Early MRD Detection after CAR-T Associated with Poor Outcome in Patients with Large B-Cell Lymphoma

Adoptive cellular therapy with chimeric antigen receptor T (CAR-T) cells targeting CD19 have transformed the treatment landscape for patients with relapsed/refractory (R/R) large B cell lymphoma (LBCL). Nevertheless, long-term remission is only achieved in ~40% of patients. Methods for early detection of disease recurrence are needed to facilitate identification of patients at high risk of treatment failure and prompt potential further therapeutic intervention. Circulating tumor DNA (ctDNA) shows great potential as a tool for sensitive and early detection of disease persistence or relapse. In patients with lymphoma, ctDNA can be assayed in peripheral blood samples and has already shown promise as a biomarker for detection of patients with LBCL at high risk of treatment failure after first-line chemo-immunotherapy (Roschewski et al, ASH 2022) and after CAR-T. Prior studies have shown that ctDNA levels are very low after therapy, so ultra-sensitive assays are needed to accurately detect MRD (Frank et al, J Clin Oncol, 2021; Sworder et al, Cancer Cell, 2023). We recently developed MAESTRO, a highly sensitive, tumor-informed, mutation enrichment sequencing MRD assay, which can detect parts-per-million levels of ctDNA using minimal sequencing (Gydush et al, Nat Biomed Eng; Parsons et al, Ann Oncol). Given its high efficiency, bespoke MAESTRO tests can also be pooled and applied to many patients at once which confers unique advantages (MAESTRO-Pool; Blewett et al Clin Chem). Here, we investigated whether MAESTRO-Pool would be able to identify patients with eventual treatment failure (i.e. progressive disease before 12 months) early after CAR-T cell therapy versus those with durable response (i.e. >12 months). We evaluated samples from 29 patients (13 non-responders [NR], 16 responders [R]) with R/R LBCL treated with axicabtagene ciloleucel at our institution between 2018-2021. Response was assessed by PET/CT at 1-, 3-, 6-, and 12-months following CAR-T, using Lugano criteria, and was classified as durable responder (R) or nonresponder/treatment failure (NR) based on PET/CT response at 12 months. Tumor DNA from pre-CAR-T FFPE biopsies were used to create a personalized tumor fingerprint for ctDNA detection by MAESTRO, and if they also had cryopreserved plasma samples (Streck) banked on at least three of the following timepoints: day 0, 7, 14, and 28 after CAR-T. Plasma samples at additional time points (e.g. 21 days, 3 months, 6 months, 1 year) were also included when available. On the day of CAR-T infusion (day 0, prior to infusion), there was a trend toward an increase in ctDNA detected in NR compared to R (p=0.07). At day 7, this difference between R and NR was even more pronounced (p<0.001). Further, R's had greater reduction in ctDNA at day 7 compared to day 0 than NRs (p=0.01). At subsequent time points, including days 14 and 28, ctDNA detection in responders approached 0 parts per million (ppm), whereas that of nonresponders continued to have detectable ctDNA (p<0.001 at day 14, p<0.001 at day 28). Nearly all responders (n=15/16, 94%) had clearance of ctDNA within 28 days of treatment; all responders had ctDNA below the LOD within 90 days of treatment. Of 7 patients with PR on the first PET/CT at day 30, 4 had PD at 3 months while 3 had CR at 3 months and subsequently sustained that response beyond 12 months. For patients with 30 day PR who later converted to CR, MAESTRO-Pool showed early clearance of ctDNA in a similar pattern to patients with CR at 30 days. Conversely, the 4 patients with 30 day PR who later progressed did not have ctDNA reduction, similar to patients with upfront PD. MAESTRO-Pool demonstrated high specificity for individual patient tumor fingerprints: of the 6000 patient-unmatched MRD tests (i.e., using plasma samples from other patients as controls), only 15 were positive for ctDNA, yielding a false positive rate of <0.5%. Notably one MAESTRO fingerprint for patient 44 was responsible for 14/29 (48.3%) of these false positive calls. Our results suggest that MAESTRO-Pool is a highly sensitive method for MRD detection by cfDNA and can already identify, especially by day 14, pts at high risk of treatment failure. If confirmed, this could support trials of early intervention in high-risk patients. Our preliminary data suggests that this approach could be more sensitive and specific than currently available MRD platforms, though head-to-head comparisons will need to be tested in future studies.