The interaction between T-cell costimulator, ICOS, and its ligand (ICOSL) expressed on B cells plays an important role in intercellular cognate interactions leading to lymphocyte activation. We found that a subset of chronic lymphocytic leukemia (CLL) B cells express ICOS and that ligation of this receptor induced leukemia-cell activation of the PI3K/AKT survival pathway. Moreover, expression of ICOS was associated with lower-level expression of ICOSL, apparently because of receptor-ligand down-modulation. We hypothesized that co-expression of ICOS and ICOSL on CLL B cells may enhance leukemia-cell stimulation and be conducive to more aggressive clinical disease. We evaluated the CLL cells from 208 patients for expression of ICOS and ICOSL. The expression of ICOS and ICOSL were characterized by the percentage of CD19+/CD3- cells that expressed ICOS and/or ICOSL using fluorescence thresholds established via parallel analyses on the same cell populations stained with isotype control mAbs. In addition, we evaluated these CLL cells for IgVH somatic mutations and for expression of ZAP-70. We then examined the relationship between expression of ICOS and ICOSL, the IgVH mutational status, ZAP-70, and the time from diagnosis to initial therapy, as per NCI working group criteria. The median proportion of cells that expressed ICOS was 3.5% (ranging from 0.1% to 99.6%), and median proportion of cells that expressed ICOSL was 62% (ranging from 0.9% to 97.4%). Ninety cases (44%) were found to use mutated IgVH genes, whereas 118 (56%) were found to use unmutated IgVH. Unmutated cases had significantly higher expression of ICOS than mutated cases (p=0.0014, Wilcoxon test), although there was no significant difference in the level of ICOSL (p=0.47). 95 cases expressed ZAP-70, and 113 cases did not. There was a significant association between ZAP-70 expression and ICOS, p=0.0048 by the Wilcoxon test, with the cases expressing ZAP-70 having higher than expected levels of ICOS. The association with ICOSL was also significant, p=0.04, with higher than expected levels of ICOSL found in the cases that did not express ZAP-70. Next we used recursive partitioning to identify the optimal threshold for distinguishing levels of ICOS and ICOSL expression that best could discriminate the time from diagnosis to initial therapy into two groups. This revealed that 35% for ICOS and 47.7% for ICOSL were the best-cut points. Median time to treatment was 5.1 years among the patients with high ICOS expression (n=25), and 6.7 years among those with low expression. The log rank p-value associated with this difference was 0.03. Median time to treatment was 4.2 years with low ICOSL (n=76), and 7.8 years among those with high expression. The log rank p-value associated with this difference was 0.0004. The stepwise model identified as significant risk factors associated with the need for early treatment: (1) expression of ZAP-70 (hazard ratio 4.53, Wald p-value < 0.0001); (2) expression of unmutated IgVH with >=98% germline sequence homology (hazard ratio 2.56, p = 0.0014), and (3) low-level expression of ICOSL (hazard ratio 1.62, p = 0.02). Because the analysis for ICOS and ICOSL by flow cytometry is relatively straightforward, low-level expression of ICOSL may be a useful surrogate marker for high-risk disease.
