Safety and efficacy of nivolumab in patients who failed to achieve a complete remission after CD19‐directed CAR T‐cell therapy in diffuse large B cell lymphoma

Chimeric antigen receptor-engineered (CAR) T cells have demonstrated their efficacy in relapsed/refractory (R/R) diffuse large B-cell lymphoma (DLBCL), but still fail to induce a durable complete remission (CR) for most patients.1, 2 One potential mechanism of immune escape is the exhaustion of CAR T cells induced by the immunosuppressive tumour microenvironment.3, 4 The PD1/PD-L1 immune checkpoint pathway is strongly associated with CAR T-cell exhaustion.5 The use of anti-PD1 inhibitor could generate a new CAR T-cell expansion and may improve the response.6 We here report on 11 patients with R/R DLBCL who did not achieve CR after CD19 directed CAR T-cell therapy by axicabtagene ciloleucel (axi-cel) and who were treated subsequently with anti-PD1 nivolumab (240 mg/m2 every 2 weeks) between 2019 and 2020. The decision to use nivolumab was made early after the initial assessment of the disease (30 days after axi-cel) or later as the disease progressed. Nivolumab was used first-line for patients (n = 5) with stable or progressive disease after axi-cel ('refractory' group), as a booster for patients (n = 2) in partial response (PR) after axi-cel ('booster' group) or as salvage therapy for patients (n = 4) who progressed following an initial PR after axi-cel ('salvage' group) (Figure 1). Nivolumab efficacy was evaluated by PET/CT after 4 injections. In the 'refractory' group, the median time to nivolumab from axi-cel was 44 days (range 37–60) and the median number of nivolumab doses was 5 (range 2–6). DLBCL histology included DLBCL not otherwise specified (DLBCL NOS) (n = 3, 60%), double-hit high-grade B-cell lymphoma (n = 1, 20%), transformed DLBCL from indolent histology (tDLBCL) (n = 1, 20%). The median pre-nivolumab LDH (pre-LDH) was 252 UI (range 187–359), the median number of prior therapies (MPT) was 4 (range 3–4), and the median maximum standardized uptake value (SUVmax) by PET was 19 kBq/mL (range 13–24). One patient developed a non-severe bacterial pneumonia. All patients experienced disease progression at day+60 of nivolumab and 4 of these 5 patients subsequently died, with a median overall survival of 527 days (IC95: 0–1231). In the two 'booster' patients, the time to nivolumab initiation from axi-cel was 36 and 54 days and the number of nivolumab doses was 8 and 9. In this group, DLBCL histology included EBV positive DLBCL and tDLBCL. The pre-LDH were 223UI and 241UI, the number of prior therapies was 4 for both and the SUVmax was 7 and 34 kBq/mL. One patient developed a non-severe bacterial pneumonia and the other one developed a non-severe bronchitis. Those two patients finally achieved CRs, which still persist at least 2 years post-nivolumab. In the 'salvage' group, the median time from axi-cel to nivolumab initiation was 110 days (range 83–144) and the median number of nivolumab doses was 5 (range 4–8). In this group, all DLBCL were DLBCL NOS. The median pre-LDH was 254 UI (range 218–412), the MPT was 5 (range 4–8), and the median SUVmax was 15 kBq/mL (range 10–27). One patient developed a non-severe bacterial pneumonia. CR was observed in two patients and progressive disease in two others resulting in a CR rate of 50%. Interestingly, CAR T cells were no longer detectable by flow cytometry at day+20 after axi-cel infusion in one of the responding patients, but we observed a new expansion of CD4+ and CD8+ CAR T cells after one dose of nivolumab (at day 144 after CAR T-cell infusion). The new peak of CAR T cells in blood reached 0.035G/L at day 202 (9% of circulating lymphocytes) (Figure 2). Immediately after the first nivolumab injection for this patient, cytokine serum analysis showed an increase of pro-inflammatory cytokines, consistent with a re-expansion of circulating CAR T cells (data not shown). The two responding patients are still alive at 2 years post-nivolumab, one subsequently received allogeneic transplant and relapsed at 1 year and the other developed a mesenteric relapse 1 year after nivolumab, successfully treated by radiotherapy. Several insights can be drawn from this experience. First, in our study, it seems that nivolumab was not effective against DLBCL refractory to CAR T-cell therapy contradicting some previously published studies of anti-PD1 pembrolizumab after CAR T-cell therapy.7 Second, we showed that nivolumab may improve upon PR, inducing a second CAR T-cell expansion (Figure 2). Importantly, neither this second expansion, and the use of nivolumab for all patients resulted in cytokine release syndrome or neurotoxicity. A possible mechanism of CAR T-cell therapy failure is the exhaustion of CAR T cells that is induced via the PD1/PD-L1 axis.3, 5 The action of anti-PD1 therapy could have reversed T-cell exhaustion, producing a new antitumoral efficacy. This study has some limitations. Single-agent nivolumab has been reported to have a moderate direct-tumour efficacy on DLBCL.8, 9 It has been shown that some patients in PR at day 30 days after CAR T cell can convert into a CR without any subsequent treatment.10 The lack of a control group makes the interpretation difficult in the 'booster' group. Also, the PDL1 status of tumours may influence the response to anti PD1 therapy but was unknown. The evolution of PDL1 status in tumour microenvironments and of tumour cells, the PD1 expression on CAR T cells and the direct impact of anti-PD1 after CAR T-cell therapy remain unknown. The timing of PD1 inhibitor initiation may explain the disparity of results across previously published studies. As an example Chong et al. reported 12 patients with pembrolizumab with a median time to first infusion of 3.3 months and had responders among refractory patients which we did not observe.7 The ZUMA 6 study, the PORTIA study and others tried to identify subgroups of patients who would benefit from anti-PD1 therapy, by using several cohorts, they also attempted to show whether the best time to use anti-PD1 was before or after CAR T-cell infusion.11-13 A recent abstract showed that patients in relapse after CAR T-cell therapy had low response rate and low median survival after being treated with anti-PD1 treatment.14 In conclusion, we reported that patients in PR after axi-cel for DLBCL, with or without subsequent progression, may benefit from nivolumab. This subgroup of patients needs to be further studied before this strategy can be broadly advised. Nicolas Gazeau and David Beauvais collected data and drafted the manuscript. Pauline Varlet and Suman Mitra acquired and analysed flow immunophenotyping data. All authors critically revised the manuscript and approved the final version. We thank Deborah E. Banker for her help in the writing process. FM, IBY, DB and SM received research grant support from AAP SIGN'IT 2020 Fondation ARC awarded to FM. The authors declare no conflict of interest.