We report the analysis of the combined miRNA/mRNA expression for the discrimination of MSI-H versus MSS human colon cancer. Because colorectal tumors characterized by MSI are distinct from MSS tumors in many molecular aspects, such the association with the methylator phenotype, which is responsible for MLH1 methylation, the higher frequency of BRAF mutations and the lower frequencies of KRAS, APC and TP53 mutations, MSI and MSS colon cancers represent tumors with a different molecular background. Thus, it is reasonable that their overall gene expression pattern (including both mRNAs and microRNAs) might be affected by any of the above mentioned mechanisms. Indeed, our analyses identified 14 miRNAs and 451 mRNAs differentially expressed between the two genetically distinct colon cancer classes. These results not only indicate the existence of a mRNA/miRNA gene expression profile able to distinguish MSS versus MSI colon cancers, but also indicate that the combination of miRNA/mRNA gene expression data can distinguish microsatellite instability status of human colorectal cancers better than each RNA class independently, suggesting that the combination of mRNA and miRNA expression may potentially represent a general approach for improving characterization and classification of bio-molecular and possibly clinical traits associated with human cancer.
We also examined the prediction power of published lists of genes differentially expressed between MSI-H and MSS colorectal cancers [34–36] against our data set. Two published lists describe differentially expressed genes (100 genes in the report from Di Pietro et al. 2005  and 542 genes from Banerjea et al. 2004 ) and one is a short list of predictors (9 genes in Kruhoffer et al. 2005 ). Although these studies were all performed on Affymetrix platforms, they generated only partially overlapping results. The list of genes from Di Pietro et al.  was the one that better correlated with our results. Twenty two of their 100 genes were present in our list of differentially expressed genes and, for all the remaining genes, expression tendency completely matched our data. A cluster analysis of our data set using the 100 genes from the Di Pietro's paper could generate a good, although not perfect, separation between MSS versus MSI-H tumors and performed well in the prediction of our samples (36 of 39 samples were correctly classified using the SVM prediction) (data not shown). These results indicate that a similar set of genes emerged as differentially expressed between MSS and MSI-H colon cancers in the Di Pietro's and our studies, with discrepancies possibly due to small statistical differences. Discrepancies between results obtained with different microarray platforms are not uncommon and are mainly determined by differences in microarray probe sequences used to detect mRNA transcripts  and in algorithms used for predictor genes identification. Hence, the parallel between Di Pietro's and our results is significant given that data were generated using a different set of samples and a different microarray platform, and suggests that the commonly identified set of genes may represent the most significant differences between MSI-H and MSS colon cancers (Additional file 6). Unfortunately, data sets from published reports were not available for cross confirmation of our list of predictive genes on published microarray data.
In this report, quantitative real time PCRs for differentially expressed mRNAs and Northern blot for differentially expressed miRNAs were performed to validate microarray expression data. These methods confirmed the differential expression detected by microarray-based expression methods. Additional evidence for the robustness of data was the finding that the MLH1 gene is among the most significantly (P < 0.001) down-regulated genes in MSI-H tumors versus the MSS cancer set (Additional file 2). Indeed, this gene represents a sort of internal control, since it is well-known that the loss of MLH1 function, which confers the microsatellite instability phenotype to tumor samples, is caused by transcriptional silencing due to promoter methylation in MSI tumors [5, 6].
In addition to their role as discriminating markers of MSS versus MSI-H tumors, some of the gene-associated functions may possibly be involved in the different phenotypes that characterize the two types of colon cancers. In fact, analysis of the functions associated with 212 (for which annotation was present) of the 451 differentially expressed genes revealed that the most frequently associated classes were cell cycle, DNA replication, recombination, repair, gastrointestinal disease and immune response (Additional file 7), suggesting that these molecular differences may be responsible for traits that distinguish MSS and MSI-H tumors. Additional studies on the molecular and biological functions of these differentially expressed genes will be required to substantiate this hypothesis.
Among the differentially expressed genes, it is interesting to note the detection of the up-regulation of several members of the mir-17-92 family in the MSS colon cancers. This family includes fourteen homologous miRNAs organized in three gene clusters . Our study revealed that, among these, miR-17-5p, miR-20, miR-25, miR-92-1, miR-92-2, miR-93-1 and miR-106a were significantly up-regulated in MSS versus MSI-H colon cancer. The chromosomes 13 and X gene clusters were previously found up-regulated in B-cell lymphoma . It was also shown that c-MYC promotes their transcription  and, interestingly, enforced expression of the mir-17-92 cluster acted with c-myc to accelerate tumour development in a mouse B-cell lymphoma model . In human solid tumors, the chromosome 13 mir-17-92 cluster was found up-regulated in small-cell lung cancer  and its ectopic over-expression enhanced lung cancer cell growth . Since these data indicate that members of the mir-17-92 family can act as oncogenes to promote cell growth and inhibit apoptosis, our data suggests that up-regulation of these miRNAs may have a role in the more aggressive clinical behavior of MSS versus MSI-H tumors.