Coordinated expression of microRNA-155 and predicted target genes in diffuse large B-cell lymphoma

Abstract

MicroRNAs (miRNAs) attenuate gene expression by pairing to the 3'UTR of target transcripts inducing RNA cleavage or translational inhibition. Overexpression of microRNA-155 (miR-155), measured either at the primary (BIC gene) or mature transcript level, was recently described in diffuse large B-cell lymphomas (DLBCL). These studies have been limited in size, however, and have not attempted to link miR-155 expression to that of putative target genes. To start to address these issues, we examined a collection of 22 well-characterized DLBCL cell lines. The expression of miR-155 is heterogeneous in these cell lines and associates with NF-kappaB activity. We found that the expression of the primary miR-155 transcript reliably reflects that of the functional mature miR-155. Because many gene array platforms include probe sets for the primary miR-155 sequences, these findings allowed us to confidently examine large array-based expression datasets of primary DLBCLs in the context of miR-155 levels. Our investigation revealed that miR-155 expression segregates with specific molecular subgroups of DLBCL and it is highest in activated B-cell (ABC)-type lymphomas. These tumors are characterized by constitutive activation of NF-kappaB signals, which supports the data derived from our cell lines. More importantly, using supervised learning algorithms, we identified a robust gene signature driven by the differential expression of miR-155. These profiles contained several gene markers, including predicted targets, consistently downregulated in tumors expressing high levels of miR-155. Our data start to unveil the genome-wide effects of miR-155 expression in DLBCL and indicate the utility of this strategy in the identification and validation of miRNA target genes.

Figure 1. Expression of mature and primary miR-155 in DLBCL cell lines

Mature and primary (unspliced and spliced) miR-155 expression was defined by northern blot (a) and q-RT-PCR (b), respectively. Densitometric quantification of the mature miRNA messages was performed with ImageQuant 5.2 (Molecular Dynamics) and normalized by the expression of the snRNA U6. The primary miR-155 expression was normalized by the expression of cyclophilin A. Pearson correlation coefficients between primary and mature miR-155 were .95 (DHL series), .94 (Ly series) and .97 (miscellaneous series).

Figure 2. NF-κB activity of DLBCL cell lines expressing distinct levels of miR-155

The NFκB activity was significantly lower in cell lines OCI-Ly1, OCI-Ly4, OCI-Ly7 and OCI-Ly18 with low/null miR-155 expression than in the high miR-155 lines, OCI-Ly3, OCI-LY10, Farage and RC-K8 (p<.001)

Figure 3. Distribution of pri-miR-155 expression by COO (left) and CCC classification (right)

In the box plots, the outliers samples (those whose expression extends beyond 1.5 times the inter-quartile range) are noted as dots. P-values (Kruskal-Wallis test) are shown in the right upper corner on the panels.

Figure 4. Gene signature of DLBCL dichotomized by the expression of miR-155

The 100 probe sets that more strongly correlate with miR-155 expression in 88 DLBCLs are shown. Gene ranks are based on signal-to-noise metrics and red or blue indicate high or low expression, respectively. Each column is a tumor and each row a gene. The genes consistently downregulated in DLBCL expressing high miR-155 levels are shown of the upper right quadrant. The predicted miR-155 targets, distributed throughout this ranked group, are indicated (in bold are those predicted by multiple models). Expectedly, expression of MiR-155, indicated on the left, drives the distinction between these two DLBCL groups.