Tumor mutational burden as a novel biomarker of resistance to CD19 CAR-T cell therapy in large B-cell lymphoma

Abstract

Abstract Background: Relapse and progression of disease (POD) remain the leading causes of treatment failure after CD19-directed chimeric antigen receptor (CAR) T-cell therapy in large B-cell lymphoma (LBCL). Tumor mutational burden (TMB) increases neoantigen load and is a biomarker of improved response to immune checkpoint blockade (ICB). Its prognostic value in relapsed/refractory (R/R) LBCL and relevance to CAR-T efficacy are unknown. Methods: We analyzed pre-treatment tumor biopsies from 119 patients with R/R LBCL treated with commercial CAR-T (53% axicabtagene-ciloleucel, 17% lisocabtagene-maraleucel, 30% tisagenlecleucel; 8% 2L, 48% 3L, 44% ≥4L). The most recent tumor samples collected within 1 year prior to CAR-T infusion underwent either matched-normal whole exome sequencing (WES; n=91), targeted next-generation sequencing (NGS; MSK-IMPACT HEME) of 400–468 genes (n=23), or both (n=61). TMB was measured as mutations per megabase (mut/Mb) across exomes or restricted to common genomic regions (targeted TMB) to allow cross-assay comparison. All samples had tumor purity >25%. Associations with relapse/POD were estimated using Fine-Gray regression with death as a competing risk; progression-free survival (PFS) and overall survival (OS) were analyzed by Cox regression. Models were adjusted for age, disease transformation status, baseline metabolic tumor volume (MTV), CAR-T product, and InflaMix clustering, a composite signature of pre-CAR-T systemic inflammation (Raj S et al, Nat Med 2025). Bulk RNA sequencing was available for 42 WES samples to assess HALLMARK pathway enrichments. Results: Median TMB by WES was 2.75 (interquartile range [IQR] 1.80–3.80) mut/Mb, consistent with prior LBCL reports. TMB was not associated with cell-of-origin, double-hit status, MYC rearrangement, TP53 mutation, age, or transformation status. Higher TMB correlated with increased risk of relapse/POD (p<0.01) after multivariable adjustment. Patients in the top TMB quartile (>3.80 mut/Mb; n=24/91) had over twofold higher risk of adverse outcomes: PFS HR 2.30 (95% CI 1.72–3.08, p<0.01), OS HR 2.21 (95% CI 1.59–3.07, p<0.05), and relapse/POD HR 2.27 (95% CI 1.67–3.06, p <0.01). Findings were reproduced in the full cohort of patients with any available NGS assay (n=119) using a targeted TMB calculated by the MSK-IMPACT HEME clinical gene panel: PFS HR 2.12 (95% CI 1.58–2.84, p <0.01), OS HR 2.22 (95% CI 1.61–3.06, p <0.05), relapse/POD HR 1.73 (95% CI 1.28–2.33, p <0.05). Notably, the high TMB group was not enriched for higher tumor burden by PET MTV or refractory pre-infusion inflammation by InflaMix. PIM1 mutation, a marker of aberrant somatic hypermutation was the most enriched gene alteration in patients with high TMB after false discovery rate adjustment (odds ratio 13.15 [95% CI 2.89-97.95], adjusted [adj.] p<0.01). To understand the functional impact of higher TMB in the LBCL microenvironment, we assessed transcriptomic profiles from bulk RNA sequencing from a subset of the tumor samples (n=42). Compared to patients in the lowest quartile by TMB, patients in the highest quartile had significantly reduced expression of inflammatory TNF-NFkB (adj. p<0.001), IL2-STAT5 (adj. p<0.05), and apoptosis (adj. p<0.05) pathways after false discovery rate adjustment, suggesting diminished immune activity. This association for TNF-NFkB pathway enrichment remained significant (adj. p<0.01) even when differential expression was analyzed across TMB as a continuous variable across all samples. Conclusion: This is the first study to demonstrate that high TMB is an independent biomarker for poor outcomes in R/R LBCL patients treated with CAR T-cell therapy, providing supplemental prognostic value over tumor burden and inflammatory markers. This finding contrasts with the role of TMB as a favorable biomarker for ICB. Our transcriptomic analysis suggests a potential mechanism involving attenuated immune-signaling in the tumor microenvironment, which may hinder CAR T-cell efficacy. Finally, TMB quantified by a clinical NGS gene panel yielded concordant findings, underscoring feasibility for bedside implementation.