Longitudinal Analyses of the Genomic, Transcriptomic, and T Cell Repertoire in Diffuse Large B Cell Lymphoma Demonstrates Changes in Signaling and Immune Recognition at Relapse

Background and Aims: Although diffuse large B cell lymphoma (DLBCL) can be cured using immuno-chemotherapy, 40% of patients experience relapsed or refractory disease. Large-scale profiling technologies have predominantly focused on the tumor at diagnosis with a limited number of longitudinal studies and no compelling biomarkers linked to relapse identified. To address this, we utilized a multifaceted approach integrating genetic, transcriptomic and T cell clonality analyses in paired diagnostic and relapse tumors to enable robust identification of signaling pathways and associated micro-environmental changes underlying treatment failure and disease recurrence.

Methods: Twenty-six patients were selected based on the availability of both the diagnostic and relapse tumor, with all patients treated with rituximab containing immuno-chemotherapy regimens. The median age was 61 years and the median time to relapse was 20 months (range 6-115). Targeted sequencing was performed using a custom capture assay (Agilent Technologies) comprising 160 genes recurrently mutated in DLBCL. DNA fingerprinting analysis confirmed that paired biopsies originated from the same patient. Gene expression (GEP) and tumor infiltrating T cell profiling were performed using the Ion AmpliSeq Transcriptome Human Gene Expression and T cell receptor beta sequencing (immunoSEQ) assays respectively. Cell-of-origin (COO) classification was completed using the Lymph2Cx assay based on NanoString technology which distinguishes two subtypes of DLBCL, activated B-cell-like (ABC) and germinal center B-cell-like (GCB) DLBCL.

Results: At diagnosis 40% of patients were classified as GCB type, 52% ABC and 8% unclassified with the COO remaining largely consistent upon relapse. While the COO did not significantly vary over time, GEP data (irrespective of COO or time to relapse) did identify temporal changes in broader signaling pathways with increased activity of mTOR, HMGB1 and integrin linked kinase signaling observed at relapse. Comparison of diagnosis and relapse T cells within the tumor microenvironment consistently demonstrated a significant loss of T cell clonotype complexity over time, despite an increased mutational load in relapsed tumors. GEP analysis of immune receptors showed that the loss of T cell clonotype complexity upon relapse correlated with down-regulation of immune receptor genes CD20 and CD58 . Genetic profiling revealed variable degrees of temporal heterogeneity, with on average 78% (range 11%-100%) of variants shared between diagnosis and relapse tumors. Genes known to regulate the epigenome including, HIST1H1E, HIST1H1D and KMT2D were predominantly shared between paired tumors suggesting early occurrence, however we also observed temporal changes in potentially actionable mutations or genes with biological relevance including STAT3, CD58, TP53 and CARD11 .

Conclusion: Collectively, while it is possible to propose individual genes or pathways on a case by case basis, there is no compelling single pathway attributable to disease relapse with integrated profiling of a larger series ongoing. However, notable findings include increased mTOR signaling and a reduction in the T cell clonotype repertoire, altering the ability to recognize tumor antigens at relapse. Whilst the COO phenotype demonstrated fairly robust temporal consistency, there were examples of potentially actionable mutations lost during the course of the disease, and this instability must be considered in clinical trials that stratify patients by mutation status.