Multiomics Analysis of Paired Diagnosis and Relapse DLBCL Biopsies Shows a Reduction in T Cell Infiltration and Function at Relapse

Introduction Despite introduction of innovative immune based therapies, many patients with relapsed/refractory (R/R) diffuse large B-cell lymphoma (DLBCL) continue to have poor outcomes. Defining immune changes that occur within tumors at relapse is crucial to understanding treatment failure. To explore the evolution of the tumor microenvironment (TME) across treatment timepoints, we performed comprehensive multiomics analysis encompassing immune gene expression, T-cell receptor (TCR) sequencing and spatial transcriptome analysis in a large Australian cohort of R/R DLBCL patients with paired diagnosis and relapse biopsies. Methods Archived biopsies from 141 clinically annotated R/R DLBCL patients were collected from 6 Australian centres. Gene expression profiling using the Nanostring PanCancer Immune panel of 770 immune-related genes was performed on 125 diagnostic and 124 R/R biopsies (first R/R episode: 102, subsequent R/R episode: 22). Lymphoma Subtype (LST) signature was used to evaluate cell of origin (COO). TCR sequencing was performed on 22 pairs using the Archer Immunoverse RNA assay. Nanostring GeoMx spatial whole transcriptome assay was performed on 17 diagnostic and 6 R/R biopsies, segmented into CD20, CD8, CD4 and CD68 regions. Results 63 R/R patients were assigned germinal centre B cell (GCB) COO, 41 activated B cell (ABC), and 19 were unclassified. Only 2 late relapsing (LR) patients (relapse >12 months post end of treatment [EOT]) had COO discordance. COO was not prognostic at relapse. Differential immune gene expression (corrected for false discovery) on 87 pairs of diagnosis and R/R biopsies demonstrated profound changes in the T cell compartment. T cell associated genes (CD3E, CD5, CD7, CD8A, CXCL9: adj p<0.044), PD-L1 pathway genes (JAK2, STAT3, LCK, RELA, MYD88: adj p<0.037) and the immune checkpoint gene TIGIT (adj p=0.005) were significantly downregulated at relapse. This aligned with Nanostring cell abundance scores demonstrating a marked reduction in overall T cell abundance, in particular total CD8 (p=0.003) and exhausted CD8 T cell subtypes (p=0.0004). Reduction in T cell and checkpoint genes were particularly marked in patients with primary refractory (RF) or early relapsing (ER) disease (relapse <12 months post EOT), whereas LR patients demonstrated limited immune gene expression changes across disease timepoints. Compared to RF/ER patients, LR patients had higher T cell abundance at diagnosis (p=0.0018). This abundance was stable during disease in LR patients, with RF/ER patients showing a further reduction at R/R timepoint (p=0.0187). Analysis of 16 patients with 3 serial biopsies demonstrated an ongoing reduction in overall T cell abundance (p=0.035) and exhausted CD8 T cells (p=0.0125) across treatment episodes. Paired analysis of the T cell repertoire showed preserved diversity across disease timepoints. However, low TCR diversity (Shannon's index) at relapse was predictive of poor overall survival (p=0.038), as we have previously demonstrated in newly diagnosed patients (Keane et al, CCR, 2017). Spatial whole transcriptome pathway analysis of CD8 enriched segments demonstrated reduced immunogenic cell death signalling at relapse, with upstream analysis predicting severe interferon-gamma and tumor-necrosis factor inhibition at relapse. Together these findings suggest reduced anti-tumoral/cytotoxic capacity within the TME at relapse. Reduction in CXCL9 at relapse was only identified in T cell segments (CD8 and CD4), noting high expression of CD8/CXCL9 has been identified as highly specific for positive response to immune checkpoint therapy in solid organ tumor cohorts, as likely markers of neo-antigen reactive T cells (Litchfield et al, Cell, 2021). Conclusions Key differences in the TME are evident in DLBCL at relapse with reduced T cell infiltration and function at relapse demonstrated by gene expression and spatial transcriptome analysis. Reduction in T cell infiltration is not seen however, in late relapsing patients, whom have improved overall survival and higher rates of treatment response. Low T cell diversity is a marker of poor overall survival at relapse. These findings suggest T cell mediated immunity is critical for chemotherapy response in DLBCL and could have implications for predicting responses to newer T cell directed therapies such as bispecific T cell and CART therapies.