We found that Dicer and Drosha mRNA expression is variable among invasive epithelial ovarian cancer specimens and in ovarian-cancer cell lines and that they are significantly associated with survival, indicating that levels of mRNA of Dicer and Drosha in ovarian-cancer cells are clinically relevant. The association was validated in independent clinical data sets for ovarian cancer. We also sought mutations in the genomic sequences of the Dicer and Drosha genes in a cohort of patients with variable gene expression and did not find a consistent trend. Furthermore, the results of our functional assays indicate that cells with low Dicer expression could not effectively silence genes when synthetic shRNA constructs were transfected.
The production of mature endogenous interfering RNA involves a cascade of events that are inextricably linked to the functions of Dicer and Drosha. For example, Lee et al.5
demonstrated that in cells with silenced Dicer or Drosha expression, precursor and mature miRNA sequences were reduced. Loss of Dicer in mice disrupts embryonic stem-cell differentiation and is lethal during early development.26
Low levels of Dicer mRNA also affect normal cellular development and immune responses in preclinical models.27-29
Furthermore, abnormalities in the copy number of DNA of the Dicer gene and the argonaute 2 gene (a component of the RNA-induced silencing complex) have been described in human melanoma, breast, and ovarian cancers.13
It is therefore possible that deregulated miRNA expression, observed in several types of tumor,13
is secondary to defective RNA silencing machinery. Decreased Dicer mRNA has also been associated with decreased survival in patients with non–small-cell lung cancer.30
In addition, Dicer expression appears to be up-regulated in noninvasive precursors of invasive lung adenocarcinoma.31
Some findings in other tumor types are inconsistent with our findings. High Dicer expression and high Drosha expression were correlated with poor prognostic factors in prostate cancer and esophageal carcinoma.32,33
Furthermore, reduction of Drosha expression by means of siRNA reduced cellular proliferation has been reported in esophageal-cancer cell lines.33
There are several plausible explanations for the divergent expression patterns of Dicer and Drosha among different solid tumors and how they relate to clinical and pathologic variables. Direct interactions with other components of the RNA-interference cascade could result in compensatory alterations of Dicer or Drosha expression in the presence of mutated cofactors such as genes for the DiGeorge syndrome critical region gene 8 (DGCR8
), exportin 5 (XPO5
), and argonaute 2 (AGO2
In addition, miRNA could have varying regulatory effects independent of alterations in the RNA-interference processing machinery.37
In support of this contention is the association of altered Dicer protein expression (both overexpression and underexpression relative to controls) in mucoepidermoid carcinomas of the upper aerodigestive tract with overall survival.38
Despite growing evidence that Dicer and Drosha mRNA levels vary among tumor types, the regulation of these genes is unclear. Dicer gene mutations have been found in Caenorhabditis elegans39
and in humans, and deletions of the Dicer gene locus were detected in some precancerous and invasive lung adenocarcinomas.31
Our mutational analyses showed that alterations of genomic DNA probably do not account for the variability in Dicer and Drosha levels. We did find that single-nucleotide mutations can occur in both genes; however, the functional role of such mutations remains unclear. In breast-cancer cell lines, there are two forms of Dicer, due to alternative splicing mechanisms, which appear to affect the stability of the Dicer protein.40
DNA methylation of the Dicer gene was not found in a small subgroup of lung-cancer specimens.30
As the function of miRNAs in the genesis of tumors becomes clearer, further studies will be needed to elucidate the regulation and stability of the RNA-interference machinery.
Our findings have implications for the development of treatments for ovarian cancer that are based on RNA interference. Highlighting this point is the differential targeting efficiency of a constitutively expressed gene that we found through a functional assay of gene silencing. The shRNA that is complementary to the target gene has been tested in animal models and found to induce robust gene silencing. However, one study showed increased mortality among mice after delivery of multiple shRNA sequences.41
These results and our data suggest that miRNA processing may be hindered in tumors with low Dicer and low Drosha expression, which could lead to a poor clinical outcome.