Comparative Gene Expression Profiling of Leukemia Cells in Peripheral Blood and Tissue Compartments Reveals a Prominent Role of the Microenvironment for CLL Cell Proliferation.

Abstract 355

In-vitro studies suggest that chronic lymphocytic leukemia (CLL) cells depend on the tissue microenvironment. Different molecules and cell types have been reported to enhance the proliferation and survival of CLL cells. The presence of CLL cells in three distinct compartments: peripheral blood (PB), bone marrow (BM) and lymph node (LN), provides a unique opportunity to investigate the effects of the microenvironment on tumor cell biology in-vivo. To this effect, we used gene expression profiling (Affymetrix HU133 plus arrays) to compare purified CLL cells sampled from PB, BM, and/or LN from 24 previously untreated patients. Initially, an unsupervised hierarchical clustering of all samples appeared to be dominated by the effect of the individual patient. However, in 12 patients where all three sites had been sampled, we used a 3-level one-way ANOVA blocked by patients to estimate patient effect and tissue effect. Three principal components of the 36 samples revealed a clear separation of the tumor cells according to their compartment of origin. Furthermore, supervised analysis with a cutoff of >2-fold change and false discovery rate <0.2 identified 151 genes that discriminated between circulating and LN resident CLL cells (n=17), most of which were more highly expressed in LN, and 27 genes that were differentially expressed in BM as compared to PB cells (n=19). Among the genes upregulated in the lymph node many are readily recognized as related to cell proliferation (e.g. Cyclin D2 and c-MYC) or NF-κB signaling. However, to use an observer independent, unbiased discovery tool to query the gene list for the presence of functional gene signatures we used gene set enrichment analysis (GSEA) and identified several gene expression signatures that were preferentially expressed in LN resident cells: a proliferation signature characterized by E2F and c-MYC regulated genes, signatures related to B-cell receptor and NF-kB signaling were prominent (FDR for all <0.02, normalized enrichment scores 1.81-2.15). A gene expression based E2F score was highest in LN, followed by BM and weakest in PB. Increased nuclear accumulation of E2F1 and c-MYC in LN compared to PB CLL cells was confirmed by Western blotting in paired samples. In general these changes were more prominent in the IgVH unmutated CLL subtype as compared to IgVH mutated CLL cases. In particular, the proliferation E2F score was higher in LN biopsies of IgVH unmutated CLL than IgVH mutated CLL (P=0.04). The E2F score was also an excellent predictor of tumor progression measured as progression free survival (PFS) from diagnosis to treatment: patients with a high E2F score had a median PFS of 16.6 months compared to a PFS in excess of 10 years for patients with a low score (P=0.015). The acquired proliferation and activation signatures in CLL cells which are more prominent in LN resident CLL cells than in cells residing in the BM, suggests that the two microenvironment niches are not identical. Possible upstream cascades driving the signature of CLL cells in the tissue appear to be related to NF-kB and B-cell receptor activation. In conclusion: proliferation and cell activation signatures are acquired in the tissue and are more prominent in LN resident CLL cells than in the BM, suggesting that these two microenvironmental niches have different effects on tumor biology. The LN E2F proliferation signature was more prominent in IgVH unmutated CLL cells and correlated with clinical disease progression.