Molecular Tumor Burden and Clonal Evolution As Factors Associated with Early Progression in Large B-Cell Lymphomas Treated with CAR-T Therapy

Introduction Anti-CD19 CAR-T cell therapy (CAR-T) provides a high initial response rate in patients with large B-cell lymphoma (LBCL). However, about half of the patients relapse after treatment, most of them before 6 months. Identifying predictive biomarkers of treatment failure may be helpful in therapeutic decision-making and sequencing strategies. Refractoriness and high tumor burden measured by PET/CT have been identified as factors related to outcome. Here, we aimed to evaluate the impact of molecular tumor burden, its dynamics (preapheresis (PA) and preinfusion (PI)) and the clonal evolution by studying circulating tumor DNA (ctDNA) using a high-depth sequencing panel (CAPPseq Lymph). Material and methods: We selected 59 patients diagnosed with LBAG consecutively treated with CAR-T (69% Axi-cel and 31% Tisa-cel) in our center, for whom we had plasma PA and PI samples (n=118 samples). Free DNA in plasma was extracted using the Maxwell® RSC ccfDNA Plasma Kit (Promega). The CAPP seq Lymph capture panel (Twist) includes 134 genes related to LBAG. Library preparation was performed using the Kappa Hyperplus kit (Roche) and sequencing was performed on NovaSeq (Illumina). The cutoff point for variant identification was >0.01 VAF (variant allele frequency) or the variant having more than 10 reads. Quantification of ctDNA was performed by averaging the VAF of each sample in % (0-100). The correlation of the different variables with progression was performed using chi-square or Mann-Whitney U test, and the determination of the best cutoff point for quantitative variables was performed using ROC curves (R Studio). Results: The 97,5% (115/118) of the samples had sufficient depth for analysis (mean 930X, 200X-2600X). The total number of selected variants was 669 belonging to 77 genes, with the most frequently altered genes were KMT2D (80%), CREBBP (50%), TP53 (43%), SOCS1 (40%), NOTCH1 (29%). Of the 59 patients, 32 (54.2%) relapsed or progressed. A correlation was detected between the following variables and progression: Patients who did not received autologous transplant (2% vs. 28% p=0.022); stable disease pre-CART (7% vs. 41% p=0.002). Patients with higher burden of PA ctDNA (mean VAF: 10% vs 5.6% p=0.012) (cutoff point 5%) and PI (mean VAF: 9.2% vs 4.6%, p=0.003) (cutoff point 2%) and a higher number of PA variants (10.3 vs 5.6, p=0.005) (cutoff point 6.5) and PI (9.7 vs 3.7, p=0.007) (cutoff point 6.5) were associated with a higher probability relapse or refractoriness. In the multivariate analysis, the presence of stable disease (OR: 0.1 (0.025-0.8) p=0.028) and a PI ctDNA level >2% (OR: 4.8 (1.2-12) p=0.025) remained significant. Interestingly, 28% (17/59) of the patients showed clone selection/evolution. We identified 91 new variants, with a mean of 7 new variants (range: 2-29) per patient between PA and PI. Likewise, it was related to a higher risk of relapse or refractoriness compared with the patients without new variants (41% vs.12% p=0.047). The genes more frequently involved in clonal evolution were TP53(35%), ARID1A (29.4%), NOTCH1 (29.4%), and NOTCH2 (23.5%). For this group, 11 had received bridging therapy (11 Pola-based regimens and 4 radiotherapies, none received chemotherapy). Strikingly, among the patients with new variants, some had achieved a response or were in stable disease by PET-CT before infusion (8 were in progression disease, 5 in partial response, 1 in stable disease, and 4 in complete remission). Conclusion This study identifies high molecular tumor burden as a factor independently related to relapse or refractoriness in patients with LBCL treated with CAR-T. Additionally, clonal selection/evolution has been identified for the first time as a possible predictor of the worst outcome. In this regard, we have identified that both the amount of ctDNA, the number of variants, and the appearance of new variants between PA and PI were associated with relapse/refractoriness after CAR-T. In this sense, even in patients in response, and with low tumor burden by PET CT, the appearance of new variants could provide differential information.