The structure of a prototypical CAR can be divided into three segments: (1) Single chain fragment of variable (scFv) which determines the specificity, (2) extracellular spacer, and (3) intracellular signaling domain(s) (Fig. 3.2). Other aspects of CAR design may impact the ability to recycle T cell effector functions in the tumor microenvironment such as the transmembrane domain [76], where the CAR binds the TAA [77], and the density of CAR on the T cell surface.

scFv

The scFv consists of the variable regions of heavy and light chains from a mAb with desired specificity for TAA. Indeed, this determines the specificity of the CAR. Many cell surface TAAs expressed on a variety of tumor cells have been targeted by CARs using scFvs from specific mAb (Table 3.1). In addition to these specific scFvs, a universal CAR has recently been described [78, 79]. Here, scFv from anti-FITC or avidin dimer was used to construct CARs. For example, after infusing T cells expressing a fluorescein-specific CAR, multiple cell surface molecules expressed on tumors can be targeted by infusion of FITC-labeled mAbs. This technology enables the target tumor antigens to be changed after infusing CAR-expressing T cells.

Spacer

The spacer domain is gaining importance in its ability to influence the ability of T cells to target tumor and protect the genetically modified T cells from activation-induced cell death [80]. A panel of spacers has been used including the CH₂–CH₃ Fc hinge domain from IgG1 or IgG4, CD8α, as well as small hinges. The IgG1 CH₂–CH₃ Fc hinge domain is widely used spacers, but the potential to bind to the Fcγ receptor raises concerns for deleterious side effects. Modification of this construct may avoid this issue [81]. Information regarding the impact of spacer is beginning to be published. For example, CAR-mediated activation is less efficient in T cells expressing a CAR containing a spacer sequence (IgG1 Fc domain) compared to CAR containing no spacer sequence [82]. This report was restricted to the use of first-generation (described later) signaling domain. The role that spacer sequence plays in the function of a second- or third-generation CAR constructs is being evaluated.

Intracellular domain

CD3ζ or Fc receptor γ has multiple ITAM motifs to transduce an activation signal in T cells and as such have been widely used as intracellular signaling domains. Construct containing just these ITAMs are referred to as a “first-generation” CARs. Although these signaling domains can induce an activation signal in T cells, they are usually suboptimal and curtail the therapeutic impact after adoptive transfer [83–87].

Co-stimulatory molecules that provide a “2nd signal” are required to trigger a fully competent T cell activation event. To improve CAR-mediated signaling, multiple co-stimulatory molecules have been combined with chimeric CD3ζ. CD28 is one of the well-known co-stimulatory molecules which are important for T cell survival [88]. Enforced expression of CD28 in CD28^neg virus-specific CTL clones restores production of IL-2 and proliferation in these cells [89]. CD28 was first tested as a single intracellular domain instead of CD3ζ [90, 91]. T cells transduced with this construct showed increased IL-2 production and proliferation. Following this observation, Maher et al. generated prostate-specific membrane antigen (PSMA) target CAR containing both CD28 and CD3ζ signaling domain (“2nd generation”) [92]. T cells expressing this CAR construct produced increased amounts of IL-2, showed increased proliferation, and maintained specificity for target PSMA. The authors also showed that the functional activity was dependent on the order of CD28 and CD3ζ within the CAR. Improved function was observed in the CAR having transmembrane and proximal intracellular domain of CD28 fused to CD3ζ. The improved function afforded by the CAR design that employs CD28 and CD3ζ CAR has been confirmed in vitro [83, 93] and in the clinical setting [94]. Co-stimulatory molecules belonging to the tumor necrosis factor (TNF) ligand superfamily (e.g., 4-1BB (CD137), OX40 (CD134), and CD27) are also important for co-stimulation of T cells and promote long-term T cell survival [95, 96]. These molecules associate with TNF receptor-associated factor (TRAF) and link to many molecules which are important for cell survival (e.g., BCL2, BCL-X, phosphoinositide 3 kinase, AKT, and NFkB pathway) [97]. As anticipated, CAR constructs containing chimeric CD134 [98, 99] or CD137 [100–102] have been shown to sustain in vivo persistence of CAR⁺ T cells. Combining signaling from multiple signaling molecules, such as CD3, CD28, and CD137 (or CD134) to form a third-generation CAR, has also been tested [103–105]. Several groups have reported that CAR containing three activation motifs have potent anti-tumor efficacy in models [106, 107]. A CAR signaling through chimeric CD27 has also been shown to result in improved function and in vivo survival, comparable to CAR containing CD137 [108]. In addition to anti-apoptotic molecules, CD27 has been reported to activate the Wnt pathway in chronic myeloid leukemia stem cells [109]. Since the Wnt pathway has recently been identified as the key signaling pathway of stem-cell-like memory T cells [110], it will be intriguing to see how the Wnt pathway is regulated in T cells expressing CAR that include the CD27 signaling domain. Although much progress has been made in the intracellular signaling domain of CARs, many of these constructs still need to be evaluated in clinical trials.

Targeting one TAA that is exclusively expressed on tumor cells is difficult, especially given the heterogeneity of antigen expression on solid tumors. An approach to increase the specificity of genetically modified T cells is to synchronously target two cell surface TAAs expressed on tumors, thereby enabling CAR⁺ T cells to distinguish tumor cells from normal cells. This can apparently be achieved using two CAR designs which activate T cells via CD3ζ on one CAR and a co-stimulatory molecule (CD28 or CD137) on the other CAR. Kloss et al. developed CARs targeting PSMA and prostate stem cell antigen (PSCA) as model antigens and showed that T cells expressing both anti-PSMA CAR (CD28 + CD137) and low-affinity anti-PSCA CAR (CD3ζ) kill only those tumor cells that express both TAAs [111]. A probable limitation is that the affinity of scFv linked with CD3ζ should be low as described by the authors. The antigen density and affinity of scFv vary with each CAR to tumor combination, and therefore, this needs to be taken into consideration when applying this strategy to the clinical setting.

Autologous T cells are infused in many current clinical trials, but T cells from heavily pre-treated patients may be damaged in quality and compromised in quantity, and (due to the time required for production and manufacture) they may not be available where and when the patient needs them. Any time delay before infusion allows disease progression to take place, and the possibility that treatment is therefore not possible for some patients. “Off-the-shelf” (OTS) therapy administering T cells expressing a CAR would resolve these issues. Toward this end, CAR⁺ T cells have been pre-prepared from healthy donors, using either peripheral blood mononuclear cells (PBMC) or umbilical cord blood (UCB) derived mononuclear cells, for infusion when the patient needs them rather than when the T cells are available. A potential problem using OTS T cells is the development of an allogeneic immune reaction (either from donor to patient or from patient to donor). Zakezewski et al. generated precursor CAR⁺ T cells by in vitro differentiation of CAR-transduced HSCs activated and propagated on Notch-1 ligand expressing OP9 cell lines [112]. When precursor CAR⁺ T cells were infused into irradiated mice they became host MHC restricted and host tolerant. This technology may be used to generate “OTS” CAR⁺ T cell precursors.

We have shown that the endogenous TCRαβ can be eliminated from CAR⁺ T cells by ZFNs targeting either TCRα constant region or TCRβ constant region [113]. These TCRαβ^neg CAR⁺ T cells can then be enriched by clinically compatible paramagnetic bead-based sorting. TCRαβ^neg CAR⁺ T cells cannot proliferate by TCRαβ stimulation while maintaining activity through CAR. We have also shown that HLA can be eliminated from CAR⁺ T cells, thus enabling these cells to evade attack from patient T cells recognizing allogeneic antigens. Although further evaluation is needed, CAR⁺ T cells may be used as an OTS “drug” in the growing field of immunotherapy.

Multiple phase I clinical trials administering T cells expressing first-generation CARs have been reported [84–87, 114, 115]. Overall clinical responses in these studies are limited presumably due to incomplete T cell activation. To provide additional stimulation to these T cells, Pule et al. reported that half the patients treated with anti-GD2 CAR expressed in EBV-specific T cells showed clinical responses. These EBV-specific T cells presumably retained sufficient stimulation through the EBV-specific TCRαβ to enable targeting of tumor via first-generation CAR. This observation reinforces that co-stimulation is important to improve anti-tumor function.

The results of early-phase clinical studies using second-generation CAR (including CD28 or CD137 intracellular domain) are promising [94, 116–121]. The improved anti-tumor responses appear to correlate with sustained in vivo persistence of CAR⁺ T cells. For example, Kalos et al. reported that all three patients with advanced chronic lymphocytic leukemia who received CD19-specific T cells expressing a CAR that signaled through chimeric CD137 showed marked clinical responses [117]. It was calculated that on average one CAR⁺ T cell can act as a serial killer and eradicate 1,000 tumor cells. Most adverse events related to CAR⁺ T cell infusion were tolerable with intensive clinical support. Expected on-target toxicity occurred as these CAR⁺ T cells do not distinguish between CD19 on malignant versus normal B cells. Thus, eradication of normal B lineage has served as a biomarker and has been observed. The loss of humoral immunity can lead to complications as one patient has died from viral infection after receiving CD19-specific T cells [119]. However, unexpected on-target related severe adverse events have also reported. In a trial targeting carbonic anhydrase IX (CAIX) CAR T cells were attributed to causing hepatotoxicity due to the low level of CAIX expression on the liver [115]. One patient with metastatic colon cancer treated with ERBB2 target CAR therapy suffered from fatal respiratory distress due to the low level of expression of ERBB2 on lung epithelial cells [122]. Another unforeseen, an adverse event was also reported in a patient with CLL treated with CD19 target CAR. Although the precise cause of death was not reported, this patient showed elevated serum cytokine levels after infusion of CAR⁺ T cells [123].

In vivo survival of infused T cells is important to eradicate residual tumor cells and maintain immune surveillance. BCL2 and BCL-XL are anti-apoptotic molecules that contribute to T cell differentiation and survival [124]. Pro-survival function of cytokines, especially cytokines that signal through the common γ chain, relies on the induction of BCL2 or BCL-XL expression [125]. BCL-XL is also an anti-apoptotic molecule reported to be associated with T cell survival mediated through CD28 co-stimulation [88]. Enforced expression of BCL2 in TAA-specific T cells has been tested in vitro and in vivo [126, 127]. These T cells have augmented resistant to apoptosis, survive in the absence of cytokine support, and exhibit enhanced anti-tumor activity in a mouse model. Genetic modification with BCL-XL protects T cells from Fas-mediated death and enhances survival in vitro and in vivo [127, 128]. Fas-mediated T cell death can be also reduced by introduction of siRNA targeting Fas [129]. hTERT is the catalytic subunit of telomerase and overexpression of hTERT prevents the shortening of a chromosomal ends after multiple cell divisions. Because of this, transduction of hTERT is used for immortalization of cells. hTERT-transduced T cells have been evaluated and found to have enhanced survival [130]. The clinical translation of all these approaches to improve T cell survival needs to be balanced against the risk of infusing a product capable of autonomous proliferation. Thus, at a minimum, these T cells will need to be engineered for conditional ablation through the expression of a suicide gene.

T cell survival is dependent on factors intrinsic to the T cells and extrinsic as exerted by microenvironment. Cytokines that signal through the common γ chain are indispensable for T cell expansion and survival [131]. These cytokines have been shown to enhance in vivo T cell survival animals and human trials [132–134]. However, the exogenous administration of supraphysiologic concentrations of recombinant human cytokines is expensive, requires repeated injection, and can cause toxicity. Genetic engineering of T cells with enforced expression of cytokines may be an alternative approach to enhancing T cell persistence. The enforced expression of IL-15 has been tested in CD19-specific CAR⁺ T cells. These T cells secreted IL-15 and proliferated upon encounter with TAA resulting in an improved anti-tumor effect in vivo [135]. However, prolonged expression of IL-15 may lead to genetic alteration and lead to malignant transformation [136]; thus, the co-expression of suicide genes is recommended. IL-7 is an important cytokine for homeostatic expansion of naïve and memory T cells, and IL-7Rα is the important cellular marker that identifies long-lived T cells. However, many T cells lose IL-7Rα after ex vivo culture. The transduction of IL-7Rα into EBV-specific CTL restores responsiveness to IL-7 [137]. Although enforced expression of IL-21 in effector T cells has not been published, IL-21 transduction into T cells capable of antigen presentation (T-APC) has been evaluated. IL-21-secreting T-APCs enhance the generation and proliferation of MART-1-specific CTLs [138]. IL-12 is a heterodimer consisting of a light chain (p35) and a heavy chain (p40) [139]. IL-12 has an important role in T cell priming, effector function, and survival [140]. Several groups have transduced TAA-specific T cells to express IL-12. This resulted in the enhanced cytotoxic function of T cells, but IL-12 alone did not sustain the long-term survival of T cells [141]. Of note, IL-12 expressing CAR T cells acquired resistance to suppression mediated regulatory T cells [142], attracted activated macrophages in tumor site [143], and also reduced tumor infiltration by myeloid suppressor cells [144]. Thus, it appears that IL-12 from T cells has the potential to disrupt a suppressive tumor microenvironment as well as to enhance effector functioning.

The long-term efficacy of adoptive T cell therapy is subject to immune checkpoint regulation. Interaction of inhibitory molecules on activated T cells and their ligand on tumor cells compromises T cell function. These checkpoints include CTLA-4, PD-1, TIM-3, and LAG3. Such molecules can be targeted by therapeutic mAbs. Promising clinical results infusing mAbs that block the interaction of CTLA-4 with B7.1 (or B7.2) [145, 146] or the interaction between PD-1 and PD-L1 (or PD-L2) [147, 148] in patients with solid tumors have been reported. Combining these mAbs with genetically engineered T cells may enhance their therapeutic efficacy. Alternatively, inhibitory molecules expressed on T cells may also be eliminated by post-transcriptional suppression with shRNA. This has been tested by suppressing PD-1 expression on T cells. Increased production of IFNγ and T cell degranulation was noted although PD-1 suppression was incomplete [149]. With the emergence of gene targeting technologies such as ZFN [150], TALEN [151], and CRISPR/Cas9 [152], these molecules may be targeted to completely disrupt their expression.

Thymidine kinase is involved in the pyrimidine salvage pathway. Herpes simplex virus type 1 thymidine kinase (HSV-TK) is desirable for gene therapy because of its broad substrate specificity (e.g., pyrimidines, pyrimidine analogs, and purine analogs). Of these potential substrates, ganciclovir (GCV) is favored for TK-mediated suicide gene therapy because of the need for a low drug concentration and that it is clinically available. The administration of GCV leads to DNA chain termination and cell death in cells expressing HSV-TK. The first use for the HSV-TK gene in T cells was to control GVHD after DLI in allogeneic HSCT [158]. Since this first study, hundreds of patients have been treated using this technology in phases I–II studies. Control of GVHD was observed in all patients who developed GVHD [159]. The ability to control GVHD with HSV-TK has also been proved in the more challenging setting of HLA-haploidentical transplantation [160]. In addition to use as a suicide gene, HSV-TK can be used for noninvasive imaging. T cells expressing variants of HSV-TK can be tracked by injection of radiolabeled 2′-fluoro-2′-deoxy-1-beta-D-arabinofuranosyl-5-iodouracil (FIAU) as imaged by positron emission tomography [161]. This is achieved by the trapping of FIAU in the T cell cytoplasm after phosphorylation by HSV-TK [162]. A drawback to HSV-TK is its immunogenicity and the myelosuppression due to GCV infusion. Mutants of HSV-TK have been developed to overcome GCV mediated myelotoxicity. These need lower concentrations of GCV than wild-type HSV-TK gene, thus reducing the potential for myelosuppression [163, 164]. Immunogenicity of HSV-TK leads to the induction of HSV-TK-specific T cells in the host which compromises persistence of the infused genetically modified T cells [165]. A variant of the less immunogenic human-derived thymidine monophosphate kinase (TMPK) can also be used to conditionally ablate T cells upon addition of 3′-azido-3′-deoxythymidine (AZT) [166]. One concern is the amount of AZT that is needed to achieve ablation of T cells if used in humans.

Another class of suicide genes is based on the use of a small molecule to dimerize recombinant Fas. This uses a variant of FK506-binding protein (FKBP) fused to Fas and AP1903 that is a chemical inducer of dimerization (CID). T cells expressing this fusion can be conditionally eliminated by administration of CID as a result of cross-linked Fas initiating caspase-dependent apoptosis. This construct has been tested in human [136] and macaque T cells [137]. The extent of drug-mediated T cell death was comparable to that of HSV-TK-transduced T cells (up to 90 %), but the pace of inducing apoptosis in this system was improved. Moreover, because this system is derived from human proteins, this fusion transgene may avoid the immunogenicity of HSV-TK. Straathof et al. generated an alternative using caspase 9 instead of Fas intracellular domain [135]. Several constructs were evaluated, and the construct of one copy of FKBP fused to large and small subunits of caspase 9 (referred to as iCasp9) was preferred. A clinically appealing approach has been developed to generate T cells that all express the iCasp9 and thus are all susceptible to conditional ablation. This was based on co-expression of iCasp9 with recombinant CD19 and using paramagnetic beads to positively select successfully genetically modified T cells with mAb against CD19. The initial clinical reports infused haploidentical CD19⁺ (iCasp9⁺) T cells after HSCT [136]. Four out of 5 patients infused with transduced T cells developed GVHD. Their GVHD resolved after one infusion of CID due to the rapid elimination of the genetically modified transduced T cells. Although further clinical experience with this suicide gene therapy is necessary, iCasp9 may be the best alternative to HSV-TK.