Diffuse Large B Cell Lymphoma Genetic Classification By Targeted Sequencing and Associations with Immunochemotherapy-Treated Patients’ Clinical Outcome

Diffuse large B-cell lymphoma (DLBCL) is an aggressive and heterogeneous disease with variable prognosis associated with clinical features, cell-of-origin and genetic aberrations. The main problems for DLBCL patients are that a substantial percentage of them (30-40%) are refractory to treatment or relapse and the lack of accurate predictive markers that adequately determine which patients will benefit from immunochemotherapy. Several recent deep-sequencing studies have proposed new genetic subtypes based on the DLBCL genomic profile (Wright et al., 2020; Lacy et al. 2020) and associated with clinical outcome. However, a consensus and validated classification is still needed. The aim was to determine whether genetic alterations of individual genes or genes clustered in pathways are associated with clinical outcome in immunochemotherapy-treated patients.

We used targeted massive sequencing to analyze 84 diagnostic samples from a multicentric cohort of patients with DLBCL treated with rituximab-containing therapies with a median follow up of 6 years. A two-step genetic classifier was built with mutation information from 27 genes and BCL2/BCL6 translocations, based on recently proposed genetic subtypes (Wright et al., 2020; Lacy et al. 2020). Logistic regression and Kaplan–Meier analyses for survival and risk of relapse were performed to assess the potential clinical value of the classifier. Moreover, this two-step classifier was validated in an external cohort and its specificity and sensitivity were tested to determine its accuracy in classifying patients compared to the previous approaches.

We found that the most frequently mutated genes were IGLL5 (43%), KMT2D (33.3%), CREBBP (28.6%), PIM1 (26.2%), and CARD11 (22.6%). Mutations in CD79B, PRDM1, and NOTCH2 were associated with a higher risk of relapse after treatment, whereas patients with mutations in CD79B, ETS1, PRDM1, and TNFAIP3 had significantly shorter survival. Analyzing the impact of gene mutations on predefined gene sets related to lymphomagenesis revealed that mutations in genes involved in B-cell development, and BCR-PI3K and MAPK-ERK pathways were significantly associated with a higher risk of relapse and/or shorter overall survival. These results confirmed the current landscape of genetic alterations in DLBCL.

According to the two-step classifier, we categorized the samples into MCD, BN2, EZB, ST2, and N1 genetic subgroups (Figure). We tested the accuracy of our classification in an external series (UK population-based Haematological Malignancy Research Network cohort [HMRN]) to determine its specificity and sensitivity compared with their own classification (Lacy et al. 2020) and the LymphGen algorithm (Wright et al., 2020). The comparison demonstrated (Wright et al., 2020; Lacy et al. 2020), a specificity and sensitivity higher than 85% for each subtype, except for BN2. This controversy may be explained by the fact that BN2 is less strongly defined than the other subtypes.

We then tested their clinical impact and found the EZB and ST2 subtypes to have a better clinical course and a lower risk of relapse. The BN2 group had the worst clinical outcome of our series, unlike other published studies (Figure). These results were validated in the HMRN cohort, in which N1 showed a higher risk of relapse and shorter overall survival, whereas ST2 is the group with the most favorable outcome. Although N1 was not included for clinical outcome analysis in our cohort due to the small number of cases, the validation of our classifier defined N1 as the most aggressive subtype.

In summary, we propose and validate a feasible genetic DLBCL classifier based on an optimized panel of genes that unifies previous genetic classification algorithms in such a way as to facilitate its implementation as part of pathology laboratories for routine patient management. This genetic classifier, combined with clinical data and other molecular characteristics, should eventually help develop improved risk models for DLBCL patients, and guide precision therapy.