The Kindlin-2 regulation of epithelial-to-mesenchymal transition in breast cancer metastasis is mediated through miR-200b

Metastasis is the main cause of death in cancer patients, including breast cancer (BC). Despite recent progress in understanding the biological and molecular determinants of BC metastasis, effective therapeutic treatments are yet to be developed. Among the multitude of molecular mechanisms that regulate cancer metastasis, the epithelial-to-mesenchymal transition (EMT) program plays a key role in the activation of the biological steps leading to the metastatic phenotype. Kindlin-2 has been associated with the pathogenesis of several types of cancers, including BC. The role of Kindlin-2 in the regulation of BC metastasis, and to a lesser extent in EMT is not well understood. In this study, we show that Kindlin-2 is closely associated with the development of the metastatic phenotype in BC. We report that knockout of Kindlin-2 in either human or mouse BC cells, significantly inhibits metastasis in both human and mouse models of BC metastasis. We also report that the Kindlin-2-mediated inhibition of metastasis is the result of inhibition of expression of key molecular markers of the EMT program. Mechanistically, we show that miR-200b, a master regulator of EMT, directly targets and inhibits the expression of Kindlin-2, leading to the subsequent inhibition of EMT and metastasis. Together, our data support the targeting of Kindlin-2 as a therapeutic strategy against BC metastasis.


Results
Kindlin-2 is required for cancer cell adhesion, spreading and focal adhesion formation. Several cell functions are required for cancer cell invasion and metastasis, including adhesion and spreading. Kindlin-2, like the other members of the Kindlin family was shown to regulate cell spreading and adhesion, downstream of integrins in different cell types. This function has, however, not been established in breast cancer cells. Our recently published study has shown that Kindlin-2 is overexpressed in aggressive human and mouse breast cancer cells and tumours, and that Kindlin-2 is required for primary tumour growth in both human and murine BC models 13 . The main goal of this study was to investigate whether the function of Kindlin-2 in BC tumour progression and metastasis is linked to the activation of cell spreading and adhesion. We used CRISPR/Cas9 gene editing to generate several clones lacking Kindlin-2 expression in MDA-MB-231 cells ( Fig. 1A and 13 ). Loss of Kindlin-2 had no evident effect on cell proliferation 13 . When Kindlin-2-deficient (Kindlin-2 CRISPR-1 or-2) or the control (Scram CRISPR) MDA-MB-231 cells were seeded on fibronectin-coated slides, the number of Kindlin-2-deficient cells that adhered was more that 6-fold lower (p < 0.01) than that of the control cells (Fig. 1B,C). Cell spreading was also significantly affected by loss of Kindlin-2; the average surface of the Kindlin-2 CRISPR MDA-MB-231 cells was more than 5-fold smaller (p < 0.01) than that of the control (Scram CRISPR) counterparts (Fig. 1D,E). Since focal adhesions are known to be required for cell adhesion and spreading, we used immuno-fluorescence staining of vinculin as a way to visualize focal adhesions in control or Kindlin-2-deficient MDA-MB-231 cells that were allowed to adhere and spread on fibronectin. Cells lacking kindlin-2 (Fig. 1F, lower panels) were unable to spread when compared to the control cells (Fig. 1F, upper panels), which confirm the results shown in Fig. 1D. We also found that in the control (Scram CRISPR) cells, Kindlin-2 to stain in a dot-like pattern (left panels), which nicely co-localized with the Vinculin structures (middle) in the merged image (Fig. 1F, right panels). These Vinculin structures that represent focal adhesions were abundant and evenly distributed throughout the surface of fully spread control cells, but were hardly detected in Kindlin-2-deficient cells (Fig. 1F, lower panels). Together, these data show that Kindlin-2 is required for cell spreading and adhesion of BC cells.

Loss of Kindlin-2 inhibits invadopodia formation and degradation of extracellular matrix. Our
recently published study showed that Kindlin-2 is required for BC motility, migration and invasion, without affecting cell proliferation 13 . The biological significance of loss of Kindlin-2 was further investigated by assessing the ability of BC cells to form invadopodia and degrade the ECM. When control (Scram CRISPR) MDA-MB-231 cells were seeded on gelatine-coated coverslips, F-actin staining, revealed invadopodia as dot-like structures (Arrow heads) or rosettes (Arrows) on the surface of the cells ( Fig. 2A, upper panel). The Kindlin-2-deficient (Kindlin-2 CRISPR-1) cells, on the other hand, formed fewer invadopodia structures ( Fig. 2A, lower panel). In fact, loss of Kindlin-2 resulted in more than 5-fold reduction (p < 0.01) in the number of invadopodia (Fig. 2B). In cancer cells and other cell types, invadopodia coincide with sites of ECM degradation. Accordingly, we utilized this gelatine-based invadopodia assay to assess the effect of Kindlin-2 on the degradation of the ECM by MDA-MB-231 cells. Control or Kindlin-2-deficient MDA-MB-231 were seeded onto fluorescent (FITC) gelatine slides. Similar to data presented in Fig. 2A, staining for F-Actin showed that the control (Scram CRISPR) cells formed significantly higher (p < 0.01) number of invadopodia than the Kindlin-2-deficient (Kindlin-2 CRISPR-1) cells (Fig. 2C, left panel and Fig. 2D). These invadopodia structures overlapped with the sites of gelatine degradation, shown as dark areas where gelatine was depleted (Fig. 2C, middle panel). Loss of Kindlin-2 also resulted in a significant inhibition (p < 0.01) of ECM degradation (Fig. 2D). Expression and activity of metalloproteases (MMPs) are known to be required for EMC degradation at the invadopodia sites. We found loss of Kindlin-2 to also significantly inhibit the expression of MMP2 and MMP9 (Fig. S1 in Supplementary information). These results demonstrate the role of Kindlin-2 in the regulation of invadopodia formation and the invadopodia-mediated degradation of the ECM that is required for cancer cell motility and invasion.
Kindlin-2 is required of BC metastasis in both the spontaneous and the experimental metastasis assays. Cancer metastasis is a complex and multistep process, requiring cancer cells to escape the confinement of the primary tumour, survive in the blood/lymph system and then to establish a new niche at a distant site. We previously showed that Kindlin-2 deficiency inhibited the growth of primary BC tumours using either human MDA-MB-231 or murine 4T1 BC cells lacking Kindlin-2 injected in mammary fat pads of mice 13 . The effect of Kindlin-2 on metastasis was, however, not reported. Therefore, we sought to investigate how the loss of Kindlin-2 influences BC metastasis. After the primary tumours were excised from the mammary sites, the mice were kept alive and monitored for an additional 3 weeks, after which mice were sacrificed, the lungs removed and metastasis nodules counted. In the MDA-MB-231 BC model, we found the number of lung nodules, which represent the metastasis foci, was reduced by more than 10-fold (p < 0.01) in mice implanted with the Kindlin-2-deficient (Kindlin-2 CRISPR) cells compared to their control (Scram CRISPR) counterparts (Fig. 3A,B). These results were replicated in murine 4T1 BC model, where CRISPR/Cas9-mediated knockout of Kindlin-2 ( Fig. 3C), also resulted in ~10-fold decrease (p < 0.01) in the number of lung metastasis foci in mice implanted with the Kindin-2-deficient 4T1 cells, compared to the control cells (Fig. 3D,E).
We also used the lung colonization assay, in which cancer cells were injected directly in the blood via the tail vein, to determine the effect of loss of Kindlin-2 on metastasis. Here again we found the lung metastasis nodules were reduced by ~10-fold (p < 0.01) when Kindlin-2-deficient MDA-MB-231 ( Fig. 3F-H) or Kindlin-2-deficient 4T1 (Figs 3I and 4L) were injected into mice compared to their control counterparts. These data show that Kindlin-2 plays a critical role in the invasion and metastasis of BC tumours as demonstrated here with two different BC cells and in both the spontaneous and experimental metastasis assays.
Loss of Kindlin-2 inhibits expression of key markers of the epithelial-to-mesenchymal transition program in BC tumours and metastases. EMT confers migratory and invasive capabilities, as well as metastatic capacities to cancer cells 3,4,7,8 . We therefore sought to determine whether the inhibition of BC metastasis that we found to be caused by loss of Kindlin-2 is a consequence of Kindlin-2 mediated regulation of EMT in the primary tumours. Among the several hallmarks of EMT are changes in expression levels of Fibronectin, Vimentin and N-Cadherin 15,16 . Immunostaining of mammary fat pad-derived tumour sections revealed a significant loss (p < 0.01) of all three EMT markers (Fig. 4A) in tumours derived from the Kindlin-2-deficient MDA-MB-231 or 4T1 (Kindlin-2-CRISPR) BC cells compared to their control counterparts (Scram CRISPR). We  extended our immunostaining analyses to the lungs of mice implanted in the mammary fat pad with the Kindlin-2-deficient BC cells or their control derivatives. Here also, we found a significant (p < 0.01) loss of Fibronectin and Vimentin in the lungs of mice implanted with the Kindlin-2-deficient MDA-MB-231 or 4T1 BC cells compared to their counterpart controls (Fig. 4B). Similar results were found when we stained sections of lungs from mice that were injected via tail vein with the Kindlin-2-deficient MDA-MB-231 or 4T1 BC cells compared to their controls (Fig. 4C).
Kindlin-2 is a direct target of microRNA miR-200b. The role of microRNAs in the regulation of the EMT program is well documented in the literature [17][18][19] . To further understand the underlying molecular and genetic or epigenetic mechanisms involved in the Kindlin-2-mediated promotion of BC metastasis through the activation of EMT, we investigated the role of microRNAs in the regulation of Kindlin-2 expression. In silico analyses, using several microRNA prediction algorithms (Target Scan, miRanda and PITA), identified microRNA miR-200b as one of the top predicted microRNAs that targets Kindlin-2 (Fig. 5A). In fact we showed that the seed sequence of miR-200b is a perfect match to a target sequence in the 3′UTR of Kindlin-2 (Fig. 5A). Further reinforcing our focus on miR-200b as a potential regulator of Kindlin-2 in BC are published findings that this microRNA is a master regulator of EMT in several cancer types 14,[19][20][21] . First, we quantified the expression levels of Kindlin-2 and miR-200b in representative BC cell lines, and found Kindlin-2 expression levels to be low in the non-aggressive MCF7 and SKBR3 cells while it was comparatively high in the more aggressive and metastatic BT549 and MDA-MB-231 cells (Fig. 5B). This observation confirmed our previously published findings 13 . Of note, expression levels of miR-200b in these cell lines showed an inverse pattern: High in MCF7 and SKBR3, and low in BT549 and MDA-MB-231 cells (Fig. 5C). In fact, miR-200b was almost undetectable in these aggressive BC cell lines, which are of mesenchymal phenotype 22,23 , while it was abundant in the MCF7 and SKBR3 which are of epithelial phenotype (Fig. 5C) 22,23 . Also in the murine BC series in which Kindlin-2 expression levels were correlated positively with aggressiveness of these cells (low in the non-tumorigenic 67NR cells and the non-invasive 4T07 cells, and high in the aggressive and highly metastatic 4T1 cells 24 . miR-200b showed a negative correlation: its expression levels were more that 5-fold lower in the 4T07 cells compared to 67NR and was almost undetectable in the 4T1 cells. (Fig. 5D). Thus, we demonstrate an inverse correlation between Kindlin-2 and miR-200b in the aggressive and mesenchymal BC cells vs their non-aggressive and epithelial counterparts, which strongly suggest the involvement of Kindlin-2 in the miR-200b-mediated regulation of the EMT and that miR-200b is involved in the regulation of Kindlin-2 during this process.
Next, using both western blot (Fig. 5E) and qRT-PCR ( Fig. 5F) analyses, we showed that transient transfection of MDA-MB-231 cells with miR-200b resulted in ~60% knockdown (p < 0.05) of Kindlin-2 protein levels ( Fig. 5E), compared to the parental cells or those transfected with non-targeting microRNA (NT-miR). On the other hand, anti-miR-200b had no significant effect on Kindlin-2 protein levels (Fig. 5E). Conversely, when we transfected MCF7, a cell line that expresses high levels of miR-200b, with anti-miR-200b oligonucleotides (Fig. 5F), we observed a ~2-fold increase (p < 0.05) in Kindlin-2 transcript levels compared to the parental MCF7 cells or MCF7 cells transfected with the non-targeting microRNA (NT-miR). Thus, inhibition of Kindlin-2 expression is induced by miR-200b. To verify that miR-200b represses expression of Kindlin-2 by binding directly to Kindlin-2 transcripts, we used a luciferase gene reporter assay (pmirGlo 25 ). MDA-MB-231 cells, which express very low levels of miR-200b, were engineered to express a pmirGlo construct where the wild-type form of Kindlin-2 3′-UTR (Wild-type K2-3′UTR) was subcloned upstream of the luciferase gene (231-wild-type K2-3′UTR). These cells remained either untreated (Ctrl) or transfected with either a non-targeting micro-RNA (NT-miR) or the miR-200b mimics (miR200). As expected, no difference in luciferase activity was found between the untreated 231-wild-type K2-3′UTR cells (Ctrl) and those same cells transfected with the non-targeting microRNA (NT-miR; Fig. 5G). In contrast, transfection of the 231-wild-type K2-3′UTR cells with miR-200b mimics (miR200) resulted in ~60% reduction (p < 0.05) in luciferase activity (Fig. 5G). Mutation of the seed sequence of miR-200b in the 3′-UTR of K2 (K2 3′UTR with scrambled miR-200 seed sequence) abrogated the effect of the exogenous miR-200b (Fig. 5H); the levels of luciferase activity did not show any significant difference when compared to the control cells or those treated with the non-targeting microRNA (Fig. 5H). These findings were replicated in the BT549 cells, confirming the generality of this observation. Thus, miR-200b specifically binds to its target sequence within the 3′UTR of Kindlin-2 and inhibits its expression.

The miR-200b-mediated downregulation of Kindlin-2 inhibits BC metastasis.
Having demonstrated that miR-200b specifically targets and inhibits Kindlin-2 expression in vitro, we extended our investigation to in vivo setting in the spontaneous metastasis assay. MDA-MB-231 cells with stable expression of miR-200b mimics or a scrambled control were injected into the mammary fat pads NSG mice and the metastasis burden was assessed 3 weeks after resection of the primary tumours by quantifying the number of lung metastasis foci. Overexpression of miR-200b slowed the growth of primary tumours; at the 5-week time point when all the primary tumours were removed, the average volume of the tumours derived from the miR-200-overexpressing MDA-MB-231 cells (miR200b) was more than 3-fold less (p < 0.01) than those derived from the control (Scram) cells (Fig. 6A). More importantly the number of lung metastasis foci was reduced by ~4-fold (p < 0.01) in the mice  (Fig. 6B,C). We also used genomic DNA PCR to specifically amplify the human genomic DNA in the lungs of the mice injected with the modified human MDA-MB-1231 cells. We found the amounts of human DNA present in the lungs of mice injected with the miR-200b-overexpressin MDA-MB-231 (miR200b) to be more that 120-fold less than in the control (Scram) counterparts (Fig. 6D), confirming the inhibitory effect of miR-200b on metastasis. qt-RT-PCR analyses of total RNA from the primary tumours showed that Kindlin-2 mRNA levels were significantly lower in the tumours derived from the miR-200b-expressing MDA-MB-231 (miR200) cells (Fig. 6E) compared to those derived from the control (Scram) counterparts. The level of Kindlin-2 downregulation was comparable to that observed in the tumours derived from the Kindlin-2-deficient MDA-MB-231 (K2-KO) cells (Fig. 6I). Finally, we assessed the effect of overexpression of miR-200b on expression of EMT markers in the tumours using qt-RT-PCR. Expression levels of N-Cadherin, Fibronectin and Vimentin were significantly downregulated (p < 0.05) in the tumours derived from the miR-200b-overexressing MDA-MB-231 (miR200) cells compared to the tumours from the control (Scram) counterparts ( Fig. 6F-H, respectively). The levels of downregulation of these EMT markers were similar to those observed in tumours from the Kindlin-2-deficient (K2KO) MDA-MB-231 cells (Fig. 6J-L). Together, these data demonstrate that Kindlin-2 modulates the growth and metastasis of BC tumours through the regulation of the EMT program and this regulation takes place downstream of miR-200b microRNA.

Discussion
Aberrant activation of the EMT program endows tumours with migratory and invasive properties, as well as metastatic capabilities. Further, the EMT program also confers cancer cells with stem cell properties allowing the tumour cells to survive in circulation and establish distant colonies that often acquire resistance to therapeutic agents. EMT has been demonstrated as an essential early step during the invasion-metastasis cascade, whereby loss of E-Cadherin in cell junctions allows the dissolution of cell-cell contacts in primary tumours, thus enabling the release of invasive cells 4,7,19,26,27 . Concomitant with EMT, an increase in motility, migratory and invasive properties and in expression levels of EMT markers such as N-Cadherin, Fibronectin and Vimentin, take place which results in the increase in metastasis potential of cancer cells. This study demonstrates that Kindlin-2 supports BC Interest in Kindlin-2 was stimulated by the identification of its integrin co-activator function of integrins by binding to the cytoplasmic tail of beta-integrin, along with talin, to enhance affinity/avidity for ligands 28 . It is well-established that integrin are key players in the pathology of several cancers (reviewed in 11 ); and the Kindlin-2-mediated co-activation of integrins was shown to regulate the function of integrins in cancers 29,30 . In fact, Kindlin-2 was found to be highly expressed in tumours from different types of cancers, including BC (reviewed in 11 ). Our recently published study identified a new Kindlin-2/TGFβ/CSF-1 signalling axis that BC cells utilize to recruit and polarize host macrophage to drive tumour progression 13 . Whether the same signalling axis was sufficient to activate BC metastasis was not clear. The present study demonstrated that Kindlin-2 function is involved in the enhancement of BC metastasis by regulating the EMT program downstream of microRNA miR-200b. The miR-200 family is no stranger to EMT. In fact, the miR-200 microRNAs have been established as master regulators of the EMT program in both physiological and pathological conditions 19,21,27,31 . miR-200b was recently shown to suppress invasiveness and to modulate the cytoskeletal and adhesive machinery by targeting Kindlin-2 in oesophageal squamous cell carcinoma cells 32 . Our study revealed similar activity of miR-200b in BC by targeting Kindlin-2 in vitro; loss of adhesion and spreading, inhibition of focal adhesion formation, and inhibition of invadopodia and extracellular degradation. In addition, targeting and inhibition of Kindlin-2 by miR-200b in BC also resulted in the regulation of the EMT program leading to the inhibition of BC invasion and metastasis in vivo. Kindlin-2 was also reported to promote BC invasion through epigenetic silencing of members of the miR-200 gene family 33 . Therefore, our study together with the data published by Yu et al. 33 show that EMT and the resultant invasion metastasis cascade in BC seems to be regulated trough a miR-200b/Kindlin-2 feedback loop.
The miR-200b/Kindlin-2/EMT signalling axis may not be the only mechanisms through which Kindlin-2 regulates cancer metastasis. Recent studies have involved Kindlin-2 in other oncogenic signalling pathways responsible for the activation of cancer metastasis, including a recent study by Lin et al. 34 , that showed Kindlin-2 promotes metastasis of hepatocellular carcinoma through Wnt/β-catenin signalling. It is possible that both signalling mechanisms may contribute together to the Kindlin-2-mediated regulation of BC metastasis, as the Wnt/β-catenin signalling has been reported to be regulated by Kindlin-2 in BC 35 .
The finding from this study on the role of Kindlin-2 in the regulation of the metastasis phenotype is a continued effort that was built upon our recently published data that demonstrated a key role for Kindlin-2 in the modulation of the tumour microenvironment to support BC growth and progression 13 , as well as its involvement in the molecular mechanisms of tumour resistance to therapy 36 . Together these studies support the function of Kindlin-2 as a major regulator of several key hallmarks of cancer 37,38 ; and make a strong argument for the development of therapeutic strategies to target Kindlin-2 in BC as well as other cancers.

Ethics statement. All animal studies were performed under protocols approved by the Institutional Animal
Care and Use Committee at Cleveland Clinic. This study used 6-to 8-week-old female NSG or BALB/C mice (Jackson Laboratory, Farmington, CT). All methods were performed in accordance with the guidelines and regulations set and approved by the Cleveland Clinic.
Cell Lines and Reagents. BC cell lines were obtained from American Type Culture Collection (ATCC), and maintained according the manufacturer's protocols. Cell lines were also routinely authenticated by STR DNA fingerprinting analysis. 4T1 cells were engineered to stably express firefly luciferase by transfection with pNifty-CMV-luciferase and selection (500 μg/ml) with Zeocin (Invitrogen, Carlsbad, CA). Kindlin-2-KO cells were generated by lentiviral transduction. We used two independent and verified Kindlin-2-specific sgRNAs for human and mouse Kindlin-2 and a scrambled sgRNA (i.e., nonsilencing sgRNA). Loss of Kindlin-2 expression was determined by Western blot.
Western blot analyses. Cells were grown in 6-well tissue culture plates until confluency reached ~80%.
Cells were then washed twice with ice-cold PBS and lysed on ice with RIPA buffer. The crude lysates were transferred to prechilled tubes and centrifuged at 13K RPM for 15 min at 4 °C. Cleared cell lysates were denatured with SDS sample buffer and resolved (~25 ug) on a 4-20 gradient SDS acrylamide gels (BioRad), followed by transfer to polyvinylidene difluoride membranes (BioRad). Membranes were incubated in 5% bovine serum albumin in PBST buffer for 1 h at room temperature, followed by incubation with the primary antibody (as specified) overnight at 4 °C. Membranes were then washed 3 times with PBST buffer and incubated in the appropriate secondary antibody at room temperature for 1 h, and the signals were developed using the Western Lights chemiluminescence detection kit (PerkinElmer Life Sciences).
Cell spreading and adhesion assays. Cell adhesion assays were carried out by seeding cells on round coverslips that were pre-coated for 1 hour with 10 μg/mL collagen I (Sigma) in 12-well plates. Cells were allowed to adhere for 1 h or spread overnight in a 37 °C CO2 incubator. Spread cells were washed twice with PBS then fixed with 4% Paraformaldehyde for 20 minutes and washed twice with PBS. Fixed cells were permeabilized with 0.1% Triton X-100 in PBS for 15 min, washed with PBS, then stained for 1 hour with Alexa Fluor 568-conjugated phalloidin (Invitrogen, Molecular Probes). After washing, the coverslips were on slides on a drop of DAPI-containing immune-fluor mounting medium (MP Biomedicals). Images were acquired with a fluorescent microscope (Leica) equipped with digital camera using Image-Pro Plus version 5.1.2.59. Cell adhesion was quantified by counting the number of cells (nuclei) and averaged to 20 fields, while cell spreading was quantified with ImageJ software version 1.43 u and the results were plotted as the total surface area covered by cells, in square micro meter, averaged to 300 cells.
Invadopodia formation and extracellular matrix degradation Assays. FITC-gelatine degradation assays were performed according to the procedures described in our previously published study 39 , following the manufacturer's instructions (Invitrogen, CA). We used fluorescence confocal microscopy to document invadopodia and ECM degradation, as described 40,41 . The ImageJ software was used to analyse the data and for 3D reconstruction of images.
SCIentIfIC REPORts | (2018) 8:7360 | DOI:10.1038/s41598-018-25373-0 Primary Tumour Growth, Bioluminescence Imaging and Metastasis assays. Parental (scram) or Kindlin-2-deficient MDA-MB-231 cells (106 cells per mouse, n = 5) were implanted into the mammary fat pads of female NSG mice. Tumour growth was followed by twice weekly monitoring of tumour volume with digital Vernier callipers. Luciferase-expressing parental (scram) or Kindlin-2-deficient 4T1 cells (10,000 cells per mouse, n = 5) were implanted into the mammary fat pads of female BALB/c mice. Tumour growth was quantified using bioluminescence imaging as described 13 . In the metastasis, cells (2 × 10 6 ) suspended in 0.15 ml of sterile PBS were injected using a 28-gauge needle into a tail vein of 6-to 8-week-old female NSG mice. Mice were sacrificed 5 weeks later and the recovered lungs were filled and stained with 15% black India ink using a 19-gauge needle, administered through the trachea. Tumour cells appear as white spots on the lung since they do not take up the ink. Metastasis foci (white spots) were counted and the results were plotted as average number of foci per lobe.
Immunohistochemistry. Tumour and lung tissues were harvested and snap-frozen in optimal cutting temperature (OCT) medium (Sakura Finetek, Torrance, CA, USA), and 8-μm sections were prepared. Tumour sections were stained with the following antibodies: rabbit anti-fibronectin (Millipore, Temecula, CA, USA), rabbit anti-vimentin (Cell Signaling Technology, Danvers, MA, USA), rabbit anti-N-cadherin (Thermo Scientific, Rockford, IL, USA), followed by Alexa 488 or Alexa 568-conjugated goat anti-rabbit IgG. Lung sections were stained with hematoxylin (Vector Laboratories, Burlingame, CA, USA) and eosin (Fisher Scientific Co., Kalamazoo, MI, USA) following the manufacturer's protocol. We used fluorescent or bright-field imaging microscopy (Leica, Wetzlar, Germany) and ImagePro Plus Capture and Analysis software (Media Cybernetics, Rockville, MD, USA) to analyse the stained sections. We used Image Pro-Plus software to quantify stained areas from 10-15 independent fields/sections. Genomic DNA Assay. The human genomic PCR was performed as previously described 42 . The human-specific ERVK6A gene can be identified in human DNA (MDA-MB-231) but not in mouse DNA (mouse lung). The presence of the human DNA sequences in the mouse lungs is an indication of the presence of human cells in these tissues. ERVK6A was amplified form DNA of lungs derived from control mice (scram 231 cells) or mice injected with 23-K2-KO cells. The mouse Wave3 gene is used as a control for integrity of the mouse DNA in each sample.
Site-directed Mutagenesis, and Luciferase promoter assay. The 3′-UTR of Kindlin-2 was PCR-amplified using K2 cDNA as a template and subcloned into pmirGlo vector downstream of the firefly luciferase expression cassette (Promega, Madison, WI). The correct sequence and orientation were verified by sequencing. We used the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA), as per the manufacturer's instructions, to generate the K2 3′-UTR variants, where seed sequences that are recognized by miR200 microRNA were deleted. pmirGlo reporter plasmids (1 μg total plasmid amount) were transfected with Lipofectamine 2000 (Invitrogen) into the specified cancer cells, seeded in 12-well plates (3 × 104 cells/well). Cells were collected after 48 h for assay using the dual-luciferase reporter assay system (Promega). For co-transfection experiments, microRNA mimics or miR negative control (5 nM) or anti-miR200b microRNA inhibitor or anti-miR negative control (20 nM) were added to the transfection mix.
Real-time quantitative RT-PCR. We used TRIzol reagent (Invitrogen, Camarillo, CA) to extract total RNA from cancer cell lines and tumour tissues using, according to the manufacturer's instructions. Quantitative real-time PCR was performed as described previously 36,41,43 . Oligonucleotide primers used) for qt-RT-PCR (Supplementary Table 1) were obtained from Qiagen (Germantown, MD).
Immunostaining. Tissues were collected at the designated times and snap-frozen in optimal cutting temperature (OCT) medium (Sakura Finetek, Japan), and 8-μm sections were prepared and stained with the following antibodies; Rabbit anti-fibronectin (1:100) (Millipore, Temecula, CA, USA), rabbit anti-vimentin (1:100) (Cell Signalling Technology, Danvers, MA, USA), rabbit anti-N-cadherin (1:1000) (Thermo Scientific, Rockford, IL, USA), as described previously 13,41 . Stained sections were analysed using fluorescent or bright-field imaging microscopy (Leica, Germany) and ImagePro Plus Capture and Analysis software (Media Cybernetics, Rockville, MD). Vimentin, Fibronectin and N-Cadherin positive areas were quantified in 15 independent fields/section using Image Pro-Plus software 41,44 . Statistical analysis. Experiments were done in triplicate and analysed using the Student's t-test. In calculating two-tailed significance levels for equality of means, equal variances were assumed for the two populations. Results were considered significant at p < 0.05. Data availability. The datasets generated and analysed during this study that are not included in the published article are available from the corresponding author upon reasonable request.