Engineering T cells to cellular factories with synthetic Notch receptors

This week journal Cell published new article from Wendell Lim's lab at UCSF (also a founder of Cell Design Labs) that reads like a science fiction story. It was a continuation of previous work that focused on developing customized molecular architecture based on Notch core regulatory domain attached to synthetic extracellular recognition and intracellular transcriptional domains (SynNotch). By changing extracellular and intracellular domains one can design T cells producing molecule of interest upon engagement with specific ligand. 

For this new paper, the authors managed to transform T cells into cellular factories which upon specific SynNotch stimulation were making and expressing (a) cytokines (b) checkpoint inhibitors (c) bi-specific antibodies (d) CARs (e) transcriptional factors (f) lytic granules.

Basically, T cells are transfected with vector containing SynNotch module linked to promoter encoding molecule of interest. The most of the experiments were done in vitro. One in vivo experiment the authors put in paper was similar in overall design to one previously reported. In short, NSG mice were injected with K562 tumor expressing GFP + CD19 or only CD19. Afterwards T cells containing SynNotch module designed to recognize GFP were infused. Upon GFP recognition these T cells could start expressing soluble Blinatumomab, a-CD19/CD3 BiTE molecule that in turn can engage conventional TCR and produce T cell activation. Indeed, the authors showed that only tumor cells expressing both GFP and CD19 were efficiently controlled by engineered T cells (though its is not clear why soluble BiTE antibodies would not diffuse freely and engage single CD19+ tumors as well).    

While all these results look very impressive, it nonetheless lack some of the critical elements that are required to properly analyze beneficial effect of SynNotch. For example, in vivo experiments were done in mice that does not express tumor antigens naturally (anti-CD19 here was human in origin). So we can't tell how such T cells would behave in "human-like" environment. Similarly, in vitro experiments with SynNotch-anti-HER2 expressing T cells are of dubious value without testing it at least in humanized mouse models.

David Usharauli

Novel combinatorial antigen-sensing chimeric antigen receptor circuit permits tumor-selective therapeutic T cell activity

Few days ago journal Cell published new study on chimeric antigen receptor (CAR) T cells that produced lots of discussion in immuno-oncology field. I read the actual paper and here below I will provide my opinion about it. 

I will point out that overall this study is of high quality. Basically, the whole premise of this paper is based on what the authors had correctly pointed out in the introduction "Few antigens are absolutely tumor specific, and T cells targeted to antigens that are also found on normal bystander tissues can cause life-threatening adverse side effects."

So how this could be avoided? According the authors "combinatorial antigen recognition T cell circuits in which a synNotch receptor for one [tissue-selective] antigen drives the inducible expression of a CAR for a second [tumor-specific] antigen" could provide such a fail-safe mechanism. In essence, in this model CAR-T cells would behave very similarly to how NK cells are sensing healthy and "abnormal" cells or tissues.

In first set of experiments, the authors tested validity of this concept on Jurkat T cells, a human T cell line. Jurkat T cells were transduced with a-CD19 synNotch receptor and a-mesothelin CAR gene and then exposed to "K562 myelogenous leukemia cells with ectopic expression of CD19, mesothelin, or both antigens". The authors detected IL-2 production [as a readout of Jurkat T cell activation] only with K562 cells expressing both CD19 and mesothelin.

Similar results were obtained with primary human T cells transduced with a-GFP synNotch receptor and a-CD19 CAR and exposed to "K562 target cells expressing CD19 only, GFP only, or both GFP and CD19". 

Importantly, in vivo experiments also indicated that T cell response were restricted to tumor cells with dual antigen expression. Only dual GFP+/CD19+ Daudi B cell tumor, but not single CD19+ Daudi cells, could induce CAR-T cell activation [as measured by luciferase signal]. Of note, there was low but detectable level of CAR T cell engagement at the single CD19+ Daudi tumor site.

And even more importantly, even in mice simultaneously transplanted with both single or dual (GFP+/CD19+) expressing K562 tumor cells, only dual antigen-expressing tumor cells were targeted. This indicated tumor-selective nature of combinatorial CAR-T cells. It showed that in this model, combinatorial CAR-T cells do not "engage a tumor expressing the synNotch ligand (GFP), become primed by expressing the a-CD19 CAR, and then migrate elsewhere to then kill single antigen (CD19+ only) bystander tissues." (though, I would admit I was surprised with these results).

In summary, these results clearly showed some of the advantages of combinatorial CAR-T cells. By expressing tumor tissue-selective SynNotch "primer" receptor, these CAR-T cells are primed and engage tumor "specific" antigens within tumor tissue only (but not in other tissue). 

Now question is how far such tumor tissue "selectivity" goes. The authors have not done more direct experiment wherein tumor and actual healthy tissues share "common" antigen to see if combinatorial CAR-T cells would maintain their "bias" towards tumor tissue. Since the main premise of such combinatorial CAR-T cells is to prevent CAR-T cell's "side effects" against healthy tissue, we need model that reliably test this scenario.   

David Usharauli