miRNAs and Cancer: a little RNA goes a long way

Introduction

Sixteen years ago in back-to-back papers in Cell, Ambros, Ruvkun and their colleagues reported that a small RNA encoded by the lin-4 locus was capable of controlling the developmental timing of the nematode C. elegans by modulating the expression of a protein-coding gene lin-14 (Lee et al., 1993; Wightman et al., 1993). At the time few would have imagined that this discovery marked the birth of a new and far-reaching field of research. Indeed, it took several more years to appreciate that small RNAs like lin-4 (now termed microRNAs or miRNAs) were not just an interesting peculiarity of the nematode but were an abundant (and pervasive) feature of all bilateria, including Homo sapiens (reviewed in Bartel, 2004).

Over the past few years, biochemical and genetic studies have begun to reveal the physiological functions of individual miRNAs. miRNAs act by modulating the expression of target genes via sequence complementarity between the so-called “seed” sequence of the miRNA and the “seed-match” present in the mRNA. Such binding inhibits the translation and reduces the stability of the target mRNA, leading to decreased expression of the target protein (reviewed in Eulalio et al., 2008)). miRNAs have been shown to control a wide array of biological processes, including differentiation, proliferation and apoptosis. As the deregulation of these very same processes are hallmarks of cancer, it has been speculated for some time that mutations affecting miRNAs and/or their functional interaction with oncogenes and tumor suppressor genes might also participate to tumorigenesis. Moreover, because miRNAs can coordinately target multiple mRNAs, they could influence the activity of numerous proteins (for example, in a particular pathway) and exert more extensive effects.

In this review, we summarize recent findings that now overwhelmingly support an important role for these tiny RNAs in controlling cell transformation and tumor progression as well as the critical questions that remain to be answered.

A plethora of possible oncogenic mechanisms

Because miRNAs act by repressing gene expression through direct base-pairing interactions with their target mRNAs, there are several possible mechanisms through which miRNAs could affect tumorigenesis. Overexpression, amplification or loss of epigenetic silencing of a gene encoding a miRNA that targets a (or multiple) tumor suppressor gene(s) could inhibit the activity of an anti-oncogenic pathway. By contrast, the physical deletion or epigenetic silencing of a miRNA that normally functions to repress expression of one or more oncogenes might lead to increased protein expression and gain of oncogenic potency. More subtle miRNA mutations could reduce or eliminate binding to key targets or even create seed sequences to target a new set of mRNAs, thereby altering the balance of critical growth regulatory proteins. Seed-match sequences of target mRNAs could also be the sites of mutation, rendering them free from the repression of a given miRNA or subject to the effects of another (Table 1).

Table 1.

Potentially oncogenic genetic and epigenetic changes involving miRNAs or their targets. The table includes changes affecting directly the miRNA gene as well as genetic lesions in protein-coding oncogenes and tumor suppressor gene that would result in reduced or increased affinity for one or more miRNAs.

Mutation/epigenetic change	Predicted Functional Consequence	Examples	Reference
Deletion of miRNA	De-repression of oncogene	miR-15a-16-1	(Calin et al., 2002)
Epigenetic silencing of miRNA locus	De-repression of oncogene	miR-29, miR-203	(Bueno et al., 2008; Wang et al., 2008)
Point mutation affecting a miRNA or a miRNA precursor	Reduced affinity for oncogene	N.E.
	Increased affinity for tumor suppressor gene	N.E.
	Reduced processing efficiency	miR-16-1*	(Calin et al., 2005; Raveche et al., 2007)
	Increased processing efficiency	N.E.
Genomic Amplification or translocation of miRNA locus	Increased repression of tumor suppressor gene	miR-17~92 miR-21	(Haverty et al., 2008; Ota et al., 2004)
Loss of epigenetic silencing of miRNA locus	Increased repression of target tumor suppressor gene	N.E.
Point mutation in oncogene	Decreased or lost affinity for miRNA	N.E.
Point mutation in tumor suppressor gene	Gain or increased affinity for miRNA	N.E.
Rearrangement of 3′UTR (translocation, deletion)	Loss of miRNA-mediate repression	HMGA-2	(Lee and Dutta, 2007; Mayr et al., 2007)
	Gain of miRNA-mediated repression	N.E.

N.E. = No know examples.

^*the mutation is a single base change in the miR-16-1 precursor RNA. The sequence of the mature microRNAs is not affected.

Although not all of these potential mechanisms have been as yet documented in human cancers, over the past half-dozen years a veritable flood of reports have linked miRNAs to tumor development in one fashion or another. These range from genomic and gene expression alterations affecting miRNA genes in human cancers to observations in genetically-engineered mouse models of the disease. Taken together, the available data provide a compelling case that alterations in miRNA-mRNA regulation can promote tumor development.

As is true for protein-coding genes associated with cancer, the most convincing evidence linking miRNAs to tumorigenesis comes from genetic alterations in cancer cells. Beginning with the work of Croce and colleagues in 2002 (Calin et al., 2002), who showed that a pair of neighboring miRNAs are frequently focally deleted in human chronic lymphocytic leukemia (CLL; see below), there are now several examples in which miRNA genes are either lost or amplified in tumors (Reviewed in Calin and Croce, 2006). Moreover, miRNA expression profiling studies comparing cancer tissue to normal tissue have revealed provocative patterns of miRNA expression, some of which have been linked to changes in methylation status of the miRNA genes (Reviewed in Saito and Jones, 2006). Functional studies performed in cancer cell lines or mouse models of the disease have provided further support for a direct role of a subset of these miRNAs in tumorigenesis. Similar to the miRNA field as a whole, the discovery of the physiologically relevant targets of these cancer-associated miRNAs is still lagging, but here, too, there has been recent progress with interesting candidate targets emerging.

miRNAs as Oncogenes

miRNAs that are amplified or over-expressed in cancer could function as oncogenes, and a number of putative oncogenic miRNAs have been proposed. An interesting case is represented by miR-155, which has been found to be upregulated in several hematopoietic malignancies and tumors of the breast, lung and pancreas (Eis et al., 2005; Kluiver et al., 2005; Metzler et al., 2004). The gene encoding the primary transcript for miR-155 had been identified well before the discovery of miRNAs, as a common proviral DNA insertion site in lymphomas induced by the avian leucosis virus (Clurman and Hayward, 1989). The absence of an obvious open reading frame remained a puzzling feature of the BIC oncogene (as it was initially named) even after it was shown that it could cooperate with Myc in lymphomagenesis and erythroleukemogenesis (Tam et al., 2002). While the observation that the BIC RNA could form extensive secondary structures (including a 145 bp stem-loop that we now know is the precursor to miR-155) suggested that the RNA itself could be the oncogenic factor (Tam, 2001; Tam et al., 1997), its mechanism of action remained unclear until the identification of miR-155. The study of genetically-engineered mice with gain- and loss-of-function alleles of miR-155 has provided valuable insights into its physiologic and oncogenic properties. Ectopic expression of miR-155 in the B cells of transgenic mice is sufficient to induce polyclonal proliferation of pre-B cells, followed by full blown B cell leukemia (Costinean et al., 2006), although the mRNA targets of miR-155 are still unknown. As for the normal functions of miR-155, the characterization of miR-155 knock-out mice indicates that this miRNA is involved in the maturation of B cells in the germinal centers (Rodriguez et al., 2007; Thai et al., 2007; Vigorito et al., 2007). Indeed, a rapid and intense induction of miR-155 is observed in germinal center B cells after antigen stimulation, and in miR-155 mutant mice, immunoglobulin isotype switching in response to T-dependent and T-independent antigens is impaired. Several targets have been proposed to be important mediators of miR-155 function in B cells, including the transcription factor Pu.1 (Vigorito et al., 2007) and the activation induced cytidine deaminase (AID) (Dorsett et al., 2008; Teng et al., 2008). In an elegant set of genetic experiments in mice, two groups have shown that mutating a single recognition site for miR-155 in the 3-UTR of AID can partially phenocopy the consequences of miR-155 deletion (Dorsett et al., 2008; Teng et al., 2008).

Another notable member of the family of oncogenic miRNAs is the miR-17~92 cluster. This cluster, which consists of six miRNAs that are processed from a single primary transcript, was initially linked to human cancer based on the observation that it maps to a chromosomal region that is frequently amplified in a subset of human B cell lymphomas (Ota et al., 2004). miR-17~92 is also frequently over-expressed in several other human cancers, including lung cancers (Hayashita et al., 2005), hepatocarcinomas (Connolly et al., 2008), neuroblastomas (Schulte et al., 2008), and colorectal cancers (He et al., 2005). In an important in vivo test of the oncogenic potential of miR-17~92, He et al. demonstrated that a truncated version of the cluster (lacking miR-92) could cooperate with c-Myc and greatly accelerate tumorigenesis in a mouse model of B cell lymphoma (He et al., 2005). Moreover, the miR-17~92 cluster itself is among the transcriptional targets of c-Myc (O’Donnell et al., 2005). While miR-17~92 deregulation does not appear to be sufficient to initiate tumorigenesis per se, transgenic mice over-expressing this cluster in lymphocyte progenitors cells develop a lymphoproliferative disorder affecting both B and T cells that eventually results in autoimmunity (Xiao et al., 2008). In contrast, mice carrying a homozygous deletion of the miR-17~92 locus exhibit premature death of B cells at the pro-B/pre-B stage, resulting in lymphopenia (Ventura et al., 2008).

While the full spectrum of genes regulated by the six miRNAs encoded by the miR-17~92 cluster is still unknown, one candidate, the pro-apoptotic gene Bim, has been proposed to be a likely mediator of the B cell phenotype in miR-17~92-null and in miR-17~92 transgenic mice (Ventura et al., 2008; Xiao et al., 2008). Bim belongs to the family of pro-apoptotic BH3 only proteins, and has long been known to be a critical regulator of B cell survival (Bouillet et al., 1999; Bouillet et al., 2002) and a potent tumor suppressor gene in the Eμ-Myc model of B cell lymphoma (Egle et al., 2004; Hemann et al., 2005). Its 3′UTR contains multiple binding sites for miRNAs encoded by miR-17~92 and, consistent with Bim being a direct target of miR-17~92, its levels are increased in miR-17~92 –null pre-B cells and reduced in B cell from mice overexpressing miR-17~92. While Bim appears to be a functionally relevant target of miR-17~92, the expression of many other genes is likely to be controlled by this cluster, and many other putative targets have been identified (Dews et al., 2006; Fontana et al., 2008; Fontana et al., 2007; Lu et al., 2007; Petrocca et al., 2008).

miRNAs as tumor suppressors

Several miRNAs have been implicated as tumor suppressors based on their physical deletion or reduced expression in human cancer. Beyond these associations, functional studies of a subset of these miRNAs indicate that their over-expression can limit cancer cell growth or induce apoptosis in cell culture or upon transplantation in suitable host animals. This increasingly long list includes (but is not limited to) miR-15a~16-1 (Cimmino et al., 2005), the let-7 family (reviewed in Bussing et al., 2008), miR-34a and miR-34-b~c (reviewed in He et al., 2007b), miR-29 (Wang et al., 2008), miR-192 and miR-215 (Braun et al., 2008; Georges et al., 2008). However, in none of these cases has the definitive in vivo loss-of-function experiment been performed via gene targeting in the mouse.

The miR-15a~16-1 cluster of miRNAs has recently emerged as an excellent candidate to be the long sought-after tumor suppressor gene on 13q14. This chromosomal region is deleted in the majority of CLLs as well as mantle cell lymphomas and prostate cancers (Calin et al., 2002). There is good circumstantial evidence that miR-15a~miR-16-1 is a bona fide tumor suppressor. MiR-15a~16-1 is located in the minimally deleted region in CLL (Calin et al., 2002), and a germ-line point mutation (a single base change) in pre-miR-16-1 has been observed in a few CLL patients (Calin et al., 2005). This mutation has been linked to a reduced expression of miR-16-1, possibly due to less efficient processing of the precursor RNA, but large scale studies are needed to determine whether this is indeed a cancer-predisposing mutation. An interesting observation in this regard is that in New Zealand Black mice (NZB), a mouse strain that shows a strong predisposition to the development of a B lymphoproliferative disease (LPD) reminiscent of human CLL, a very similar base change in pre-miR-16-1 has been linked to the development of LPD (Raveche et al., 2007).

It appears that the tumor suppressive activity of miR-15a~16-1 is not limited to B cells. Over 50% of human prostate cancers show deletion of 13q14. Accordingly, a recent study has shown that inhibition of miR-15a and miR-16 activity leads to prostatic hyperplasia in mice and promotes survival, proliferation and invasion of primary prostate cells in vitro (Bonci et al., 2008). In the same study, the therapeutic potential of reconstituting expression of this cluster was illustrated by the significant regression of prostatic tumor xenografts upon intra-tumoral delivery of miR-15a and miR-16-1. While the identity of the critical targets of these two miRNAs is still unknown, the list of oncogenes that are directly regulated by miR-15a and miR-16-1 include BCL2, CyclinD1 and WNT3A (Bonci et al., 2008; Cimmino et al., 2005).

Among the most actively studied of the putative tumor suppressive miRNAs are the members of the let-7 family (reviewed in Bussing et al., 2008)). The human genome contains a dozen of let-7 family members, organized in eight different loci (http://microrna.sanger.ac.uk/cgi-bin/sequences/mirna_summary.pl?fam=MIPF0000002). The first member of the let-7 family was discovered in C. elegans, where it induces cell cycle exit and terminal differentiation of a particular cell type at the transition from larval to adult life (Reinhart et al., 2000). Consistent with a role in inhibiting tumor development in humans, reduced levels of multiple members of the let-7 family are frequently observed in lung cancers, where they correlate with poor prognosis (Takamizawa et al., 2004; Yanaihara et al., 2006). In addition, various let-7 genes are located at chromosomal sites deleted in a variety of human cancers (Calin et al., 2004). Let-7 genes can also be directly repressed by the c-Myc oncoprotein (Chang et al., 2008) and their precursor RNAs subjected to inhibition of further processing by lin-28 (Newman et al., 2008; Rybak et al., 2008; Viswanathan et al., 2008). Functionally, let-7 has been shown to repress members of the Ras family of oncogenes (Johnson et al., 2005) as well as the oncogene HMGA2 (Lee and Dutta, 2007; Mayr et al., 2007) and even c-Myc itself (Sampson et al., 2007). In the best example of an oncogenic mutation affecting a miRNA binding site, translocations involving the HMGA2 oncogene remove functional let-7 seed-match sequences, causing over-expression of the oncoprotein (Lee and Dutta, 2007; Mayr et al., 2007). Finally, over-expression of let-7 miRNAs can suppress tumor development in mouse models of breast and lung cancer (Esquela-Kerscher et al., 2008; Kumar et al., 2008; Yu et al., 2007). Mouse knock-out studies have not been reported for any let-7 family members, and, given the potential for functional overlap within this family, it may be a while before it is clear whether loss of let-7 function in the mouse can promote tumorigenesis.

A series of recent reports has explored the regulation of miRNAs by tumor suppressor genes. These studies have focused on the miRNAs regulated by p53, a tumor suppressor gene that is frequently inactivated in human cancers (reviewed in He et al., 2007b). This approach has lead to the identification of the miR-34 family as being an important mediator of p53 activity (Bommer et al., 2007; Chang et al., 2007; Corney et al., 2007; He et al., 2007a; Raver-Shapira et al., 2007; Tarasov et al., 2007; Wei et al., 2006). This family consists of three highly related miRNAs expressed from two separate loci: miR-34a from chromosome 1p36 and miR-34b/miR-34c as a cluster from 11q23. The transcription of both loci appears to be directly regulated by p53 via binding to conserved sites in the respective promoters (Chang et al., 2007; Corney et al., 2007; He et al., 2007a; Raver-Shapira et al., 2007; Tarasov et al., 2007; Wei et al., 2006). Similar to p53 itself, the expression of miR-34 can induce cell cycle arrest (Bommer et al., 2007; Cole et al., 2008; Corney et al., 2007; Tarasov et al., 2007; Tazawa et al., 2007) or apoptosis (Chang et al., 2007; Cole et al., 2008; He et al., 2007a; Raver-Shapira et al., 2007; Tarasov et al., 2007; Welch et al., 2007). Reduced expression of miR-34b/miR-34c has been reported in breast and non-small cell lung cancer cell lines (Bommer et al., 2007; Calin et al., 2004). Furthermore, miR-34a is located on 1p36, a region of frequent hemizygous deletion in human neuroblastomas and a variety of other cancers (Versteeg et al., 1995). Interestingly, this region includes another candidate tumor suppressor gene, CDH5, that acts by inducing p53 expression via p19Arf (Bagchi et al., 2007). Thus, a deletion of 1p36 can in principle impair the p53 pathway simultaneously upstream and downstream of p53.

miRNAs as modulators of tumor progression and metastasis

In addition to their role in promoting the development of primary tumors, miRNAs have also been implicated in affecting the tumor progression process, including the lethal, metastatic phase of the disease. Several cell biological processes, including those controlling adhesion, migration and invasion, are involved in allowing primary tumor cells to leave their original location and move to another site in the body. Not surprisingly, miRNAs help to regulate these processes as well and, as such, alterations in miRNA function can influence metastatic potential (reviewed in Ma and Weinberg, 2008).

Among several putative pro-metastatic miRNAs, miR-10b and miR-373 are of particular interest. miR-10b is a direct transcriptional target of Twist1 (Ma et al., 2007), a known inducer of the epithelial to mesenchymal transition (EMT) and metastatic progression (Yang et al., 2004). Ectopic expression of miR-10b in non-metastatic breast cancer cell lines promotes cellular invasiveness and the metastatic spread of transplanted tumors, at least in part as a consequence of the direct repression of the homeobox protein HOXD10 (Ma et al., 2007). miR-373 was identified in a functional screen for miRNAs that could promote cell migration in vitro (Huang et al., 2008) and its pro-metastatic potential has been validated in tumor transplantation experiments (Huang et al., 2008) using breast cancer cells. Of note, miR-373 had been previously identified as a potential oncogene (together with miR-372) in testicular germ-cell tumors (Voorhoeve et al., 2006), although it has been proposed that the prometastatic and the oncogenic properties of this miRNA are due to the regulation of different genes (CD44 and LATS2, respectively).

Studies of breast cancers have also revealed a series of miRNAs that are both under-expressed in advanced cancers and capable of inhibiting cell migration and metastatic spread. Members of the miR-200 family of miRNAs target the ZEB transcription factors, known inducers of the EMT, and, thus, reduce cellular migration and invasiveness (Burk et al., 2008; Gregory et al., 2008; Korpal et al., 2008; Park et al., 2008). Based on their differential expression in non-metastatic versus metastatic breast cancer cell lines, miR-126, miR-206 and miR-335 were also proposed to be inhibitors of tumor progression (Tavazoie et al., 2008). Indeed, over-expression of these miRNAs can inhibit metastasis in a cell transplantation model and reduced expression of miR-126 and miR-335 correlates with poor metastasis-free survival of breast cancer patients (Tavazoie et al., 2008).

Global deregulation of miRNAs in cancer

In this review, we have focused on the role of specific miRNAs in tumorigenesis, an already extensive and rapidly expanding list. However, recent work has also revealed intriguing changes in the global state of miRNA expression in cancer. Specifically, miRNA expression profiling experiments have demonstrated that most (although not all) miRNAs are under-expressed in tumor tissue compared to normal (Lu et al., 2005). While it is possible that this phenomenon reflects the less differentiated state of the tumor cells and/or their higher proliferation rate, one alternative explanation is that reduced miRNA levels are selected for during tumorigenesis because this itself provide some proliferative or survival advantage. These two possibilities are not necessarily mutually exclusive and indeed there is experimental evidence for both. For example, a significant increase in miRNA levels is observed upon induction of differentiation of the cancer cell line HL60 (Lu et al., 2005), consistent with the function of miRNAs to reinforce transcriptional programs and help maintain the differentiated state. On the other hand, working with experimental models of lung cancer, Kumar and colleagues have shown that genetic or RNAi-based inhibition of miRNA biogenesis can promote tumor formation and progression (Kumar et al., 2007). Finally, a recent study has demonstrated a widespread transcriptional silencing of miRNAs by c-Myc (Chang et al., 2008), suggesting that this might contribute to its potent oncogenic activity.

Independent of the functional consequences of miRNA expression patterns in cancer, miRNA profiles have value as diagnostic and prognostic markers of the disease. For example, it is sometimes impossible to determine the tissue of origin of a metastasis in patients with unknown primary tumors. Because many miRNAs display exquisite tissue specificity, miRNA profiling of these lesions might prove useful. Indeed, Lu and colleagues have shown that microRNAs profiling is more efficient at classifying poorly differentiated cancers than mRNA profiling (Lu et al., 2005), and Rosenfeld et al. have demonstrated an accuracy of ~90% in identifying the tissue of origin of primary and metastatic cancers using a classifier based on 48 miRNAs (Rosenfeld et al., 2008). miRNA profiling of human cancer might also help the oncologist identify the best treatment strategy by providing prognostic information (reviewed in Barbarotto et al., 2008)). Indeed, in the two most common forms of non-small cell lung cancers (adenocarcinomas and squamous cells carcinomas), high expression of miR-155 and low expression of let-7 correlate with poor prognosis (Takamizawa et al., 2004; Yanaihara et al., 2006). Similarly, in colon cancers, high miR-21 levels is associated with poor survival (Schetter et al., 2008), while in chronic lymphocytic leukemias a miRNA “signature” composed of 13 miRNAs is associated with disease progression (Calin et al., 2005). The results of these and other analogous reports are promising, but it is important to note that larger scale studies will be required to validate the usefulness of miRNA profiling in a clinical setting.

A view to the future

It is certainly too early to predict the full extent of the functions of miRNAs in cancer. A clearer picture will likely emerge as the efforts to re-sequence the cancer genome reveal the true frequency of mutations in miRNAs and in their target sequences in protein-coding genes (although the latter will require specific analysis of the 3′ UTR regions). At the same time, more sophisticated in vivo models will likely help determine the oncogenic and tumor suppressive potential of individual miRNAs and miRNA families. Also, improved experimental and computational methods to identify miRNA targets will provide a more comprehensive understanding of their mechanism of action and of the pathways that they modulate. It is difficult to overestimate the potential impact of these findings. As increasingly effective pharmacological means to modulate miRNA activities are currently being developed (Elmen et al., 2008a; Elmen et al., 2008b; Krutzfeldt et al., 2005), identifying miRNAs that are essential for tumor maintenance or for metastasis might provide exciting new therapeutic opportunities. What begun sixteen years ago as a peculiar discovery in the simple worm has already gone a long way to change the way we think of gene regulation. It might change the way we understand and treat cancers as well.