CAR19 Therapy Drives Expansion of Clonal Hematopoiesis and Associated Cytopenias

Background: For patients with relapsed/refractory large B-cell lymphomas (rrLBCL), CD19-directed chimeric antigen receptor T cells (CAR19) improve survival compared to autologous hematopoietic cell transplantation (HCT). However, major toxicities of CAR19 therapy include prolonged cytopenias and associated infections. To better understand the impact of CAR19 on such toxicities, we studied a cohort of LBCL patients achieving durable remissions to assess immune recovery after CAR19 treatment. We (1) compared immune reconstitution in HCT vs CAR19 treated patients, (2) performed molecular analysis of post-CAR19 cytopenic marrow specimens, and (3) profiled emergent mutations in cell free DNA (cfDNA) before and after CAR19. Methods longitudinal regression day 28-800 ANOVA p0.005), but similar recovery of hemoglobin and platelets. CAR19-treated patients experienced significantly more Grade ≥3 infections in competing risk analyses (HR 2.61, CI 1.26-5.42), and higher non-relapse mortality (log-rank p=0.027). Next, to characterize mechanisms underlying cytopenias after CAR19, we profiled marrow aspirates from cases requiring biopsy due to prolonged cytopenia or clinical concern for myeloid neoplasms. No patient had evidence of lymphoma relapse at the time of biopsy, and all were 60d from CAR19 infusion (median 377d, range 86-1561d). Among 21 patients profiled by DNA sequencing, 18 (86%) had evidence of ≥1 somatic mutation in genes canonically associated with clonal hematopoiesis (CH) in the marrow ( Fig. 1A), with the most recurrently mutated genes being DNMT3A (38%), TP53 (33%), PPM1D (24%), ASXL1 (14%), and TET2 (14%). Among 6 patients diagnosed with and requiring therapy for a myeloid malignancy, all (100%) had a canonical myeloid karyotypic aberration detected by FISH, as compared with 0 in the 14 other evaluable cases ( Fig. 1A). CAR19 cells were proportionately enriched in the marrow relative to paired peripheral blood by flow cytometry (n = 12, median 2.5-fold enrichment, Wilcoxon p0.001). Using 10x scRNA-Seq, viable CAR19 was detectable in 7 of 8 marrows (88%) and were predominantly CD4+. We hypothesized that CH clones detectable after CAR19 would also be present in pre-CAR19 blood samples and detectable in cfDNA. To test this, we profiled 786 longitudinal samples from 141 patients using CAPP-seq and serially monitored CH clones in cfDNA. Among 141 eligible cases, 40 (28%) were evaluable for CH, in having both pre- and post-CAR19 infusion specimens, while in durable remission (median follow up 967d). Of 51 CH clones detected in post-infusion cfDNA, 46 (90.1%) were also detectable prior to CAR19, typically expanding post-infusion (median = 1.8-fold, Fig. 1B). We next tested whether expanding CH clones showed evidence for selection after CAR19. Strikingly, among 48 recurrently mutated genes tested, DNMT3A and PPM1D showed significant evidence for clonal selection and expansion after CAR19 (Q-value 0.05). When compared to patients without emerging CH clones, patients with ≥5-fold clonal expansions of canonical CH genes (23% of cases) had reduced ANC, but not ALC, recovery (median ANC=1.43 vs 2.43 at 6-mo, Wilcoxon p0.001). This association remained significant in multivariate longitudinal regression that included age, sex, prior lines of therapy, and pre-infusion ANC (ANOVA p = 0.007). Conclusion: CAR19 treated patients experience reduced immune reconstitution and higher infection rates than HCT treated counterparts. Bone marrow findings reveal prolonged CAR19 cytopenia to commonly be associated with clonal cytopenia of undetermined significance (CCUS). Despite durable lymphoma remissions, canonical myeloid clonal expansions are common post-CAR19, and associated with myelosuppression. Non-invasive genotyping could aid in early identification of patients at risk for prolonged cytopenias after CAR19 and inform interventions to decrease associated complications.