Molecular profiling of EBV associated diffuse large B-cell lymphoma

Abstract

Epstein-Barr virus (EBV) associated diffuse large B-cell lymphoma (DLBCL) represents a rare aggressive B-cell lymphoma subtype characterized by an adverse clinical outcome. EBV infection of lymphoma cells has been associated with different lymphoma subtypes while the precise role of EBV in lymphomagenesis and specific molecular characteristics of these lymphomas remain elusive. To further unravel the biology of EBV associated DLBCL, we present a comprehensive molecular analysis of overall 60 primary EBV positive (EBV+) DLBCLs using targeted sequencing of cancer candidate genes (CCGs) and genome-wide determination of recurrent somatic copy number alterations (SCNAs) in 46 cases, respectively. Applying the LymphGen classifier 2.0, we found that less than 20% of primary EBV + DLBCLs correspond to one of the established molecular DLBCL subtypes underscoring the unique biology of this entity. We have identified recurrent mutations activating the oncogenic JAK-STAT and NOTCH pathways as well as frequent amplifications of 9p24.1 contributing to immune escape by PD-L1 overexpression. Our findings enable further functional preclinical and clinical studies exploring the therapeutic potential of targeting these aberrations in patients with EBV + DLBCL to improve outcome.

Fig. 1. Mutational landscape.

A Venn diagram illustrating number of samples analyzed by targeted sequencing (n = 46), OncoScan (n = 46), and NanoString (n = 37), respectively. B Pie chart showing the distribution of 37 primary EBV + DLBCLs with respect to ABC DLBCL (n = 16), GCB DLBCL (n = 10), or unclassifiable DLBCL (n = 11) based on NanoString nCounter FLEX gene expression profiling. C All detected non-synonymous mutations are depicted for each primary EBV + DLBCL sample per column (n = 46) and are sorted by cohort frequency (see Supplementary Table 4 for all results). Sample order is based on waterfall sorting by binary gene mutation status. Genes reaching significance according to dNdScv (q < 0.1) are highlighted (*). Bar graphs at the left show the ratio of non-synonymous (blue)/ synonymous mutations (green) detected per gene. Bar graphs at the top depict the number of called mutations per sample. Detected mutations per gene are shown at the right. Type of mutations are color-coded. Gender, EBV status (EBER > 50% or EBER 10–40%), cell of origin (COO), and genetic subtype according to the LymphGen classifier are indicated for each sample.

Fig. 2. Recurrent mutations on protein level.

Needle plots illustrating the distribution of detected mutations on protein level for the selected cancer candidate genes and their particular isoforms SOCS1 (NM_003745.1), STAT3 (NM_139276), KMT2D (NM_003482.3), KMT2C (NM_170606.2), CD58 (NM_001779.2), B2M (NM_004048.2), NOTCH2 (NM_024408.3), NOTCH1 (NM_017617.4), FOXO1 (NM_002015.3), and TP53 (NM_001126112.2). Selected protein domains are shown in color and types of mutation are indicated. Exon boundaries are visualized by dashed lines.

Fig. 3. Co-occurrence of selected mutations and SCNAs affecting the same biological pathways.

Alterations belonging to one sample are shown per column. Sample order is based on waterfall sorting by binary gene mutation status.

Fig. 4. Recurrent copy number alterations.

A GISTIC v2.0.23 defined copy number amplifications (red) and deletions (blue) are visualized for each chromosome, arm-level alterations on the left, and focal lesions on the right. Insignificant lesions (q_G2.0  >  0.1) are shaded in gray. Selected potential driver genes are shown with corresponding cohort frequencies (cosmic cancer genes are shown in bold letters). B, C Representative images of immunohistochemical PD-L1 stainings of a PD-L1 negative (B) and a PD-L1 positive (C) EBV + DLBCL case.

Fig. 5. Molecular DLBCL clusters.

A Bar graph showing the sensitivity (blue) and precision (green) for prediction of molecular clusters in our cohort of EBV + DLBCLs applying the Full Model according to the LymphGen classifier 2.0. B Pie chart depicting the distribution of molecular subtypes in our cohort of EBV + DLBCLs (n = 46) and (C) in a cohort of DLBCLs NOS (n = 839) according to the LymphGen classifier 2.0 [25, 38].