Molecular Pathways: Breaking the Epithelial Cancer Barrier for Chimeric Antigen Receptor and T-cell Receptor Gene Therapy

TCRs but not CARs can target intracellular antigens

The antigen recognition mechanisms of CARs and TCRs differ in crucial ways that affect which antigens they can target and how tumors might evade attack. CARs, through their ScFv domain, recognize membrane-bound cell surface antigens (Fig. 1). In contrast, TCRs recognize peptides that are generated by intracellular processing of proteins and presented to T cells by histocompatibility complex (MHC) molecules (Fig. 1). CAR recognition of targets is not MHC restricted; all tumors with the target antigen, regardless of human leukocyte antigen (HLA) haplotype, can be targeted. Because CARs recognize antigens that are not presented by MHC molecules, tumors cannot evade recognition through downregulation or loss of antigen processing and presentation genes such as transporter associated with antigen processing, proteasome subunits, or HLA molecules. However, CARs can bind to soluble antigens, which, if present in blood, can inhibit the intended recognition of membrane-bound molecules. In addition, most CARs have some degree of inherent ScFv clustering that can trigger tonic signaling, premature T-cell exhaustion, and loss of antitumor activity (8). In contrast, TCRs are not inhibited by soluble antigen because their targets are presented by MHC molecules, and they do not exhibit substantial tonic signaling because they are composed of physiologic TCR signaling proteins (Figure 1).

Advantages to targeting tumor-specific antigens

Antigen receptor gene therapy is based on the infusion of high numbers of T cells directed against a tumor antigen. The strength of this T-cell attack can be further augmented by host conditioning with chemotherapy that decreases negative regulatory cells and increases homeostatic cytokines (9, 10). The potency of the treatment can be additionally increased by the administration of exogenous cytokines (10). The intent of the treatment is to completely eliminate every cell expressing the target antigen with a single infusion of T cells, and in some cases this goal appears to be achieved. However, the ability of antigen receptor gene therapy to destroy target cells located throughout the body has important consequences for target antigen selection.

A review of clinical experience with various target antigens is informative. CD19 (a B-cell antigen) CAR therapy has been reported to eliminate not only malignant B cells but also normal B cells, which have broad tissue distribution (3). Melanoma antigen recognized by T cells 1 (MART1) and gp100 (melanocyte antigens) TCR therapy were shown to mediate not only melanoma regression but also sometimes-severe injury to melanocytes populating skin, eyes, and ears, which cause rash and vision and hearing loss (11). Carcinoembryonic antigen TCR therapy, which targeted an antigen expressed by both healthy colonic epithelial cells and colon cancers, was found to induce severe diarrhea that led to discontinuation of a clinical trial after three patients (12). Finally, carbonic anhydrase IX (CAIX) CAR therapy, which was directed against an antigen present on the surface of renal cell carcinomas and biliary epithelium, was found to cause cholangitis by CAR-mediated recognition of CAIX on the epithelial cells of normal bile ducts (13, 14). From these studies it is evident that antigen receptor gene therapy can induce destruction of diverse target cells regardless of their location. One implication of this finding is that, for treatments that are intended to induce complete regression of metastatic epithelial tumors, it may be important to target antigens that are expressed by tumors but not by vital healthy tissues.

Targeting tumor-specific antigens that are shared between patients

Few antigens are shared by the tumors of different patients and not expressed by healthy tissues (15). Cancer germline antigens are expressed in development but silenced in nongermline adult tissues, and reexpressed by some cancers. Successful treatment of some patients with melanoma or synovial sarcoma with TCR gene therapy targeting the germline cancer antigen NY-ESO-1 has been reported (7). Mutated gene products can also be attractive targets, especially when they are oncogenic drivers. Certain oncogenic Kirsten rat sarcoma viral oncogene homolog (KRAS) mutations occur at relatively high frequency in pancreatic, colorectal, lung, endometrial, ovarian, and prostate cancers. A TCR recognizing an HLA-A11 restricted epitope of the G12D KRAS variant has been described and may permit the first KRAS-targeted TCR gene therapy (16). Perhaps the most attractive targets for T-cell therapy in solid tumors are the viral antigens of tumors such as cervical, oropharyngeal, and anal cancer [caused by human papillomavirus (HPV)] or undifferentiated nasopharygeal carcinoma [uNPC; caused by Epstein–Barr virus (EBV)]. In a recent clinical trial, adoptively transferred tumor-infiltrating T cells from cultures selected for their recognition of HPV oncoproteins mediated durable, complete tumor regression in some patients with HPV⁺ cervical cancer (17). The magnitude of HPV reactivity of the infused cells correlated with objective tumor response, but “bystander” cells with non-HPV specificities were also infused and may have played a role in the clinical responses. A TCR targeting an HLA-A2-restricted epitope of HPV-16 E6 has been identified, and T cells expressing this receptor demonstrate recognition of cervical and oropharyngeal cancer cell lines (18). Results of an ongoing clinical trial of TCR gene therapy with this receptor in cervical, oropharyngeal, anal, vaginal, vulvar, and penile cancers will more directly interrogate the HPV-16 E6 oncoprotein as a therapeutic target (NCT02280811). As with HPV, EBV can drive oncogenesis, and adoptive transfer of EBV-specific T cells has been studied and clinical activity reported in hematologic malignancies (19, 20) and possibly undifferentiated nasopharyngeal carcinoma (21).

Inhibition of negative regulators of T-cell function

Effective treatment of metastatic epithelial cancers by CAR or TCR genetically engineered T cells may require enhancement of T-cell function. Enhanced T-cell function can be achieved through reduced expression or blockade of inhibitory molecules, or through increased expression of stimulatory molecules. Decreased gene expression can be accomplished by strategies that target either genomic DNA or mRNA. Emerging genome editing technologies can precisely introduce indels to prevent gene expression, or replace nucleotide sequences to modify gene expression (Fig. 2; ref.22). At least four types of programmable nucleases for genome editing have been described: meganucleases, zinc finger nucleases (ZFN), transcription activator-like effector nucleases (TALEN), and CRISPR-associated nuclease Cas9. Although programmable nuclease technologies hold great promise for manipulating gene expression in therapeutic T cells, their clinical development is in its infancy. A small clinical trial has tested ZFN editing of CCR5 for the treatment of patients with human immunodeficiency virus (HIV), suggesting the clinical feasibility of this approach in humans (23). Recombinant Cas9 protein complexed with an in vitro transcribed single-guide RNA (RNPs) has been reported to efficiently edit primary human CD4⁺ T-cell CXCR4 and PD-1 genes (24). A megaTAL nuclease introduced with adeno-associated virus-mediated delivery of a CCR5-targeting template has also been described to efficiently modify CCR5 in human T cells (25). Genome editing strategies have the advantage that they can completely eliminate expression of a functional gene product in some cells but the disadvantages that the platforms for high-efficiency editing and scaled up clinical application may require further development.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Emerging technologies for gene regulation and potential target genes for modified expression in therapeutic T cells. Expression of genes with inhibitory effects on T-cell function can be reduced or eliminated from some cells by RNA silencing or gene editing technologies. Programmable nucleases (light blue), such as CRISPR-Cas9, TALENs, ZFNs, and meganucleases, can edit a particular gene, changing the genomic sequence and permanently preventing expression. Inhibitory RNAs (magenta) can target an mRNA transcript for reduced expression. The target genes for these systems may be inhibitory receptors (green), such as PD-1 or CTLA-4, or antagonists of TCR or cytokine signals, such as CISH, CBL-B, or SHP-1. Other technologies, such as ribozyme switches, transcriptional activators, and dimerizing receptors, use small molecules to activate expression of genes that stimulate T-cell function, such as cytokines or engineered stimulatory receptors. Small-molecule dimerizers can also be used as safety switches to induce cell death in the event of life-threatening T cell–mediated toxicity (i.e., iCaspase 9).

Another approach to inhibiting gene expression is to increase degradation of a target mRNA through RNA interference with short hairpin RNA or artificial microRNA (Fig. 2). These technologies are easily adapted to clinical application by integrating expression of the targeting RNA into established clinical gene transfer systems. They also have the advantage that tandem hairpin designs may permit simultaneous targeting of multiple genes. A disadvantage is that this approach can decrease but not completely eliminate expression of a gene.

Potential targets for gene silencing or knockdown, or mAb blockade

A host of molecules have been reported to inhibit T-cell function, some of which have been studied in mouse models or clinical trials of T-cell-based cancer therapy. Much of this work has centered on inhibitory receptors expressed by T cells. mAbs that block interactions of the inhibitory receptor programmed death 1 (PD-1), with its ligands, programmed death ligand 1 (PD-L1) and programmed death ligand 2 (PD-L2), have clinical activity in melanoma, non–small cell lung cancer, renal cell carcinoma, urothelial cancer, head and neck squamous cell carcinoma, and other tumors (26). PD-1 axis blockade with mAbs also has been reported to improve adoptive T-cell therapy in mouse models of CAR and TCR therapy (27–29). Hence, PD-1 is an attractive molecule to target in combination with antigen receptor gene therapy. Another T-cell inhibitory receptor that has been targeted in cancer therapy is CTL antigen 4 (CTLA-4). Inhibition of CTLA-4 binding to its ligands, CD80 and CD86, can induce regression of melanoma and renal cell carcinoma (30, 31). A clinical trial for melanoma that combines CTLA-4 blockade with tumor-infiltrating lymphocyte (TIL) infusion is ongoing (NCT01701674); tumor response in 5 of 11 patients has been reported (32). Checkpoint blockade with the combination of anti-PD-1 and anti-CTLA-4 mAbs has greater clinical activity than blockade with either agent alone in melanoma (33). Dual PD-1 and CTLA-4 blockade combined with adoptive T-cell therapy is a potentially interesting area for further exploration (34). Further study in animal models and in clinical trials will be required to determine the optimal combinations of inhibitory receptors to antagonize. Emerging data also support strategies to inhibit intrinsic regulators of TCR and cytokine signaling, such as Src Homology Region 2 Domain-Containing Phosphatase 1 (SHP-1; refs.35, 36), cytokine inducible SH-2-Containing Protein (CISH; ref. 37), or E3 ubiquitin-protein ligase CBL-B (38, 39).

Controlled overexpression of genes that stimulate T-cell function

The function of antitumor T cells for adoptive transfer may be improved by transgenic expression of molecules that enhance T-cell activation and proliferation. It may be important to have an element of control over the timing and magnitude of expression of these molecules and the survival of the cells that express them. For example, constitutive IL15 transgene expression enhances the antitumor function of T cells in a mouse model of TCR gene therapy, but some mice die from delayed hyperproliferation of the infused cells (40), and human T cells transduced to constitutively express IL15 can display uncontrolled proliferation in vitro (41). Thus, although transgenic IL15 expression can improve T-cell function, its expression may need to be controlled. As another example of increased treatment risk from transgenic cytokine expression, in a clinical trial of TIL for melanoma in which T cells were transduced to express single-chain IL12 under a nuclear factor of the activated T-cell (NFAT) promoter, tumor regression occurred in some patients, but IL12-related toxicities prevented further development of the approach (42).

Technologies to modulate transgene expression or activation, or to terminate T-cell survival in vivo, might be required. One system for controlling T-cell stimulatory signals is to administer T cells that express an engineered costimulatory receptor that is reversibly dimerized by a small molecule (i.e., rimiducid) that is given systemically to the patient (Fig. 2; ref. 43). Another strategy is to use the same type of system as a “suicide gene” to induce cell death through dimerization of inducible caspase 9 (iCasp9; Fig. 2). This approach was reported to eliminate donor-derived iCasp9-engineered T cells in patients with graft-versus-host disease in a stem cell transplantation clinical trial (44). Other technologies in which transgene expression can be controlled by small molecules include the RheoSwitch inducible promotor system (45) and synthetic ribozyme switches; these technologies require vetting in clinical trials but have the potential to allow titration of transgene expression in vivo (Fig. 2; refs. 46, 47).