Elegant laboratory science beginning more than 30 years ago led to the design of gene-engineered chimeric antigen receptor T cells (CAR-Ts) aimed at eradicating malignancy.1 Multiple centers individually advanced clinical development by creating second-generation anti-CD19 CAR-T therapies, that showed dramatic responses in in chemo-refractory chronic lymphocytic leukemia, acute lymphocytic leukemia (ALL), and diffuse large B-cell lymphoma (DLBCL).2,3 There have since been three U.S. Food and Drug Administration approvals for CD19-targeted CAR-T therapy for ALL (tisagenlecleucel) and DLCBL (axicabtagene ciloleucel).4-6
The pace of CAR-T clinical development has been remarkable given the regulatory complexities of gene therapy clinical trials and the operational intricacies of cellular therapy handling and administration. In particular, management of toxicities has been an area of noteworthy consideration. Significant clinician and allied health effort is required following therapy to monitor for, and treat, the most common CAR-T cell toxicities — cytokine release syndrome and immune effector cell–associated neurotoxicity syndrome.7
In their article, Dr. Jennifer Brudno and colleagues describe a clinical trial using a new autologous CD19 CAR-T therapy in relapsed/refractory B cell lymphomas to test two hypothesis: 1) Would a human scFv be less immunogenic than a CAR with a murine scFv, and 2) would CAR-Ts having a different CAR hinge and transmembrane domain secrete less cytokines and have subsequent lower toxicity rates? The new CAR they tested, Hu19-CD828Z, contains a CD19-binding domain derived from a human ScFv (Hu19) and CD8 hinge and transmembrane domains (CD8). This is in contrast to FMC63-28Z, the CAR their group previously developed, now marketed as axicabtagene ciloleucel, which contains a murine ScFv (FMC63) and CD28 hinge and transmembrane domains (28).8 Both contain CD28 costimulatory signaling domains (28Z). Additionally, with clear in vitro experiments, they dissect each piece of the CAR molecule’s contribution to cytokine excretion, and by extension, toxicity.
This feasibility trial enrolled 26 patients, with 20 eventually receiving CAR-T infusion following standard fludarabine (30 mg/m2 × 3 days) and cyclophosphamide (300 mg/m2 × 3 days) conditioning chemotherapy. Patients had relapsed/refractory DLBCL or transformed follicular lymphoma (n=14), follicular lymphoma (n = 3), Burkitt lymphoma (n = 1), or Mantle cell lymphoma (n = 2). The objective response rate was 70 percent, and complete response rate was 55 percent, with 40 percent of patients in ongoing response at the time of last follow-up (range of ongoing CR duration, 17-35 months). Responses and durable responses were seen across all three CAR-T dose levels. Toxicity rates were compared to the groups prior feasibility trial testing of FMC63-28Z, which had a similar design and patient population (n = 22 patients).8 The rate of grade 3 or higher neurotoxicity by CTCAE v3 was 5 percent (1 in 20 patients) in this trial compared to 50 percent (11 of 22 patients) with FMC63-28Z. The rate of grade 2 or higher neurotoxicity was also lower with Hu19-CD828Z, at 20 percent (4 of 20 patients) compared with 77 percent (17 of 22 patients). CAR-Ts as a percentage of the final infusion product and the CD4:CD8 CAR-T ratio were similar between the two trials; however, the Hu19-CD828Z CAR-T products had fewer differentiated CAR-T phenotypes as measured by CCR7 and CD45RA, compared with the FMC63-28Z trial.
Higher peak CAR-T expansion in the blood after infusion has been associated with severe neurotoxicity.5 There was no statistical difference in peak CAR levels between the two trials; however, there were significantly more detectable CAR-Ts at one month after infusion in this trial testing Hu19-CD828Z. Additional ex vivo testing showed that by interferon-γ ELISPOT assay, 17 percent (3of 18 patients) of Hu19-CD828Z–treated patients, versus 42 percent (8 of 19 patients) of FMC63-28Z–treated patients, developed T-cell responses against peptide pools derived from the respective CAR constructs. Taken together, these data indicate that the humanized CAR-T therapy may have decreased immunogenicity as compared to the murine-based CAR.
Elevated serum levels of proinflammatory cytokines are known to be associated with severe neurotoxicity.5 Multiple immunologic proteins were tested in the serum of study patients. The serum levels of Granzyme-A, monocyte chemoattractant protein-1, interleukin (IL) -2, tumor necrosis factor α, MIP-1alpha, IL-6, IL-7, interferon-γ, and IL-8 were all lower in the patients treated on this trial as compared to the former. Finally, the authors conducted in vitro studies comparing FMC63-28Z with Hu19-CD828Z, and a different CAR that contained the humanized scFv but the CD28 hinge and transmembrane domain (Hu19-CD28Z). They found that CAR-Ts with the CD28 hinge plus transmembrane domains secreted more cytokines than those with the CD8, though contribution by the retroviral construct used to transduce the T cells (gammaretrovirus vs. lentivirus) was not completely ruled out as a contributing factor. These data together indicate that decreased secretion of cytokines may be responsible for the lower neurotoxicity seen with the humanized CAR containing the CD8 hinge and transmembrane domain.
In Brief
This study demonstrates that modification of the CAR design can affect CAR T-cell function and clinical outcome in a significant way. Dr. Brudno and colleagues conclude that neither the fact that a CAR targets CD19, nor inclusion of a CD28Z costimulatory domain, are intrinsically linked to high incidence of neurotoxicity. Levels of cytokine release in Hu19-CD828Z were decreased both in vitro and in vivo, and were linked to hinge and transmembrane domain differences, providing a plausible explanation for the lower rates of neurotoxicity. This is consistent with a separate recent report showing complementary results in CAR-Ts containing 4-1BB costimulatory domains.9 Compared to a prior trial using a different CAR with higher toxicity rates, peak CAR blood levels were not different; however, persistent CAR levels at one month were higher, indicating that differences in CAR expansion in vivo were not responsible for differences in toxicity, and hinting that the humanized scFv may have decreased immunogenicity impacting persistence.
There is a rapid pace, and crowded space, in CD19 CAR-T clinical development. The degree to which the incremental CAR design changes described here will fit in with emerging allogeneic CARs, on-off switchable CARs, redirectable CARs, and bispecific CARs is uncertain. Despite these caveats, a CAR-T therapy with equivalent efficacy yet low complication rates would allow for outpatient treatment and a reliable expectation for avoidance of hospital admission. Ultimately, these findings have the potential to increase the adoption rate of the therapy and to impact more lives.
References
Competing Interests
Dr. Locke indicated no relevant conflicts of interest.
