Research ArticleCancer

β1 Integrin Inhibition Elicits a Prometastatic Switch Through the TGFβ–miR-200–ZEB Network in E-Cadherin–Positive Triple-Negative Breast Cancer

See allHide authors and affiliations

Science Signaling  11 Feb 2014:
Vol. 7, Issue 312, pp. ra15
DOI: 10.1126/scisignal.2004751


Interactions with the extracellular matrix (ECM) through integrin adhesion receptors provide cancer cells with physical and chemical cues that act together with growth factors to support survival and proliferation. Antagonists that target integrins containing the β1 subunit inhibit tumor growth and sensitize cells to irradiation or cytotoxic chemotherapy in preclinical breast cancer models and are under clinical investigation. We found that the loss of β1 integrins attenuated breast tumor growth but markedly enhanced tumor cell dissemination to the lungs. When cultured in three-dimensional ECM scaffolds, antibodies that blocked β1 integrin function or knockdown of β1 switched the migratory behavior of human and mouse E-cadherin–positive triple-negative breast cancer (TNBC) cells from collective to single cell movement. This switch involved activation of the transforming growth factor–β (TGFβ) signaling network that led to a shift in the balance between miR-200 microRNAs and the transcription factor zinc finger E-box–binding homeobox 2 (ZEB2), resulting in suppressed transcription of the gene encoding E-cadherin. Reducing the abundance of a TGFβ receptor, restoring the ZEB/miR-200 balance, or increasing the abundance of E-cadherin reestablished cohesion in β1 integrin–deficient cells and reduced dissemination to the lungs without affecting growth of the primary tumor. These findings reveal that β1 integrins control a signaling network that promotes an epithelial phenotype and suppresses dissemination and indicate that targeting β1 integrins may have undesirable effects in TNBC.


The cell-cell adhesion receptor E-cadherin is an established suppressor of cell invasion (1, 2). Yet, highly aggressive tumors, including infiltrating ductal breast carcinomas, metastasize in its apparent presence (3). In such cancers, a transient loss of E-cadherin during local epithelial-to-mesenchymal transitions (EMT) is believed to promote tumor dissemination (4). The reverse process, MET, may promote subsequent clonal outgrowth of disseminated tumor cells at metastatic sites (5). The balance between EMT and MET is controlled by transforming growth factor–β (TGFβ) signaling through a double-negative feed-forward interaction between the microRNA (miR)–200 family and zinc finger E-box–binding homeobox (ZEB) transcriptional repressors targeting CDH1 transcripts, encoding E-cadherin (68). Thus, understanding how alterations in this network are triggered in tumor cells is critical to improving strategies to treat or prevent metastatic disease.

Integrins, another class of adhesion receptors, mediate adhesion to extracellular matrix (ECM) proteins and promote cell survival and proliferation (9, 10). Hence, drugs that disrupt integrin-mediated adhesion of cancer cells or cells in the tumor microenvironment, including endothelial cells, are under clinical investigation (11, 12). Antibody blocking and gene deletion in mouse models suggest that members of the β1 subfamily of integrins represent therapeutic targets in breast cancer (1316). However, correlations of β1 integrin abundance with cancer progression or patient survival are ambiguous, and data obtained in some mouse models for breast or prostate cancer indicate that β1 integrins may, in fact, suppress cancer progression (13, 1719).

Integrins also control cell migration on two-dimensional (2D) ECM surfaces or within 3D ECM scaffolds (2023). Thus, integrin-mediated ECM contacts influence multiple aspects of tumor cell biology ranging from survival to proliferation to migration. This occurs through modulation of various signaling networks (2427) that are rewired by mutations that differ between cancer types and patients. Hence, the consequence of altered integrin expression profiles as observed with cancer progression (18) or the effect of integrin-targeted therapies may be highly context-dependent. Here, we find that loss of β1 integrins has remarkably distinct effects on tumor growth versus migration, dissemination, and metastasis in the context of E-cadherin–positive triple-negative breast cancer (TNBC), indicating that the use of β1 integrins as drug targets may need reconsideration.


Knockdown of β1 integrins in E-cadherin–positive TNBC attenuates tumor growth but triggers enhanced dissemination and lung metastasis

The role of β1 integrins in breast cancer growth and metastasis was investigated in a mouse orthotopic transplantation model, using triple-negative (lacking expression of genes encoding estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2), E-cadherin–positive 4T1 cells injected into the mammary fat pad. Knockdown of β1 integrins using two distinct short hairpin RNAs (shRNAs) in this model (4T1-shβ1) suppressed the outgrowth within the mammary gland compared with tumors derived from wild-type or control shRNA-transduced cells (Fig. 1, A and B). However, tumors in which β1 integrins were silenced showed a marked increase in the number of lung colonies, indicating that tumor growth and metastatic potential were inversely affected (Fig. 1C). Like the primary tumor, lung metastases that originated from 4T1-shβ1 tumors were smaller than those derived from control tumors (Fig. 1C). The number of circulating tumor cells produced by the small β1 integrin–depleted tumors detected 3 to 4 weeks after implantation was also significantly increased (Fig. 1D). As in mice, knockdown of β1 integrins increased the ability of 4T1 cells to migrate away from the primary tumor cell mass in xenografts (Fig. 1, E and F), in which tumor cells were injected into the yolk of zebrafish embryos and monitored for distance traveled throughout the embryo (28). Thus, a suppressive role of β1 integrins in tumor cell dissemination was observed in two in vivo models.

Fig. 1 Integrin control of breast cancer dissemination.

(A) Relative β1 integrin surface abundance by fluorescence-activated cell sorting (FACS) in 4T1 cells transduced with the indicated shRNA (n = 3 experiments; shc, control shRNA). MFU, mean fluorescence units. (B to D) Analysis of primary tumor growth (B), metastatic lung colonization (C), and circulating tumor cells (D) after orthotopic transplantation of shRNA-transduced 4T1 cells. Data are means ± SEM in more than 25 mice per condition from three independent experiments. ns, not significant. (E and F) Graphic representation (E) and analysis (F) derived from automated image analysis of dissemination of labeled, shRNA-transduced 4T1 cells in zebrafish embryos 5 days after intrayolk injection. Each color in (E) depicts tumor cells from a single embryo; data in (F) are the means ± SEM cumulative distance migrated by cells in about 40 zebrafish embryos per condition from two independent experiments. (G) β3 surface abundance by FACS in shRNA-transduced 4T1 cells in a representative of two independent experiments. (H) β3 surface abundance by FACS in cells expressing control vector (4T1) and human β3 complementary DNA (cDNA) (hu-ITGβ3) in a representative of two independent experiments. (I) Number of metastatic lung colonies after orthotopic transplantation of 4T1 cells or 4T1 cells transduced with β1 shRNA or with β3 cDNA. Data are means ± SD from eight mice per condition. *P < 0.05; **P < 0.01.

Compensatory increase in β3 integrins, induced by the loss of β1 integrin, does not explain enhanced metastatic potential

We and others have shown that genetic deletion of β1 integrins can induce the increased abundance of other integrin types, including β3 integrins (29, 30). A similar response was observed in 4T1 cells when the expression of β1 integrins was attenuated (Fig. 1G). Because αvβ3 integrin can play a critical role in the growth and progression of different cancer types (3033), we investigated whether this compensatory increase in αvβ3, rather than decreased β1 integrin abundance, caused enhanced metastasis. Attempts to generate stable β13 double-knockdown cells to test αvβ3 dependency failed because of an apparently strong negative selection against stable silencing of both subunits. However, in the converse experiment, overexpression of αvβ3 by itself did not lead to enhanced lung metastasis of 4T1 cells, indicating that the loss of β1 integrins, rather than the increase in αvβ3, was required for enhanced metastatic spread (Fig. 1, H and I).

Disruption of β1 integrin–mediated adhesion triggers a switch from collective invasion to individual cell migration in mouse and human E-cadherin–positive TNBC

On the basis of these in vivo results, we asked how the loss of β1 integrins affected cell migration strategies. We noticed that β1 integrin–depleted 4T1 cells lost cell-cell contacts and displayed a scattered morphology in 2D culture (Fig. 2A). This effect was reversible: Reexpression of β1 integrins fully restored cell-cell contacts and island formation (Fig. 2A). Tumor cell spheroids were generated in 3D collagen ECM scaffolds, as previously described, to assess cell migration in 3D culture (34). Under these conditions, 4T1 cells displayed collective invasion, with E-cadherin–positive cell-cell contacts remaining intact (Fig. 2B and movie S1).

Fig. 2 Integrin control of migration strategies.

(A) Microphotographs showing morphology in 2D culture of wild-type 4T1, 4T1-shβ1, or 4T1-shβ1 cells ectopically expressing human β1 in a representative of three experiments. (B) E-cadherin staining of 4T1 spheroids 4 days after injection into collagen gel in a representative of two experiments. Green, E-cadherin; blue, Hoechst; arrowhead, collective invading strands. (C) Representative images (one spheroid of at least four from at least three experiments each) of 4T1, BT20, MCF10A, and MDA-MB-435S cells expressing control (top) or β1 integrin shRNA (bottom) imaged 4 days after injection into collagen gels. Arrowheads, as in (A); arrows, individual cell migration. (D) F-actin staining (phalloidin) of 4T1 spheroids incubated with a control antibody [control immunoglobulin G (cIgG)] or an antibody against β11 Ab) 4 days after injection into collagen gel. Images are representative of one spheroid of three from two experiments per condition. (E) Relative β1 surface abundance by FACS in shRNA-transduced cells. Data are means ± SEM from three experiments (BT20) or are representative of two experiments (MCF10A and MDA-MB-435S). Scale bars, 50 μm (B to D).

Silencing β1 integrins in 4T1 cells blocked collective invasion and triggered a switch to effective single-cell migration (Fig. 2C, fig. S1, and movie S2). Treatment of 4T1 cells with a function-blocking antibody against β1 integrin also prevented collective invasion and promoted individual cell migration (Fig. 2D and fig. S1). Moreover, an identical switch after β1 integrin inhibition was observed in BT20 human E-cadherin–positive TNBC cells. Again, decreased abundance of β1 integrins induced a shift from invasive outgrowth to the migration of individual cells (Fig. 2, C and E, and fig. S1). Collective invasion of MCF10A nontumorigenic breast epithelial cells was also blocked in response to β1 integrin silencing, but these cells were unable to switch their migration strategy, and, instead, motility was arrested. MDA-MB-435S E-cadherin–negative TNBC cells already migrated as individual cells, and the inhibition of β1 integrins in these cells had little effect, other than inducing a more rounded morphology (Fig. 2, C and E), which is in agreement with other reports indicating that mesenchymal, but not amoeboid, individual cell movement depends on integrin-mediated ECM attachments (21). Together, these findings indicate that E-cadherin–positive TNBC cells can effectively bypass the inhibition of collective invasion when ECM contacts through β1 integrins are lost by switching their migration strategy to individual cell migration.

Knockdown of β1 integrins triggers a loss of E-cadherin, which is critical for the prometastatic switch in cell migration

To analyze the signal rewiring that is induced by the loss of β1 integrin, we compared gene expression profiles of two independent shβ1 lines with wild-type and shRNA-treated control 4T1 cells. With a false discovery rate (FDR) <0.001 and 1.5-fold difference as a cutoff, 1230 differentially expressed genes were shared between both shβ1 lines (table S1). In this set, ingenuity pathway analysis predicted “cellular movement” as the process most significantly affected by β1 integrin silencing and also predicted “cell-to-cell signaling.” These processes showed a significant decrease in the expression of the Cdh1 gene, encoding E-cadherin, in the shβ1-treated lines (fig. S2).

Quantitative polymerase chain reaction (qPCR) analysis confirmed the decrease in Cdh1 transcript abundance, and, in agreement, about 75% less E-cadherin was detected on the cell surface of 4T1-shβ1 cells (Fig. 3, A and B). Likewise, silencing β1 integrins in BT20 cells strongly suppressed the abundance of E-cadherin mRNA and cell surface protein (Fig. 3, A and B). Silencing β1 integrins also diminished total E-cadherin protein abundance in 4T1 and BT20 cells (Fig. 3C). The connection between β1 integrins and E-cadherin abundance was further confirmed in two other human TNBC cell lines, MDA-MB-468 and HCC1806 (Fig. 3D). Decreased E-cadherin abundance in response to the knockdown of β1 integrins was also observed in vivo: E-cadherin staining in cell-cell contacts in primary tumors derived from 4T1-shβ1 cells was markedly reduced as compared to that in control tumors (Fig. 3E). Finally, in agreement with reversibility of the cell scattering phenotype (Fig. 2A), reexpression of β1 integrins restored the surface abundance of E-cadherin in 4T1-shβ1 cells (Fig. 3F).

Fig. 3 Reversible decrease in E-cadherin expression in response to β1 integrin depletion.

(A to C) Cdh1 mRNA by qPCR (A), surface abundance by FACS (B), and total protein abundance by Western blot (C) in shRNA-transduced 4T1 and BT20 cells. Data are means ± SEM (*P < 0.05; **P < 0.001; ***P < 0.0001) or a representative blot from at least three experiments. (D) FACS analysis of E-cadherin surface abundance in response to β1 integrin knockdown in MDA-MB-468 and HCC1806 cells. Data are representative of two experiments. (E) E-cadherin staining in 4T1 and 4T1-shβ1 orthotopic breast tumors. Arrowheads, skin (a positive control in both images). Images are from 1 of 10 tumors per condition. Scale bar, 50 μm. (F) Surface abundance of mouse or human β1 and mouse E-cadherin in 4T1, 4T1-shβ1, or 4T1-shβ1 cells ectopically expressing human β1.

We next tested if the reduction in E-cadherin abundance was critically involved in the prometastatic switch in cell migration strategy triggered by β1 integrin loss. In support of this, ectopic expression of E-cadherin in 4T1-shβ1 cells at a similar surface abundance as that in controls restored cell-cell adhesion and collective invasion (Fig. 4, A and B, and fig. S1). Ectopic E-cadherin expression also blocked the enhanced ability of 4T1-shβ1 tumors to metastasize to the lungs (Fig. 4C). Conversely, silencing E-cadherin in wild-type 4T1 cells (4T1-shCdh1) decreased cell cohesion and induced the migration of single, elongated cells into collagen (Fig. 4, A and B). However, in vivo, the number of lung colonies produced by 4T1-shCdh1 cells was similar to that observed for wild-type 4T1 cells (Fig. 4C). These data indicate that the reduction in E-cadherin is essential for the prometastatic switch in the cell migration strategy triggered by β1 integrin inhibition, but that it is not sufficient by itself to increase metastatic spread. Moreover, tumor growth was not affected by E-cadherin knockdown and was suppressed in response to β1 integrin silencing, irrespective of the absence or presence of E-cadherin, further demonstrating that β1 integrins control tumor growth and metastasis through separate pathways (Fig. 4D).

Fig. 4 Suppression of E-cadherin drives a migration switch and is required but not sufficient for enhanced metastasis after integrin depletion.

(A) E-cadherin surface abundance by FACS in 4T1 cells expressing E-cadherin (Cdh1) shRNA or integrin β1 shRNA in the absence or presence of E-cadherin (CDH1) cDNA. Data are representative of two experiments. (B) Migration patterns of 4T1 cells expressing indicated shRNA or cDNA expression vectors (or both) 4 days after injection into collagen gels. Arrows, individual cell migration; arrowheads, collective invasion strands. Images are from one spheroid of at least four from two experiments per condition. Scale bar, 50 μm. (C and D) Analysis of spontaneous lung metastasis (C) and primary tumor growth (D) of orthotopically transplanted 4T1 cells expressing indicated shRNAs and cDNAs. Data are means ± SEM from 20 mice per condition from two independent experiments. **P < 0.01.

Enhanced ZEB2 expression in response to knockdown of β1 integrins blocks Cdh1 transcription and causes enhanced metastasis

Having established a connection between β1 integrins and E-cadherin abundance that critically affects metastatic behavior, we investigated the mechanism by which loss of β1 integrin–mediated ECM attachment triggers a block in Cdh1 expression. Luciferase reporter assays showed that β1 integrin silencing decreased transcription from the Cdh1 gene by about 80% (Fig. 5A). This prompted us to investigate the β1 integrin–associated regulation of a group of E-cadherin transcriptional repressors, including members of the Snail, basic helix-loop-helix (bHLH), and zinc finger homeodomain (ZFH) families that are implicated in EMT (4). Analysis of the microarray data showed that of these repressors, only the expression of the gene encoding ZEB2 (also known as Sip1) was significantly increased in both 4T1-shβ1 lines (fig. S2). qPCR analysis confirmed the induction of ZEB2 expression upon β1 integrin silencing in 4T1 cells and showed a similar, albeit more marked, induction of ZEB2 in BT20 cells (Fig. 5, B and C).

Fig. 5 Increased ZEB2 expression suppresses Cdh1 expression and metastatic potential in response to β1 integrin knockdown.

(A) Luciferase reporter assay for Cdh1 promoter activity in shRNA-transduced 4T1 cells. Data are means ± SEM from three experiments. (B and C) Quantification of ZEB1 and ZEB2 mRNA by qPCR in shRNA-transduced 4T1 (B) or BT20 (C) cells. Image shows ZEB2 and actin (control) PCR products loaded on gel. Data in (A) to (C) are means ± SEM from at least three experiments. (D) ITGβ1, Zeb2, and Cdh1 expression by qPCR in 4T1 cells expressing control shRNA or shβ1 in the absence or presence of shZeb2. Data are means ± SEM from at least three experiments. (E and F) Analysis of primary tumor growth (E) and spontaneous lung metastasis (F) of orthotopically transplanted shRNA-transduced 4T1 cells. Data are means ± SEM from more than 16 mice per condition from two independent experiments. *P < 0.05; **P < 0.01.

To test if the increase in ZEB2 expression was implicated in blocking Cdh1 transcription and enhancing metastatic capacity triggered by β1 integrin loss, we generated β1 integrin and ZEB2 double-knockdown 4T1 cells using each of three shRNAs against ZEB2 (Fig. 5D). All three independent shRNA lines showed that reverting the enhanced ZEB2 expression in β1 integrin–depleted cells restored Cdh1 expression to the same degree or higher than that observed in wild-type cells (Fig. 5D). As observed with E-cadherin reexpression (Fig. 4D), silencing ZEB2 did not affect the reduced growth of 4T1-shβ1 tumors (Fig. 5E). However, again in good agreement with the E-cadherin rescue experiment (Fig. 4C), two independent ZEB2 shRNAs blocked the enhanced metastasis of β1 integrin–depleted cells (Fig. 5F). Together, these data indicate that the transcriptional repression of Cdh1 induced by β1 integrin loss is mediated by activation of ZEB2, and further support that loss of β1 integrins affects tumor growth and metastasis through different pathways.

Alterations in the balance between ZEB and miR-200 after the loss of β1 integrins cause E-cadherin suppression and a switch from collective invasion to single-cell migration

ZEB transcription factors and miR-200 miRNAs establish a negative feed-forward loop that is implicated in EMT, and alterations in the balance between ZEB and miR-200 may underlie progression of a number of different types of cancer, including breast carcinomas (6, 7). We assessed how β1 integrin silencing affected miR-200 miRNAs. A strong decrease in the abundance of all five members of the miR-200 family (miR-200a, miR-200b, miR-200c, miR-141, and miR-429) was observed in 4T1, as well as BT20 cells after β1 integrin depletion (Fig. 6, A and B). To explore the role of alterations in the ZEB/miR-200 balance in the observed loss of E-cadherin and the altered migration strategy triggered by loss of β1 integrins, we used synthetic and lentiviral expression systems to reexpress each of the miR-200 miRNAs. Expression of any individual miR-200 family member was sufficient to restore cell-cell adhesion in 2D cultures of 4T1-shβ1 cells and to induce a reversal from single-cell movement back to collective invasion in 3D culture (Fig. 6C and figs. S1 and S3, A to D). These data point to overlapping functions of the miR-200 family members in this system and demonstrate that the observed decrease of all five miR-200 family members may be required for the inhibition of cell-cell adhesion upon depletion of β1 integrins. Notably, there appeared to be a selection against the expression of miR-200 in the context of β1 integrin loss: No long-term, stable cell lines could be generated where the knockdown of β1 integrins was combined with high expression of miR-200 family members. Restored cell-cell adhesion in 2D and 3D cultures upon expression of miR-200 family members in 4T1-shβ1 cells was also accompanied by decreased expression of ZEB2, concomitant with increased expression of Cdh1 (Fig. 6D). Together, these findings demonstrate that the loss of β1 integrins triggers a shift in the balance between ZEB and miR-200 family members, which induces a decrease in E-cadherin and a prometastatic switch from collective invasion to single-cell migration.

Fig. 6 Decreased miR-200 abundance in response to β1 integrin knockdown controls Zeb2 and Cdh1 expression and cell migration strategy.

(A and B) Relative miRNA amounts in shRNA-transduced 4T1 (A) and BT20 cells (B) using qPCR Exiqon primers. Data are means ± SEM from at least three experiments. Gels show miR-200 abundance by PCR. (C) 4T1 cohesion after β1 integrin knockdown without or with synthetic miR-200C MIMIC expression in 2D culture (left) and in 3D collagen gels (right). Images are representative of at least four from at least three experiments per condition. Scale bars, 50 μm. (D) Cdh1, Zeb2, and Zeb1 mRNA amounts in 4T1-shβ1 cells cotransfected with synthetic miRNA Mimics. Data are means ± SEM from at least three experiments. *P < 0.05; **P < 0.01.

Alterations in the TGFβ signaling network drive the prometastatic switch induced by β1 integrin down-regulation

Increased TGFβ signaling has been implicated in the dissemination of subsets of individually migrating cells from breast tumors (35). Autocrine TGFβ signaling can stabilize an EMT phenotype through the repression of miR-200 (8), but it is not known what may trigger local TGFβ signaling to induce EMT-like changes and dissemination. The findings described thus far suggested that disruption of β1 integrin–mediated cell adhesion might represent one such trigger. Indeed, examination of the 1230 differentially expressed genes that were shared between both 4T1-shβ1 lines identified alterations in genes encoding proteins in the TGFβ signaling or the related bone morphogenetic protein (BMP) signaling pathway (table S1 and Fig. 7A). qPCR analysis verified these events and demonstrated that several of the changes (such as increased expression of the gene encoding Smad6 or decreased expression of genes encoding inhibin and BMP7) were similarly observed in BT20 cells after β1 integrin knockdown (Fig. 7B). We used immunostaining of phosphorylated Smad2 and 3 (p-Smad2/3) to assess basal TGFβ signaling. Nuclear staining of p-Smad2/3 was significantly increased in response to β1 integrin knockdown, indicating increased TGFβ signaling in these cells (Fig. 7C).

Fig. 7 Loss of β1 integrins alters TGFβ/BMP signaling, which mediates decreased E-cadherin expression and enhanced metastasis.

(A) Cartoon depicting TGFβ/BMP signaling pathway proteins encoded by differentially expressed genes identified in the microarray analysis (table S1). Blue, proteins encoded by genes with increased expression in response to β1 integrin silencing; red, proteins encoded by genes with deceased expression. (B) Inhibin, Smad6, or BMP7 transcript amounts by qPCR in shRNA-transduced 4T1 and BT20 cells. Data are means ± SEM from at least three experiments. (C) p-Smad2/3 immunostaining in 4T1 and 4T1-shβ1 cells and quantification (means ± SEM) of relative nuclear localization by automated image analysis of 11 images, each containing more than 15 cells. (D) ITGβ1, TGFβR3, and Cdh1 expression by qPCR in shRNA-transduced 4T1 cells as indicated. Data are means ± SEM from at least three experiments. (E) 2D multicellular island formation in response to TGFβR3 knockdown in 4T1-shβ1 cells. Data are representative of three experiments. (F and G) Analysis of primary tumor growth (F) and metastatic lung colonization (G) after orthotopic transplantation of shRNA-transduced 4T1 cells as indicated. Data are means ± SEM from at least 10 mice per condition from two independent experiments. Scale bars, 50 μm (C and E). *P < 0.05, **P < 0.01.

We next tested if such increased TGFβ signaling could explain the decreased Cdh1 transcription and the enhanced metastatic capacity triggered by β1 integrin loss. For this purpose, the enhanced expression of the gene encoding TGFβ receptor 3 (TGFβR3; also known as betaglycan) identified by gene expression profiling and verified by qPCR analysis (table S1 and Fig. 7D) was blocked by targeted shRNA. Two independent TGFβR3 shRNAs restored Cdh1 expression in 4T1-shβ1 cells (Fig. 7D). This was accompanied by a restored capacity of the cells to form islands in 2D culture (Fig. 7E). Moreover, interfering with the enhanced abundance of TGFβR3 also prevented the enhanced lung metastasis seen in β1 integrin–depleted cells, whereas tumor growth remained suppressed (Fig. 7, F and G). Thus, interfering with alterations in the TGFβ/BMP signaling network at the level of TGFβR3 phenocopied the effects of ZEB2 inhibition or E-cadherin reexpression by restoring epithelial characteristics and mitigating enhanced metastasis in β1 integrin–depleted cells. Together, these results indicate that disruption of β1 integrin–mediated cell adhesion can trigger alterations in the TGFβ signaling network that drive EMT-like changes and increased dissemination in TNBC.


This study demonstrates that β1 integrin–mediated cell-ECM interactions support orthotopic breast cancer growth but that inhibition of β1 integrins can trigger a rewiring of signaling pathways that lead to enhanced metastatic spread in TNBC (Fig. 8). The role of β1 integrin–mediated ECM attachments in tumor growth agrees with earlier studies (10, 13, 14). Moreover, the smaller size of lung metastases produced from the β1 integrin–depleted tumors indicates that, besides primary tumor growth, β1 integrins also support outgrowth of metastatic colonies in the lung. This confirms and extends earlier reports where outgrowth of breast cancer cells arrested in the lung after intravenous injection was also shown to require integrin signaling (36, 37).

Fig. 8 Cartoon depicting the consequences of disrupting β1 integrins for breast tumor growth and metastasis.

Blocking or depleting β1 integrins in various studies, including here, reduces growth-promoting signaling (1). It also leads to enhanced abundance of αvβ3, which provides compensatory growth signals (2). The result of these changes is deceased primary tumor growth and metastatic outgrowth. Our study shows that in E-cadherin–positive TNBC cells, blocking or silencing β1 integrins induces a switch from collective to single cell migration. β1 integrin knockdown causes enhanced TGFβ signaling and a consequential shift in the balance between miR-200 and ZEB2 that leads to the loss of E-cadherin (3). Increased αvβ3 expression may support this pathway by increasing TGFβR signaling. The result of these changes is enhanced dissemination and lung colonization. Additionally, currently unidentified changes acting in parallel to E-cadherin loss further stimulate lung colonization (4). Thus, inhibition of β1 integrins in E-cadherin–positive TNBC can have opposite effects on tumor growth and dissemination. Dotted lines indicate relationships that are hypothesized, not proven experimentally.

Preclinical studies have implicated β1 integrins in cancer initiation, growth, and progression, but correlations of β1 integrin abundance with overall or metastasis-free patient survival are ambiguous, and in some mouse models for breast or prostate cancer, β1 integrins, in fact, suppress cancer progression (1014, 1719). Here, we show that β1 integrin depletion can cause increased metastatic spread that is related to a marked switch in migration strategy of TNBC cells. The effect on cell migration is context-dependent. E-cadherin–negative cells that move as single cells are mildly affected by the disruption of ECM attachments, whereas tumor cell types that move as collective strands may either be arrested or undergo a prometastatic switch toward individual cell movement. Invasion of T4-2 cells derived from HMT-3522 breast epithelial cells in Matrigel (38) or invasion of MCF10A cells in collagen (Fig. 2C) is prevented when β1 integrins are functionally blocked or knocked down. However, mouse and human E-cadherin–positive TNBC cells have sufficient plasticity to respond to β1 integrin inhibition by switching to an alternative migration strategy through loss of cohesion, resulting in enhanced dissemination and lung colonization.

Whereas E-cadherin is almost invariably lost in invasive lobular breast carcinomas, its expression is retained in many other types including the common ductal invasive carcinomas (3). In those cases, changes in β1 integrin abundance in a population of tumor cells may drive a local transient EMT-like switch. It remains to be studied whether disruption of ECM interactions, for instance, through altered proteolytic ECM degradation, can have similar consequences, but this is supported by our finding that β1 antibodies can induce a similar migration switch. In agreement, previous work by us and others has shown that conditions that disturb ECM adhesion in 2D culture (39) or in 3D ECM matrices of low rigidity (34) can disrupt cell-cell contacts.

Integrin-mediated ECM adhesion can modulate cell-cell adhesion; both positive and negative regulations have been reported, but crosstalk at the level of E-cadherin (CDH1) expression has not been previously demonstrated (29, 3841). EMT and loss of E-cadherin have been implicated in initiating the outgrowth of dormant metastatic breast cancer lesions through activation of pro-proliferative β1 integrin–mediated signaling (42). Conversely, although EMT may drive early steps of the metastatic cascade, the inverse process, MET is also shown to be critical for outgrowth of breast cancer metastases (5). Our findings indicate that suppressed outgrowth of β1 integrin–depleted breast tumors or metastatic lesions was unrelated to the expression of E-cadherin and consequent cell cohesion. On the other hand, the switch from collective invasion to individual cell migration and enhanced dissemination to the lungs was fully dependent on the loss of E-cadherin. Thus, β1 integrins control migration and growth during breast cancer progression through separate pathways.

TGFβ is a key inducer of EMT, and autocrine TGFβ signaling can stabilize an EMT phenotype by repressing miR-200 (8). A transient increase in TGFβ signaling has also been associated with the propensity of a subset of individually migrating cells in breast cancers to disseminate (35). Our findings indicate that loss or disruption of β1 integrin–mediated ECM adhesion may serve as one trigger for such local increased TGFβ signaling, leading to EMT and enhanced dissemination. Certain integrins can activate latent TGFβ, and effects of integrins on TGFβR expression have been reported (43). Another study shows that β3 integrins stimulate TGFβ responsiveness in 4T1 cells after inactivation of β1 integrins and are required for primary tumor growth and total lung tumor burden when β1 integrins were knocked down (44). Our observation of a strong negative selection against β13 co-silencing, precluding us to directly test αvβ3 dependency in our model, is in line with the notion that an increase in αvβ3 can compensate for the loss of β1 to promote proliferation and tumor growth. In addition to its apparently critical role in stimulating growth, αvβ3 can stimulate TGFβ signaling through its ability to bind the RGD (Arg-Gly-Asp) motif in the TGFβ latency-associated protein, LAP (43), and may well be involved in the prometastatic TGFβ signaling–associated cascade triggered by the loss of β1 integrin. However, enhanced expression of αvβ3 was not sufficient to drive this switch, thus placing responsibility with the suppressive activity of β1 integrins.

Our findings demonstrate that the inhibition of β1 integrins can trigger extensive rewiring of the TGFβ-controlled ZEB/miR-200 signaling network. The affected genes within this network are, to a large extent, shared between mouse 4T1 and human BT20 cells. We showed that restoring β1 integrin expression or preventing the induction of ZEB2 or TGFβR3 reinstates Cdh1 expression and cell cohesion. Each of these strategies was also sufficient to block the increased dissemination capacity. ZEB2 forms a transcriptional control circuit with miR-200 members in breast cancer that regulates EMT and cancer stem cell characteristics (7, 8, 45). The role of TGFβR3 in breast cancer appears to be highly dependent on other components of the network: TGFβR3 can either promote or interfere with TGFβ and BMP signaling, for example, membrane-bound TGFβR3 can activate TGFβR1 and TGFβR2 through ligand presentation, whereas cleaved soluble TGFβR3 can sequester TGFβ (46). Moreover, TGFβR3 can mediate TGFβ-induced EMT in some cases, but it is suppressed during TGFβ-induced EMT in others (47).

Together, our study demonstrates that the inhibition of β1 integrins can rewire the TGFβ/ZEB/miR-200 signaling network, driving loss of cohesion and increased dissemination of E-cadherin–positive TNBC cells. This process is fully reversible, and local alterations in β1 integrin abundance or ECM-integrin crosstalk might represent one mechanism for a transient, TGFβ-mediated EMT, enabling a subset of E-cadherin–positive breast cancer cells to disseminate to distant organs. These findings also raise concerns with respect to the use of β1 integrins as drug targets to sensitize tumors to radio- or chemotherapy. Although tumor shrinkage is achieved and the growth of metastatic colonies is attenuated, in certain breast cancer types (E-cadherin–positive triple-negative variants), reprogramming of the surviving cells may aggravate metastatic spread.


Cell lines and animals

4T1 mouse breast cancer cells, MCF10A human mammary epithelial cells, and BT20, MDA-MB-435S, MDA-MB-468, and HCC1806 human breast cancer cells were obtained from the American Type Culture Collection and cultured according to the provided protocol. Rag2−/−c−/− mice were housed in individually ventilated cages under sterile conditions. Housing and experiments were performed according to the Dutch guidelines for the care and use of laboratory animals. Sterilized food and water were provided ad libitum. Zebrafish were maintained according to standard protocols ( Embryos were grown at 28.5° to 30°C in egg water (Instant Ocean salts, 60 μg/ml). During injection with tumor cells, embryos were kept under anesthesia in 0.02% buffered 3-aminobenzoic acid ethyl ester (tricaine; Sigma).


For FACS, primary antibodies were those against mouse integrin β1 (HMβ1-1; BD Pharmingen), human integrin β1 (AIIB2; BD Transduction Laboratories), human E-cadherin (36/E-cadh; BD Transduction Laboratories), and mouse E-cadherin (DECMA; Sigma-Aldrich); secondary antibodies were allophycocyanin (APC)–conjugated anti-rat (Jackson ImmunoResearch Laboratories Inc.) and phycoerythrin (PE)–conjugated anti-mouse (Jackson ImmunoResearch Laboratories Inc.). For Western blotting, primary antibodies were those against mouse integrin β1 and mouse or human E-cadherin (36/E-cadh; BD Transduction Laboratories), and α-tubulin (B-5-1-2; Sigma); secondary antibodies were horseradish peroxidase (HRP)–linked rabbit, rat, or mouse IgG (Jackson ImmunoResearch Laboratories Inc.). An antibody for integrin β1 (HA2/5; BD Pharmingen) was used in function blocking studies. For immunohistochemistry, primary antibodies were those against mouse or human E-cadherin as for Western blotting and p-Smad2/3 (Cell Signaling Technology); secondary antibodies were Alexa 488–linked mouse IgG (Invitrogen) and Alexa 546–linked rabbit IgG (Invitrogen). F-actin was stained with Alexa 633 phalloidin (Molecular Probes).

shRNA, cDNA, and miRNA transfection

For stable gene silencing, cells were transduced using lentiviral shRNA vectors (LentiExpress; Sigma-Aldrich) according to the manufacturer’s procedures and selected for in medium containing puromycin (2 μg/ml). Control vectors included shRNA targeting TurboGFP (shc#1) and shRNA targeting enhanced green fluorescent protein (eGFP) (shc#2). Targeting sequences for shRNAs are provided in table S2. For stable expression of human β1 and β3 integrin subunits, cells were transduced with LZRS.MS.neo-β1 retroviral expression vector (29) or with pPT.PGK-β3 lentiviral cDNA expression vector (provided by D. Novack, Washington University, St. Louis, MO) (48), respectively. For the expression of mouse E-cadherin, cells were transduced with pCSCG/mECAD lentiviral cDNA expression vector (provided by P. Derksen, University Medical Center, Utrecht, the Netherlands). For the expression of miR-200 miRNAs, cells were transduced using miRIDIAN shMIMIC lentiviral miRNAs (nontargeting control, miR-200a, miR-200b, miR-200c, miR-141, and miR-429, provided by A. Amiet, Thermo Fisher Scientific) according to the manufacturer’s protocol. In all cases, sorting of cells after antibiotic selection occurred by two rounds of bulk sorting for eGFP (miRNAs) or reduction/increase in cell surface expression (integrins, E-cadherin shRNA/cDNA).

For transient transfection of synthetic miRIDIAN miRNA Mimics (provided by A. Amiet, Thermo Fisher Scientific; control nontargeting, miR-200a, miR-200b, miR-200c, miR-141, and miR-429), cells were seeded at 5 × 105 cells per well in 12-well plates and transfected at a final concentration of 50 nM using DharmaFECT2 (Thermo Fisher Scientific). Cells were replated 24 hours after transfection and used 48 hours later for E-cadherin FACS, qPCR analysis, or collagen invasion.

Luciferase reporter assay

4T1 wild-type and 4T1-shβ1 cells were transiently transfected with 10 ng of an E-cadherin firefly luciferase reporter plasmid (provided by G. Berx, VIB, Gent, Belgium) (49) and 2 ng of a cytomegalovirus-Renilla luciferase reporter using Lipofectamine PLUS (Invitrogen) and analyzed using the Dual Luciferase Reporter Assay System (Promega) 3 days later, according to the manufacturer’s procedures.

3D invasion assays

Tumor cell suspensions were injected into collagen type I gels, resulting in cell spheroids overnight as described (34). Spheroids were monitored for 1 week. The number of single cells around spheroids at 3 days (4T1) or at 7 days (BT20) after injection was quantified using ImageJ. Edge detection and bandpass filter were used to remove background and noise, or background was subtracted from the image using a rolling ball filter. Subsequently, a mask was generated including collective invasion area, and single cells outside the mask were analyzed using automated thresholding for intensity and object size. Identification of single cells was verified manually for each image.

For immunostaining 4 days after injection, gels were incubated for 30 min with collagenase (5 μg/ml) (from Clostridium histolyticum; Boehringer Mannheim) at room temperature, fixed with 4% paraformaldehyde, permeabilized in 0.2% Triton X-100, and blocked with 10% fetal bovine serum. Gels were incubated with rhodamine-conjugated phalloidin or with E-cadherin antibody followed by Alexa 488–conjugated secondary antibody and Hoechst nuclear staining. Preparations were mounted in Aqua-Poly/Mount solution (Polysciences Inc.) and analyzed using a Nikon TE2000 confocal microscope. Z-stacks (50 × 1 μm) were obtained using a 20× dry objective and converted into a single Z projection using the “extended depth of field” plug-in from ImageJ software.

For antibody blocking studies, cells were injected in collagen gels containing control IgG (10 μg/ml) or HA2/5 β1 integrin antibody and overlaid with medium containing the same antibody concentrations.

For real-time imaging, about 3-hour time-lapse movies of spheroids were obtained starting at 48 hours after injection. Image acquisition was performed with a Nikon TE2000 confocal microscope with a temperature- and CO2-controlled incubator. Differential interference contrast time-lapse videos were recorded using a charged-coupled device camera controlled by NIS Elements software. Images were converted into a single AVI (audio video interleave) file in Image-Pro Plus (v5.1; Media Cybernetics).

Mouse orthotopic transplantation experiments

Tumor cells (1 × 105) in 0.1 ml of phosphate-buffered saline were injected into the fat pad of 8- to 12-week-old female Rag2−/−c−/− mice. Size of the primary tumors was measured using calipers. Horizontal (h) and vertical (v) diameters were determined, and tumor volume (V) was calculated: V = 4/3π(1/2[√(h × v)]3). After 3 to 4 weeks, animals were anesthetized with pentobarbital, and primary tumor and lungs were excised. Primary tumor and left lung were divided into two pieces that were snap-frozen in liquid nitrogen for E-cadherin immunostaining or fixed in 4% paraformaldehyde for hematoxylin and eosin staining. To quantify lung metastases, right lungs were injected with ink solution, destained in water, and fixed in Feketes [4.3% (v/v) acetic acid, 0.35% (v/v) formaldehyde in 70% ethanol]. To analyze circulating tumor cells, blood was drawn from the right atrium of some mice by heart puncture after anesthetization but before excision of primary tumor and lungs. Blood (0.2 ml) was plated into 60-mm tissue culture dishes filled with growth medium. After 5 days, tumor cell clones were stained with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Sigma) and counted with ImageJ.

Zebrafish xenotransplantation experiments

CM-DiI–labeled tumor cells were transplanted into the yolk of dechorionated Fli-GFP transgenic zebrafish embryos, and after 6 days of incubation at 34°C, embryos were fixed and tumor cell dissemination was quantified by automated confocal imaging in 96-well plates followed by automated image analysis of Z-stacks as described (28).

mRNA and miRNA analyses

For qPCR analysis of mRNA, RNA was extracted using TRIzol (Invitrogen) and cDNA was randomly primed from 50 ng of total RNA using iScript cDNA Synthesis Kit (Bio-Rad). For detection of miRNAs, RNA was isolated using miRCURY RNA Isolation Kit (Cell & Plant), and cDNA was made using the Universal cDNA Synthesis Kit (Exiqon). Real-time qPCR was performed in triplicate using SYBR Green PCR (Applied Biosystems) on a 7900HT Fast Real-Time PCR System (Applied Biosystems). qPCR primers for mRNA are given in table S3; specific primers for miRNAs were obtained from Exiqon. qPCR data were collected and analyzed using SDS2.3 software (Applied Biosystems). Relative mRNA amounts after correction for β-actin control mRNA were expressed using 2−ΔΔCt method.

For microarrays, total RNA was extracted using mirVana RNA Isolation Kit (Ambion Inc.). RNA quality and integrity were assessed with the 2100 Bioanalyzer system (Agilent Technologies). The Affymetrix 3′ IVT Express Labeling Kit was used to synthesize biotin-labeled complementary RNA, which was then hybridized to an Affymetrix MG-430 PM Array Plate. Data quality control was performed with Affymetrix Expression Console v1.1, and all raw data passed the Affymetrix quality criteria. Median normalization of raw expression data and identification of differentially expressed genes using a random-variance t test were performed with BRB-ArrayTools Version 4.1.0 Beta 2 Release (developed by R. Simon and BRB-ArrayTools Development Team members; Annotation was done according to the NetAffx annotation date release 2009-11-23. Corrections for multiple testing were performed as described by calculating the FDRs.

Western blot, flow cytometry, and immunohistochemistry

For Western blot, cells were lysed with modified radioimmunoprecipitation assay buffer [150 mM NaCl, 1.0% Triton X-100, 0.5% Na deoxycholate, 0.1% 50 mM tris (pH 8), and protease cocktail inhibitor (Sigma-Aldrich)]. Samples were separated by SDS–polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes (Millipore), incubated with primary antibodies then HRP-labeled secondary antibodies (Jackson ImmunoResearch Laboratories Inc.), and developed with enhanced chemiluminescence substrate mixture (ECL Plus, Amersham, GE Healthcare). Blots were scanned on a Typhoon 9400 (GE Healthcare).

For FACS, cells were detached either with trypsin/EDTA (in the case of integrin surface expression) or with 0.02% EDTA only (in the case of E-cadherin surface expression). Surface expression levels were determined using primary antibodies, followed by fluorescence-conjugated secondary antibodies, and analysis on a FACSCanto or sorting on a FACSCalibur (Becton Dickinson).

For immunohistochemistry, cells were fixed in 4% paraformaldehyde, permeabilized in 0.2% Triton X-100, blocked with 2% BSA, and incubated with indicated antibodies followed by fluorescently labeled secondary antibodies in combination with Hoechst nuclear staining. Preparations were mounted in Aqua-Poly/Mount solution (Polysciences Inc.) and analyzed using a Nikon TE2000 confocal microscope.

Statistical analysis

Data are presented as means ± SEM of at least three independent biological replicates unless otherwise stated. Student’s t test (two-tailed) was used to compare groups.


Fig. S1. Quantification of single cell migration in 3D collagen gels.

Fig. S2. Pathways affected by β1 integrin knockdown from microarray data.

Fig. S3. Restored cohesion in 4T1-shβ1 cells after miR-200 expression.

Table S1. Genes affected by β1 integrin knockdown from microarray data.

Table S2. shRNA sequences.

Table S3. qPCR primer sequences.

Movie S1. 4T1 spheroid collectively invading into 3D collagen matrix.

Movie S2. 4T1-shβ1 spheroid with individual cells migrating in 3D collagen matrix.


Acknowledgments: We thank D. Novack for providing the integrin β3 expression vector; P. Derksen for sharing the E-cadherin expression vector; G. Berx for providing the E-cadherin luciferase reporter; P. ten Dijke for help with p-Smad2/3 immunofluorescence; A. Amiet for miR-200 expression constructs; and W. van Roosmalen, R. Lalai, and C. Pont for assistance with mouse experiments. Funding: Support for this work came from the Dutch Cancer Society (UL-2010-4670), the Netherlands Organization for Scientific Research (FOM 09MMC03), and EU FP7 (HEALTH-F2-2008-201439). Author contributions: H.H.T., J.X., E.N., L.H., S.E.L.D., H.E.B., and E.H.J.D. designed and performed all the in vitro and mouse experiments. V.P.S.G., S.H., and B.E.S.-J. designed and performed the zebrafish experiments. E.V., J.H.N.M., B.v.d.W., and E.H.J.D. analyzed and interpreted the data. H.H.T., J.X., and E.H.J.D. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Microarray data are deposited at ArrayExpress (accession number E-MTAB-2185).
View Abstract

Stay Connected to Science Signaling

Navigate This Article