Therapeutics & platforms
NK-cell and CAR-NK therapy
Missing-self recognition, the balance of activating and inhibitory receptors, and the absence of a clonal TCR: what makes NK cells inherently allogeneic and what limits their persistence.
A T lymphocyte recognises a peptide presented by an HLA molecule through a clonally rearranged receptor. That rearrangement is exactly what makes donor T cells dangerous: in another person a fraction of clones react against foreign HLA, producing graft-versus-host disease. An NK cell is built differently, and the whole clinical logic of the platform follows from that.
The decision is made by subtraction
NK cells do not rearrange receptors. They carry germline-encoded activating and inhibitory receptors, and the decision to kill is taken as the balance of signals at a contact.
On the inhibitory side sit the killer immunoglobulin-like receptors and NKG2A/CD94, which read HLA class I and HLA-E. A healthy cell displays enough of these and the NK cell moves on. A tumour or infected cell often downregulates HLA class I — precisely to escape T cells — and in doing so releases the inhibition. This is missing-self recognition: the NK cell kills for the absence of a marker, not its presence.
On the activating side are NKG2D, which reads the stress-induced ligands MICA/MICB and ULBP; the natural cytotoxicity receptors NKp30, NKp44 and NKp46; and DNAM-1. Separately there is CD16a (FcγRIIIa), which binds antibody Fc and turns the NK cell into the effector arm of antibody-dependent cellular cytotoxicity.
The consequence that matters: with no clonal TCR, a donor NK cell does not see foreign HLA as antigen and does not cause GvHD. The product can therefore be allogeneic by design — from cord blood, from donor peripheral blood, or from a clonal iPSC bank — and manufactured in advance rather than per patient.
A CAR adds an address, not a solution
A chimeric antigen receptor gives the NK cell directed specificity against a chosen antigen. The signalling domains have to be rethought, though: costimulation that is optimal in a T cell is not necessarily optimal in an NK cell, and constructs frequently use DAP10, DAP12 or 2B4 in place of CD28. The toxicity profile differs too — NK cells do not release the same cytokine spectrum, and severe cytokine release syndrome and neurotoxicity are reported less often than with CAR-T.
Persistence is the binding constraint
A mature circulating NK cell lives on the order of one to two weeks. An allogeneic product is also subject to rejection: host T cells recognise the donor’s foreign HLA, and once lymphocytes suppressed by conditioning recover, the infused cells disappear. This puts the platform in direct contrast with CAR-T, where years of persistence by a single clone is part of the curative mechanism. Hence constructs carrying their own IL-15 to support survival in an autocrine or paracrine loop, and repeat dosing in place of a single infusion.
The second constraint is the regulation of activity itself. An activated NK cell sheds CD16 from its surface through the metalloproteinase ADAM17, losing ADCC capacity at the moment it is working — which is why non-cleavable CD16 variants are engineered in. Cryopreservation, unavoidable for an off-the-shelf product, itself reduces cytotoxicity after thaw.
The third is solid tumours. The difficulty is not recognition but what happens after entry: TGF-beta suppresses NKG2D and NKp30, hypoxia and adenosine acting through the A2A receptor blunt killing, and migration into dense stroma is limited. None of these barriers is removed by adding a CAR.