Large-Scale Proteomics Identifies Distinct Signatures for Richter Syndrome and De Novo Diffuse Large B-Cell Lymphoma: A French Study from the Filo Group

Richter syndrome (RS) occurs during the disease course of 2 to 10% of Chronic Lymphocytic Leukemia (CLL). Diffuse Large B-Cell Lymphoma (DLBCL) subtype accounts for 90-95% of RS cases. While presenting with the same morphology as de novo DLBCLs, DLBCL subtype of RS is associated with a very poor outcome. Proteins are the primary cellular biological effectors. Proteome composition is highly dependent on regulatory mechanisms located both upstream and downstream translation (transcriptional regulation, post-translational modifications, protein metabolism). Thus, the analysis of the genome and the transcriptome only allows a putative extrapolation of the expressed proteome. Proteomic studies have been performed in the context of de novo DLBCLs (Fornecker et al. Sci Rep. 2019), unravelling a set of proteins associated with refractoriness. In CLL, it showed different profiles according to IGHV mutational status after B-Cell Receptor activation (Perrot et al. Blood 2011). No proteomic study of RS has been carried out to date. RS sample selection was performed across 7 French institutions affiliated to the FILO (French Innovative Leukemia Organization). A total of 49 fresh frozen biopsies were collected, including 28 de novo DLBCLs and 21 RS, mostly treated with first line R-CHOP. All biopsies were centrally reviewed. RS samples were characterized, with data on CLL-RS clonal relationship and mutational status for a 13-gene panel representing the most frequently mutated genes in CLL. Only DLBCL subtype RS samples with at least 50% tumor purity (range 50-95%) and a minimum 10 mg weight were selected. Peptide measurements were performed using liquid chromatography coupled with tandem mass spectrometry, according to published methods (Muller et al. Sci Rep. 2018). Stringent quality controls were applied to ensure sample integrity, abundance accuracy and overall reproducibility. Proteome reconstruction at peptide and at protein level was achieved with a specifically devised pipeline involving conditional filtering, full normalization, categorization and imputation of missing values. These tools made use of the R/Bioconductor DEP package (Zhang et al. Nat Protoc. 2018). Supervised (Bayesian linear models) and unsupervised (hierarchical clustering, K-means, PCA) analyses were further applied to identify differential protein signatures. These were functionally annotated with ReactomePA (Yu et al. Mol Biosyst. 2016) for pathways and STRING (Szklarczyk et al. Nucleic Acids