When targeting multiple antigens, CAR-T cells can be designed to activate via binding of either one antigen, both antigens, or one antigen but not the other. According to Boolean logic, the former can be referred to as “OR” gate CARs. While “OR” gate CARs address the challenge of variable antigen expression and tumor heterogeneity, the risk of on-target off-tumor effects remains a concern due to their ability to bind two antigens that may be expressed on normal tissue. To mitigate this toxicity, multi-target CAR-T cells have been engineered to activate only when both anti gens are recognized, referred to as “AND” gate CARs. Dual activation can be achieved by using a split-CAR construct in which the scFv targeting one antigen is linked to the CD3ζ domain and the scFv targeting the other antigen is linked to CD28 and 4-1BB. Only when both scFvs are bound to their target antigens will the CARs cluster together and transduce an activation signal (Fig. 1e).
Fig1. Next generation CAR-T cell modifications. Chimeric antigen receptor (CAR)-T cells can be designed to evade tumor antigen escape and mitigate off tumor effects. (a) Antigen escape can be evaded by infusing a mixture of CAR-T cell products targeting different antigens or multi-targeted CAR-T cells designed to target multiple tumor antigens either via the same or different CAR constructs or (b) CAR-T cells engineered to secrete bi-specific T cell engagers which recruit endogenous T cells to engage in tumor killing. (c) To improve safety, one approach is to turn “OFF” the CAR by attaching a zinc finger degron motif to the C-terminus, which binds lenalidomide. When lenalidomide is added, a ubiquitin ligase is recruited, which targets the CAR for degradation and inactivates the CAR-T cell. (d) Triggered assembly CARs improve safety by enabling dynamic control of CAR-T cell activation via exogenously administered agents. In one example, lenalidomide-binding (CRBN) and zinc finger degron (IKZF3) domains can be engineered into SPLIT CARs to enable dimerization and thus activation only in the presence of lenalidomide. (e) Logic-gated CAR-T cells can mitigate on-target off-tumor effects by introducing another layer of specificity. Multi-targeted SPLIT CARs activate only when two tumor antigens are bound by separate scFvs. (f) Conditional expression of a CAR against one antigen is activated by binding of a separate synthetic Notch (synNotch) receptor against a different tumor antigen. (g) CAR-T cells can inactivate upon binding to an antigen on non-malignant cells. Thus, activation depends on binding to one tumor antigen but not the other in a “NOT” logic gate approach. (iCAR, inhibitory chimeric antigen receptor; ITIM, immunoreceptor tyrosine- based inhibitory motif; PD-1, programmed cell death 1)
Another strategy for regulation of CAR-T cells is the synNotch system, which utilizes a synthetic notch receptor (synNotch) with a tumor-specific extracellular antigen binding domain linked to an intracellular transcription factor (Fig. 3f). Upon binding to its target antigen, the receptor is cleaved, and the transcription factor is triggered to move into the nucleus to induce transcription of a separate construct, encoding a CAR against a different tumor antigen. Implementation of this complex “AND” gate exhibited safe anti-tumor efficacy in a solid tumor model but had significant on-target off-tumor effects in a hematologic tumor model, felt to be due to co-localization of dual and single antigen expressing cells. Recently, synNotch receptors have been used to create a three-antigen “AND” gate that is composed of three sequentially linked CARs. This system is capable of selectively eliminating tumor cells expressing all three tumor antigens, while ignoring tumors expressing only two of the target antigens in a bilateral tumor mouse model.
Inhibitory CARs use a “NOT” gate approach, through which CAR-T cell activation depends on the presence of one antigen and the absence of another (Fig. 1g). In this system, a canonical CAR construct targeting a tumor-associated antigen is paired with a separate CAR tethered to an inhibitory intracellular domain, usually from CTLA-4 or PD-1. CAR-T cell binding to the latter target antigen transduces downstream dephosphorylation (and therefore inactivation) of the intracellular domain of the first CAR. This results in CAR-T cell inactivation upon binding to antigen on non-malignant cells.
CAR-T cell activation can also be dynamically regulated using drug-controlled systems to rapidly turn off downstream activity. It was recently discovered that dasatinib, which is FDA-approved for the treatment of chronic myeloid leukemia and ALL, inhibits LCK binding to CD3ζ and therefore inhibits CAR intracellular signaling. Although the use of dasatinib for treatment of CRS in patients has not been reported, in mouse models of lymphoma, dasatinib reduced mortality from CRS by reversibly limiting CAR-T cell activation. CAR-T cells can also be engineered to express non-signaling receptors, such as the extracellular and transmembrane domains of EGFR or CD20, which can be targeted with monoclonal antibodies cetuximab and rituximab respectively. Administration of the corresponding drug results in antibody-mediated CAR-T cell destruction. The clinical utility of this approach is limited by its slow kinetics, which is unable to rapidly reverse CRS symptoms.
More rapid off switches have been designed, including the suicide switch inducible caspase 9 system (iCasp9), which uses the pro-apoptotic caspase 9 protein fused to a modified form of the FK506 binding protein FKBP1A. Administration of a biologically inert small molecule (AP1903) triggers dimerization of Caspase 9, resulting in CAR-T cell apoptosis. Use of this switch in genetically modified donor lymphocytes enabled reversal of acute graft versus host disease (GVHD) in leukemia patients after allogeneic hematopoietic stem cell transplants and was shown to eliminate T cells within 30 minutes.
