A New Safety Approach Allowing Reversible Control of CAR T Cell Responses

Main Text

Adoptive immunotherapy involving the genetic-modification of allogeneic or autologous T cells expressing chimeric antigen receptors (CARs) targeting tumor-associated antigens has enabled the transfer of a large number of tumor-specific T cells into patients. Toxicities can occur in patients and are currently managed by immunosuppressive drugs such as corticosteroids and the anti-interleukin-6 (IL-6) receptor antibody, tocilizumab. However, these current treatments compromise the long-term activity of CAR T cells, necessitating a real clinical need for development of better strategies that enable the reversible functional control of CAR T cells. In this issue of Molecular Therapy, Richman et al.¹ have developed a new strategy for reversible and tunable inhibition of CAR T cell responses by fusion of a CAR to a ligand-induced degradation (LID) domain, where the addition of a small ligand results in proteasomal degradation of the CAR-LID protein. The promising preclinical data shown in the study by Richman et al.¹ warrants further investigation of the CAR-LID approach in clinical trials.

The results of CAR T cell therapy in refractive B cell malignancies, such as acute lymphoblastic leukemia, targeting the CD19 antigen have been unprecedented, with up to 90% response rates being achieved, leading to recent US Food and Drug Administration (FDA) approval. Thus, CAR T cells possess enormous potential for rapid proliferation and activation after antigen stimulation, and these functions remain largely uncontrolled after adoptive transfer. Many patients experience toxicities ranging from mild to severe following CAR T cell therapy. The most frequent of these toxicities is cytokine release syndrome, which is characterized by the production of inflammatory cytokines from activated CAR T cells and other immune cells, such as interferon γ (IFN-γ) and IL-6, which can lead to organ failure in extreme cases. Other manifestations include neurologic toxicity and on-target/off-tumor toxicity of healthy tissue by CAR T cells.

In their new study, Richman et al.¹ have developed an innovative strategy that enables reversible and tunable inhibition of CAR T cell activity in vitro and in vivo. This approach involves fusing the CAR to a LID domain, developed by Bonger et al.,² which comprises the mutant human protein FKBP12 with a F36V substitution, with a C-terminal attachment of a 19 amino acid peptide that functions as a cryptic degron. The administration of a small molecule ligand (shield-1) that binds to FKBP12 F36V triggers exposure of the cryptic degron, resulting in proteasomal degradation of the CAR-LID fusion protein and loss of CAR from the surface of T cells. Richman et al.¹ demonstrate that CAR-LID T cells can be transiently inactivated in vitro and CAR T cell-mediated toxicity can be delayed in vivo following the addition of the ligand (Figure 1). Importantly, they found that the degree of loss of CAR T cell function is proportional to the ligand dose, suggesting that an intermediate dose of ligand may be effective enough to control CAR T cell toxicity in patients. Furthermore, once the danger period has been subverted, the subsequent withdrawal of the ligand would permit CAR T anti-tumor responses to continue.

Figure 1.

Fusion of CAR to a Ligand-Induced Degradation (LID) Domain Is a Novel Strategy for Reversible and Tunable CAR Expression

Various approaches to dampen CAR T cell activity in vivo to minimize potentially life-threatening toxicities have been explored, the earliest of which were non-reversible. These strategies, including the co-expression of targetable surface molecules such as truncated EGFR and CD20 or caspase 9 fused to a dimerizable FK506 binding protein (FKBP), enable the depletion of CAR T cells in vivo. More recent studies have favored strategies that permit modulable control of CAR expression and activity, including the inhibition of CD3ζ signaling, using the tyrosine kinase inhibitor dazatanib, and engineering of a “split CAR,” where a functional CAR is assembled only in the presence of rapamycin. Richman et al.¹ have generated a CAR with a LID domain containing a 19 amino acid degron, which is unexposed in the native conformation of the LID domain. Upon the addition and binding of the ligand shield-1, the degron is exposed, leading to the degradation of the CAR. CAR expression was shown to be proportional to the concentration of shield-1, and decreased CAR expression led to reduced toxicities. Shield-1-mediated CAR degradation was also demonstrated to be reversible, with restoration of CAR expression and an increase in anti-tumor activity upon removal of shield-1.

The attractive feature of the CAR-LID approach is its reversible nature. Other strategies have been developed that are effective in controlling toxicity but lead to complete elimination of the CAR T cells (Figure 1). This includes the use of suicide genes, such as herpes simplex virus tyrosine kinase,³ human thymidine kinase,⁴ or inducible caspase.⁵ Another approach involves the co-expression of cell surface elimination markers, such as CD20 or EGFR in T cells, which can be targeted by the clinically approved antibodies rituximab and EGFR, respectively, to effectively eliminate CAR T cells.^6,7 However, this approach may be compromised by limited penetration and distribution of the antibodies in some cancers and tissues.

There have been a number of other reversible strategies developed for controlling CAR T cell responses (Figure 1), and it would be of interest in future studies to compare the relative effect of these approaches with CAR-LID T cells. One approach involves the use of the tyrosine kinase inhibitor dasatinib, which inhibits phosphorylation of the protein tyrosine kinase LCK, consequently preventing CD3ζ signaling,⁸ although this approach is not exclusively selective for CAR T cells. The use of universal CARs is another promising approach that links an adaptor molecule recognized by the CAR to an antibody or ligand that recognizes tumor antigen. Full activity of the CAR T cells is only permitted in the presence of the adaptor.⁹ Other approaches to control CAR T cell activity involve the development of split CARs, where the extracellular antigen-binding domain is separated from the intracellular signaling domain on the cell membrane and is only functional in the presence of a dimerizing molecule.10, 11, 12 A further strategy is the co-expression of inhibitory CARs (iCARs), which allow discrimination between healthy cells and tumor cells.¹³ Although these approaches are promising, they cannot control the intensity of activity mediated by CAR T cells following antigen stimulation by tumors.

An important feature of CAR-LID T cells is its ligand versatility. Besides the ligands shield-1/aquashield-1 (AS-1), the LID system is regulatable by clinically approved regents, such as rapamycin and everolimus, that bind to the FKBP variant domain.¹⁴ The additional anti-cancer properties of these compounds could offer further therapeutic benefit. The CAR-LID approach may potentially also be used with other regulatable systems, including the tetracycline (Tet)-on system, which regulates CAR expression at the transcriptional level, allowing further control of CAR T cell activity.¹⁵

Although the CAR-LID system offers a promising and reversible method to control CAR T cell activity for preventing toxicity in patients, there are several aspects that require further investigation in the future. One potential limitation is that the CAR-LID sequence may be immunogenic in some patients, although this risk is reduced owing to the small size of the degron. Another potential issue is the time taken for downregulation of the CAR following shield-1 ligand administration, where other approaches may be more rapid. The data by Richman et al.¹ suggest that 12–24 h would be required to reduce CAR expression. Thus, this approach would be most effective when patients show early signs of toxicity. Once the patient has stabilized, withdrawal of the ligand would allow rapid recovery of CAR T cell activity. Another potential issue for translation of the CAR-LID approach is the relatively short half-life of the shield-1/AS-1 ligands, requiring continuous infusion of the ligands to achieve effective control of CAR T cells. Future modifications of the ligands to increase the half-life and further biodistribution studies will be needed to fully capitalize on the potential of this new approach.

Another interesting observation emerging from the study by Richman et al.¹ was the lower expression of the CAR-LID transgene in T cells compared to the control CAR. This led to a significant decrease in tumor control in the models tested in this study. Although this may be related to the particular CAR used in these experiments, future efforts may be directed at improving CAR-LID expression by increasing the multiplicity of infection or by protein engineering methods to increase the level of LID activity, resulting in improving CAR-LID transgene expression in T cells.

Current CAR T cell trials do not require safety switches to control CAR T cell activity. However, the transfer of a “living drug,” such as CAR T cells, into patients has the potential for rapid proliferation after antigen activation and may lead to severe toxicity and even death. It is very likely that, with the advent of new approaches to increase the persistence and function of CAR T cells, the frequency and severity of these toxicities may rise. Hence, innovative strategies, such as that used in this study, involving CAR-LID T cells and ligand administration offer a great opportunity to be able to control CAR T cell activity in a reversible manner, therefore preventing severe toxicity while still maintaining anti-tumor function. The CAR-LID system may also provide an important tool in the future to investigate CAR T cell biology.

Conflict of interest

The authors declare no conflicts of interest.