Identification of outcome-correlated cytokine clusters in chronic lymphocytic leukemia

Individual cytokines and groups of cytokines that might represent networks in chronic lymphocytic leukemia (CLL) were analyzed and their prognostic values determined. Serum levels of 23 cytokines were measured in 84 patients and 49 age-matched controls; 17 levels were significantly elevated in patients. Unsupervised hierarchical bicluster analysis identified 3 clusters (CLs) of highly correlated but differentially expressed cytokines: CL1 (CXCL9, CXCL10, CXCL11, CCL3, CCL4, CCL19, IL-5, IL-12, and IFNγ), CL2 (TNFα, IL-6, IL-8, and GM-CSF), and CL3 (IL-1β, IL-2, IL-4, IL-15, IL-17, and IFNα). Combination scores integrating expression of CL1/CL2 or CL1/CL3 strongly correlated (P < .005) with time-tofirst-treatment and overall survival (OS), respectively. Patients with the worst course had high CL1 and low CL2 or CL3 levels. Multivariate analysis revealed that CL1/CL2 combination score and immunoglobulin heavy chain variable region mutation status were independent prognostic indicators for time-to-first-treatment, whereas CL1/CL3 combination score and immunoglobulin heavy chain variable region mutation status were independent markers for OS. Thus, we identified groups of cytokines differentially expressed in CLL that are independent prognostic indicators of aggressive disease and OS. These findings indicate the value of multicytokine analyses for prognosis and suggest therapeutic strategies in CLL aimed at reducing CL1 and increasing CL2/CL3 cytokines.