Loss of ERα induces amoeboid-like migration of breast cancer cells by downregulating vinculin

Oestrogen receptor alpha (ERα) is a well-known target of endocrine therapy for ERα-positive breast cancer. ERα-negative cells, which are enriched during endocrine therapy, are associated with metastatic relapse. Here we determine that loss of ERα in the invasive front and in lymph node metastasis in human breast cancer is significantly correlated with lymphatic metastasis. Using in vivo and in vitro experiments, we demonstrate that ERα inhibits breast cancer metastasis. Furthermore, we find that ERα is a novel regulator of vinculin expression in breast cancer. Notably, ERα suppresses the amoeboid-like movement of breast cancer cells by upregulating vinculin in 3D matrix, which in turn promotes cell–cell and cell–matrix adhesion and inhibits the formation of amoeboid-like protrusions. A positive association between ERα and vinculin expression is found in human breast cancer tissues. The results show that ERα inhibits breast cancer metastasis and suggest that ERα suppresses cell amoeboid-like movement by upregulating vinculin.

T he pathogenesis of breast cancer is associated with oestrogen receptor alpha (ERa), which is activated by sex hormones and contributes to the aberrant proliferation of breast cancer cells 1,2 . The classical mechanism of ERa action involves regulating the transcription of oestrogen-responsive genes by binding to the oestrogen-responsive element (ERE) within the promoters of the target genes 3,4 .
Endocrine therapy with selective oestrogen receptor modulators, such as tamoxifen 5 , has been widely used to antagonize ERa in breast cancer tissues 6 . However, tamoxifen appears to decrease the risk of ERa-positive contralateral breast tumours and to increase the risk of ERa-negative contralateral tumours 7 . Therefore, lost expression of ERa during adjuvant endocrine treatment for ERa-positive breast cancer allows for resistance to common adjuvant endocrine therapies and is associated with ERa-negative metastatic relapse 7 . Nevertheless, how ERa loss is associated with metastasis remains to be elucidated, particularly in a three-dimensional (3D) environment, which can better mimic human breast cancer metastasis in vivo.
The migration of invasive cells in vivo consists of the mesenchymal mode, in which invasive cells are elongated and require pericellular matrix proteolysis 8 , and the amoeboid mode, in which carcinoma cells with low adhesion and round morphology require subcellular localization of myosin II behind the cell nucleus to drive actomyosin contractility, independently of matrix metalloproteinase [9][10][11] . Moreover, mesenchymal cells usually burst after entering the bloodstream, whereas highly metastatic cancer cells efficiently penetrate the blood vessels through amoeboid-like migration with high actomyosin contractility, which provides cells with mechanical strength to resist shear forces in the circulation 9,10,12 . In addition, cells undergoing amoeboid migration have higher velocity than those undergoing mesenchymal migration 13 . Despite this observation, the molecular mechanisms that affect the amoeboid motility in a 3D environment require more in-depth study.
Vinculin (VCL), a membrane cytoskeletal protein found in focal adhesion plaques, is involved in the linkage of the ECM to the actin cytoskeleton 14 . Vinculin-null cells transfected with vinculin cDNA show markedly decreased motility and tumorigenicity 15 . In addition, downregulation of vinculin was found in several highly metastatic cancer cells 16,17 . Loss of vinculin induced protection from apoptosis in an anchorageindependent manner and enhanced cell motility 18 . However, few studies have described the involvement of vinculin in the regulation of breast cancer amoeboid movement.
In this study, we show that loss of ERa promotes tumour metastasis through in vitro experiments, in vivo tumour xenograft assays and the analysis of clinical breast cancer samples. Furthermore, we find that ERa is a novel regulator of vinculin expression in breast cancer and that loss of ERa induces amoeboid-like migration of breast cancer cells by regulating vinculin in a 3D matrix.

Results
Loss of ERa is correlated with breast cancer metastasis. Metastasis occurs when tumour cells detach from their primary location and move to the lymph nodes and then to distant organs 19,20 . Therefore, the lymph node usually serves as a bridge allowing the metastatic dissemination of tumours 21 . We examined ERa expression in human primary breast cancer tissues and the corresponding lymph node metastasis from 124 ERa-positive breast cancer patients. This analysis demonstrated that, in contrast to the abundant expression of ERa in the primary tumour, 54.8% of samples lost the expression of ERa in the corresponding lymphatic metastasis (Fig. 1a,b) (Supplementary Table 1). In addition, loss of ERa expression in lymphatic metastasis was also positively associated with the clinical stage (P ¼ 0.007), the number of lymph node metastases (P ¼ 0.011) and the loss of progesterone receptor expression (P ¼ 0.045). However, there was no significant association with age, tumour size or HER2 expression (Supplementary Table 2).
We further examined the expression of ERa at the invasive front where the infiltration of CD68-positive tumour-associated macrophages is found 22 and where at least 50% of the cell surface of tumour cells contacts the matrix 23 and the non-invasive front of the primary tumour tissues. The results showed that ERa was minimally expressed in breast cancer cells at the invasive front, whereas an increased intensity of ERa staining was observed at the non-invasive front (Fig. 1c-e). Collectively, these findings suggest that decreased ERa expression might promote breast cancer metastasis.
ERa inhibits breast cancer metastasis in vivo and in vitro. To investigate whether ERa might inhibit the metastasis of breast cancer cells in vivo, we first chose ERa-positive MCF-7 and Cas9-ERa MCF-7 cells to examine the lung metastasis in a tail vein injection model, which mimics the process of loss of ERa during metastasis in patients. The expression of ERa in the two cell lines was detected by western blot analysis (Supplementary Fig. 1a,b). The cell proliferation assay confirmed that the Cas9-ERa MCF-7 cells grew more slowly in vitro ( Supplementary  Fig. 1c). Biofluorescence was examined at different time points to monitor the location and growth of tumour xenografts in the lungs. We found that the normalized photon flux of Cas9-ERa MCF-7 cells, which represents the pulmonary metastasis focuses, was significantly higher than that of control cells, as evidenced by H&E staining (Fig. 2a,b). Therefore, the Cas9-ERa MCF-7 cells showed a more profound increase in metastatic potential to the lungs.
Furthermore, control MDA-MB-231 cells and ERa-expressing MDA-MB-231 cells were also used to examine the metastasis in an orthotopic injection model (Fig. 2c). The expression levels of ERa in these two cell lines were quantified in vitro ( Supplementary Fig. 1d,e). The normalized photon flux of control cells was lower than that of ERa-expressing cells (Fig. 2d), and the cell proliferation assay also verified that the control cells grew more slowly in vitro ( Supplementary Fig. 1f). To observe the tumour metastasis, we covered the primary tumour site to avoid strong signal interference from the primary tumour. We found that the tumours derived from control cells were more metastatic than the tumours from ERa-expressing cells ( Fig. 2e; Supplementary Fig. 1g). Further histological examination of the ipsilateral side of axillary lymph nodes showed lymphatic metastases in mice inoculated with control cells (Fig. 2e). Because metastasis accounts for B90% of breast cancer-related deaths 24,25 , we recorded the survival rate of nude mice in two additional groups. The results showed that the overexpression of ERa significantly extended the lifespan of the mice (Fig. 2f).
Transwell assays were performed with MDA-MB-231 and MCF-7 cells with gain or loss of ERa, and the results showed that ERa inhibited breast cancer cell invasion in vitro ( Supplementary  Fig. 1h-j; Fig. 2g). We also conducted transwell assays with MCF-7 cells that were treated with different concentrations of fulvestrant, an ERa downregulator. The results showed that, although 0.29 and 10 nm fulvestrant significantly downregulated ERa expression and inhibited cell proliferation, it promoted cell invasion ( Fig. 2h; Supplementary Fig. 1k-m). In addition, we found that the invasive capacity of tamoxifen-resistant MCF-7 (MR) cells with lower ERa expression was significantly stronger than that in parental MCF-7 cells ( Fig. 2i; Supplementary  Fig. 1n,o). These observations indicated that loss of ERa expression promotes the invasive and metastatic ability of breast cancer cells.
Loss of ERa induces the amoeboid migration of MCF-7 cells. The migration of invasive cells in the 3D environment consists of the mesenchymal mode and the amoeboid mode. Next, we tracked and analysed the migration of different MCF-7 cell populations of control, Cas9-ERa and GM6001 (matrix metalloproteinase inhibitor)-treated Cas9-ERa in the 3D matrix by using phase holographic imaging assays. The result showed that loss of ERa in MCF-7 cells promoted the invasive ability of tumour cells in the 3D matrix and that GM6001 did not impair the migration speed of Cas9-ERa cells (Fig. 3a,b). In addition, GM6001 did not affect the migration speed of MCF-7 cells either ( Supplementary Fig. 2a,b).
We also found that the control and Cas9-ERa MCF-7 cells did not express active MMP-2 or MMP-9 enzymes by zymography assay (Supplementary Fig. 2c). Using phase holographic imaging assays, we observed the following five phenotypes in the different populations of MCF-7 cells: 'stable adhesion', 'unstable lamellipodia', 'bleb', 'stable bleb' and 'unstable pseudopod' (Supplementary Movies 1-5). The 'bleb', 'unstable pseudopod' and 'stable bleb' phenotypes are known as amoeboid-like protrusions, which are associated with enhanced amoeboid movement 26,27 . The incidence of a 'stable adhesion' phenotype was significantly reduced, and the amoeboid-like protrusions were more predominant in Cas9-ERa-or GM6001-treated Cas9-ERa cells compared with the control (Fig. 3c). Amoeboid migration requires high levels of phosphorylation of myosin light chain (p-MLC) and subcellular localization of p-MLC behind the cell nucleus to drive actomyosin contractility 11,23 . Confocal assays showed an increased expression of the active form (phospho-Ser19) of myosin light chain (p-MLC) in Cas9-ERa MCF-7 cells, and more Cas9-ERa MCF-7 cells showed p-MLC rear distribution ( Fig. 3d-f). In addition, the roundness index of Cas9-ERa cells was higher than that of control cells (Fig. 3g). Together, these results indicated that loss of ERa induces the amoeboid-like migration of MCF-7 cells.
ERa is a transcriptional promoter of vinculin. To elucidate the molecular mechanism of ERa action on the metastasis of breast cancer, RNA-sequencing was performed using MDA-MB-231-ERa cells and MDA-MB-231-vector cells. The main metastasis-associated genes that showed significantly altered expression levels are shown in Supplementary   Fig. 3a). Using bioinformatics analysis, we found four potential ERa-binding sites in the promoter of vinculin by which ERa might regulate vinculin transcription. Subsequently, MCF-7 cells cultured in oestrogen-depleted medium were treated with 0 or 5 nm oestrogen, and the luciferase activity of the vinculin promoter was determined in these MCF-7 cells. We found that, when the cells were treated with oestrogen, the luciferase activity increased significantly ( Supplementary Fig. 3b). Next, we performed chromatin immunoprecipitation (ChIP) assays in MCF-7 cells to detect whether endogenous ERa formed a transcription initiation complex at the vinculin promoter. From quantitative PCR analysis, we found that the antibody against ERa pulled down higher amounts of vinculin promoter DNA than did the IgG control and that the most efficient binding activity was localized within the À 998 to À 1,238 region of the vinculin promoter (Fig. 4a). Then, we truncated the vinculin promoter according to the four ERa candidate-binding sites, as shown in Fig. 4b, seeking to identify which region of the promoter is important for vinculin transcription by using a dual luciferase reporter assay. The results showed that the essential region for ERa transcriptional activity in the promoter was À 940 to À 1,155 bp in both MCF-7 and MDA-MB-231 cells (Fig. 4c,d). Sequence analysis of this region revealed a 23-bp ERE located in the À 940 to À 1,155 bp region whose core sequence was GGCC (Fig. 4e). To verify the effect of the ERE region, three mutants were constructed, as shown in Fig. 4e and used in luciferase reporter assays. The results showed that all three mutants exhibited decreased luciferase activities in MDA-MB-231 and MCF-7 cells (Fig. 4f,g). Collectively, these results demonstrated that ERa directly facilitates vinculin transcription by specifically binding to the ERE region of the vinculin promoter.
ERa upregulates vinculin expression in breast cancer cells. MCF-7 cells cultured in oestrogen-depleted medium were treated with 0 or 5 nm oestrogen. We observed that the mRNA levels of vinculin were higher in MCF-7 cells treated with 5 nm oestrogen than in control cells ( Supplementary Fig. 4a). In addition, the mRNA and protein levels of vinculin were higher in ERa-positive breast cancer cell lines than in ERa-negative cell lines ( Fig. 5a,b; Supplementary Fig. 4b). To further investigate the role of ERa in regulating vinculin transcription and expression, real-time PCR and western blotting were performed to assess the effect of gain or loss of ERa expression on the level of vinculin mRNA and protein. The results showed that silencing endogenous ERa by using short interfering RNAs (siRNAs) led to a downregulation of vinculin in MCF-7 and ZR-75-1 cells and that ectopic expression of ERa by using the pcDNA3.1( À )-ERa vector in MDA-MB-231 and SK-BR-3 cells induced vinculin upregulation both at the mRNA and protein levels ( Fig. 5c-e; Supplementary Fig. 4c-f). Importantly, using a confocal assay, we further confirmed that the expression of vinculin decreased after knockdown of nuclear expression of ERa in MCF-7 cells ( Fig. 5f; Supplementary  Fig. 4g). In addition, in MDA-MB-231 cells, vinculin expression was concomitantly promoted by the ectopic nuclear expression of ERa ( Supplementary Fig. 4h). Together, these results indicated that ERa is a promoter of vinculin expression in breast cancer cells.
Vinculin downstream of ERa is important for metastasis. To investigate the role of vinculin, downstream of ERa, in breast cancer metastasis, we performed tail vein injections of control or Cas9-vinculin MCF-7-luc2 cells in athymic mice and examined the expression levels of vinculin by western blotting ( Supplementary Fig. 5a,b). We found that, compared with control cells, the Cas9-vinculin cells showed a more profound metastatic potential to the lungs (Fig. 6a,b). In addition, extravasation is a crucial step during tumour metastasis. To test whether vinculin involved in this process, the lung extravasation assay was performed. Similar numbers of control or vinculin-depleted MCF-7 cells lodged in the lung capillaries 0.5 h after injection (Fig. 6c). However, after 24 h the number of vinculin-depleted MCF-7 cells that remained in the lung parenchyma was more than the number of control cells (Fig. 6c,d).
We further studied an orthotopic mouse model of breast cancer, using MDA-MB-231 cells stably expressing ERa with or without vinculin knockdown ( Supplementary Fig. 5c). Vinculin expression levels were detected by western blotting (Supplementary Fig. 5d,e). The results showed that tumours formed by vinculin knockdown cells were more metastatic compared with those of the control cells, and, after 4 weeks, lymph node metastasis was observed in the mice injected with vinculin knockdown cells ( Supplementary Fig. 5f,g). The mice inoculated with vinculin knockdown cells had shorter lifetimes, probably because of the distant metastasis ( Supplementary  Fig. 5h). In addition, a transwell assay also showed that vinculin knockdown in MDA-MB-231 cells stably expressing ERa rescued the invasive capacity of the cells (Fig. 6e).
Furthermore, it has been reported that MDA-MB-231 cells can migrate through either the mesenchymal or amoeboid mode, and the amoeboid morphology might be predominant after protease inhibition 28 . Using zymography on conditioned two-dimensional (2D) or 3D medium, we observed that MDA-MB-231 cells expressed active MMP-2 or MMP-9 enzymes ( Supplementary  Fig. 5i). We also observed that GM6001 indeed decreased the invasive capacity of MDA-MB-231 cells in transwell assays ( Supplementary Fig. 5j,k). Moreover, we observed that GM6001 induced round cell morphology and that depleted vinculin expression reversed the decreased invasive capacity of the GM6001-treated MDA-MB-231 cells to some degree ( Supplementary Fig. 5j,k). Next, we performed orthotopic injection of control, GM6001-treated control or GM6001treated Cas9-vinculin MDA-MB-231 cells in athymic mice. Tumours from GM6001-treated cells were less metastatic than those from control cells, and tumours from GM6001-treated Cas9-vinculin cells and control cells had similar levels of metastasis ( Fig. 6f,g). In addition, GM6001-treated cells in primary tumours showed round cell morphology with a higher average roundness score than control cells in vivo (Fig. 6h). In addition, the mice inoculated with control or GM6001-treated Cas9-vinculin cells had shorter lifetimes than those inoculated with GM6001-treated control cells ( Supplementary Fig. 5l). These results showed that GM6001 can induce the rounded-amoeboid morphology of cells and loss of vinculin in GM6001-treated breast cancer cells is associated with increased metastatic potential.

Loss of vinculin promotes amoeboid features of cancer cells.
The amoeboid features contained membrane blebbing, cell rounding, high actomyosin contractility and increased invasion 29 . To investigate whether loss of vinculin could promote the amoeboid features of cells, we tracked and analysed the migration speed of different MCF-7 cell populations of control and Cas9-vinculin in the 3D matrix by using phase holographic imaging assays (Fig. 7a). We found that loss of vinculin in MCF-7 cells promoted the instantaneous speed of tumour cells in the 3D matrix (Fig. 7b). In addition, after depletion of vinculin, the activity of MLC increased, thus indicating that loss of vinculin promoted actomyosin contractility in breast cancer cells (Fig. 7c,d). We also found that the percentage of amoeboid-like protrusions, including 'bleb', 'stable bleb' and 'unstable pseudopod', in Cas9-vinculin cells significantly increased ( Fig. 7e; Supplementary Movies 6-10). Confocal assays further showed that the control MCF-7 cells formed actin stress fibres, which were concomitant with p-MLC being evenly distributed in the cytoplasm, whereas in vinculin-depleted MCF-7 cells, few stress fibres were observed, and p-MLC was mostly polarized to the rear part of the cell (Fig. 7f). Moreover, Cas9-vinculin cells exhibited cell rounding and dynamic blebbing (Fig. 7f,g) and, compared with control cells, more Cas9-vinculin cells displayed p-MLC rear distribution (Fig. 7h).
ERa correlates with vinculin in breast cancer tissues. To investigate in a clinical setting whether ERa suppresses breast cancer metastasis by facilitating vinculin expression, we measured the expression of vinculin in the same 124 human primary breast   cancer tissues and human breast cancer lymph node metastases that were used in the ERa analysis. The results showed that vinculin abundance was also significantly higher in human primary breast cancer tissues than in lymphatic metastases (Fig. 8a). A positive association was found between vinculin and ERa in both primary tumour (Po0.001, R 2 ¼ 0.528) and lymphatic metastasis (Po0.001, R 2 ¼ 0.366; Supplementary Tables 4 and 5). The expression levels of vinculin are shown in Supplementary Fig. 6a. We further compared the expression of these two molecules at the invasive and non-invasive fronts in primary breast cancer tissue. Similarly to the trend of ERa expression level, diminished ARTICLE vinculin expression was also detected at the invasive front, whereas increased staining of vinculin was observed at the non-invasive front (Fig. 8b-e). These findings indicated that ERa might suppress human breast cancer metastasis by facilitating vinculin expression.

Discussion
Among all breast cancers, ERa-positive (ER þ ) tumours constitute the largest proportion, B70% (ref. 30). Although sporadic publications have shown that the expression of ERa in MDA-MB-231 cells inhibits proliferation in vitro 31 , a plethora of laboratory and epidemiological data have demonstrated that ERa, by binding to oestrogen, is the major driving factor for growth in ERa þ breast cancers 4,32 . In the present study, we also found that expression of ERa indeed promoted proliferation of MDA-MB-231 cells in vitro and in vivo. ERa induces cell proliferation by increasing the expression of MYC and cyclin D1 (refs 33,34). Therefore, ERa has been used as a key target for endocrine therapy of ERa-positive breast cancer to block the proliferation of cancer cells 35 . However, an increasing number of the clinical epidemiological investigations show that ERa-positive primary breast cancer patients have an increased frequency of ESR1 mutations in metastatic ERa þ breast cancer tissues and even ERa-negative metastatic relapse after receiving endocrine therapy 1,7 . Although a decline in ERa levels has been detected in invasive breast cancers 4 , the relationship between ERa and tumour metastasis is still far from clear. In the present study, we observed that, compared with those in the primary tumour, breast cancer cells in lymph node metastases expressed lower levels of ERa. More importantly, loss of ERa in lymphatic metastases was also positively associated with clinical stages and lymph node metastases. Even in the primary tumour, the expression of ERa at the invasive front was lower than that at the non-invasive front. Therefore, these findings indicated that ERa expression is inversely correlated with breast cancer metastasis. We further demonstrated that ERa is indeed capable of inhibiting breast cancer metastasis by using athymic mouse models and transwell assays. This finding is very important because it suggested that breast cancer cells with loss of expression of ERa should be not only be resistant to common adjuvant endocrine therapy but also have stronger invasion and metastasis capabilities. Consistently with this possibility, we found that an appropriate concentration of fulvestrant promoted cell invasion in vitro by downregulating ERa and that the invasive capacity of tamoxifen-resistant MCF-7 cells with lower ERa expression was stronger than that of parental MCF-7 cells. Collectively, we believe that repeated biopsies are necessary to reassess the receptor status in metastatic disease to guide endocrine therapy with greater precision.
Several studies have reported that ERa suppresses cellular motility and invasion 31 by inhibiting the epithelial-mesenchymal transition in 2D conditions [36][37][38][39] . However, with regard to cell shape and movement, the 3D environment resembles the in vivo condition 40 . Some tumour cells can utilize amoeboid-like migration as an alternative to mesenchymal migration during migration in 3D cultures and in vivo 41,42 . To date, few studies have focused on the relationship between the amoeboid-like migration of breast cancer cells in a 3D environment and ERa expression. We demonstrated that ERa-positive MCF-7 cells expressed few active MMP-2 and MMP-9. Loss of ERa also promoted a more round shape and increased cell motility in the 3D matrix in vitro, and these effects were not impaired by the MMP inhibitor GM6001. The formation of amoeboid-like protrusions, such as blebs, often corresponds with the amoeboid phenotype and enhanced invasion and metastasis and blebbing movement has been widely promoted as a cancer cell migration strategy [43][44][45][46][47] . We observed the significantly increased formation of amoeboid-like protrusions in ERa-depleted MCF-7 cells in the present study. During the initiation of blebs, local dissociation of the membrane from the cortex or a local rupture of the actin cortex causes the bleb to form in the direction of the desired flow 48,49 . Furthermore, it has been reported that high contractility of the cell rear triggers bleb initiation 43,50 . Indeed, we observed that loss of ERa in MCF-7 cells induced the polarization of p-MLC at the cell rear, thereby resulting in high contractility of the cell rear, the formation of amoeboid-like protrusions and enhanced amoeboid-like migration. Even in the MDA-MB-231 cells that use mixed migration modes in a 3D matrix, ERa suppressed amoeboid-like migration after MMP inhibition, although overexpression of ERa did not alter the mesenchymal-like morphology of MDA-MB-231 cells in a 2D substrate (Supplementary Fig. 6b,c). Together, our results show that ERa inhibits the amoeboid-like migration of breast cancer cells in a 3D matrix.
With respect to the mechanisms by which ERa inhibits cancer matastasis, we focused on the regulation of ERa to vinculin expression based on transcriptome sequence analysis. Our data show that ERa is a novel regulator of vinculin expression in breast cancer and that vinculin is involved in ERa-mediated inhibition of breast cancer cell metastasis. As above, vinculin is a key regulator of focal adhesions and loss of vinculin in cells promotes focal adhesion turnover 51 . It was reported that focal adhesions could promote the formation of persistent stress fibres, which prevent the transition to amoeboid migration by competing with the cell cortex for recruitment of the actomyosin contractile machinery 52 . In the present study, our results showed that the depletion of vinculin in MCF-7 cells induced decreased cell-matrix adhesion ( Supplementary Fig. 6d). More importantly, we found that in a 3D matrix vinculin depletion leads to decreased stress fibres and p-MLC mostly polarizes to the rear part of the cell, thus resulting in high contractility of the cell rear and induced bleb initiation.
Metastasis occurs when tumour cells detach from the epithelial sheets and invade surrounding tissue. It was reported that vinculin was important for cadherin-mediated cell-cell junctions 53 and might suppress metastasis formation in vivo by promoting cadherin-mediated retention of tumour cells in primary tumours 16 . Meanwhile, we found that depletion of vinculin in MCF-7 cells induced loss of cell-cell contact ( Supplementary Fig. 6e,f). However, we also found that loss of vinculin in MCF-7 cells induced cell rounding, increased actomyosin contractility, fast shape changes and increased invasion. These observations indicated that the effect of vinculin on amoeboid features of cells might contribute to the process of local invasion or even the metastasis formation besides the effect on cell-cell adhesion. Moreover, it was reported that high levels of actomyosin contractility in cancer cells could promote and efficient lung colonization or seeding 54,55 and rapid extravasation was associated with the amoeboid features 10,43,56 . Our in vivo extravasation assay showed that the number of vinculin-depleted cells that extravasated into the lung parenchyma after 24 h was more than the number of control cells. These observations indicated a potential involvement of vinculin in efficient retention in the lungs via its regulation of amoeboid features. More importantly, we do not discard the possibility that reduced cell-cell adhesion and anoikis resistance that resulted from loss of vinculin may be involved in metastasis; however, we believe that we have identified amoeboid features regulated by vinculin, which are associated with metastasis.
In summary, our results demonstrate that ERa suppresses breast cancer metastasis by regulating vinculin. Our findings provide novel insight into the association of ERa loss during endocrine therapy with enhanced invasive and metastatic ability of breast cancer cells and should aid in a more comprehensive Serial sections (4 mm) of paraffin-embedded samples were conventionally dewaxed and hydrated with gradient ethanol. After the inactivation of endogenous peroxidase with 3% H 2 O 2 -methanol for 10 min, the sections were washed three times in PBS and blocked with goat serum for 20 min. Then, the sections were coated with primary antibodies and incubated in a wet box at 4°C overnight. After the addition of PowerVision complex (ZSGB-BIO, SP-9001, sp-9002), tumour sections were incubated at 37°C for 20 min followed by 3, 3 0diaminobenzidine (DAB) (ZSGB-BIO, ZLI-9031) colouring and wood grain redyeing. PBS instead of antibodies was used as a negative control. With respect to ERa and vinculin staining, the percentage of stained cells was categorized as follows: À , no staining; þ , 1-25%; þ þ , 25-50%; and þ þ þ , 450% staining. In parallel, corresponding haematoxylin-eosin (H&E) staining was reviewed to confirm diagnosis by a pathologist using the immunohistochemistry preparations. CRISPR/Cas9 system. GENECHEM designed and cloned the corresponding single guide RNAs (sgRNAs) into GV392 plasmid. Lentivirus was used to deliver the corresponding GV392 plasmid into breast cancer cell lines. The sequence of relative sgRNA was listed in Supplementary Table 8. The respective locus was amplified using the primers (ESR1 fwd: 5 0 -TTGTAATGCATATGAGCTCG-3 0 rev: 5 0 -CTGCTGTCCAGGTACACCTC-3 0 and VCL fwd: 5 0 -CTTGTCCAGGCAGC TCAGA-3 0 rev: 5 0 -CACCACCTCTGCCACTGT-3 0 ). Heterodimerization and digestion were performed with the Knockout and Mutation Detection Kit (GENECHEM) according to the manufacturer's instructions. Cleavage products were separated on a 2% agarose gel and stained with ethidium bromide. Images were captured with the SmartGel Image Analysis System (SAGECREATION). The corresponding results were listed in Supplementary Fig. 6g,h. Animal studies. Female athymic mice, 4-week of age, were selected. The animal study was performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee of the FMMU. For mammary fat pad injection, 3 Â 10 6 viable breast cancer cells were injected into the mammary fat pads of athymic mice (n ¼ 5). For tail intravenous injection, 1 Â 10 6 viable breast cancer cells were injected into athymic mice from tail vein (n ¼ 5). Then, the mice were anaesthetized and injected with luciferin and imaged using an IVIS-100 System at the UAB Small Animal Imaging Core Facility. Light emission from animal tissue was measured using a software provided by the vendor (Xenogen). In parallel, an equivalent number of tumour cells were inoculated into the same sites of other athymic mice to record their survival time. The operator who performed injection of tumour cells was blinded with the group allocation. At least two independent experiments were performed.
Cell proliferation assay. The proliferation ability of the cells was measured by Cell Counting Kit-8 (CCK-8) solution (7sea biotech, Shanghai, China). Briefly, breast cancer cells were seeded on 96-well plates (Corning, USA) at a concentration of 2 Â 10 4 cells per well and incubated at 37°C overnight. The Cell Counting Kit-8 reagents were then added to a subset of wells when cells grew for 24, 48, 72 or 96 h. After the cells were incubated for 2 h at 37°C, we quantified the absorbance at 450 nm using a microplate reader (Bio-Rad). Each group was made in quintuplicate.
Zymography. The assessment of MMP-2 and MMP-9 activity was performed using a Gelatin Zymography Kit (Xin Sails Biotechnology, Shanghai, China). Briefly, conditioned serum-free medium of breast cancer cells grown on plastic for 36 h was loaded into the polyacrylamide gels containing 10% gelatin and then electrophoresed at 20 mA per gel for 90 min. Following electrophoresis, gels were rinsed twice with 10 ml 1 Â Buffer A for 30 min at room temperature and incubated with 10 ml 1 Â Buffer B for 3 h at 37°C. Gels were then stained with 0.5% Coomassie Blue for 2 h and destained five times for 20 min with destaining solution (methanol:acetic acid:water ¼ 5:7:88). The protein molecular weight of MMP-2, MMP-9 and proMMP-9 is 66B72, 92 and 130 kD.
Phase holographic imaging assay. Overall, 3 Â 10 5 MCF-7 cells of different groups were embedded in Matrigel (diluted with serum-free medium) at 37°C overnight. Half an hour before the experiment, the transwell insert containing Matrigel (diluted with 10% FBS) was seeded on 3D matrix as a chemoattractant, which was in the right side of the observation point. Then, HoloMonitorTM M4 (Phiab, Sweden) was used to track and record related parameters of the movement of cells in 3D matrix in 4 h. Motility and protrusion data were obtained using Â 40 objective. The cells were imaged every 2 min and the movies were played back at 15 frames per second. The main criteria used to group cells into the different phenotypes were the size of protrusions (o2 mm or not) and the lifetime of protrusions (o2 min or not).
Calculation of cell roundness. It was assessed by dividing the shortest diameter of each cell by the longest one (ratio-b/ratio-a) to produce a score between 0 and 1, with perfectly round cells having a score of 1 (ref. 9).
ChIP assay. ChIP experiments were performed using the EZ ChIP Chromatin Immunoprecipitation Kit (Millipore, Billerica, USA). Four primer sets were designed to flank-related putative ERa-binding sites in the promoter region of vinculin. Details of the primer sequence are listed in Supplementary Table 9. Briefly, MCF-7 cells were fixed with 1% paraformaldehyde and sonicated seven times for 10 s each using a sonicator with a microtip in a 1.5-ml tube. Anti-ERa antibody or control human IgG was applied to pull down the chromatin associated with ERa. The chromatin-antibody complexes were collected with Protein G-Agarose. After washing and elution of the complexes from the beads, the DNA-protein crosslinks were reversed at 65°C overnight. The amounts of the specific DNA fragment were then quantified by real-time PCR and normalized against the genomic DNA preparation from the same cells. Each group was made in triplicate.
Luciferase reporter assay. Briefly, breast cancer MCF-7 or MDA-MB-231 cells were seeded in 24-well plates at 50% confluence and transfected with either ERa siRNAs or ERa plasmid using Lipofectamine 2000 and then co-transfected with pRL-Tk and vinculin promoter (pGL3-basic-vinculin). Thirty-six hours later, the cells were lysed in a passive lysis buffer (Promega, San Luis Obispo) and the luciferase activity was measured. Each group was made in triplicate.
Quantitative real-time PCR. Total RNA was isolated from cultured cells with RNAiso Plus (Takara, Dalian, China), and cDNA was synthesized with the PrimeScript RT Reagent Kit (Takara). Then, 2 ml of cDNA was used for real-time PCR reactions in a Prism 7500 real-time thermocycler (Applied BioSystems, Foster City, CA, USA) with SYBR Green Ex Taq (Takara) according to the manufacturer's instructions. The primer sequences are provided in Supplementary Table 10. Each group was made in triplicate.
Western blot analysis. In brief, proteins were transferred to polyvinylidene difluoride membranes after SDS-PAGE using a Bio-Rad Semi-Dry electrophoretic cell. Western blot analyses were performed using specific antibodies followed by an horseradish peroxidase (HRP)-conjugated IgG antibody. Enhanced chemiluminescence (Pierce) for HRP was used for immunoreactive protein visualization. Uncropped scans of the blots are shown in Supplementary Figs 7 and 8. 4°C with corresponding antibody after being blocked for 30 min with goat serum. Then, the cells were incubated for 2 h with secondary FITC-conjugated or Cy3-conjugated antibodies. Nuclei were counterstained with 4,6-diamidino-2phenylindole. Images were captured using a confocal microscope (FluoView FV1000, Olympus, Tokyo, Japan).
Lung extravasation assay. Overall, 5 Â 10 5 Cas9-vinculin MCF-7 cells (enhanced green fluorescent protein, eGFP) and 5 Â 10 5 control MCF-7 cells (monomeric red fluorescent protein, mRFP) were mixed and injected into the tail vein of nude mice (female, 4 weeks old). Mice were killed after 0.5 or 24 h and the lungs were fixed (4% formaldehyde for 24 h) and examined for fluorescently labelled cells under a confocal microscope (FluoView FV1000, Olympus). Lung retention was represented as fluorescence signal (eGFP or mRFP from MCF-7 cells) per field. Each experiment had four mice per condition, and experiments were replicated three times.
Transwell assay. A total of 3 Â 10 5 cells were placed into chambers coated with 100 ml of Matrigel (BD Biosciences, 356234). The chambers were then inserted into a 24-well plate and incubated for 24 h in a corresponding medium with 10% FBS before examination. The cells remaining on the upper surface of the membranes were removed, whereas the cells that migrated to the lower surface were fixed with 95% ethanol and stained in 4 g l À 1 crystal violet solution. Finally, the invasive cells distributed in eight randomly selected views were counted under a microscope ( Â 20) and averaged.
Cell adhesion assay. In all, 1 Â 10 4 cells were seeded on Matrigel-coated 96-well plate and cultured for 2 h. After washing, adhesion cells were counted using optical microscope (Olympus). The adhesion cells distributed in a randomly selected view were counted under a microscope ( Â 40). Each group was made in octuplicate.
Binding of cells to recombinant E-cadherin. A 96-well plate was coated with 100 ml human recombinant E-cadherin Fc chimera (648-EC-100, R&D Systems)/ well at 1.5 mg ml À 1 in PBS at 37°C for 1 h. The plate was washed with PBS 3 times and blocked by adding 100 ml per well 1% BSA in PBS at 37°C for 30 min. Then 3 Â 10 4 cells per well are added to Recombinant Human E-Cadherin Fc Chimera coated plates and the plate was kept at 37°C for 90 min. After washing, adhesion cells were counted by optical microscope (Olympus, Tokyo, Japan). The adhesion cells distributed in a randomly selected view were counted under a microscope ( Â 40). Each group was made in octuplicate.
Statistical analysis. Statistical analysis was performed using the SPSS statistical software (SPSS16.0, Chicago, CA, USA). A value of Po0.05 was considered statistically significant. A random number table was used to randomize the mice into control and treatment groups. The numbers of mice (in vivo) were determined on the basis of our pre-tests and previous experience with similar experiments. Sample size was chosen to ensure adequate and statistically significant results. Investigators who determined the expression levels of ERa and vinculin and cell-cell adhesion were blinded with respect to the treatment allocation. The in vitro experiments were repeated at least three times. The statistical tests were two-sided.
Data availability. The authors declare that the main data supporting the findings of this study are available within the article and its Supplementary Information Files. Extra data are available from the corresponding authors upon request.