Subset-Specific Recurrence of Mutations and Identification of Functional Modules Provides New Clues about the Pathogenesis of Large Granular Lymphocyte Leukemia

Background: Large granular lymphocyte (LGL) leukemia is a rare disease characterized by a clonal persistence of cytotoxic T cells or natural killer (NK) cells. Patients usually suffer from cytopenias and other organ-related autoimmune phenomena. These are putatively mediated by the cytotoxic LGL cells constitutively activated following an antigen-driven immune response. In addition to gain-of-function mutations in the STAT3 gene, which occur in 40-50% of patients, recurrent alterations only in the STAT5b and TNFAIP3 tumor suppressor genes have been described thus far. However, based on gene expression analyses, JAK/STAT pathway activation and deregulation of several pro-apoptotic (sphingolipid and FAS/FAS ligand) and pro-survival signaling pathways (PI3K/AKT and RAS) are common features of LGL leukemia.

In this project, we aimed to characterize the genomic landscape of LGL leukemia using exome sequencing and systems genetics approaches in a patient cohort including both T- and NK-LGL cases and patients without known STAT mutations.

Methods: The study cohort included 19 patients diagnosed with LGL leukemia that underwent exome sequencing analysis with matched germline controls. 13 patients had CD8+ T LGL and 3 patients CD4+ T LGL phenotype and 3 patients were NK LGL cases. From the T LGL leukemia cases CD8+ or CD4+ T cells were sorted (according to the dominant phenotype) and used as the tumor sample. In NK LGL leukemias, sorted CD3neg,CD16/56+ NK cells constituted the tumor fraction that underwent exome sequencing. Polyclonal blood lymphocytes depleted from LGL cells were used as germline controls. The exome was captured with the Nimblegen SeqCap EZ Exome Library v2.0 and the sequencing was performed with the Illumina HiSeq2000 sequencing platform. All bioinformatics steps were carried out using a custom bioinformatics pipeline. Putative somatic variants were identified by subtracting, for each patient, the ones called in the normal samples from those found in the tumor sample. After filtering by call quality and allele frequency in ExAC database, somatic variants were prioritized according to the predicted impact from the SnpEff software. Genes hit by variants putatively altering their function were finally mapped to Kegg and Reactome to generate pathway-derived meta gene networks for the identification of affected functional components.

Results: 4 patients had STAT3 mutations and 4 additional cases had STAT5B mutations. In addition to STAT mutations, a number of novel somatic variants, which were recurrently mutated were discovered. These included the tumor suppressor gene FAT4, the epigenetic regulator KMT2D, as well as genes involved in the control of cell proliferation (CDC27 and ARL13B).

With the systems genetics approach based on integration of pathway-derived mutated gene network topologies for identification of connected components we were able to discover affected functional modules. The main network component included key genes, which either directly interact (such as the FLT3 tyrosine kinase) or are functionally connected (such as ADCY3, ANGPT2, CD40LG, PRKCD, PTK2, KRAS, and RAB12 genes) with STAT proteins. Additional modules with putative pathogenetic relevance in LGL leukemia and mutated in the absence of STAT mutations were cell cycle control (CDC27, PLK1, CDC25B, RAD21), Notch signaling (NOTCH2, NOTCH3 and MAML3) and epigenetic regulation through histone-lysine methyltransferase activity (KMT2D and ASH1L).

The comparison of various LGL leukemia subtypes revealed that the mutation burden was especially high among the CD4+ T LGL leukemia cases. Part of the genes and modules affected were shared between the different subtypes of LGL leukemia, but for example KIR2DL1 mutations were only found in CD8+ and NK LGL leukemia cases.

Conclusions: With the exome sequencing and systems genetic approach we were able to discover specific gene networks, which are recurrently mutated in LGL leukemia and particularly in patients without STAT mutations. As several mutated genes are directly or indirectly connected with the STAT pathway, the data strengthen the key role of JAK/STAT signaling in LGL leukemia. The novel identified pathway modules beyond STAT networks provide intriguing insights into the pathobiology of LGL leukemia.