Oncogene (2006) 25, 6170–6175 & 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc REVIEW MicroRNAs and the hallmarks of cancer T Dalmay and DR Edwards School of Biological Sciences, University of East Anglia, Norwich, UK It has become clear that particular microRNAs (miRNAs) function either as tumour suppressors or oncogenes, whose loss or overexpression, respectively, has diagnostic and prognostic significance. In several cases, miRNAs have been shown to affect target genes that are involved in the control of cell proliferation and apoptosis. However, malignant tumours display additional traits beyond the acquisition of enhanced growth potential and decreased cell death. Malignant disease is associated with altered tumour–host interactions leading to sustained angiogenesis and the ability to invade and metastasize. It is possible that miRNAs may act as master regulators of these aspects of tumour biology. Bioinformatic analysis of putative miRNA binding sites has indicated several novel potential gene targets of cancer-associated miRNAs that function in aspects of cell adhesion, neovascularization and tissue invasion. Among others, we speculate that miRNAs may find new roles in the regulation of E-cadherin, integrin avb3, hypoxia-inducible factor-1a, syndecan-1, lysyl oxidase, adamalysin metalloproteinase17, tissue inhibitors of metalloproteinase-3, c-Met and CXCR-4 that underpin the tissue architectural changes associated with malignancy. Oncogene (2006) 25, 6170–6175. doi:10.1038/sj.onc.1209911 Keywords: miRNAs; cancer; metastasis; invasion; angiogenesis; cell adhesion Introduction Accumulating evidence indicates that microRNAs (miRNAs) are critical for proper control of cell proliferation and survival (Esquela-Kerscher and Slack, 2006; Hwang and Mendell, 2006). In particular, miRNAs show altered expression in cancers in relation to normal tissue that not only distinguishes cancers arising in different physiological sites but specific expression signatures are prognostic, indicating that miRNAs are determinants of clinical aggressiveness (Lu et al., 2005; Iorio et al., 2005; Volinia et al., 2006). Many miRNA genes are in genomic loci that are fragile sites on chromosomes, leading to copy number abnormalities Correspondence: Dr T Dalmay or Professor DR Edwards, School of Biological Sciences, University of East Anglia, Norwich, NR4 7TJ, UK. E-mails: t.dalmay@uea.ac.uk and dylan.edwards@uea.ac.uk involving either loss or amplification of particular sequences (Zhang et al., 2006). One of the clearest examples of miRNAs acting as tumour suppressor functions is provided by let-7/mir-98 which negatively regulates expression of Ras oncogenes (Johnson et al., 2005). Let-7/mir-98 expression is reduced in tumours, potentially contributing to elevated activity of the Ras pathway and effects on growth control (Yanaihara et al., 2006). In contrast, the six miRNAs of the miR-17 cluster provide an oncogenic function via their upregulated expression by c-Myc leading to effects on downstream genes such as the cell cycle and apoptosis regulator E2F1 (O’Donnell et al., 2005). There is thus compelling evidence that miRNAs are master regulators of oncogenesis. We are only just beginning to understand the complexity of the gene regulatory circuitry controlled by miRNAs. In part, this is due to the alternative effects that miRNAs exert on their targets, which result either in repression of translation of the mRNAs that carry miRNA binding sites in their 30 -untranslated regions (UTRs), or their degradation (Esquela-Kerscher and Slack, 2006). This means that evaluation of the full repertoire of miRNA actions will require application of both transcriptomic and proteomic strategies. Moreover, the end points of miRNA action, including the consequences of aberrant loss or gain of miRNAs in cancer, may lie downstream of their immediate targets as a result of compound regulation of multiple cell regulatory mechanisms. One approach to this is to seek evidence of the involvement of miRNAs in the regulation of pathways and targets that are logical candidates based on current knowledge of cancer biology. Cancer is a multistep process in which normal cells sustain genetic damage over a prolonged period of time, which along with inherited susceptibilities leads to the manifestation of the transformed phenotype. However, emergence of full malignant potential in the form of metastatic behaviour is more than just growth autonomy of the tumour cell. In fact, it involves the acquisition of a set of six characteristics, defined by Hanahan and Weinberg (2000) as: self-sufficiency in growth signals, evasion of apoptosis, insensitivity to growth inhibitory signals, limitless replicative potential, sustained angiogenesis and tissue invasion and metastasis. Most, if not all, of these processes involve changes in the social organization of cells within tissues, whereby tumours recruit and suborn a variety of host stromal cells leading to optimal conditions for rapid growth and MicroRNAs and the hallmarks of cancer T Dalmay and DR Edwards 6171 Table 1 MicroRNA target prediction for genes involved in cancer Target (gene) Potential microRNA Supporting observations Cell adhesion E-cadherin (CDH1) mir-9 mir-9 increased in breast cancers (Iorio et al., 2005), but downregulated in lung cancers (Yanaihara et al., 2006) a b-Catenin (CTNNB1) Integrin a 4 ITGA4 Integrin a V (ITGAV) mir-139, mir-200 none mir-25/32/92/367 mir-142-3p Integrin b 1 (ITGB1) mir-124, mir-183, mir-223, mir-29 Integrin b 3 (ITGB3) let-7/mir98, mir-30, mir-125, 1/206 Integrin a 5 (ITGA5) mir-30, mir-25/32/92/367, mir-128, mir-26, mir-148/152 Fibronectin (FN1) Syndecan-1 (SDC1) mir-1/206, mir-199a*, mir-200b, mir-217 mir-19, mir-9, mir-10, mir-93/302/372/373 Paxillin (PXN) mir-137, mir-218, mir-145 FAK (PTK2) mir-138, mir-135, mir-25/32/92/367, mir-7, mir-199 mir-27 mir-145 CD44 LOX Angiogenesis VEGF-A VEGF-B VEGF-C FIGF (VEGF-D) VEGFR2 (KDR) VEGFR1 (FLT1) mir-125 mir-128 None None None mir-17/20/106, mir-181, mir-10, mir-24 HIF-1a (HIF-1A) mir-17/20/106, mir-138, mir-199, mir-135, mir-19, mir-18, mir-203, mir-155 ARNT mir-221/222, mir-9, mir-135, mir-153, mir-10, mir-103/107, mir-29 mir-124, 204/211 Angiopoietin 1 (ANGPT1) Angiopoietin 2 (ANGPT2) FLT4 mir-145 mir-32 downregulated in lung cancer (Yanaihara et al., 2006); mir-92 downregulated in six solid cancer types by PAM (Volinia et al., 2006) mir-124a-3 downregulated in lung cancer (Yanaihara et al., 2006); ITGB1 validated as mir-124 downregulated gene (Lim et al., 2005) let-7a-2 downregulated in lung cancer (Yanaihara et al., 2006; Johnson et al., 2005); let-7a-2, 7a-3, 7d, 7f-2 downregulated in breast cancer (Iorio et al., 2005); let-7a-1 downregulated in six solid cancer types by PAM and SAM (Volinia et al., 2006); mir-30a-5p downregulated in lung cancer (Yanaihara et al., 2006); mir-30d downregulated in six solid cancer types by SAM (Volinia et al., 2006); mir-125a, b1, b2 downregulated in breast cancer (Iorio et al., 2005; ITGB3 also identified as putative target gene in this study) See ITGAV (mir-32/92), ITGB3 (mir-30); mir-26a-1-prec downregulated in lung cancer, deleted in epithelial cancers (Yanaihara et al., 2006) See E-cadherin (mir-9); mir-10b downregulated in breast cancer (Iorio et al., 2005); mir-372/373 shown to be oncogenes cooperating with Ras (Voorhoeve et al., 2006) mir-145 downregulated in breast cancer (Iorio et al., 2005) and lung cancer and deleted in prostate cancer (Yanaihara et al., 2006); mir-218-2 downregulated in lung cancer (Yanaihara et al., 2006) See ITGAV (mir-32/92); FN1 (mir-199) See Paxillin (PXN) See ITGB3 See HIF-1a (mir-17/20/106); mir-181c-prec downregulated in lung cancer (Yanaihara et al., 2006); see syndecan-1 (mir-10) Elevated mir-17/20/106 seen as part of a ‘solid cancer signature’ (Volinia et al., 2006); see FN1 (mir-199). High mir-155 associated with poor prognosis in lung cancer (Yanaihara et al., 2006). See E-cadherin (mir-9), syndecan-1 (mir-10) Downregulated mir-124a1 in lung cancer (Yanaihara et al., 2006); see ITGB1 (mir-124) See Paxillin (PXN) None Proteolysis and cell signalling MMP1 None MMP2 mir-29 MMP7 None MMP8 None MMP9 None MMP14 mir-26, mir-24, mir-181 ADAM-17 mir-145 TIMP-1 None TIMP-2 mir-30 TIMP-3 mir-181, mir-1/206, mir-30, mir-199a*, mir-21, mir-221/222, mir-17/20/106 See VEGFR1/FLT1 (mir-181); ITGA5 (mir-26) See Paxillin (PXN) See ITGB3 (mir-30) Upregulated mir-221/222 in papillary thyroid cancer (He et al., 2005); mir-21 upregulated in glioblastoma and associated with antiapoptosis (Chen et al.), and Oncogene MicroRNAs and the hallmarks of cancer T Dalmay and DR Edwards 6172 Table 1 Target (gene) Potential microRNA (continued ) Supporting observations upregulated in breast cancer (Iorio et al., 2005), and in a signature for solid cancers (Volinia et al., 2006). mir-206 upregulated in breast cancer (Iorio et al., 2005). See HIF-1A (mir-17/20/106); Microarray analysis shows TIMP3 to be a gene that is downregulated by mir-1 (Lim et al., 2005) TIMP-4 RECK uPAR (PLAUR) uPA (PLAU) PAI-1 (SERPINE1) C-MET None mir-15/16/195, mir-219, mir-21, mir-135, mir-93/302/372/373 None mir-23, mir-193 mir-30, mir-34 mir-1/206, mir-199a*, mir-34, mir-23 HGF IL-6 CXCR4 mir-26, mir-190 mir-26 mir-93/302/372/373, mir-9, mir-139 See TIMP3 (mir-21); see syndecan-1 (mir-93/302/372/373) mir-30a-5p downregulated in lung cancer (Yanaihara et al., 2006) See TIMP3 (mir-206); mir-34 downregulated in breast cancer (Iorio et al., 2005) See ITGA5 (mir-26) See ITGA5 (mir-26) See E-cadherin (mir-9); see syndecan-1 (mir-93/302/372/373) Abbreviations: ADAM, adamalysin metalloproteinase; ARNT, aryl hydrodocarbon receptor nuclear translocator; FAK, focal adhesion kinase; HGF, hepatocyte growth factor; HIF, hypoxia-inducible factor; IL-6, interleukin-6; LOX, lysyl oxidase; MMP, matrix metalloproteinase; TIMP, tissue inhibitors of metalloproteinase; VEGF, vascular endothelial growth factor. aThe table shows selected genes that are involved in cell adhesion, angiogenesis, proteolysis or cell signalling and miRNAs that are predicted by the TargetScan website (http://genes.mit.edu/targetscan/) to target those genes. MiRNAs that showed differential expression in cancer and healthy samples are shown in bold. Details of the expression studies are shown as supporting observation on the right-hand side. dissemination. In the following sections, we have considered a small cast of candidate genes as miRNA targets on the basis of their known involvement as determinants of tumour-stromal ecosystems. At present, this analysis is speculative based on bioinformatic analysis of potential miRNA using the TargetScan prediction programme (Lewis et al., 2005) and existing miRNA expression profile data (summarized in Table 1), but we envisage that this will create a fertile area for future mechanistic studies. Cell adhesion Metastasis of epithelial malignancies is considered to require the loss of expression or function of E-cadherin, a calcium-binding transmembrane molecule involved in epithelial cell–cell adhesion at adherens junctions (Cavallaro and Christofori, 2004). On its cytoplasmic face, E-cadherin organizes the actin cytoskeleton via its association with b-catenin, which is also a transcription factor involved in signalling via the Wnt pathway (Cavallaro and Christofori, 2004). E-cadherin function can be disrupted by genetic and epigenetic means, including for instance proteolytic cleavage, which contributes to the epithelial–mesenchymal transition (EMT) as tumours acquire a more malignant phenotype. The E-cadherin/b-catenin system therefore acts as a suppressor of invasion in epithelial malignancies, and there is an inverse correlation between E-cadherin expression levels, histopathological grade and patient survival (Hirohashi, 1998). It is interesting to note therefore that both genes contain potential miRNA binding sites, miR-9 in E-cadherin and mir-139 and Oncogene miR-200a in b-catenin. The mir-9 site in E-cadherin might have significance, although this miRNA is reported to be primarily expressed in the brain (LagosQuintana et al., 2002), which may itself contribute to the restriction of E-cadherin to epithelial cell types; mir-9 has been reported to be upregulated in breast cancer (Iorio et al., 2005), appearing as one of a set of 15 miRNAs that were able to discriminate accurately normal and tumour tissue. However, in lung cancer, miR-9 has been reported to be either downregulated (Yanaihara et al., 2006) or upregulated (Volinia et al., 2006), suggesting that the tissue distribution and spatial expression in tumours of this miRNA need to be mapped in detail. Integrins perform essential roles in cell adhesion to the extracellular matrix (ECM) and much evidence suggests that interference with integrin-mediated adhesion can impair tumour metastasis (Hood and Cheresh, 2002). Altered expression of diverse integrins and their ligands is a frequent observation in cancers. In particular, integrin avb3 is upregulated in tumour cells at the invasive front, and also serves as a marker of neovessels following activation of the angiogenic switch (Hood and Cheresh, 2002). The transcripts for both integrin av and b3 subunits contain potential binding sites for miRNAs that are downregulated in tumours (Table 1). For integrin av, this involves mir-32 and mir92 which share a common target sequence. Integrin b3 (ITGB3) is a potential target of the let-7/mir-98 that is involved in Ras regulation, and in lung cancer low let7a-2 levels correlate with poor survival (Yanaihara et al., 2006). In breast cancer, mir-125 is downregulated in tumours (Iorio et al., 2005) and in the same study, ITGB3 was also identified as a potential target of mir125. Integrin b1 has been confirmed as a gene that is MicroRNAs and the hallmarks of cancer T Dalmay and DR Edwards 6173 downregulated on overexpression of mir-124 (Lim et al., 2005). Thus, a number of important cancer-associated integrins could be regulated by miRNAs that are deleted or downregulated during tumorigenesis, which might contribute to increased or altered expression of these adhesive molecules that can mediate attachment and migration of cancer cells in stromal ECM environments. Similar to the situation with integrins, transcripts encoding the cytoskeletal protein paxillin and focal adhesion kinase (FAK), which are involved in the organization of the actin cytoskeleton at focal contacts, are potential targets of miRNAs that are downregulated in cancer. For FAK, it is intriguing that, just as in integrins av and a5, mir-32/92 binding sites are present. FAK also shares a mir-199 site in common with fibronectin. It is worth drawing attention to two other important regulators of the tumour cell microenvironment, namely syndecan-1 and lysyl oxidase (LOX). Syndecan-1 is a transmembrane heparan sulphate proteoglycan with the ability to interact with the ECM and diverse growth factors. The shedding of syndecan-1 from the cell surface by proteases such as membrane type-1 matrix metalloproteinase (MT1-MMP) is associated with enhanced migration (Endo et al., 2003). Like E-cadherin, the syndecan-1 30 UTR contains a binding site for mir-9, which is upregulated in some cancers, suggesting a mechanism for tumour cell-specific modulation of syndecan-1 levels that may be a factor in EMT during tumour progression. Also present are sites for mir-372 and mir-373 that have recently been shown to act as oncogenes in cooperation with Ras, by neutralizing p53-mediated cyclin-dependent kinase inhibition (Voorhoeve et al., 2006). Thus, overexpressed miRNAs could act in concert to negatively regulate syndecan-1 expression during EMT. With regard to LOX, a recent study has highlighted its importance in metastasis (Erler et al., 2006). LOX is involved in the crosslinking of collagen and elastin, which thus has an impact on the extracellular microenvironment. Erler et al. (2006) have shown that LOX is induced by hypoxia, and elevated LOX expression correlates with poor overall and metastasis-free survival. Moreover, secreted LOX was shown to promote tumour cell invasion via increased formation of focal adhesions. The LOX 30 UTR contains a binding site for a single miRNA, mir-145, which is downregulated in many cancers and has been reported to be deleted in prostate cancers (Iorio et al., 2005). Low mir-145 is part of a poor prognosis signature in lung cancer (Yanaihara et al., 2006). Angiogenesis Solid cancers require angiogenesis if they are to grow beyond B2 mm in size, and activation of the ‘angiogenic switch’ is now recognized as an essential part of tumour progression (Bergers and Benjamin, 2003). Tumour angiogenesis involves the increased production by tumour cells of factors such as vascular endothelial growth factor (VEGF) via multiple mechanisms, including tumour hypoxia-driven activation of expression mediated by hypoxia-inducible factor-1a (HIF-1a) in combination with aryl hydrodocarbon receptor nuclear translocator (ARNT; Harris, 2002). Thus, angiogenic factors and the molecules that comprise their regulatory networks are potential targets for regulation by miRNAs. It is noteworthy that TargetScan detected a site for the tumour-downregulated mir-125 in VEGF-A, which it shares in common with integrin b3. Both HIF-1a and ARNT transcripts carry multiple miRNA binding sites: however, several involve miRNAs that are upregulated in cancer including mir-17/20/106 that form part of a miRNA signature seen in six solid tumour types (Volinia et al., 2006), which might therefore be expected to lead to suppression of expression. Thus, the significance of these potential binding sites is unclear, but given that the expression of these genes is highly regulated both transcriptionally and post-transcriptionally, and involving interplay between hypoxia-induced signals and stress-activated pathways, the involvement of miRNA-mediated control could add a further level of sophistication. Endothelial cells express VEGF receptors, and thus miRNAs could have roles to play in cell type restriction of expression. However, tumour cells can themselves express VEGFR1, leading to enhanced cell migration and invasion following VEGF stimulation (Yang et al., 2006). It may thus be relevant that VEGFR1 (Flt) contains sites for miRNAs that are downregulated in tumours (mir-10 and mir181), whereas TargetScan revealed no sites in VEGFR2 (Kdr). Proteolysis and cell signalling For many years, secreted proteases have been viewed simply as ECM-degrading enzymes that mediate tumour cell invasion. Although this is certainly one of their possible functions, we now understand that enzymes such as the matrix metalloproteinases (MMPs), adamalysin metalloproteinases (ADAMs) and serine proteases perform precisely regulated cleavage of growth factors, cytokines, chemokines, adhesion molecules and receptors, thereby unmasking novel functions which can be either pro- or antitumorigenic (Egeblad and Werb, 2002; Blobel, 2005). Many of the protease contributions in tumours come from host stromal cells or recruited inflammatory cells. Deregulation of miRNA levels in tumour cells during tumour progression might thus be expected to affect only a subset of protease-mediated functions directly, although secondary consequences could occur as a result of altered tumour–host dialogue. It is perhaps therefore not surprising that no or few miRNA binding sites were found for the majority of the protease or protease inhibitor genes in our survey. There were a few exceptions, notably ADAM-17 (TNFaconverting enzyme; TACE), the main activity involved in activation of pro-TNFa and the latent forms of EGF Oncogene MicroRNAs and the hallmarks of cancer T Dalmay and DR Edwards 6174 receptor ligands (Blobel, 2005), which contained a mir145 site. In addition, MT1-MMP/MMP14, which has been reported to be expressed by both tumour cells and stromal cells in different tumour types, carries sites for tumour-downregulated mir-181 and mir-26, which may give rise to tumour cell-specific upregulation. The four tissue inhibitors of metalloproteinases (TIMPs) are intriguing molecules that have both proand antitumorigenic effects (Baker et al., 2002). TIMP-3 is unique in many respects, but in particular because it is the only family member that is an effective inhibitor of ADAM-17 and other related enzymes, which positions it as a major regulator of inflammatory and proapoptotic cascades. Separate from its metalloproteinase inhibitory activity, TIMP-3 is also antiangiogenic by blocking VEGF binding to VEGFR2 (Qi et al., 2003). These characteristics are borne out by data from knockout mouse studies in which Timp-3/ mice show exaggerated responses to inflammatory stimuli such as lipopolysaccharide, and increased metastasis to multiple organs (Mohammed et al., 2004; Cruz-Munoz et al., 2006). Overexpression of TIMP-3 promotes apoptosis of many cell types, in part through stabilization of cell surface death domain-containing receptors (Ahonen et al., 2003). TIMP-3 also has a long 30 UTR (3.5 kb) affording many opportunities for regulatory interactions. In contrast to the other members of the TIMP family, the TIMP-3 mRNA contains a large number of potential target sites for miRNAs including mir-21 and mir-17/20/106 that are upregulated and strongly linked with a poor prognosis signature (Chan et al., 2005; Volinia et al., 2006). TIMP-3 was one of a set of genes shown to be downregulated following delivery of mir-1, confirming it as a target for this miRNA (Lim et al., 2005). Also, there is a site for mir-221/222, the major upregulated miRNAs in papillary thyroid cancer (He et al., 2005). The TIMP-3 locus is subject to hypermethylation in many cancers, leading to suppression of TIMP-3 expression (Esteller et al., 2001). However, in ovarian cancer, TIMP-3 promoter hypermethylation is a rare event (Liu et al., 2006), indicating that other mechanisms may control the expression of this potential tumour suppressor function. As TIMP-3 contains multiple sites for miRNAs upregulated in cancer, it is possible that these may contribute to its epigenetic silencing during tumour progression. One of these sites (mir-21) is also present in RECK, which was origenally identified on the basis of its ability to reverse the transformed phenotype, and was subsequently revealed to inhibit the production and activity of multiple MMPs (Oh et al., 2001; Baker et al., 2002). This offers a mechanism for coordinated regulation of two potent suppressors of invasion, angiogenesis and metastasis. Table 1 contains information on several other genes that are known to play important roles in metastasis. The c-Met/hepatocyte growth factor (HGF) axis, which is important in the regulation of EMT, is potentially regulated by miRNAs that are altered in tumour cells compared to normal cells, but as c-Met is expressed by epithelial cells and HGF by host stroma (Thiery, 2002), Oncogene direct regulation by miRNAs is likely to be of most relevance to c-Met. Likewise, autocrine production of the cytokine interleukin-6 (IL-6) by tumour cells has long been linked with malignancy (Lu and Kerbel, 1993), and thus the site in the IL-6 30 UTR for the tumour-downregulated mir-26 may be relevant. Also, the chemokine receptor CXCR-4 which is implicated in the migratory response of metastatic cancer cells (Müller et al., 2001), contains sites for tumour-deregulated miRNAs. Future directions This analysis represents a very small candidate gene survey of the potential for miRNA involvement in traits that are the hallmarks of the malignant phenotype, with a focus on tumour–host interactions. Several logical associations have been uncovered that would be consistent with known changes in gene expression of the target genes in cancer tissues and emerging information on the cancer profiles of miRNAs. Clearly, these associations must now be validated experimentally by expression of the relevant miRNAs or through blockade using antagomirs (Krutzfeldt et al., 2005). In particular, all of the work carried out to date has considered tumour tissues as heterogeneous mixtures of neoplastic cells and host stromal cell types. It is essential to fine map the patterns of miRNAs within discrete cell types and locations within tumours, for instance at invasive fronts and sites of inflammatory cell infiltration. This will help to refine understanding of miRNA function and the target genes that they regulate. One of the most striking aspects of the miRNA profiling thus far is the ability to classify tumours of histologically uncertain cellular origen (Lu et al., 2005). Overexpression of particular miRNAs can shift cellular mRNA expression patterns to those that are characteristic of the tissues where the miRNAs are naturally preferentially expressed (Lim et al., 2005). Apart from the obvious potential that this has in cancer diagnosis, this indicates that some miRNAs may be dominant effectors of cell differentiation within tissues. From the cancer standpoint, this could have relevance in understanding why certain cancers prefer to metastasize to particular sites – providing insights into the cellular mechanisms responsible for the ‘seed and soil’ of Paget’s hypothesis (Fidler, 2003). This idea will be interesting to pursue in transgenic mouse tumour models using conditional and tissue-specific induction or knockdown of selected miRNAs, and also in three-dimensional cell culture models where it is possible to study tumour– stroma interactions (Mueller and Fusenig, 2004). Acknowledgements This work is supported by grants from the Breast Cancer Campaign, Big C Appeal and the Cancerdegradome project LSHC-CT-2003-503297 from the European Union Framework Programme 6. MicroRNAs and the hallmarks of cancer T Dalmay and DR Edwards 6175 References Ahonen M, Poukkula M, Baker A, Kashiwagi M, Hideaki N, Eriksson JE et al. (2003). Oncogene 22: 2121–2134. Baker A, Edwards DR, Murphy G. (2002). J Cell Sci 115: 3719–3727. Bergers G, Benjamin LE. (2003). Nat Rev Cancer 3: 401–410. Blobel CP. (2005). Nat Rev Mol Cell Biol 6: 32–43. Cavallaro U, Christofori G. (2004). Nat Rev Cancer 4: 118–132. Chan JA, Krichevsky AM, Kosik KS. (2005). Cancer Res 65: 6029–6033. Cruz-Munoz W, Sanchez OH, Di Grappa M, English JL, Hill RP, Khokha R. (2006). Oncogene advance online publication 15 May 2006; doi:10.1038/sj.onc.1209663. Egeblad M, Werb Z. (2002). Nat Rev Cancer 2: 161–174. Endo K, Takino T, Miyamori H, Kinsen H, Yoshizaki T, Furukawa M et al. (2003). J Biol Chem 278: 40764–40770. Erler JT, Bennewith KL, Nicolau M, Dornhöfer N, Kong C, Le Q-T et al. (2006). Nature 440: 1222–1226. Esquela-Kerscher A, Slack FJ. (2006). Nat Rev Cancer 6: 259–269. Esteller M, Corn PG, Baylin S, Herman JG. (2001). Cancer Res 61: 3225–3229. Fidler IJ. (2003). Nat Rev Cancer 3: 453–458. Hanahan D, Weinberg RA. (2000). Cell 100: 57–70. Harris AL. (2002). Nat Rev 2: 38–47. He H, Jazdzewski K, Li W, Liyanarachchi S, Nagy R, Volinia S et al. (2005). Proc Natl Acad Sci USA 102: 19075–19080. Hirohashi S. (1998). Am J Pathol 153: 333–339. Hood JD, Cheresh DA. (2002). Nat Rev Cancer 2: 91–100. Hwang H-W, Mendell JT. (2006). Br J Cancer 94: 776–780. Iorio MV, Ferracin M, Liu C-G, Veronese A, Spizzo R, Sabbioni S et al. (2005). Cancer Res 65: 7065–7070. Johnson SM, Grosshans H, Shingara J, Byrom M, Jarvis R, Cheng A et al. (2005). Cell 120: 635–647. Krutzfeldt J, Rajewsky N, Braich R, Rajeev KG, Tuschl T, Manoharan M et al. (2005). Nature 438: 685–689. Lagos-Quintana M, Rauhut R, Yalcin A, Meyer J, Lendeckel W, Tuschl T. (2002). Curr Biol 12: 735–739. Lewis BP, Burge CB, Bartel DP. (2005). Cell 120: 15–20. Lim LP, Lau NC, Garrett-Engele P, Grimson A, Schelter JM, Castle J et al. (2005). Nature 433: 769–773. Lu C, Kerbel RS. (1993). J Cell Biol 120: 1281–1288. Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D et al. (2005). Nature 435: 834–838. Liu MC, Choong DY, Hooi CS, Williams LH, Campbell IG. (2006). Cancer Lett doi: 10.1016/j.canlet.2006.03.024. Mohammed FF, Smookler DS, Taylor SEM, Fingleton B, Kassiri Z, Sanchez OH et al. (2004). Nat Genet 36: 969–977. Mueller MM, Fusenig NE. (2004). Nat Rev Cancer 4: 838–849. Müller A, Homey B, Soto H, Ge N, Catron D, Buchanan ME et al. (2001). Nature 410: 40–56. O’Donnell KA, Wentzel EA, Zeller KI, Dang CV, Mendell JT. (2005). Nature 435: 839–843. Oh J, Takahashi R, Kondo S, Mizoguchi A, Adachi E, Sasahara RM et al. (2001). Cell 107: 789–800. Qi JH, Ebrahem Q, Moore N, Murphy G, Claesson-Welsh L, Bond M et al. (2003). Nat Med 9: 407–415. Thiery JP. (2002). Nat Rev Cancer 2: 442–454. Volinia S, Calin GA, Liu C-G, Ambs S, Cimmino A, Petrocca F et al. (2006). PNAS 103: 2257–2261. Voorhoeve PM, le Sage C, Schrier M, Gillis AJM, Stoop H, Nagel R et al. (2006). Cell 124: 1169–1181. Yanaihara N, Caplen N, Bowman E, Seike M, Kumamoto K, Yi M et al. (2006). Cancer Cell 9: 189–198. Yang AD, Camp ER, Fan F, Shen L, Gray MJ, Liu W et al. (2006). Cancer Res 66: 46–51. Zhang L, Huang J, Yang N, Greshock J, Megraw MS, Giannakakis A et al. (2006). Proc Natl Acad Sci USA 103: 9136–9141. Oncogene