Making Better Chimeric Antigen Receptors for Adoptive T-cell Therapy

Over the past few decades, our understanding of the role of the immune system in cancer has grown considerably and so has the technology to purify and manipulate specific immune cell types with a goal of treating disease. The transfer of genetically engineered immune cells as a form of cellular therapy has been investigated as a treatment option for HIV and cancer (1, 2). Chimeric antigen receptor (CAR) T-cell therapy uses gene transfer technology to reprogram a patient's T cells to express CARs (Fig. 1), thereby directing cytotoxic potential of T cells against tumor cells that would otherwise be ignored (3). CARs are engineered fusion proteins that contain an extracellular antigen-binding domain composed of a single-chain variable fragment derived from an antibody and intracellular signaling domains, which are involved in the initiation of T-cell signaling and downstream T-cell effector functions (4). First-generation CARs consisted of only the T-cell receptor complex CD3ζ chain domain and antigen recognition domains, showed minimal clinical success, and were characterized by very low levels of engraftment in patients (5, 6). Second-generation CARs containing costimulatory domains, typically either CD28 or 4-1BB, were hypothesized and shown to augment CAR T-cell survival and proliferation (7–9). The inclusion of a costimulatory domain dramatically increased the antitumor efficacy and persistence of CAR T cells (3, 10, 11). Interest and investment in the development of CAR T-cell therapy is rapidly increasing in both academia and industry, with multiple ongoing clinical trials as well as many expectations for the future of the field. Although CAR T-cell therapies are on a fast track to approval by the FDA for B-cell malignancies, there is active investigation into building better CAR T cells for treating hematologic malignancies and solid tumors.

The most clinical data using CAR T-cell therapy have been generated with CD19-specific CAR T cells in patients with relapsed or refractory B-cell malignancies, many of whom have no curative option other than hematopoietic stem cell transplant. CD19 is highly and uniformly expressed on B cells, starting early in development and continuing through all mature stages except plasma cells. CD19 is also expressed on B-cell malignancies developing in the bone marrow [B-cell acute lymphoblastic leukemia (ALL)] and secondary lymphoid organs (chronic lymphocytic lymphoma, diffuse large B-cell lymphoma, and follicular lymphoma; ref.12).

The CD19 CAR T-cell field has become highly competitive in recent years, with several pharmaceutical companies developing partnerships with academic institutions. Complete response rates of approximately 90% have been observed in both pediatric and adult patients with relapsed or refractory ALL who were treated with CD19 CAR T cells expressing either a CD28 or a 4-1BB costimulatory domain (13–17). Overall response rates of 50% to 100% have been observed recently in patients with diffuse large B-cell lymphoma, follicular lymphoma, or chronic lymphocytic lymphoma who were treated with the CD19 CAR T-cell therapy CTL019 (18, 19). While CD19 is not highly expressed on terminally differentiated plasma cells, CD19 CAR T cells may have clinical benefit in patients with multiple myeloma; this is hypothesized to be due to the continual repopulation of malignant plasma cells from a malignant B-cell precursor (20, 21). Although CD19-specific CAR T cells have demonstrated considerable efficacy in B-cell malignancies, treatment of other hematologic malignancies will require further CAR T-cell target identification and validation. Targets currently under investigation are summarized in Table 1. 

A major differentiating factor among the current CARs that have been investigated in clinical trials has been the level of CAR T-cell persistence. Early studies using CAR T cells did not include a lymphodepletion step, which may have contributed to their very short persistence and poor antitumor activity. Lymphodepletion is now included in most CAR T-cell therapy protocols; however, variable levels of persistence are observed both between and within clinical trials. One strategy to improve lymphodepletion and CAR T-cell persistence is through increasing the intensity of lymphodepletion, which leads to depletion of regulatory T cells and greater engraftment of the infused T cells (22–24). The selective depletion of regulatory T cells along with administration of cytokines to support CAR T-cell function is also under investigation (25). Further research is needed to determine which lymphodepletion methods result in optimal CAR T-cell persistence and antitumor benefit.

An important question in the field is the degree to which the costimulatory domains included in the CAR affect CAR T-cell persistence. A recent study showed that certain CAR T cells containing a CD28 costimulatory domain increased the expression of T-cell exhaustion-related genes, while the 4-1BB (TNFSF9) costimulatory domain with the same antigen specificity ameliorated this exhausted phenotype (26). This may explain why in recent clinical trials of patients with relapsed or refractory ALL, CAR T cells expressing a CD28 domain have been reported to persist for up to 3 months, while CAR T cells with a 4-1BB domain persist for up to 5 years, and more than 6 months in most cases (13, 17, 19). Other costimulatory domains have also been investigated for inclusion in CARs, most notably OX40 (TNFRSF4) and ICOS, which both resulted in improved target cell lysis in vitro when compared with CARs lacking a costimulatory domain (8, 27). Indeed, third-generation CARs containing two costimulatory domains have recently entered into clinical trials. While it is possible that future CARs incorporating multiple costimulatory domains will result in increased CAR T-cell antitumor activity, more studies are required to better understand the kinetics of each of the costimulatory domains and their relative effects in the clinic. Incorporation of large numbers of signaling domains may also lower vector titers due to large transgenes, lower the expression of the CAR at the cell surface, or result in otherwise decreased functionality due to the requirement for accessory signaling apparatus proximal to the cell membrane.

The optimal duration of persistence of CAR T cells is unknown, and may in fact be different for CD19-directed CAR T cells than for CAR T cells directed to other malignancies. Long-term persistence of CD19-directed CAR T cells has both the advantage of ongoing disease surveillance and the disadvantage of long-term B-cell aplasia. In addition to the CAR-encoded signaling domains, other factors may affect the persistence of CAR T cells, such as the cell culture system used during manufacturing, the mode of gene transfer and associated promoters, and the functionality and phenotype of the input T cells, which in turn may be affected by age, disease, and prior therapies.

A method to potentially increase the proliferation and long-term activity of CAR T cells in vivo involves using T cells specific for an antigen associated with chronic viral infection as the starting cell population for CAR T-cell manufacturing. In a murine model using cytomegalovirus (CMV)-specific T cells that were engineered to express a CD19-specific CAR, vaccination of the mice with CMV peptide resulted in enhanced proliferation and antitumor activity of the CAR T cells due to stimulation through the endogenous CMV-specific T-cell receptor (28). Similar approaches have been investigated using CAR T cells derived from Epstein–Barr virus- or adenovirus-specific T cells (29, 30). Several trials are ongoing to investigate the safety and efficacy of virus-specific CAR T cells in patients, including NCT00709033, NCT01430390, and NCT01109095.

In some cases, particularly in the development of novel targets, transient CAR T-cell persistence may be desired. CARs can be expressed for a short duration through transfection of T cells with mRNA encoding the CAR, instead of using a viral vector, which permanently integrates the CAR into the genome (Fig. 2A). RNA transfection is a fast and efficient procedure, requiring only 1 day of T-cell activation before transfection and resulting in very high (>40%) transfection efficiency (31). As mRNA is unable to integrate into the host genome, mRNA transfection results in a short-lived population of CAR T cells, which would require multiple doses to achieve an AUC that could be considered a therapeutic dose. Suboptimal scheduling of doses may result in generation of an immune or allergic response against the CAR, especially if a murine single-chain variable fragment is used (32). CARs transfected into T cells using mRNA are currently being investigated in early clinical trials at the University of Pennsylvania (Philadelphia, PA; NCT02624258, NCT01837602, NCT02277522, NCT02623582).

One of the beauties of CAR T cells is that they are “living drugs”: once infused, physiologic mechanisms maintain T-cell homeostasis, memory formation, and antigen-driven expansion. However, imperfect human intervention may lead to T cells that target an undesired tissue or proliferate to greater levels than necessary and therapeutic. As CAR T cells become incorporated into standard therapies, it may be useful to design them with patient- or physician-controlled persistence mechanisms, either “ON” switches or “suicide” switches. For technical reasons, suicide switches are easier to incorporate into T cells. One of the fastest acting and clinically tested suicide gene strategies is the inducible caspase-9 (iCasp9) system (Fig. 2B; ref.33). Cells transduced with iCasp9 can be depleted by administration of a synthetic small molecule that dimerizes iCasp9 promolecules, triggering activation of the apoptotic pathway (34). Induction of iCasp9 dimerization and T-cell depletion via the administration of the small-molecule AP1903 is a strategy that has been used in patients with graft-versus-host disease, demonstrating the feasibility of this approach (35).

CAR T cells may also be depleted through the coexpression of a protein for which a depleting antibody is already in clinical use, that is, CD20 or EGFR (Fig. 2C). Administration of the depleting antibody is expected to deplete target-expressing CAR T cells if therapy-related toxicity arises or if “cure” has been achieved and maintained (36). To our knowledge, neither of these strategies have been clinically tested as a CAR T-cell–depleting method in patients, but CARs containing an EGFR transgene are currently under clinical investigation at several centers (NCT02028455, NCT02159495, NCT01865617). Most investigators have preferred to manage toxicities with either cytokine blockade or corticosteroids, or both, rather than permanently ablate a potentially curative (and expensive) therapy.

A recent article demonstrated proof-of-concept for the first “ON-switch”-based CAR T cells; here, the signaling apparatus of the CAR was separated and each end was fused to a dimerizing domain similar to the basis of iCasp9, where the full CAR is reconstituted only in the presence of a tacrolimus-based drug (37). The clinical feasibility of such a system is likely to yield interesting results.

Despite the fast-track regulatory pathway for CD19-directed CAR T cells, there are still many areas of CAR design and genetic modifications of T cells that could broaden the therapeutic window and applicability of genetically modified T cells for adoptive immunotherapy. For example, CARs have typically contained a single-chain variable fragment sequence derived from a mouse antibody, but humanized antibody fragments may be less immunogenic; this may be particularly important for CAR T cells directed to any antigen other than B cells, because serologic immune responses to the CAR could limit their functionality (38). A particularly exciting approach will be to target multiple antigens, so that Boolean gating such as “antigen 1 AND 2” or “antigen 1 NOT 2” can be employed to more specifically target tumor tissues. Although these approaches are in very early stages, one could imagine engineering T cells by using either several distinct CAR constructs in one T cell or a single bispecific construct (39–41).

While CAR T-cell therapy has demonstrated high response rates in patients with leukemia or lymphoma, solid tumors present unique challenges for the use of cellular therapies. Several trials investigating CAR T-cell therapy in solid tumors have been initiated, but efficacy has been low. The best responses recently reported in trials using CAR T cells specific for the solid tumor antigens mesothelin, PSMA, or ERBB2 were stable disease in 24% to 67% of the patients (42–44). The efficacy of CAR T-cell therapy in solid tumors may be reduced due to several factors. In contrast to hematologic malignancies, the solid tumor microenvironment is composed of immune cells, endothelial cells, fibroblasts, extracellular matrix molecules, and cytokines. This microenvironment not only reduces access of modified T cells to the entire mass of a solid tumor, but also plays a role in negative regulatory signaling that may limit CAR T-cell efficacy (Fig. 3). For example, tumor stroma cells often produce molecules such as TGFβ, IL10, and indoleamine-2,3-dioxygenase, which promote suppression of an effector T-cell response by regulatory T cells (see article by Zarour in this CCR Focus; ref.45). Expressing TGFβ dominant-negative receptor II (DNRII) in T cells may be a method to reduce the influence of TGFβ on CAR T-cell therapy for solid tumors. Preclinical studies showed that cells expressing TGFβ DNRII had increased function, survival, and antitumor activity than cells that were not modified to express TGFβ DNRII (46, 47).

Solid tumor cells also often upregulate immune checkpoint ligands such as PD-L1, which dampens an effector T-cell response when engaged with its receptor PD-1 (48) and could lead to inhibition of CAR T-cell therapies in the tumor microenvironment. It is possible that the combination of CAR T cells and PD-1 blockade may yield the greatest benefit for CAR T-cell therapy in solid tumors, as CAR T cells generally target only one (or at most two) surface molecules, whereas checkpoint blockade has the potential to unleash the endogenous T-cell response, which is better equipped to sense the entire range of neoantigens resulting from tumor-specific mutations (see article by Türeci and colleagues in this CCR Focus; ref. 49). Antibodies blocking the PD-1/PD-L1 pathway have recently been approved by the FDA for use in certain solid tumors, and combination of one of these antibodies with CAR T-cell therapy may enhance the efficacy of CAR T cells in solid tumors.

An additional strategy for reducing the immunosuppressive effects of the solid tumor microenvironment involves using T cells redirected for universal cytokine-mediated killing (TRUCK), which are CAR T cells engineered to secrete the proinflammatory cytokine IL12, which can activate an innate immune response against the tumor (50). In addition, IL12 inhibits the action of regulatory T cells and myeloid-derived suppressor cells, which block antitumor T-cell responses (51, 52). However, high levels of IL12 can be very toxic, as demonstrated in clinical trials studying the use of recombinant human IL12 as a therapy (53). Therefore, if large amounts of IL12 are produced endogenously, as in the case of an infection that triggers T cells through their endogenous T-cell receptors, IL12-secreting CAR T cells may contribute to pathologic levels of IL12 in the patient.

Solid tumors can also be difficult for immune cells to penetrate if the right combination of chemokines and their receptors is not present to facilitate T-cell migration into tissues, and this may be a primary reason for difficulty in using CAR T-cell therapy for solid tumors. In an alternative CAR engineering approach, CAR T cells may be directed to the tumor through coexpression of chemokine receptors such as CXCR2 or CCR4 (54, 55). In addition, a recent preclinical study showed that macrophages residing outside of the tumor microenvironment regulate infiltration of T cells into pancreatic tumors in mice (56); strategies to reduce the immune privilege provided by intratumoral and extratumoral macrophages may enhance the success of CAR T-cell therapy in some solid tumor types.

The single greatest challenge in targeting solid tumors is the identification of suitable target antigens. Many antigens identified in solid tumors are also expressed at low levels on healthy tissues, and the negative effects of long-term CAR T-cell–mediated attack of these tissues may outweigh the antitumor benefits provided by the therapy. While these on-target, off-tumor effects are also a problem when treating B-cell malignancies, the risks of B-cell aplasia are less severe than T-cell–mediated attack of vital organs. Long-term strategies for accurately and effectively targeting solid tumors may require further engineering of the CAR or combinatorial antigen recognition approaches (40). Alternatively, the identification of antigens that arise from mutations present in the tumor but not in healthy tissues may enable the development of CARs with greater specificity for solid tumors. Table 1 shows a list of many investigational targets for CAR T cells, and details of selected targets are described below.

Signaling through EGFR promotes cell proliferation, motility, and adhesion (57). A common deletion in EGFR creates EGFRvIII, a constitutively active variant (57). EGFRvIII is not expressed by normal tissues, but is expressed in some cases of glioblastoma (58). CAR T cells targeting EGFRvIII are currently investigated in clinical trials at the University of Pennsylvania and the National Cancer Institute (NCI). This trial is examining the use of EGFRvIII-specific CAR T cells in patients with residual or recurrent glioma (NCT02209376). This study estimates an enrollment of 12 patients, with a primary completion date in 2016. Preliminary results indicate that CAR T-cell manufacturing is feasible in this patient population, and that infusions are well-tolerated in most patients. The phase I/II trial under way at the NCI is studying EGFRvIII-targeted CAR T cells for patients with refractory malignant glioma or glioblastoma (NCT01454596). The NCI study estimates an enrollment of up to 107 patients over a 7-year period, with a primary completion date in 2018. It is too early to determine clinical benefit, an endpoint that is particularly challenging to assess in patients with diseases like glioblastoma, for whom imaging-based responses are difficult to interpret and routine tumor biopsies could entail significant risk.

ERBB2 is part of the EGFR family and is overexpressed in several types of cancer. Three phase I dose-escalation trials using ERBB2-specifc CAR T cells are ongoing at Baylor College of Medicine (Houston, TX; NCT00902044, NCT01109095, NCT00889954). These trials are investigating ERBB2-specific CAR T cells in patients with ERBB2-positive malignancies, including glioblastoma multiforme and sarcoma (59). In patients with sarcoma, ERBB2-specific CAR T cells persisted for at least 6 weeks in 7 of the 9 patients treated with the highest doses of CAR T cells (42). In addition, no dose-limiting toxicity was observed in any of the patients (42), suggesting that ERBB2-specific CAR T cells may be a good therapeutic approach in patients with metastatic or recurrent sarcoma. However, the levels of engraftment and penetration into tumor were low, and clinical benefit is not obvious.

Mesothelin is a surface glycoprotein of unknown function that is overexpressed by mesotheliomas, pancreatic cancers, ovarian cancers, and lung cancers (60), and CAR T cells targeting mesothelin have also been studied in several clinical trials. Because mesothelin is expressed in several tissues, targeted treatments such as CAR T-cell therapy have the potential to cause off-target effects and should be used with caution. In a phase I clinical trial conducted at the University of Pennsylvania, patients with mesothelin-expressing tumors whose disease had progressed after first-line therapy were treated with T cells transiently transfected with mRNA for a mesothelin-targeted CAR (61). The mesothelin-specific CAR used in this study was a second-generation CAR containing a CD3ζ domain and a 4-1BB costimulatory domain. These short-lived mesothelin-specific CAR T cells displayed modest antitumor activity in 2 patients, demonstrating the feasibility of transient mRNA transfection and the utility of mesothelin as an antigen for CAR T-cell recognition (61).

An additional phase I trial ongoing at the University of Pennsylvania involves a lentiviral-transduced mesothelin-specific CAR (NCT02159716). This clinical trial, which began in June 2014, is currently enrolling patients with chemotherapy-refractory malignant pancreatic adenocarcinoma, epithelial ovarian cancer, and malignant epithelial pleural mesothelioma. Among early results from 6 patients in this trial, 4 patients have shown stable disease at day 28 following CAR T-cell infusion (43). There were no acute adverse events associated with CAR T-cell infusion, and the lentiviral-transduced mesothelin-targeted CAR T cells showed improved persistence compared with the mesothelin-targeted CARs expressed through mRNA transfection (43).

A recent preclinical study at Memorial Sloan Kettering Cancer Center (New York, NY) compared intrapleurally administered and systemically administered mesothelin-specific CAR T cells in a model of pleural malignancy (62). The intrapleurally administered CAR T cells were found to have superior antitumor activity, persistence, and intratumoral accumulation compared with the systemically administered CAR T cells (14, 62). A phase I clinical trial at Memorial Sloan Kettering Cancer Center will soon begin testing the safety of intrapleural administration of mesothelin-specific CAR T cells in patients with pleural malignancies (NCT02414269; refs.14, 62). In addition, the NCI is conducting a clinical trial using retrovirally transduced mesothelin-specific CAR T cells in patients with metastatic pancreatic cancer, mesothelioma, and ovarian cancer (NCT01583686).

Prostate stem cell antigen (PSCA) and prostate-specific membrane antigen (PSMA) have been investigated as CAR T-cell targets in several preclinical studies. PSCA is a cell surface protein that is overexpressed on several solid tumor types, including prostate, pancreatic, and kidney cancers (63). PSCA-specific CAR T cells showed efficacy in a pancreatic cancer xenograft model and were reactive against prostate tumor cells in vitro and in vivo (64, 65). Therefore, PSCA represents an antigen that may be used to target multiple tumor types; however, targeting PSCA may result in off-tumor, on-target toxicity in organs unaffiliated with the tumor that express low levels of PSCA, such as the placenta and kidney (66, 67). CAR T cells recognizing PSMA have also shown efficacy in vitro and in vivo (68, 69). In addition, both PSCA and PSMA were targeted in a preclinical study that showed that CAR T cells can be engineered to recognize only tumors expressing both antigens, thereby increasing CAR T-cell specificity and reducing off-tumor effects (40). A phase I clinical trial (NCT01140373) using PSMA-specific CAR T cells in patients with metastatic prostate cancer resulted in stable disease in 2 of 4 patients treated (44).

Immune-based therapy for cancer is undergoing rapid growth both in academic research laboratories and in industry-sponsored clinical trials. CAR T-cell therapy is now global, although most open trials are located in the United States and China (Fig. 4). This promises to be an exciting advance in the fields of cancer immunotherapy and cellular therapy. The impressive response rates observed in clinical trials of CD19 CAR T cells have led to the rapid proliferation of preclinical studies testing new targets and methods to make CAR T-cell therapy safer and more broadly applicable to various tumor types. Strategies to genetically modify T cells to improve their targeting, enhance their tissue penetration, and control their expansion and persistence are all ways to make better CAR T cells.

M.V. Maus and C.H. June report receiving a commercial research grant from Novartis Pharmaceuticals and have ownership interest in patents on CAR T cells, which are owned by the University of Pennsylvania and licensed to Novartis Pharmaceuticals. No other potential conflicts of interest were disclosed.

Conception and design: M.V. Maus, C.H. June

Writing, review, and/or revision of the manuscript: M.V. Maus, C.H. June

Financial support for medical editorial assistance was provided by Novartis Pharmaceuticals. The authors thank Judith Murphy, PhD, for providing editorial assistance with this manuscript.

M.V. Maus is supported by the NCI of the NIH under award number K08166039.