NK Cell Licensing and Missing-Self
An NK cell kills what has stopped looking like self. That is the missing-self rule, and it is the exact inverse of the T-cell rule \u2014 which is why the two systems together close the escape route that losing MHC class I would otherwise open. But missing self is only half of the arithmetic. The cell also reads induced self: stress ligands such as ULBP2 that appear on damaged or transformed cells and engage NKG2D. A target that keeps its class I can still be killed if enough activating ligand has appeared, and a target that has lost it can still be spared if nothing has. The second thing worth carrying away is that licensing runs the opposite way to intuition: an NK cell that has never engaged self class I through an inhibitory receptor is hyporesponsive rather than autoreactive. Inhibition is what makes the cell competent in the first place. Click any protein for the matching Assay Genie In Vivo antibody or ELISA kit.
The rule is subtraction, and it has two terms. An NK cell integrates inhibitory input from receptors reading MHC class I against activating input from receptors reading induced ligands, and kills when the balance tips. Losing class I removes inhibition; acquiring ULBP2 or another stress ligand adds activation. Either alone can be sufficient, which is why the field's shorthand \u2014 "NK cells kill cells that have lost MHC" \u2014 is only ever half the story, and why a tumour with normal class I can still be an NK target. NKG2A adds a further wrinkle: it reads HLA-E, which itself presents peptides derived from other class I leader sequences, so it is monitoring class I synthesis by proxy rather than directly. A tumour can therefore escape NKG2A while losing a classical class I allele, and the two events need not travel together.
Licensing inverts the intuition, and it is testable here. An NK cell that has never engaged self class I through an inhibitory receptor does not become autoreactive \u2014 it becomes hyporesponsive. Inhibitory signalling during development is what confers functional competence, which is why MHC class I-deficient mice have NK cells that are present, normal in number, and poorly responsive. Ly49C is the mouse receptor at the centre of this, and it is available in functional grade (clone 4LO3311) \u2014 which is what turns licensing from a model into an experiment. It also explains why haplotype matters so much on this map: Ly49 recognition is allele-specific, and a class I reagent matched to the wrong strain reports nothing.
One ligand, two receptors, opposite directions. CD155 is read by both TIGIT, which inhibits, and DNAM-1, which activates \u2014 with TIGIT binding at higher affinity. The practical consequence is that CD155 expression on a tumour predicts nothing on its own; what predicts is the DNAM-1 to TIGIT ratio on the effector, and that ratio shifts with activation state. This is the cleanest available illustration of why single-molecule biomarkers underperform on this pathway. The same logic applies to the Fc receptors, which sit on the activating side and are the route by which every depleting therapeutic antibody works: CD16 on human NK cells carries ADCC, and the FcγRIIIa 158V/F polymorphism predicts clinical response to rituximab and trastuzumab \u2014 about as direct a line from this map to a clinical outcome as immunology provides.
The In Vivo tie-in, and where the map is thin. 26 of the 38 nodes carry a functional-grade antibody, and unusually for a checkpoint pathway the inhibitory receptors themselves are among them: Ly49C and NKG2A are both blockable in vivo, and TIGIT is blockable in both species. Three mouse class I haplotypes are stocked alongside two human specificities, which is what makes haplotype-matched design possible. Around them sit human CD16 (clone 3G8), Fc\u03b3RIV (clone 9E9, which carries most antibody-dependent killing in mouse models), CD155, CD200, NK1.1, CD2, CD11a, CD18, ICAM-1 in both species, CD44, CD62L, CD70, IL-2, IFN-\u03b3 and TNF. Stated plainly, the activating arm is the thinner half: NKG2D, KLRG1, CD27, CD20, CD38 and CD52 are biosimilar only; DNAM-1 is stocked as a recombinant protein rather than a blocking antibody; and ULBP2, IL-15, perforin and granzyme B are ELISA only. That asymmetry is worth knowing before designing around NKG2D, and it is why the productive experiments on this pathway tend to interrupt inhibition and read the activating side out, rather than the reverse. One reagent caution: TNF is stocked ultra-low endotoxin, which is not a formality \u2014 LPS activates NK cells indirectly through myeloid IL-12 and would otherwise be scored as the cytokine's own effect. For research use only; not for use in diagnostic or therapeutic procedures.
Every protein node links to a product — ELISA kit, In Vivo antibody or research antibody.