No abstract

No abstract

microRNAs (miRNAs) are small noncoding RNAs that regulate gene expression by targeting messenger RNA (mRNA) transcripts. Recently, a miRNA expression profile of human tumors has been characterized by an overall miRNA downregulation1–3. Explanations for this observation include a failure of miRNA post-transcriptional regulation4, transcriptional silencing associated with hypermethylation of CpG island promoters5–7 and miRNA transcriptional repression by oncogenic factors8. Another possibility is that the enzymes and cofactors involved in miRNA processing pathways may themselves be targets of genetic disruption, further enhancing cellular transformation9. However, no loss-of-function genetic alterations in the genes encoding these proteins have been reported. Here we have identified truncating mutations in TARBP2 (TAR RNA-binding protein 2), encoding an integral component of a DICER1-containing complex10,11, in sporadic and hereditary carcinomas with microsatellite instability12–14. The presence of TARBP2 frameshift mutations causes diminished TRBP protein expression and a defect in the processing of miRNAs. The reintroduction of TRBP in the deficient cells restores the efficient production of miRNAs and inhibits tumor growth. Most important, the TRBP impairment is associated with a destabilization of the DICER1 protein. These results provide, for a subset of human tumors, an explanation for the observed defects in the expression of mature miRNAs.

Epithelial-mesenchymal (EMT) and mesenchymal-epithelial (MET) transitions occur in the development of human tumorigenesis and are part of the natural history of the process to adapt to the changing microenvironment. In this setting, the miR-200 family is recognized as a master regulator of the epithelial phenotype by targeting ZEB1 and ZEB2, two important transcriptional repressors of the cell adherence (E-cadherin) and polarity (CRB3 and LGL2) genes. Recently, the putative DNA methylation associated inactivation of various miR-200 members has been described in cancer. Herein, we show that the miR-200ba429 and miR-200c141 transcripts undergo a dynamic epigenetic regulation linked to EMT or MET phenotypes in tumor progression. The 5′-CpG islands of both miR-200 loci were found unmethylated and coupled to the expression of the corresponding miRNAs in human cancer cell lines with epithelial features, such as low levels of ZEB1/ZEB2 and high expression of E-cadherin, CRB3 and LGL2, while CpG island hypermethylation-associated silencing was observed in transformed cells with mesenchymal characteristics. The recovery of miR-200ba429 and miR-200c141 expression by stable transfection in the hypermethylated cells restored the epithelial markers and inhibited migration in cell culture and tumoral growth and metastasis formation in nude mice. We also discovered, using both cell culture and animal models, that the miR-200 epigenetic silencing is not an static and fixed process but it can be shifted to hypermethylated or unmethylated 5′-CpG island status corresponding to the EMT and MET phenotypes, respectively. In fact, careful laser microdissection in human primary colorectal tumorigenesis unveiled that in normal colon mucosa crypts (epithelia) and stroma (mesenchyma) already are unmethylated and methylated at these loci, respectively; and that the colorectal tumors undergo selective miR-200 hypermethylation of their epithelial component. These findings indicate that the epigenetic silencing plasticity of the miR-200 family contributes to the evolving and adapting phenotypes of human tumors.

The authors wish to note, "The colony formation assay for SNU-1 upon enoxacin treatment in Fig. 1B is incorrect because of inadvertent duplication with the SNU-638 sample during the preparation of the figure. We have now replaced it with the correct assay. The data for RKO.shTRBP in Fig. 3C were erroneously graphed because the mean fold change was derived from an incorrect Fig. S5A where the formula used for quantitative RT-PCR analysis was ΔctAssay/ΔctControl rather than the correct formula 2-(ΔctAssay-ΔctControl). The data for RKO.shTRBP in Fig. S5A and CRC56 and CRC43 in Fig. S9B were erroneously graphed because of the same error with the formula. The corrected figures and their legends appear below. The figures in the supplemental information have also been corrected. These errors do not affect the conclusions of the article. We sincerely regret these mistakes. The error bars on the graphs used throughout the article indicate standard deviation (SD)."The corrected Fig. 1, Fig. 3, Fig. S5, and Fig. S9 appear below, along with their corresponding legends. The SI has been corrected online.