Metastasis is a complex, multi-step process by which primary tumour cells invade adjacent tissue, enter the systemic circulation (intravasate), translocate through the vasculature, arrest in distant capillaries, extravasate into the surrounding tissue parenchyma, and finally proliferate from microscopic growths (micrometastases) into macroscopic secondary tumours1. In recent years, studies have been carried out to investigate the genes and gene products that drive the metastatic process. For instance, work in a number of laboratories has revealed several transcription factors that can program many of the cell-biological changes needed to execute the initial steps of the invasion–metastasis cascade2, 3, 4, 5, 6, 7, 8.
Lately, it has become evident that, in addition to alterations in protein-encoding genes, abnormalities in non-coding genes can also contribute to cancer pathogenesis9, 10. In particular, a class of small cellular RNAs, termed microRNAs (miRNAs), acting as agents of the RNA interference pathway, can lead to silencing of their cognate target genes, doing so either by cleaving mRNA molecules or by inhibiting their translation11. Indeed, miRNAs have been implicated in the regulation of a variety of cellular processes, including apoptosis12, haematopoietic differentiation13, metabolism14, skin morphogenesis15 and neural development16.
More than 50% of annotated human miRNA genes are located in fragile chromosomal regions that are susceptible to amplification, deletion, or translocation during the course of tumour development17. Moreover, recent evidence indicates that some miRNAs can function either as oncogenes or tumour suppressors10, 18, 19, and expression profiling analyses have revealed characteristic miRNA signatures in certain human cancers9, 20, 21. However, the precise parts played by the expressed miRNAs in specific steps of malignant progression, including metastasis, are still unknown. For these reasons, we undertook to associate specific miRNAs with specific stages of malignant progression, with the hope that such associations might provide insights into the causal mechanisms of cancer cell invasion and metastasis.
miR-10b is highly expressed in metastatic breast cancer cells
To identify miRNAs that regulate breast cancer metastasis, we selected candidate miRNAs on the basis of previously reported miRNA microarray profiling; these analyses had identified 29 miRNAs that are differentially expressed between primary breast carcinomas and normal mammary tissue, without regard to the eventual metastatic progression of these tumours22. We investigated the expression of these candidate miRNAs in a series of human mammary epithelial cells and tumour cell lines (Supplementary Table 1). Out of a total of eight selected miRNAs, three (miR-155, miR-9 and miR-10b) were found to be markedly upregulated in breast cancer cells when compared with either primary human mammary epithelial cells (HMECs) or with the spontaneously immortalized MCF-10A cells (Fig. 1a and Supplementary Table 1). Unlike miR-155 and miR-9, the expression of which was not specific to metastatic cells (Fig. 1a, b), miR-10b was highly expressed only in metastatic cancer cells. For example, the expression level of miR-10b was 50-fold higher in cells of the MDA-MB-231 line, which are capable of metastasizing, than in cells of the MCF-7 human breast cancer line, which have little if any metastatic powers (Fig. 1c). This correlation indicated that miR-10b might well have a causal role in breast cancer metastasis.
Figure 1: miR-10b is highly expressed in metastatic breast cancer cells and positively regulates cell migration and invasion.
a, RT–PCR of miR-155, miR-9 and miR-10b in a series of human mammary epithelial cells. b, c, Real-time RT–PCR of miR-9 (b) and miR-10b (c) in four different human mammary epithelial cells. d, e, Transwell migration assay (d) and Matrigel invasion assay (e) of MDA-MB-231 cells transfected with the inhibitor for miR-9 (anti-miR-9) or miR-10b (anti-miR-10b) (quantified below). Magnification in e, 
200. f, RT–PCR of miR-10b in HMECs and SUM149 cells infected with the miR-10b-expressing or empty vector. g, h, Transwell migration assay and Matrigel invasion assay of HMECs (g) and SUM149 cells (h) infected with the miR-10b-expressing or empty vector. A representative experiment is shown in triplicate along with s.e.m. in b–e, g and h.
miR-10b positively regulates cell migration and invasion in vitro
We first performed in vitro loss-of-function analyses by silencing the miRNAs with antisense oligonucleotides23. We assessed the level of miRNA silencing by a reporter assay, in which the predicted miRNA binding site was cloned into the 3' untranslated region (UTR) of a luciferase reporter gene24. We found that transfection of the antisense inhibitor for miR-9 or miR-10b in MDA-MB-231 cells caused a two- to threefold increase in the luciferase activity (Supplementary Fig. 1a, b), suggesting that each of the transfected antisense RNAs achieved a greater than 50% inhibition of the actions of its cognate miRNA. Although neither of these two antisense RNAs affected the motility of MDA-MB-231 cells (Fig. 1d), silencing of miR-10b led to a more than tenfold reduction in the invasive properties of these cells, as gauged by an in vitro invasion assay in which miR-9 inhibition had only a marginal effect (Fig. 1e). This reduction was not due to impairment of cell viability (Supplementary Fig. 1c). Taken together, these observations suggested that miR-10b function is required for in vitro invasiveness but not for viability or motility of these metastatic cells.
To determine whether miR-10b overexpression would increase the basal levels of cell migration or invasion, we cloned the genomic sequence of the human mir-10b gene into a green fluorescent protein (GFP)-expressing, murine stem-cell retrovirus (MSCV)-derived vector13. We then used the resulting vector to express miR-10b in immortalized HMECs25, and in the SUM149 cell line, a line of non-metastatic human breast cancer cells26, 27. The miR-10b expression level was gauged by PCR with reverse transcription (RT–PCR; Fig. 1f).
In both cell lines, ectopic expression of miR-10b had no effect on their proliferation in vitro (Supplementary Fig. 1d, e), but did result in a four- to sixfold increase in cell motility and invasiveness (Fig. 1g, h). These results indicated that overexpression of miR-10b is sufficient to promote both migration and invasion in vitro.
miR-10b initiates tumour invasion in vivo
A key question was whether miR-10b could induce metastatic behaviours in vivo. To address this possibility, we overexpressed miR-10b in otherwise non-metastatic human breast cancer cells. To begin, we implanted miR-10b-transduced or mock-infected SUM149 cells into the mammary fat pads of NOD-SCID mice. The host mice displayed visible mammary tumours within two weeks after injection and became moribund at week 11 owing to primary tumour burden, at which point this experiment was terminated. Importantly, GFP expression was maintained in the tumour cells throughout this time period (Fig. 2a).
Figure 2: miR-10b induces tumour invasion.
a, GFP imaging of the SUM149 primary tumours at week 11 after orthotopic injection of NOD-SCID mice. Magnification, 8. b, Growth curves of primary mammary tumours formed by SUM149 cells infected with the miR-10b-expressing or empty vector. Each data point represents the mean 
s.e.m. of 3–4 mice. c, Haematoxylin and eosin (H&E)-stained sections of primary mammary tumours formed by SUM149 cells infected with the miR-10b-expressing or empty vector, at week 6 after orthotopic transplantation. Arrows in panels C and D indicate areas of stromal invasion. Panels E and F demonstrate muscular invasion and vascular invasion, respectively. Magnification: panels A and C, 100; panels B and D–F, 400. d, Ki-67- (panels A, B, E, F) and MECA-32- (panels C, D, G, H) stained sections of primary mammary tumours formed by SUM149 cells infected with the miR-10b-expressing or empty vector, at week 6 after orthotopic transplantation. Arrows in panel E indicate areas of invasion. N, necrosis. Magnification: panels A, C, E and G, 40; panels B, D, F and H, 200. e, f, Quantification of Ki-67 staining (percentage of Ki-67+ carcinoma cells among total carcinoma cells; e) and vessels (using MECA-32-stained sections; f) at the centre and the edge of the SUM149 tumours. n = 3 mice at 6 weeks after implantation. Error bars in e and f indicate s.e.m. g, Prominent intratumoural vessels are associated with the invasion front of miR-10b-overexpressing tumours, as demonstrated by MECA-32 staining of primary mammary tumours formed by SUM149 cells infected with the miR-10b-expressing or empty vector, at week 6 after orthotopic transplantation. Magnification, 400.
At 6 weeks post implantation, the control tumours and the miR-10b-overexpressing SUM149 tumours were of comparable size (Fig. 2b), indicating no effects of this miRNA on primary tumour formation. As anticipated, the control SUM149 tumours were strictly non-invasive, as shown by their confinement within fibrotic capsules (Fig. 2c, panels A, B). In stark contrast, the miR-10b-overexpressing SUM149 tumours displayed a massive desmoplastic reaction, with islands of epithelial cancer cells that had invaded the stroma (Fig. 2c, panels C, D). In addition, we observed apparent muscular and vascular invasion by the miR-10b-overexpressing tumour cells (Fig. 2c, panels E, F). Hence, ectopic expression of miR-10b could confer invasiveness on SUM149 cancer cells that were previously non-invasive in vivo.
To determine whether miR-10b expression in the primary tumours would also affect cell proliferation and tumour angiogenesis, we performed immunohistochemistry with the Ki-67 proliferation marker and the MECA-32 endothelial cell marker. We found that the distribution, but not the total number, of Ki-67+ cells in the miR-10b-overexpressing SUM149 tumours was distinct from that seen in the control tumours: in the miR-10b-expressing tumours, large necrotic centres were apparent that lacked proliferating cells, whereas the invasion fronts were enriched in Ki-67+ cells and were highly vascularized (Fig. 2d, panel E–H, and Fig. 2e, f); this contrasted with the appearance of the control tumours, which exhibited an essentially even distribution of Ki-67+ cells (interspersed with Ki-67- cells) and poor vascularization (Fig. 2d, panel A–D, and Fig. 2e, f). Furthermore, in contrast to the control tumours, the vessels associated with the invasion fronts of the miR-10b-overexpressing tumours were seen not only in the stroma (peritumoural), but also within the epithelial tumour masses (intratumoural, Fig. 2g). Hence, the invasion fronts of miR-10b-overexpressing tumours exhibited very high levels of both cell proliferation and angiogenesis.
At later times, those mice injected with miR-10b-overexpressing cells carried larger tumour burdens (P = 0.03, Fig. 2b). We concluded that the enhanced tumour growth of these miR-10b-overexpressing cells was likely to be a consequence rather than a cause of their invasiveness, because (1) the substantial stromal invasion by the miR-10b-expressing cells preceded by several weeks the observed differences in overall tumour growth rates, and (2) the miR-10b-transduced cells and their control counterparts proliferated at similar rates in vitro (Supplementary Fig. 1d, e) and in vivo (as gauged by overall levels of Ki-67, Fig. 2d, e). Thus, we speculate that vascularization might be a rate-limiting determinant of overall tumour growth, and that increased invasiveness might afford the miR-10b-overexpressing cells better access to the vasculature.
miR-10b initiates distant metastasis
We asked whether expression of miR-10b would also result in distant metastasis. As early as week 6, haematoxylin and eosin staining revealed the presence of miR-10b-overexpressing SUM149 cells in the primary tumour-associated vessels (Fig. 2c, panel F). Immunohistochemical analyses further revealed significant numbers of hyperproliferative Ki-67+ tumour cells in the lumina of some of the vessels associated with the miR-10b-overexpressing tumours (Fig. 3a). In contrast, no significant evidence of intravasating or intravasated cancer cells was found in the control tumours.
Figure 3: miR-10b induces distant metastasis.
a, MECA-32- and Ki-67-stained sections of a primary mammary tumour formed by miR-10b-transduced SUM149 cells, at week 6 after orthotopic transplantation. Red arrows in panel A indicate tumour cells within a vessel, and black arrows in panel D indicate endothelial cells. Magnification: panels A and C, 100; panels B and D: 400. b, H&E- and AE1/AE3-stained sections of lungs isolated from mice that received orthotopic injection of miR-10b-transduced or mock-infected SUM149 cells, at week 9 after transplantation. Circles indicate clusters of metastatic cells. The arrow indicates normal bronchial epithelium. Inset, AE1/AE3 staining of a SUM149 primary tumour. Magnification, 600. c, Numbers of lung micrometastases (micromets) per section in individual mice that received orthotopic injection of miR-10b-transduced or mock-infected SUM149 cells, at week 9 (left panel) and week 11 (right panel) after transplantation, respectively. d, Bright field, GFP imaging, and H&E staining of lungs isolated from mice that received orthotopic injection of miR-10b-transduced or mock-infected SUM159 cells, at week 11 after transplantation. Magnification, 8 for bright field and GFP imaging; 600 for H&E staining. e, GFP imaging and H&E staining of a macroscopic peritoneal metastasis in a mouse that received orthotopic injection of miR-10b-transduced SUM159 cells, at week 11 after transplantation. Magnification, 8 for GFP imaging; 400 for H&E staining. f, Incidence of lung metastasis and macroscopic peritoneal metastasis in mice that received orthotopic injection of miR-10b-transduced or mock-infected SUM159 cells.
We also examined other tissues for the presence of disseminated tumour cells. At week 6 after implantation, we observed occasional single GFP+ cells in a few regions of the lung cryosections, indicating relatively early dissemination of the miR-10b-overexpressing cells from primary tumours; the lungs of mice that carried control tumours lacked such GFP+ cells (Supplementary Fig. 2). At 9 weeks after implantation, the lungs from mice implanted with miR-10b-overexpressing SUM149 cells exhibited clusters of dense hyperchromatic cells that were positive for cytokeratins, as demonstrated by staining with AE1/AE3 (a cocktail of two distinct anti-cytokeratin monoclonal antibodies, Fig. 3b). On average, we found 
1 micrometastasis per 5-
m section (Fig. 3c, left panel). At week 11, there was a further increase in the number of such micrometastatic clusters (4 micrometastases per section, Fig. 3c, right panel). In the lungs of hosts bearing control SUM149 tumours, however, the only AE1/AE3-positive cells were normal bronchial epithelial cells (Fig. 3b).
The ability of ectopically expressed miR-10b to elicit metastasis was also examined in the SUM159 cell line—a second line of human breast cancer cells. These cells are invasive but non-metastatic27, allowing us to determine whether these cells too would respond to miR-10b by acquiring metastatic potential. In contrast to the SUM149 cells described above, control SUM159 cells exhibited a high level of motility and invasiveness in vitro (data not shown). Moreover, the xenograft mammary tumours formed by the control SUM159 cells displayed local invasion (data not shown). Importantly, no metastases were found in 10 mice that had been transplanted with mock-infected SUM159 cells (Fig. 3d, f). Strikingly, however, 8 out of 10 mice that had received orthotopic transplantation of miR-10b-overexpressing SUM159 cells exhibited numerous lung metastases, which were readily detectable both by GFP fluorescence and by histological analysis (Fig. 3d, f). Furthermore, 30% of these mice (3/10) developed macroscopic peritoneal metastases (0.5–2 cm in diameter, Fig. 3e, f). Taken together, these observations indicate that ectopic miR-10b expression can drive tumour invasion and metastasis in otherwise non-metastatic breast tumours, thereby acting as a potent pro-metastatic agent.
miR-10b is directly regulated by the transcription factor Twist
We undertook to determine how miR-10b expression is regulated. The observation that miR-10b is highly expressed in metastatic breast cancer cells indicated that the gene encoding this miRNA might be the target of certain transcription factors that are activated specifically in metastatic cells. Of note, recent studies have demonstrated that several transcription factors previously known as master regulators of embryogenesis, are highly expressed in metastatic cells and seem to have causal roles in tumour metastasis, ostensibly by inducing epithelial–mesenchymal transitions (EMTs) in cancer cells2, 3, 4, 5, 6, 7, 8; this process is thought to contribute to the invasiveness and dissemination of epithelial tumour cells28. These various observations led us to ask whether any of these EMT-inducing factors might function to activate miR-10b expression.
We first assessed miR-10b expression in the four lines of mouse mammary tumour cells that had been used in the identification of Twist as a metastasis-promoting gene6. We discovered that the level of miR-10b correlated with the known metastatic potentials of these cell types, with the lowest expression level in 67NR cells, which are unable to intravasate from the primary tumour, whereas the highest expression level was seen in 4T1 cells, which are capable of generating macroscopic metastases (Fig. 4a, b). This closely paralleled the expression pattern of Twist in these cell lines6, indicating that miR-10b expression might well be upregulated by this transcription factor. We therefore expressed either Twist1 (also known as Twist) or, as a control, a second EMT-inducing transcription factor—snail (SNAI1)—in the non-tumourigenic, immortalized HMECs, which had been found to express a low level of miR-10b (Fig. 1f). In contrast to the behaviour of SNAI1, which reduced miR-10b expression by 57%, ectopic expression of Twist1 led to a 4.5-fold increase in the level of this miRNA in these HMECs (Fig. 4c, d).
Figure 4: miR-10b is regulated by Twist.
a, b, RT–PCR (a) and real-time RT–PCR (b) of miR-10b in 67NR, 168FARN, 4TO7 and 4T1 cells. c, d, RT–PCR (c) and real-time RT–PCR (d) of miR-10b in HMECs transduced by SNAI1, Twist1 or the empty vector. e, Upper panel, human mir-10b genomic locus. The two short blue lines represent two PCR amplicons. Lower panel, ChIP assay in HEK293T cells transfected with a vector expressing Myc–TWIST1 or the Myc tag alone. PCR was performed with primers specific for human mir-10b E-box 1 and E-box 2, respectively. f, Transwell migration assay and Matrigel invasion assay of Twist1-transduced or mock-infected HMECs that were transfected with the inhibitor for miR-10b or the control oligonucleotide. A representative experiment is shown in triplicate along with s.e.m. in b, d and f.
High resolution image and legend (127K)We next performed chromatin immunoprecipitation (ChIP) assays to determine whether TWIST1 controls miR-10b expression by binding directly to the mir-10b gene. This basic helix–loop–helix transcription factor has been shown to bind to E-box sequences (CANNTG) present in the genes that it regulates29, 30. We examined the 4-kb genomic sequence upstream of the human mir-10b stem-loop and identified two conserved E-boxes, at -313 bp (E-box 1) and -2,422 bp (E-box 2), respectively (Fig. 4e). In addition, the putative promoter of human mir-10b spans between -111 bp and -460 bp31, which encompasses E-box 1, the most proximal E-box. We designed two PCR amplicons to assay for the presence of these two putative binding sites in chromatin immunoprecipitates. The experiments revealed that TWIST1 bound to E-box 1, but not to E-box 2 (Fig. 4e). Thus, TWIST1 specifically binds to the putative promoter of mir-10b, providing strong evidence that miR-10b can be directly regulated by this transcription factor.
Although miR-10b does not seem to be essential for cell motility in highly metastatic cancer cells (Fig. 1d), we asked nonetheless whether this miRNA is required for Twist1-induced migration and invasion in otherwise poorly motile cells. To this end, we introduced the antisense oligonucleotide for miR-10b into Twist1-overexpressing HMECs. In accord with a previous report6, overexpression of Twist1 led to a strong increase in the motility and invasiveness of these cells (Fig. 4f). Strikingly, miR-10b inhibition consistently led to a fivefold reduction in the motility and invasiveness of Twist1-overexpressing HMECs (Fig. 4f). Whereas Twist1 is capable, on its own, of inducing an EMT6, miR-10b is not (data not shown). Instead, it seems to be essential to one element of the multi-component, Twist1-induced EMT program—increased cell motility and invasiveness.
HOXD10 is a direct and functional target of miR-10b
To understand the mechanisms by which miR-10b induces tumour invasion and metastasis, we used several computational methods to help identify miR-10b targets in humans. Among the approximately 100 targets predicted by both the TargetScan32 and PicTar33 search programs, two genes—homeobox D10 (HOXD10) and RB1CC1 (also named FIP200)—were previously implicated in suppression of cell migration and/or invasion. HOXD10 was of particular interest, because its expression has been found to be progressively lost in breast tumours showing increasing degrees of malignancy34, 35. More importantly, restored expression of HOXD10 in MDA-MB-231 cells has been found to impair migration and invasion in vitro as well as tumour progression in vivo35. The HOXD10-encoded mRNA contains a 3' UTR element that is partially complementary to miR-10b and carries the identical sequence in the human, mouse and rat mRNA orthologues (Fig. 5a).
Figure 5: miR-10b suppresses HOXD10, leading to induction of RHOC.
a, Upper panel, predicted duplex formation between human HOXD10 3' UTR and miR-10b. Lower panel, sequence of the miR-10b binding site within the HOXD10 3' UTR of human (H.s.), and within the HoxD10 3' UTR of mouse (M.m.), and rat (R.n.). b, Real-time RT–PCR of HOXD10 in HMECs and SUM149 cells infected with the miR-10b-expressing or empty vector. Data were normalized to the level of GAPDH mRNA. c, Luciferase activity of wild-type (UTR-WT) or mutant (UTR-mut.) HOXD10 3' UTR reporter gene in SUM149 cells infected with the miR-10b-expressing or empty vector. d, Immunoblotting of HOXD10 and RHOC in HMECs and SUM149 cells infected with the miR-10b-expressing or empty vector. e, f, Upper panel, transwell migration assay and Matrigel invasion assay of miR-10b-transduced or mock-infected SUM149 cells with transient transfection of HOXD10 (e) or RHOC siRNA (f). Lower panel, immunoblotting of HOXD10 (e) or RHOC (f). A representative experiment is shown in triplicate along with s.e.m. in b, c, e and f.
High resolution image and legend (162K)Although miR-10b overexpression did not cause degradation of HOXD10 mRNA (Fig. 5b), it did, however, reduce the activity of a luciferase reporter gene fused to the wild-type HOXD10 3' UTR (48% reduction, P = 0.003, Fig. 5c), indicating that miR-10b targets HOXD10 through translational inhibition. The action of miR-10b on HOXD10 depends on the presence of a single miR-10b cognate binding site within the 3' UTR, because the activity of a luciferase reporter that carries a mutant HOXD10 3' UTR—with substitution of four nucleotides within the miR-10b binding site (Fig. 5a)—was not reduced by expression of miR-10b (Fig. 5c). In support of these results, we observed a clear reduction in the level of the endogenous HOXD10 protein in miR-10b-overexpressing cells (tenfold and threefold reduction in HMECs and SUM149 cells, respectively, Fig. 5d).
Others have demonstrated that HOXD10 represses expression of genes that are involved in cell migration and extracellular matrix remodelling, including RHOC, 
3 integrin, matrix metalloproteinase-14, and urokinase-type plasminogen activator receptor36. Among these, RHOC has been identified as an especially important player in metastasis37, 38, and its expression correlates with metastatic spread of various types of carcinomas39, 40, 41. Indeed, we found that miR-10b-transduced cells exhibited robust expression of RHOC protein, whereas RHOC expression in the control cells was barely detectable (Fig. 5d).
We next ascertained whether reduction of HOXD10 levels might provide an explanation for the induction of cell motility and invasiveness observed following miR-10b overexpression. We overexpressed miR-10b in SUM149 cells together with a construct expressing HOXD10 constitutively; this construct encodes the entire HOXD10 coding sequence but lacks the 3' UTR of HOXD10-encoding mRNA, yielding an mRNA that is resistant to miR-10b-mediated inhibition of translation. Strikingly, the resulting constitutive expression of HOXD10 completely abrogated miR-10b-induced cell motility and invasiveness (Fig. 5e), without affecting the proliferation or viability of these cells (data not shown), suggesting that this HOX gene is indeed a functionally important target of miR-10b. Furthermore, transfection of RHOC siRNA (small-interfering RNA), which caused a > 90% reduction in the level of the RHOC protein (Fig. 5f), led to a strong but not complete suppression of miR-10b-induced cell migration (by 81%) and invasion (by 87%, Fig. 5f). Hence, RHOC seems to be a key downstream effector of miR-10b.
miR-10b expression is elevated in metastatic breast tumours
Paradoxically, a recent microarray study reported that miR-10b is among the miRNAs found to be downregulated in primary breast tumours (independent of their clinical aggressiveness) compared with normal breast tissue22. To address this apparent paradox and determine whether miR-10b expression correlates with clinical outcome in patients, we measured its levels in primary tumour samples from 23 breast cancer patients. When compared with normal breast tissue, miR-10b expression level was lower in all of the breast carcinomas from metastasis-free patients (5/5). In contrast, 50% of the metastasis-positive patients (9/18) had elevated miR-10b levels in their primary tumours (P = 0.03, Fig. 6a).
Figure 6: miR-10b expression level is associated with the metastasis outcome in breast cancer patients.
a, Real-time RT–PCR of miR-10b in 23 primary breast tumours from patients with indicated status of metastasis. N, normal breast tissue. Error bars indicate s.e.m. of triplicate experiments. The numbers above the columns indicate fold increase relative to normal tissue. b, Model for the regulation and function of miR-10b in cancer metastasis. RISC, RNA-induced silencing complex.
High resolution image and legend (98K)These results are in consonance with the expression pattern of miR-10b in cultured human mammary cells (Fig. 1c). In addition, we observed that ectopic miR-10b expression in non-tumourigenic, immortalized EpH4 mouse mammary epithelial cells42 did not transform them to a tumourigenic state, whereas overexpression of Ras indeed enabled their tumourigenic growth (data not shown). Collectively, our findings indicate that miR-10b plays a part specifically in the metastatic process but not in primary tumour formation.
Discussion
The present work has led to the identification of a Twist-induced miRNA (miR-10b) that inhibits synthesis of the HOXD10 protein, permitting the expression of the pro-metastatic gene product, RHOC; this favours, in turn, cancer cell migration and invasion (Fig. 6b). Importantly, overexpression of this miRNA in otherwise non-metastatic breast cancer cells enables them to acquire invasive and metastatic behaviour. Conversely, silencing of miR-10b inhibits Twist-mediated cell migration and invasion. It remains to be seen whether miR-10b inhibition in highly malignant cells will reverse the metastatic phenotype in vivo. Such analysis is hampered at present by the lack of an experimental strategy for stably silencing miRNAs over extended periods of time.
Of particular interest, the mir-10b gene is located within the HOXD gene cluster (Fig. 6b). In fact, a number of HOX genes have been predicted to be targets of miRNA regulation, both in vertebrates32 and in flies43, and some of these associations have been validated experimentally: (1) miR-196, which has near perfect complementarity with the HOXB8 mRNA 3'UTR region, downregulates HOXB8 expression through cleavage of this mRNA44; and (2) miR-iab-4-5p downregulates ultrabithorax (UBX, a Drosophila HOX protein), leading to a homeotic phenotype in flies45. Accordingly, miR-10b represents a third 'HOX miRNA' that regulates expression of a HOX gene within the same HOX gene cluster that carries its encoding gene. This indicates that miR-10b is also likely to be involved in regulating certain steps of embryogenesis in which HOXD10 is known to participate. In addition, a close relative of miR-10b, miR-10a (the gene for which is located within the HOXB gene cluster), has been recently reported to target HOXA1 (ref. 46), a gene that plays an oncogenic role in human mammary carcinoma cells47, indicating that miR-10a might have an opposite rather than similar function in breast cancer.
In silico analyses predict that yet other mRNAs are direct targets of miR-10b-mediated silencing. Because the invasion–metastasis cascade involves a number of distinct steps1, it is therefore possible that miR-10b regulates additional targets that are involved in several of these steps. A future challenge will be to identify the entire complement of miRNAs and their mRNA targets to elucidate more fully the contributions of these miRNAs to high-grade malignancy.
Methods Summary
Quantification of miRNAs was performed by real-time RT–PCR. Activity of miRNAs was measured by a luciferase reporter assay. Silencing of miRNAs was done by transfection of a chemically modified antisense oligonucleotide that was complementary to the endogenous miRNA being targeted. Ectopic expression of miR-10b was achieved by retroviral infection. Cell motility and invasiveness were gauged by transwell migration assay and Matrigel invasion assay, respectively. Chromatin immunoprecipitation was performed with a ChIP assay kit, followed by PCR with primers specific for mir-10b. Metastatic activity was assayed by mammary fat pad injection of cancer cells into NOD-SCID female mice. GFP imaging, histological analysis and cytokeratin immunohistochemistry were used to detect distant metastasis. Human materials were obtained from the Memorial Sloan-Kettering Cancer Center along with pathology reports and radiological reviews.
Full methods accompany this paper.