Molecular Landscape of Primary Refractory DLBCL

Background: Diffuse large B-cell lymphoma (DLBCL) is the most common aggressive lymphoma and can be cured with frontline immunochemotherapy (IC, e.g., R-CHOP or similar) in approximately 60% of patients. DLBCL that is progressive during frontline treatment or fails to achieve a complete response (CR) by the end of treatment (EOT) is referred to as primary refractory DLBCL (prDLBCL) and has a poor prognosis. It is not possible to identify such patients at initial diagnosis. Molecular predictors are highly needed. Large scale genomic profiling has led to a broader understanding of the molecular drivers in newly diagnosed DLBCL (ndDLBCL) and have identified unique molecular subtypes. However, no classification clearly identifies prDLBCL, in part due to the small number of cases analyzed to date. This study examined the genomic landscape of prDLBCL and compared it to that of non-primary refractory DLBCL (non-prDLBCL). Methods 444 patients with ndDLBCL between 2002 and 2015 and enrolled in the Molecular Epidemiology Resource (MER) of the Mayo/Iowa Lymphoma SPORE were included. RNA and DNA were isolated from FFPE tumor samples at diagnosis and evaluated for whole exome sequencing (WES, n=404) and RNA sequencing (RNAseq, n=321). WES and RNAseq data were processed using the Bristol Myers Squibb (BMS) or Mayo in house standard pipelines, and copy number alterations (CNAs) were identified by OncoScan or WES using Nexus copy number software. PrDLBCL was defined as less than a CR by EOT and all patients received complete treatment with frontline IC. Available data for prDLBCL included WES (N=24), RNAseq (N=21), and CNAs (N=23). All statistical comparisons made were between prDLBCL and non-prDLBCL (WES N=380, RNAseq N=300, CNA N=364). Pathway analysis was performed using pathfindR. Genetic classification was performed using LymphGen and HMRN. The tumor microenvironment (TME) was analyzed using Lymphoma Microenvironment Classification (LME) and Lymphoma EcoTyper. Results Primary refractory DLBCL accounted for 6% (N=24) of ndDLBCL cases. Cell of origin was available on 22 cases, 10 (45%) were GCB, 9 (41%) were ABC, and 3 (14%) were unclassified. MYC and BCL2 and/or BCL6 rearrangements (DH/THL) were available on 16 cases, 2 (12.5%) were DH/THL and 14 (87.5%) of cases were negative (N=14). TP53 (38%, N=9/24), KMT2D (29%, N=7/24) and IGLL5 (25%, N=6/24) were the most frequent mutations in prDLBCL (Figure 1). TP53 mutations were more frequent in the prDLBCL cohort compared to non-prDLBCL cohort (15.5%, Odds ratio (OR) 3.25, CI 1.1-8.3, p=0.01). Deletions at 17p13.1 (TP53) were the most frequent copy number loss alteration, detected in 48% (N=11/23) of prDLBCL cases, compared to 24% (N=89/364) of ndDLBCL cases (OR 2.9, CI 1.1-7.2, P=0.01). Together, TP53 mutation and/or deletion was observed in 65% of prDLBCL cases (N=15/23). There was an increased frequency of gains/amplifications at 8q24.21 (MYC, OR 2.3, CI 0.82-5.9, P=0.09) and 9p24.1 (PDL1, OR 2.9, CI 1.0-7.6, P=0.03) in prDLBCL. Using the LymphGen classification, only 2 (8%) of prDLBCL cases were in the A53 subtype (hallmarked by TP53 alterations), despite a high frequency of TP53 alterations. The EZB subtype (hallmarked by EZH2 and BCL2 alterations) was the most frequent (N=7, 30% of prDLBCL cases). To identify transcriptomic signatures of prDLBCL, differential gene expression analysis was performed comparing prDLBCL to non-prDLBCL followed by pathway analysis. This analysis identified 396 upregulated and 252 downregulation genes (P<0.01). PathfindR analysis identified activation of metabolic, inflammatory, and MAPK pathways. Analysis of the TME using Lymphoma EcoTyper and LME found no dicernable differences between groups. Conclusion: DLBCL patients with primary refractory disease represent a high-risk group of patients with poor survival outcomes. While our sample size is small, we found a high frequency of TP53 alterations (mutation and deletion) and MYC copy number gain in prDLBCL cases. These findings suggest TP53 and MYC alterations may mediate primary treatment resistance and could be two important mechanisms contributing to the poor outcomes. Molecular profiling at the time of diagnosis is not yet able to identify patients at high risk of frontline treatment failure, however, these results contribute to our understanding of the genomic landscape of prDLBCL. Figure 1: Oncoplot of Mutations and Copy Number Alterations in prDLBCL Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal