Evidence from multiple tumor samples reveal a marked change in miRNA expression profiles, indicating widespread alterations in miRNA networks in tumorigenesis4
. As miRNAs regulate key processes involved in tumor evolution such as transcriptional regulation, differentiation, proliferation and apoptosis, such alteration may be of crucial importance to cancer formation. The first detailed example of this link came from the study of the minimally deleted region of a common chromosomal 13q14 abnormality in chronic lymphocytic leukemia, showing that deletion of miR-15a and miR-16-1 is associated with this lymphoproliferative disorder5
. The causative role and the mechanism of this miRNA deletion were also confirmed and elucidated in a mouse model6
, and it was also suggested that these alterations may be acquired early in the disease process.
Following this report, numerous studies have revealed the details of deregulated miRNA expression in various human malignancies, and the list of the potential and recognized tumor-suppressive miRNAs and oncogenic miRNAs is growing7
. Moreover, it is now evident that some characteristics of cancer-related biological processes; tumor angiogenesis, maintenance of cancer stem cell, and metastasis, are associated with miRNA switches8–11
. The complexity by which different cancer cells acquire miRNA variation is ever expanding. For example, miR-223 regulation in leukemia demonstrates how gross genomic alterations may be involved in the deregulation of miRNA function. It was shown that the t(8;21) translocation product AML1/ETO recruits chromatin remodeling enzymes at an AML1-binding site on the pre-miR-223 gene, and induces heterochromatic silencing of miR-22312
. A high frequency of genomic alterations was also confirmed in a study of genomic hybridization on 227 human ovarian, breast and melanoma cancer cells, with a strong correlation between genomic changes such as copy number alteration, and miRNA expression13,14
. Thus, it appears that miRNAs form a lucrative and commonly affected target for genomic instability, leading to significant perturbations of principal regulatory systems, and allowing tumor progression.
2.1 Fragile sites may contribute to genomic instability-derived microRNA alterations
Chromosomal fragile sites are specific loci prone to breakage and rearrangements when cells are exposed to partial inhibition of DNA synthesis. Many genes involved in cancer-specific recurrent translocations are located within fragile sites (reviewed in Drukin and Glover15
). In 2004, Calin et al
, performed the first systematic analysis of miRNAs in fragile sites, based on the known or predicted miRNAs at the time16
. They mapped 186 miRNAs and compared their location to the location of known nonrandom genetic alterations, with the observation that miRNA genes and miRNA clusters are frequently located at or around fragile sites, as well as in regions of loss of heterozygosity, amplification, or common breakpoint regions. Overall, 98 of 186 (52.5%) of miRNA genes are in cancer-associated genomic regions or in fragile sites, which correlated with the lower measured expression of miRNAs. It may be hypothesized that this association is a result of a detection bias or that the use of fragile sites as integration sites for viral genomes and/or transposable elements may contribute to this intriguing finding. Although the evolutionary basis for this finding is still perplexing, this first report suggested that genomic instability may preferentially target miRNAs. Additional studies demonstrated a similar association between miRNAs and fragile sites or cancer susceptibility loci, in murine cancer models17,18
. Furthermore, based on computational models to predict miRNA promoters (either host gene promoter for intragenic miRNAs or independent promoters), it was also hypothesized that in addition to direct hits to the miRNA genes, promoter deletion can also affect miRNA expression17
. In a more recent report, utilizing current miRNA libraries Lagana et al19
have found that fragile sites are particularly enriched in miRNAs but also in protein coding genes. The mapping of human miRNA genes revealed that 242 of 715 miRNAs (33.8%) were located in chromosome fragile sites, considering that fragile sites account for about 25.8% of the length of all sites considered, with a similar finding for protein coding genes (33.9% are located in fragile sites). They have also shown that there is a significant chromosomal variability, for example chromosome 19 has the highest number of miRNAs in fragile regions, while no miRNAs have been yet found in the fragile sites of chromosome 20.
2.2 Epigenetic instability and microRNAs
Changes in miRNA expression can also be achieved by epigenetic modification, indeed, many miRNAs are found near CpG islands20
. In a landmark report by Saito et al
, miR-127, a tumor suppressor miRNA targeting BCL-6
, was found to be embedded in a CpG island and silenced by DNA methylation21
. Furthermore, the same group also demonstrated that therapy with chromatin modifying agents successfully reactivated the silenced miRNA in a bladder cancer cell line22
. This work lead to the identification of multiple other examples of epigenetic silencing of miRNAs in cancer23–27
. Specific examples include the silencing of miR-9-1 through hypermethylation as a frequent and early event in breast cancer, and the role of miR-1 silencing in hepatocellular carcinoma demonstrated convincingly in DNMT1 knock out mice28
. More systematic methods of observing epigenetic modification of miRNAs were attempted by differential profiling of miRNA expression patterns in DNMT1
double-gene knockouts in HCT116 cells29
, and in metastatic lymph node cancer cells after 5-aza-2′-dC-induced demethylation26
. Conversely, in a study of miRNAs in ovarian cancer, three miRNAs (miR21, miR-203, miR-205) were found to be over-expressed, suggesting that at times epigenetic activation through hypomethylation may lead to miRNA over-expression in vivo30
. Thus epigenetic changes may also play a key role in miRNA perturbation secondary to genomic instability.
2.3 MicroRNAs, beyond gene copy number
Layers of post transcriptional regulation of astonishing complexity are being unraveled in miRNA biology, all demonstrating that miRNA function is dictated only in part by the actual transcription of the miRNA gene. Although far less well studied, these layers of complexity may play an important part of miRNA alteration by genomic instability in cancer, far beyond copy number changes. One of the most intriguing examples of miRNA single base alteration leading to decreased processing is in the case of the SNP rs2910164 in miR-146a which contributes to the mispairing in the hairpin of the precursor associated with a predisposition for thyroid carcinoma31
. A striking aspect of this finding is that heterozygocity carried a higher susceptibility than homozygocity, an unusual occurrence in cancer genes, with further demonstration of homozygous to heterozygous somatic mutations at rs2910164 in several patients with papillary thyroid tumors. rs2910164 was also associated with a younger age of breast cancer diagnosis in familial breast cancer, which in this case is thought to involve increased binding affinity to the target mRNA, BRCA132
. Another example is an SNP in the terminal loop of pre-miR-27a, rs895819, which confers a reduced risk of developing breast cancer in families with a history of BRCA and non-BRCA-related disease33,34
, caused perhaps by changes in miRNA maturation. The location of the rs895819 SNP in the centre of the terminal loop of mir-27a may act to decrease the size of the loop and affect the binding of DROSHA or alternatively, affect the binding affinity of several DROSHA inhibitors (e.g lin28) (Reviewed in Ryan et al20
In addition to mutations in the miRNA, alterations of the 3′ UTR of a gene may create or abolish, a miRNA binding site. This has been explored in detail in the case of let-7
binding to the 3′ UTR of the KRAS
oncogene, where a a SNP (rs61764370) was correlated with an increase risk of developing lung and oral cancer35,36
, with the possible presence of a negative feedback mechanism leading to further decrease of let-7
In addition, miRNAs may have target sites in the 5′ UTR and open reading frames. This is further complicated by occasional intron interruption which requires splicing for effective binding38
, or removal of binding sites by splicing affecting relative abundance of splice variants39
. Another aspect of miRNA heterogeneity has been recently identified and termed isomiRs40
. These variations are thought to arise from variable DROSHA and DICER cleavage sites in the hairpin loop, leading to 3′ deletion/addition, 5′ deletion/addition and internal modifications. An additional layer of complexity of miRNA transcriptional regulation was recently discovered, whereby pseudogenes participate in the regulatory interaction between miRNAs and coding mRNAs by competing with their sister functional genes over binding with miRNA41
. The magnitude of this effect on miRNA biology in cancer remains to be fully explored, but this first publication provided valuable insights with regards to pseudogene contribution to miRNA regulation of key players such as the PTEN and KRAS proteins.
Thus in addition to the more crude copy deletion and amplification already demonstrated in miRNA deregulation in cancer, it is plausible that genomic instability may lead to mutations that can affect miRNA regulation by altering processing, target binding sites and pseudogene binding in a highly complex fashion.
2.4 Global modification of microRNA processing
In addition to hits to specific miRNAs, more global changes to miRNA processing can also contribute to tumorigenesis. While miRNA maturation is a tightly regulated event with biogenesis of mature miRNAs involving several enzymatic steps, experimental observation of cancer cells showed widespread alterations in miRNA expression. While some miRNAs are elevated, most miRNAs have significantly reduced expression4
. Furthermore, The global repression of miRNA maturation promotes cellular transformation and tumorigenesis42
. While the mechanisms remain to be fully elucidated, it suggests that miRNAs might have intrinsic function in tumor suppression.
This global suppression may be a result of a block in intranuclear processing by Drosha as recently suggested43
and demonstrated in ovarian cancer to be associated with poor survival44
. Thomson et al further studied this by comparing expression maps of mature miRNA and of their primary transcript, where they demonstrated the loss of correlation between pri-miRNA and mature expression in the tumor samples43
. Alternatively, global miRNA decreases may result from disruption of pre-miRNA’s cytoplasmic export after Drosha processing. XPO5 is responsible for miRNA nuclear export, and was found to be down-regulated in lung cancer45
Following export from the nucleus, Dicer is responsible for the cleavage of pre-miRNA into mature miRNA. Conflicting results have been reported regarding its expression in cancer cells with some indicating up-regulation or amplification as in the case of prostate, lung and ovarian cancer14,45,46
, while others have shown that reduced expression of Dicer is associated with a poor prognosis in lung cancer47
. Kumar et al demonstrated in a mouse knock out model that DICER functions as a haploinsufficient tumor suppressor in cancer48
. Frameshift mutations in TRBP, a regulator of Dicer stability, that introduce premature stop codons, resulted in reduced TRBP expression, reduced DICER expression and lower miRNA production as well as higher cellular proliferation levels49
An exciting study by Suzuki et al, demonstrated an important link between genomic instability and global miRNA regulation through the role of the most frequently mutated protein p53 in regulating miRNA processing50
. The authors revealed that DNA damage and p53 activation increased expression of not only a known p53 target, miR-34a, but also a broad set of mature miRNAs, and their accumulation is mediated transcriptionally and post-transcriptionally. This was found to be a result of a direct interaction between p53 and endogenous Drosha/DGCR8, modulated through the p68/p72 RNA helicase. Collectively, their results suggested that the p53 pathway promotes the recruitment of the p68–Drosha complex to the target pri-miRNAs, enhancing Drosha-mediated processing of several miRNAs. Furthermore, mutant p53 proteins with oncogenic activities hindered miRNA processing in a transcription-independent manner, through interference with miRNA biogenesis by negative titration of Drosha microprocessor components and interference with Drosha/p68 accessibility to pri-miRNAs.
In summary, avoidance of tight gene regulation by miRNAs might be a general feature of cancer cells. Genomic instability may contribute to global miRNA escape through mutations affecting different parts of the processing and maturation machinery either directly or indirectly as in the case of p53. The loss of a miRNA-controlled gene regulatory layer may serve as a global program for cellular proliferation as it may hold an intrinsic tumor-suppressive function.