equivalent in a total-cell-count readout. They are not: they fail at different steps, in different vascular beds, with different kinetics, and with different consequences for the cells already in tissue.

Redundancy is why single-chemokine blockade almost never works

The ligand layer is deliberately degenerate. CXCL9 and CXCL10 both signal through CXCR3. CXCL8 uses both CXCR1 and CXCR2 — CXCR1 is the high-affinity, CXCL8-selective one that mediates the late, high-concentration neutrophil response. CCL19 and CCL21 both use CCR7, and CCL21 is the one immobilised on stroma, which is what makes it a real gradient rather than a soluble cloud.

Remove one ligand and the others cover for it. Blocking at the receptor is therefore consistently more effective than blocking at the ligand — the same lesson the VEGF axis teaches, arrived at from a different direction. CCL2 and CCR2 are the cleanest counter-example, because CCR2 defines the inflammatory monocyte so tightly that CCR2-deficient monocytes are made normally and simply cannot leave the marrow. CCL5 and CCR5 are the best-drugged pair in the family, for reasons that have nothing to do with immunology.

ACKR1 — the Duffy antigen — is on the map as the exception that proves the design. It binds a wide range of chemokines and signals through none of them, acting as a scavenger and a transcytosis platform. Measuring chemokine concentration in a tissue where ACKR1 is abundant tells you about sequestration, not about signalling. The same caution applies to CXCL12, which is a retention signal rather than a recruitment signal: blocking CXCR4 mobilises cells out of a niche, the opposite direction to every other block on this map.

Arrest is where the drugs are, and where tissue selectivity lives

The integrin determines the destination, and it does so because the ligand is expressed regionally. CD11a paired with CD18 forms LFA-1, which binds ICAM-1 and is general-purpose — loss of CD18 causes leukocyte adhesion deficiency type I, where cells circulate normally and never arrest anywhere. Efalizumab targeted CD11a and was withdrawn for PML.

Now take the same α chain and change its partner. CD49d (integrin α4) paired with β1 binds VCAM-1 and gets cells into brain and inflamed tissue. Natalizumab blocks exactly this, which is why it is effective in multiple sclerosis and why it carries a PML risk: it also blocks immune surveillance of the CNS. Pair the same α4 with integrin β7 instead and you get α4β7 (LPAM-1), which binds MAdCAM-1 — expressed on gut endothelium and almost nowhere else. That is vedolizumab's target, and the reason it is effective in inflammatory bowel disease with none of the CNS liability.

One shared α chain, two partners, two completely different safety profiles. It is the cleanest illustration in immunology that specificity can live in the heterodimer rather than in either subunit, and it is why blocking the pair is a different drug from blocking the chain.

Then the cell has to stay

Arrival is not residency. Egress from tissue requires S1PR1, and CD69 physically binds and sequesters it to prevent exit. That is why CD69 is the canonical residency marker rather than merely an activation marker — it is doing mechanical work, not reporting a state. Fingolimod exploits the same node from the other direction, removing S1PR1 so that lymphocytes are trapped in node. CD103 binds E-cadherin and holds the cell against epithelium, and CCR9 reads CCL25 from small-intestinal epithelium; both are induced alongside α4β7 by retinoic acid from gut dendritic cells, which is how the gut address is written in the first place.

Which step does your phenotype depend on?

The In Vivo angle

Thirteen of the 37 nodes carry a functional-grade In Vivo antibody, and unusually for this library they sit exactly where the interventions are rather than one layer away from them: CD11a and CD18 for LFA-1; CD49d, integrin β7 and α4β7 for the α4 pair; ICAM-1 in mouse (ultra-low endotoxin) and in human; L-selectin in both species; plus CD44, CXCR4, CXCR3 and CXCL9. Every single step of the cascade can be blocked in vivo with a validated reagent.

The measurable side covers the rest. The chemokines and the endothelial ligands — CXCL8, CCL2, CCL5, CXCL10, CXCL12, CCL19, CCL21, VCAM-1, MAdCAM-1, soluble P- and E-selectin, PSGL-1 and ACKR1 — are quantitative analytes, which is the right way to use them given the ligand redundancy described above. The natural design is therefore: block one step at a time, measure the gradient, and find out which step your model actually depends on. Running anti-CD62L, anti-CXCR3 and anti-CD49d in parallel arms of the same experiment answers that question in one study.

The one-line version

Roll, activate the integrin, arrest, then stay. The chemokine chooses the direction, the integrin pair chooses the tissue, and the drug that works is almost always the one that blocks arrest — because that is the only step with no redundancy.

For the adhesion cascade on its own, see the leukocyte adhesion cascade; for what happens after arrival, see mucosal homing and tissue-resident memory; and for the vascular side of the same problem in tumours, see tumour angiogenesis and the VEGF axis.

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