PF510 A 14‐GENE SIGNATURE ASSOCIATED TO CHOLESTEROL METABOLISM IDENTIFIES M1‐LIKE TUMOR‐INFILTRATING MACROPHAGES AND PREDICTS PATIENT SURVIVAL IN DIFFUSE LARGE B CELL LYMPHOMA

Background: Diffuse large B cell lymphoma (DLBCL) is a heterogeneous disease with high variability in clinical outcome, genetic features and cells of origin (COO). Gene expression profiling (GEP) studies have identified stromal signatures reflecting extracellular matrix deposition and macrophage infiltration as predictive of favorable outcome to standard immune‐chemotherapy. However, while a reliable gene panel has been recently developed to detect stromal compartment, no reproducible macrophage biomarkers applicable to clinical practice are currently available. Tumor‐associated macrophages (TAM) can display a M1‐like phenotype with an antitumor activity, or act as M2‐like subpopulation promoting angiogenesis and tumor progression. As described for many immune cells, environmental and metabolic signals including oxygen, cytokine and nutrient gradients induce profound reprogramming also in macrophages and their precursors. In particular, the cholesterol metabolism, mainly regulated by the nuclear liver X receptors (LXRα and LXRβ), is known to control functional polarization, inflammatory response and apoptosis processes. Aims: We investigated whether LXRs may play a role in DLBCL biology and modulate the function of macrophages in tumor microenvironment, ultimately affecting prognosis. Methods: We first examined the expression level of LXRα and LXRβ, encoded by NR1H3 and NR1H2 genes respectively, throughout several B‐cell lymphomas GEP from different tissues and used CIBERSORT deconvolution algorithm to analyze DLBCL microenvironment composition according to NR1H3 expression. Furthermore, Gene set enrichment analyses (GSEA), in vitro studies and in situ RNA hybridization were used to correlate NR1H3 with a particular TAM subpopulation. Then, we identified a panel of 50 genes closely related to LXRα by using a machine learning approach (Random Forest), and explored their prognostic potential in a validation cohort of 210 fresh frozen DLBCLs. Finally, we analyzed the expression level of 14 genes functionally selected from the whole LXRα panel by NanoString technology in an independent cohort of 148 formalin fixed paraffin embedded (FFPE) specimens. Results: LXRα but not LXRβ was found upregulated in non‐malignant cells showing a strong association with M1‐like macrophages by combining GSEA, in situ and in vitro results. Hierarchical clustering analysis of the 50‐gene signature associated to LXRα identified three patient groups with heterogeneous clinical behavior in the cohort of 210 fresh/frozen DLBCLs. Notably, this difference was strikingly evident when the panel of 14 functionally‐selected genes was used to stratify an independent cohort of 148 DLBCLs by NanoString, identifying two main prognostic subgroups. In particular, cases with low expression of the selected genes showed a significant adverse outcome in term of overall survival and progression free survival compared to those with intermediate and high expression (Figure 1). Moreover, this difference was independent from COO, and identified a group of GCB‐DLBCLs with unexpected less favorable outcome. Summary/Conclusion: We characterized macrophages functionally and demonstrated that the metabolic regulator LXRα may represent a reliable functional biomarker of M1‐like macrophages in DLBCL. Furthermore, we identified a robust prognostic panel of 14 genes closely related to cholesterol metabolism and routinely applicable to FFPE samples. The panel recognizes distinct prognostic subgroups of DLBCLs independently from their COO, providing the rationale to design clinical trials exploring new immunomodulatory drugs.