Background: Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous disease, whose response to standard chemotherapy is variable. Stratification of DLBCL into GCB (germinal center B-cell)- or ABC (activated B-cell)-like type, so-called cell-of-origin (COO) classification, has been performed based on gene expression signatures and/or immunohistochemistry. Since COO classification may not necessarily retain prognostic power, a more accurate risk stratification model is warranted.
Purpose: The goal of this study is to identify a gene expression signature(s) that accurately predicts clinical outcomes of DLBCL patients.
Hypothesis: During the process of normal B-cell development in germinal center (GC), a variety of microenvironment cells, including T cells, fibroblasts, dendritic cells and macrophages, coordinately regulate clonal selection and proliferation of B-cells. Therefore, we hypothesized that the differences in the contexts of lymphoma microenvironment could affect clinical outcomes of DLBCL.
Methods: Standard genome-wide gene expression analysis, including microarray and RNA-seq, can only measure abundant transcripts derived from lymphoma cells. In contrast, cells constituting lymphoma microenvironment are minor fractions in lymphoma tissues, and transcripts derived from these cells are expressed at low levels. To quantitatively measure those lowly-expressed transcripts, we employed the nCounter system, which enables an accurate measurement of rare and highly-fragmented transcripts from formalin-fixed paraffin-embedded (FFPE) tissues.
Results: We first examined expression levels of 1900 genes related to immune-, cancer- and kinase-pathways using tissues from 30 DLBCL cases. We then identified a set of genes that were differentially expressed between cases exhibiting good prognosis (alive > 2 years without relapse) and those with poor prognosis (disease progression < 1 year). Strikingly, differentially-expressed genes were not B cell-related genes, but those related to microenvironment cells, including T cells, fibroblasts and macrophages. Furthermore, high levels of microenvironment-related genes dictated favorable prognosis. To identify subtypes of microenvironment cells, we next performed single-cell q-PCR assays (C1 system) and a multi-color immunofluorescence analysis of DLBCL tissues (Mantra system). We found that presence of GC microenvironment components such as follicular helper T cells, CD11c+ macrophages and FGFR1+ stromal cells was an accurate predictor for favorable prognosis.
We defined the GC microenvironment signature based on the expression levels of ICOS, CD11c and FGFR1 mRNA using nCounter and established a novel scoring system. We then validated the system using an independent DLBCL cohort consisting of 170 cases. Our scoring system enabled precise prognostic stratification among patients treated with R-CHOP independently of IPI status and traditional COO-based prognostic models (Figure 1). Finally, to explore potential correlation between the microenvironment signature and mutational status of DLBCL cells, we performed target capture sequencing of 100 DLBCL cases using a clinically-validated sequencing platform (OncoPanel). Our interim analysis did not reveal a correlation between specific lymphoma mutations and the microenvironment signature, suggesting that GC microenvironment signature, which we have identified, predicts clinical outcomes independently from mutational status of DLBCL cells.
Conclusions: We identified a GC microenvironment signature that accurately dictates clinical outcomes and established a novel nCounter-based scoring system, which may inform prognostic stratification and the selection of chemotherapeutic regimen of DLBCL.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.