CAR T-cell Integration of Multiple Input Signals Allows for Precise Targeting of Cancer

T cells transduced with chimeric antigen receptors (CAR) can be effectively redirected to and activated by a cell-surface native target antigen of choice. The adoptive transfer of CD19 (a pan B-cell antigen)-targeting CAR T cells has demonstrated sustained clinical responses in patients with hematologic malignancies, leading to the recent FDA approvals of tisagenlecleucel (Kymriah) and axicabtagene ciloleucel (Yescarta) for relapsed/refractory cases of acute B-cell lymphoblastic leukemia and adult large B-cell lymphoma, respectively. CAR T-cell targeting of solid cancers faces several obstacles. First, almost every proposed cancer antigen has some expression on normal tissue, which risks “on-target, off-tumor” toxicity. Another obstacle is the strongly immunosuppressive microenvironment characteristic for solid tumors, which impairs the effector function, proliferative capacity, and in vivo persistence of adoptively transferred T cells.

Most engineering solutions reported to date address individual problems posed by target specificity or the tumor immunosuppressive microenvironment. In this issue of Cancer Discovery, Sukumaran and colleagues describe an engineering approach that not only increases CAR T-cell specificity to the tumor through pattern recognition of three ligands but in doing so renders the CAR T cells resistant to immunosuppressive components (IL4) and TGFβ) of the tumor milieu (1). The authors devised a three-component split CAR system, comprising of chimeric receptors that recognize prostate stem cell antigen (PSCA), TGFβ, and IL4 and whose endodomains recapitulate a physiologic T-cell signaling (providing signals 1, 2, and 3; Fig. 1). Using pancreatic cancer as a model system, they demonstrated that a functional T-cell response is reconstituted in tumor-mimicking conditions (target cells expressing PSCA, TGFβ, and IL4) leading to PSCA⁺ target cell lysis accompanied by survival and expansion of genetically modified T cells in TGFβ- and IL4-rich milieu. Conversely, the presence of antigen alone or either cytokine alone was insufficient to promote expansion of the transgenic T cells, highlighting their dependence on all three input signals. This system, termed SmarT cells (tumor-specific molecular-pattern activated and regulated T cells), showed increased tumor selectivity both in vitro and in vivo. The authors generated an animal model recapitulating both healthy and tumor-mimicking conditions by injecting NSG mice subcutaneously in opposite flanks with cells expressing PSCA only (mimicking normal tissue) and tumor cells expressing PSCA, TGFβ, and IL4, a signature present in pancreatic tumors. Upon administration, SmarT cells “sensed” the tumor site, distinguishing it from healthy antigen-positive tissue, resulting in selective elimination of PSCA⁺TGFβ⁺IL4⁺ tumors. Furthermore, substantial tumor site–restricted SmarT-cell expansion was observed, attesting to the reliance of the genetically modified T cells on the presence of all three input signals for adequate T-cell response. In subsequent rechallenge experiments, SmarT cells reexpanded only at the tumor site in the presence of PSCA, IL4, and TGFβ and led to effective tumor eradication, demonstrating retained selectivity and potency of this approach.

Over the last few years, several strategies to overcome “on-target, off-tumor” toxicities by more refined discrimination between normal and malignant tissues have been described. Among the proposed strategies, the most promising ones involve splitting of the conventional single-input CAR into two independent receptors, capable of Boolean logic signal transduction (e.g., OR gate, AND gate, and AND–NOT gate). An early example of the AND strategy involves separating CD3ζ chain (signal 1) and costimulation (signal 2) into two independent chimeric receptors, whose signaling is complementary upon recognition of distinct targets expressed by tumor cells (MUC1 and HER2 CARs, ref. 2; and later PSMA and PSCA CARs, ref. 3). A more sophisticated strategy to engineer multi-input control of T cells involves the building of two antigen tumor recognition circuits based on a synthetic Notch (synNotch) receptor-inducible system, whereby engagement with one tumor antigen induces expression of a second CAR, resulting in biphasic signaling response (4).

Although Boolean-gated CAR approaches are appealing in their refinement of tumor selectivity, they do not address any of the hurdles that the immunosuppressive microenvironment poses for adoptively transferred CAR T cells. Most solid tumors have insufficient amounts of activating costimulatory ligands and immunostimulatory cytokines, which contributes to the lack of expansion and persistence of adoptively transferred CAR T cells that have trafficked to the tumor site. Preclinical studies aiming at improving persistence of T cells at the tumor site have shown that restoration of signal 3 through genetic modification of T cells to express secreted or tethered cytokines enhances T-cell antitumor activity (5). However, systemic or locally provided cytokines exhibit substantial toxicities. To circumvent this effect, a constitutively active IL7 cytokine receptor has been described, which delivers an IL7 signal in the absence of the secreted cytokine (6). An alternative cytokine support system is one provided by chimeric cytokine receptors that translate a negative signal produced by an immunosuppressive cytokine (IL4) within the tumor microenvironment to an immunostimulatory signal (IL2 receptor or IL7 receptor; ref. 7). Similarly, a soluble immunosuppressive signal (TGFβ) has been successfully used to stimulate CAR T-cell effector functions, demonstrating that CARs can be specifically engineered to respond to freely soluble ligands (8).

Building onto these concepts, the authors of this paper have engineered T cells to recognize a pattern exclusive to the tumor site using three independent receptors that recognize the tumor cells (PSCA⁺) but also invert the effects of two soluble immunosuppressive molecules (IL4 and TGFβ) into immunostimulatory signals, thereby recapitulating fully functional physiologic T-cell signaling (Fig. 1). Furthermore, by using a three-input CAR design incorporating soluble as well as membrane-bound antigens, Sukumaran and colleagues sought to minimize the risk of immune escape due to antigen loss, an event already observed in the clinic (9).

In summary, the current study by Sukumaran and colleagues demonstrates that T cells can be engineered to recognize a specific pattern present exclusively at the tumor site, thereby enhancing both the antitumor efficacy and safety profile of the transgenic T cells. By limiting T-cell activity exclusively to the tumor site, this genetic approach reduces “on-target, off-tumor” toxicity. In addition, this strategy not only has the benefit of rendering genetically engineered T cells resistant to immunosuppressive cytokines (TGFβ and IL4) present in the tumor milieu, but also allows SmarT cells to engraft within the tumor microenvironment. The described system can readily be extended to other soluble and membrane-bound inhibitory molecules present at the tumor site (e.g., IL10, PD-1, and CTLA4).

Increasingly complex synthetic biology approaches as described by Sukumaran and colleagues represent the “killer apps” of T-cell engineering. It is inconceivable that such complexity could be incorporated into a small molecule or a protein therapeutic; consequently, T-cell engineering will find increasing application in cancer therapy.

No potential conflicts of interest were disclosed.