**3.2 Modified CAR T cells**

CAR T cells combine the single-chain variable fragment (scFV) of monoclonal antibodies with the internal component of the T cell receptor. There are three main generations of CAR T cells—first-generation CAR T cells include an scFV as well as CD3ζ endodomain. The second generation built upon this by adding a costimulatory molecule such as CD28 or 4-1-BB to promote expansion. Finally, third-generation CAR T cells consist of an scFV, CD3ζ, as well as two or more costimulatory molecules. CAR T cells, especially third-generation CAR T cells, have had great success in patients with B cell malignancies [133].

However, single-agent CAR T cells have had limited success in patients with CNS malignancies. This is likely due to several factors, including a high degree of heterogeneity in the tumor microenvironment (TME), loss of antigen during tumor progression, exhaustion of the CARs within the TME, and finally upregulation of immunosuppressive molecules that inhibit CAR T cell killing [134].

Bielamowicz *et al.* utilized human GBM cells to identify three antigens expressed on human glioma cells: HER2, IL-13Rα2, and EphA2. Single-agent CAR T cells, bispecific CAR T cells targeting IL-13Ra2 and EphA2, as well as trivalent CAR T cells specific for all three antigens were developed and tumor-killing was first assessed *in vitro*. Upon coculture with target human glioma cells, secretion of IL-2 and IFNγ were significantly higher upon treatment with trivalent CAR T cells relative to nontransduced T cell controls. In addition, the specific lysis of target cells was significantly greater when co-cultured with trivalent CAR T cells relative to controls. Efficacy was also evaluated using intracranially injected human glioma cells followed by intracranial injection of single CAR T (targeting IL-13Rα2), bivalent CAR T (targeting EphA2 and IL-13Rα2), trivalent CAR T cells (targeting HER2, IL-13Rα2, and EphA2), or nontransduced T cell controls. The authors found the trivalent CAR T cells provided superior anti-tumor efficacy relative to controls [135].

Several other modified CARs have shown increased efficacy relative to their first-generation counterparts. Krenciute *et al.* modified IL-13Rα2-specific CAR T cells to secrete IL-15, a cytokine that promotes activation, proliferation, and cancer cell lysis. Relative to IL-13Rα2 CAR T cells that did not secrete IL-15 (first-generation), these second-generation CAR T cells showed increased lysis of target tumor cells and increased proliferation *in vitro*. In addition, when evaluated *in vivo*, mice treated with the second-generation CAR T cells had significantly increased PFS and OS relative to those treated with the first-generation CAR T cells. The authors found mice that succumbed to the tumor after treatment CAR T cell therapy downregulated the expression of IL-13Rα2 [136].

In the context of neuroblastoma, disialoganglioside (GD2) represents a promising tumor-associated target for CAR T cell therapy. GD2 has been shown to promote malignant phenotypes such as proliferation, migration, and invasion [137]. In a phase I clinical trial, GD2-specific CAR T cells were evaluated in neuroblastoma patients in combination with cyclophosphamide and fludarabine as well as the checkpoint inhibitor, anti-PD-1 [138]. Although the therapy was found to be safe,

only modest anti-tumor responses were observed [139]. To improve upon the efficacy of these CAR T cells, Moghimi *et al.* modified GD2-specific CAR T cells to express B7H3 and found enhanced anti-tumor responses both *in vitro* and *in vivo* relative to untreated controls. They go on to determine the enhanced efficacy is likely due to improved metabolic function [140].

Another promising CAR T cell target for brain tumors is CD70. In terms of normal immunology, CD70 is a co-stimulatory molecule expressed in activated immune cells. However, Jin *et al.* found CD70 to be overexpressed in tumor samples isolated from IDH wild-type low-grade glioma and GBM patients [141]. Using a model of high-grade glioma, modified CD70 CAR T cells that express CXCR1 or CXCR2 improved T cell migration to the tumor site. In addition, survival of tumorbearing mice was improved when treated with CXCR1 or CXCR2 modified CD70 CAR T cells relative to unmodified CD70 CAR T cells [142]. Collectively, these results suggest modified CAR T cells may hold promising anti-tumor responses relative to their first-generation counterparts.

A huge limitation of CAR T cells is the eventual expression of exhaustion molecules, leading to a lack of anti-tumor efficacy. Weber *et al.* recently utilized transient periods of rest using a small molecule as well as dasatinib, a tyrosine kinase inhibitor that inhibits T cell signaling. The authors utilized GD2.CD28ζ CAR T cells in a model of human osteosarcoma. The use of rest in pre-exhausted CAR T cells redirected their fate from a state of exhaustion toward a memory-like state. Furthermore, in T cells that already acquired markers of exhaustion, the use of rest reversed the exhaustion phenotype and caused epigenetic remodeling similar to non-exhausted controls. CAR T cells subjected to intermittent rest through alternating CAR expression or dasatinib-treatment demonstrated superior anti-tumor effects. These findings have profound translational implications to improve therapeutic response using CAR T cells [143].
