In this issue of Blood Advances, Hirayama et al1 demonstrate that the timing of initiation of programmed cell death ligand 1 (PD-L1) blockade affects both efficacy and toxicity associated with anti-CD19 chimeric antigen receptor T-cell (CAR-T) therapy.1 The authors elucidate the potential mechanism for these outcomes and provide rationale for further studies of dual checkpoint blockade and CAR-T therapy.
Although CAR-T therapies have revolutionized the treatment of large B-cell lymphoma (LBCL), more than half of patients will ultimately experience progressive disease.2-4 Several potential mechanisms of resistance have been implicated, including dysregulation of the tumor microenvironment and CAR-T exhaustion.5,6 Specifically, activation of the programmed cell death protein 1 (PD-1)/PD-L1 axis is thought to contribute to immune dysfunction, thus providing a potential therapeutic target to augment CAR-T efficacy. However, despite encouraging preclinical data, early reports of clinical trials combining anti-CD19 CAR-T therapy with checkpoint blockade have been disappointing thus far.7,8 The reasons why this combination has been not more successful remains poorly described but is likely due to the multifactorial nature of the pathways involved in immune regulation. Furthermore, the impact of timing of PD-1/PD-L1 inhibition in relation to CAR-T infusion on outcomes remains unknown.
Here, Hirayama et al report the results of an investigator-initiated, phase 1 clinical trial of anti-CD19 autologous CAR-T therapy (JCAR014), composed of equal parts CD4+:CD8+ T cells, in combination with the PD-L1 inhibitor, durvalumab, for patients with relapsed/refractory LBCL. Interestingly, durvalumab was administered in 2 groups with variable timing of infusion. Group 1 (n = 11) received durvalumab no earlier than 21 days after JCAR014 infusion, and Group 2 (n = 18) received durvalumab 1 day before JCAR014 infusion. Patients were treated with durvalumab every 4 weeks for up to 10 cycles.
As has been seen in previous trials combining CAR-T therapy and checkpoint inhibitors,7-9 JCAR014 in combination with durvalumab did not appear to increase rates of cytokine release syndrome (CRS) or immune effector cell–associated neurologic syndrome (ICANS) as compared with prior experience with CAR-Ts alone. The rates of CRS and ICANS were also similar between the 2 treatment groups. The authors noted, however, that CRS occurred later (median of 6 vs 4 days, P = .05) and for a shorter duration (median of 3 vs 8 days, P = .08) in the group that received durvalumab before CAR-T infusion.
The best overall response rate and complete response (CR) rate were 38% (95% confidence interval [CI], 22-57) and 35% (95% CI, 19-54), respectively, which appeared lower than outcomes with JCAR014 alone in the group’s prior phase 1/2 trial. Although the authors had expected that durvalumab infusion before CAR-Ts would result in more robust CAR-T counts in the peripheral blood and, thus, improved efficacy, they found the opposite to be true. Despite a lower tumor burden, patients who had received durvalumab before CAR-T infusion had lower response rates than patients receiving durvalumab after CAR-T infusion. In vivo CAR-T pharmacokinetics (PK) also demonstrated that these patients had a longer time to maximum CAR-T counts in the peripheral blood. As compared with Group 1, Group 2 also had decreased levels of inflammatory cytokines.
The authors subsequently aimed to explain the potential for inferior response and delayed CAR-T expansion observed in patients treated with prior durvalumab. They found that soluble PD-L1 (sPD-L1), which is known to increase early after checkpoint blockade administration and correlate with poor outcomes,10 was higher in Group 2 than in Group 1. For patients in Group 2, the rise in sPD-L1 also appeared to coincide with the period of CAR-T expansion. In vitro studies further demonstrated that sPD-L1 inhibits CAR-T–induced cytokine production in a dose-dependent fashion, suggesting that higher levels of sPD-L1 seen with early durvalumab treatment may be responsible for impaired CAR-T response.
Among responders treated with dual JCAR014 and durvalumab, the duration of response appeared superior to JCAR014 alone, suggesting a potential advantage of ongoing treatment with checkpoint blockade. The authors also observed cases of late responses, with 2 patients achieving a CR ∼1 year after JCAR014 infusion. It was hypothesized that these findings may be because of high rates of CAR-T re-expansion that was appreciated in the peripheral blood among patients receiving durvalumab.
Although there is rationale to combine CAR-T therapy with checkpoint inhibitors, this study provides a cautionary tale that therapies should not be combined based on assumptions alone. It is not always the case that a novel combination hypothesized to be better than either therapy alone will in the least be no worse. There are variables and biologic sequelae that are both unknown and unpredictable despite our best preclinical modeling, and so, trials of combinatorial therapies need to be designed thoughtfully and carefully to understand the impact of the addition of “drug X” on the PK and mechanism of action of “drug Y.” This is never truer than with studies involving CAR-T therapy in which there is a single attempt for a definitive outcome in a curable disease. This study demonstrates the unforeseen effect of pre–CAR-T dosing of durvalumab on CAR-T PK and response rates to CAR-Ts, perhaps through an increase in sPD-L1 levels after durvalumab dosing, and as such that timing of therapy may serve as a key variable that has implications on both toxicity and efficacy. Studies adding drugs and other therapies to standard CAR-T therapies should therefore consider designs with small patient cohorts and pauses built in, to analyze efficacy, safety, and PK and biomarker kinetics before expanding into larger patient cohorts.
Given the small sample size and cross-trial comparisons, further studies will be required to confirm the potential clinical implications of the timing of checkpoint blockade on CAR-T outcomes. However, despite underwhelming results of combination trials to date, there appears to be promise to maintenance strategies of checkpoint inhibition after CAR-Ts and a rationale for ongoing studies of this approach in LBCL. The likely heterogenous mechanisms of resistance to CAR-T therapy may dilute any positive impact checkpoint inhibition could have on distinct patients and tumor subtypes, and a deeper understanding of exceptional responders may allow us to identify a group for whom this is a viable and worthwhile strategy.
Conflict-of-interest disclosure: J.L.C. has performed consulting activities for ADC Therapeutics, Karyopharm, Regeneron, Kite, MorphoSys/Incyte, and Seagen, and has received research funding from Bayer, AbbVie, Genentech/Roche, and Merck. C.A.J. has performed consulting activities for Kite/Gilead, Novartis, Bristol Myers Squibb/Celgene, Ipsen, Abintus Bio, Caribou Bio, Miltenyi, MorphoSys, AbbVie, AstraZeneca, ADC Therapeutics, Daiichi-Sankyo, Sana, and Synthekine, and has received research funding from Kite/Gilead and Pfizer.
