Authors: Matthew Loya, Omar Jabado, Manling Ma-Edmonds, Angelo Harris, Anantharaman Muthuswamy, Suzana Couto, David Soong, Brandon W. Higgs, Mahesh Bachu, Christopher Chiu, Maria Jure-Kunkel, Kate Sasser, Mark Fereshteh
Published: 2022-06-06
DOI: 10.1200/jco.2022.40.16_suppl.7576
Source: Full article
7576 Background: Diffuse large B-cell Lymphoma (DLBCL) is a highly heterogenous disease. Gene microarrays were initially used to classify DLBCL into germinal center B-cell-like (GCB) or activated B-cell-like (ABC) Cell of Origin (COO) subtypes. ABC is associated with shorter overall survival. In newly diagnosed patients, COO classification by RNA profiling is a validated prognostic. A simpler immunohistochemical (IHC) staining of CD10, MUM1 and BCL6 is a proxy used in clinical practice in lieu of transcriptomics due to its expense, complexity and tissue requirements. Recent advances in the HTG EdgeSeq platform allow genome-scale profiling with minimal tissue. We successfully applied this novel technology to perform simultaneous COO classification, immune cell enrichment and tumor pathway analysis using a single FFPE slide. Methods: Accuracy of the HTG EdgeSeq panel (19,000 genes) was assessed in a head-to-head comparison with RNAseq using FFPE tumor samples (n = 8). DLBCL resections and core needle biopsies were commercially sourced and COO typed using Han’s algorithm into GCB or non-GCB (n = 65). Tumor locations included: lymphoid organs, gastrointestinal tract, testes, and the pleural cavity. EdgeSeq was performed on single slides with an average tissue area of 40mm2. Transcriptomic COO classification was performed using a linear combination of genes as described in Wright et al., PNAS 2003, substituting HTG platform-specific weights. Validated COO gene sets from literature and commercial diagnostic assays were tested. Immune cell gene signature enrichment analysis was performed using xCell (Aran et al., Genome Biol 2017); pathway analysis was performed with GSEA (Subramanian et al., PNAS 2005). Results: Gene expression levels estimated from whole transcriptome EdgeSeq on single slides were highly correlated to whole transcriptome RNAseq. Differential expression analysis of GCB vs non-GCB showed that key prognostic genes were detectable and enriched in the expected subtypes. Using these pre-established signatures, subtyping accuracy was ̃93% on the training set and 89% on the test set. Immune cell enrichment analysis identified class-switched memory B-cells as more prevalent in non-GCB subjects. This is consistent with emerging evidence that memory B-cells are the primary source of ABC DLBCL and not plasma cells (Venturutti & Melnick, Blood 2020). Pathway analysis identified genes regulated by the oncogene ic transcription factor MYC were enriched in non-GCB samples; MYC protein was found to be overexpressed in ABC in a large study (Hu et al., Blood 2013). Conclusions: Combined COO typing and whole transcriptome analysis from a single slide efficiently uses precious patient tissue. Longitudinal core needle sampling may yield insights into tumor evolution and therapeutic mechanisms of action across the DLBCL treatment landscape.