A computational method to identify cell types within a complex tissue, based on analysis of gene expression profiles, is described in this paper. We introduce CIBERSORT, a method for characterizing cell composition of complex tissues from their gene expression profiles. When applied to enumeration of hematopoietic subsets in RNA mixtures from fresh, frozen and fixed tissues, including solid tumors, CIBERSORT outperformed other methods with respect to noise, unknown mixture content and closely related cell types. CIBERSORT should enable large-scale analysis of RNA mixtures for cellular biomarkers and therapeutic targets ( http://cibersort.stanford.edu/ ).

Classical Hodgkin lymphoma (cHL) and mediastinal large B-cell lymphoma (MLBCL) are lymphoid malignancies with certain shared clinical, histologic, and molecular features. Primary cHLs and MLBCLs include variable numbers of malignant cells within an inflammatory infiltrate, suggesting that these tumors escape immune surveillance. Herein, we integrate high-resolution copy number data with transcriptional profiles and identify the immunoregulatory genes, PD-L1 and PD-L2, as key targets at the 9p24.1 amplification peak in HL and MLBCL cell lines. We extend these findings to laser-capture microdissected primary Hodgkin Reed-Sternberg cells and primary MLBCLs and find that programmed cell death-1 (PD-1) ligand/9p24.1 amplification is restricted to nodular sclerosing HL, the cHL subtype most closely related to MLBCL. Using quantitative immunohistochemical methods, we document the association between 9p24.1 copy number and PD-1 ligand expression in primary tumors. In cHL and MLBCL, the extended 9p24.1 amplification region also included the Janus kinase 2 (JAK2) locus. Of note, JAK2 amplification increased protein expression and activity, specifically induced PD-1 ligand transcription and enhanced sensitivity to JAK2 inhibition. Therefore, 9p24.1 amplification is a disease-specific structural alteration that increases both the gene dosage of PD-1 ligands and their induction by JAK2, defining the PD-1 pathway and JAK2 as complementary rational therapeutic targets.

Purpose: Programmed cell death ligand 1 (PD-L1) is a molecule expressed on antigen-presenting cells that engages the PD-1 receptor on T cells and inhibits T-cell receptor signaling. The PD-1 axis can be exploited by tumor cells to dampen host antitumor immune responses and foster tumor cell survival. PD-1 blockade has shown promise in multiple malignancies but should be directed toward patients in whom it will be most effective. In recent studies, we found that the chromosome 9p24.1 amplification increased the gene dosage of PD-L1 and its induction by JAK2 in a subset of patients with classical Hodgkin lymphoma (cHL). However, cHLs with normal 9p24.1 copy numbers also expressed detectable PD-L1, prompting analyses of additional PD-L1 regulatory mechanisms. Experimental Design: Herein, we utilized immunohistochemical, genomic, and functional analyses to define alternative mechanisms of PD-L1 activation in cHL and additional EBV⁺ lymphoproliferative disorders. Results: We identified an AP-1–responsive enhancer in the PD-L1 gene. In cHL Reed–Sternberg cells, which exhibit constitutive AP-1 activation, the PD-L1 enhancer binds AP-1 components and increases PD-L1 promoter activity. In addition, we defined Epstein–Barr virus (EBV) infection as an alternative mechanism for PD-L1 induction in cHLs with diploid 9p24.1. PD-L1 was also expressed by EBV-transformed lymphoblastoid cell lines as a result of latent membrane protein 1–mediated, JAK/STAT-dependent promoter and AP-1–associated enhancer activity. In addition, more than 70% of EBV⁺ posttransplant lymphoproliferative disorders expressed detectable PD-L1. Conclusions: AP-1 signaling and EBV infection represent alternative mechanisms of PD-L1 induction and extend the spectrum of tumors in which to consider PD-1 blockade. Clin Cancer Res; 18(6); 1611–8. ©2012 AACR.