Refining the genomic landscape of pediatric and young adult aggressive B-cell lymphomas

Abstract

Aggressive B-cell non-Hodgkin lymphomas in children and young adults (CAYA) include the genetically well-defined Burkitt lymphoma (BL) and diffuse large B-cell lymphoma (DLBCL). Additionally, a group of highly proliferative lymphoma in CAYA with intermediate morphological features between BL and DLBCL, named high grade B-cell lymphoma (HGBCL, NOS) do exist. Differently to BL and DLBCL, the whole genetic profile of this heterogeneous group has only been studied in a limited number of cases, and its genetic profile has not been completely elucidated. To characterise the molecular profile of these tumors, we developed a pipeline to identify potential driver mutations characteristics of each subgroup regardless the material, classify DLBCL and HGBCL, NOS samples according to the previously defined genetic subtypes, and develop a predictor that can be used to molecularly classify CAYA HGBCL, NOS into BL or DLBCL regardless of MYC status information. To generate the classification learner model, two datasets were recruited from literature and own data consisting of 52 BL and 52 DLBCL under 25 years old. An additional 15 WES data set derived from FFPE DLBCLs (under 18 years old) was also evaluated. For the cohort to test the model, we studied 18 aggressive B- cell lymphomas diagnosed in CAYA patients younger than 23 years old using a custom designed panel for Target Sequencing. In this sense, MYC- positive samples in the HGBCL, NOS patients presented recurrent BL-related mutations in ID3, DDX3X, ARID1A, and SMARCA4. In contrast, MYC-negative samples were genetically closer to GCB DLBCL containing mutations in CARD11, KMT2C, and EBF1. In turn, the latter had a mutational profile similar to that described in the CAYA-DLBCL samples studied by Reddy and collaborators as well as the five WES CAYA-DLBCL samples further analysed in this project. Finally, the classification learner was introduced as a step forward to allow independent prediction of samples without regard to MYC status. This resulted in high levels of accuracy and a reduction in the number of predictors that displayed agreement with the prior literature.