Not all CAR-T cells are created equal

CAR-T cell-based adoptive immunotherapy against B cell-derived malignancies shows remarkable effectiveness in clinical trials. But, for some reason, other CAR-T cells, specific for diverse tumor-associated antigens, are not as effective as α-CD19 CAR-T cells. So why is that?

New paper in Nature Medicine may provide some clues to this mystery. The authors, led by Crystal Mackall, showed that unlike α-CD19 CAR-T cells, several CAR-T constructs displayed antigen-independent basal CD3zeta signaling leading to early exhaustion and lack of anti-tumor in vivo effectiveness.   

When comparing two CAR constructs α-CD19 and α-GD2 (sarcoma antigen), the authors showed that while both CAR-T cells displayed similar in vitro cytotoxicity, their in vivo anti-tumor activity against sarcoma cell line double-positive for CD19 and GD2 were vastly different: only α-CD19 CAR-T cells could control tumor growth.  

These results suggested that CAR construct, rather than tumor targets, was responsible for this difference. Indeed, T cells double-transduced with both α-CD19 and α-GD2 CAR constructs displayed diminished in vivo anti-tumor effectiveness against CD19⁺sarcoma (compared to CD19-only CAR-T cells).  

Further analysis showed that unlike CD19 CAR-T cells, GD2 CAR-T cells showed spontaneous, basal CD3zeta activity. Such basal CD3zeta activity were absent in GD2 CAR-T cells with mutations in CD28/CD3 signaling domains indicating that GD2 CAR was responsible for this spontaneous signaling.

Furthermore, the authors showed that spontaneous clustering of GD2 CAR construct (oligomerization of scFv fragments) probably was responsible for basal signaling in GD2 CAR-T cells (and not vice versa).

Finally, the authors showed that substitution of CD28 domain with 4-1BB (CD137) domain in CAR-T construct could partially rescue GD2 CAR-T cells exhaustion phenotype and improve its in vivo anti-tumor effectiveness.

In summary, data provided in this paper indicate that biochemical property of CAR constructs impart their anti-tumor effectiveness. This suggests that design of effective CAR-T construct would take more than just specificity. In my view the issue is that scFv are of BCR origin, but precise nature of BCR signaling is not clear even today. One model suggests that BCR signaling requires their oligomerization and other model suggests exactly opposite: BCR signaling requires disruption of BCR oligomers. But since assembly of TCR singaling complexes may differ from that of BCR signaling complex assembly, simply transplanting BCR domain into T cells may not be as successful as initially thought.

David Usharauli