PTEN suppresses epithelial–mesenchymal transition and cancer stem cell activity by downregulating Abi1

The epithelial–mesenchymal transition (EMT) is an embryonic program frequently reactivated during cancer progression and is implicated in cancer invasion and metastasis. Cancer cells can also acquire stem cell properties to self-renew and give rise to new tumors through the EMT. Inactivation of the tumor suppressor PTEN has been shown to induce the EMT, but the underlying molecular mechanisms are less understood. In this study, we reconstituted PTEN-deficient breast cancer cells with wild-type and mutant PTEN, demonstrating that restoration of PTEN expression converted cancer cells with mesenchymal traits to an epithelial phenotype and inhibited cancer stem cell (CSC) activity. The protein rather than the lipid phosphatase activity of PTEN accounts for the reversal of the EMT. PTEN dephosphorylates and downregulates Abi1 in breast cancer cells. Gain- and loss-of-function analysis indicates that upregulation of Abi1 mediates PTEN loss-induced EMT and CSC activity. These results suggest that PTEN may suppress breast cancer invasion and metastasis via dephosphorylating and downregulating Abi1.


Results
pten expression correlates with an epithelial phenotype of breast cancer cells isolated from primary ductal carcinomas. We and others have found that deletion of the Pten gene in mouse embryonic stem (ES) cells prevents their differentiation into polarized epiblast epithelial cells in embryoid bodies. Ablation of PTEN also limits the contribution of the mutant ES cells to tissues derived from the three germ layers in chimeric mice 43,45 . To determine whether PTEN is required for the maintenance of epithelial characteristics in breast cancer cells, we analyzed the phenotype of PTEN-positive BT474 and PTEN-negative BT549 human breast cancer cells, both of which were derived from primary ductal carcinomas 46,47 . BT474 cells are wild-type for PTEN and displayed an epithelial morphology (Fig. 1A). They expressed the epithelial marker E-cadherin, but not the mesenchymal marker vimentin (Fig. 1B). By contrast, BT549 cells have homozygous truncating mutation of PTEN (premature termination at the codon of 274), which resulted in the loss of the PTEN protein 48 . These cells assumed a fibroblast shape and expressed vimentin but not E-cadherin. In addition, they also expressed higher levels of c-Myc, an oncogene that reprograms cellular metabolism to promote cancer development 49 . RT-PCR analysis revealed higher mRNA levels of the EMT-inducing transcription factors Snail1, Slug, ZEB1 and Twist2 in BT549 cells (Figs. 1C, D). Immunoblot analysis confirmed that expression of Snail1 was increased at the protein level (Fig. 1B). These results suggest that increased expression of these EMT drivers may underlie the mesenchymal phenotype of BT549 cells. In line with their mesenchymal properties, BT549 cells expressed a higher level of CD44 and a lower level of CD24 at the population level as detected by semi-quantitative RT-PCR and immunoblotting ( Fig. 1B-D). The CD44 high /CD24 low expression pattern is characteristic of breast CSCs 50,51 . Similarly, reduced E-cadherin and CD24 and increased vimentin, CD44, and Snail were also observed in MDA-MB-468 cells -another PTEN-negative breast cancer cell line with a 44-bp deletion in the PTEN gene, which results in frameshifting and loss of the PTEN protein ( Fig. 1E) 18,52 . These results suggest that loss of PTEN correlates with a mesenchymal phenotype and the expression pattern of cell surface markers characteristic of breast CSCs. (D) Ethidium bromide-stained PCR products were quantified by densitometry and plotted as a ratio to 18S. N = 3, *P < 0.05 versus BT474. (E) BT474 and MDA-MB-468 cells were cultured to confluence and subjected to immunoblot analysis. Actin was used as a loading control.  Fig. 2A-C). This PTEN-regulated phenotypical change in cancer cells had not previously been reported. As expected, PTEN restoration suppressed cell growth and Akt activation as assessed by growth curve analysis and phosphorylation of Akt at S473, respectively ( Figures S1 and 2D). Notably, it also inhibited the expression of c-Myc, Sox2 and vimentin, and induced the expression of E-cadherin. RT-PCR analysis demonstrated reduced mRNA expression of the EMT-inducing transcription factors Snail1, Slug, ZEB1, and Twist2 (Fig. 2E,F) in these cells. The decreased expression of Snail1 after PTEN reconstitution was confirmed by immunoblotting (Fig. 2D). Furthermore, stable PTEN expression in BT549 cells increased CD24 and decreased CD44 at both mRNA and protein levels. (Fig. 2D-F). To further explore if this phenotypic switch correlates with reduced invasive activities, we performed a Matrigel invasion assay. As shown in Fig. 2G and H, stable PTEN transfection markedly reduced BT549 cell invasion into Matrigel. Taken together, these results provide evidence that restoration of PTEN suppresses the EMT and limits the invasive activity in PTEN-deficient breast cancer cells.
the protein phosphatase activity of pten confers an epithelial phenotype to breast cancer cells. PTEN is a dual-specificity phosphatase which can hydrolyze phospho-serine, -threonine, and -tyrosine www.nature.com/scientificreports/ residues in phosphoproteins as well as phosphoinositides 53,54 . The cysteine residue at the catalytic motif (C124) acts as a nucleophile to attack the phosphate group of its substrates and is essential for the phosphatase activity. The cancer-related mutation of C124 to serine (C124S) abolishes both lipid and protein phosphatase activities. By contrast, the G129E mutation initially discovered in Cowden disease, only affects PTEN binding to phosphoinositides and therefore, selectively inhibits its lipid phosphatase activity 53,54 . To elucidate the relative contribution of the lipid and protein phosphatase activity to the EMT and CSC activity in breast cancer, we stably transfected BT549 cells with wild-type PTEN, the G129E, or the C124S mutant ( Fig. 3A,B). Of note, live phase microscopy showed that expression of PTEN G129E, but not C124S, converted BT549 cells from a mesenchymal to an epithelial morphology, similar to the effect of reconstitution with wild-type PTEN (Fig. 3A). This is corroborated by immunoblot analysis demonstrating upregulation of E-cadherin and downregulation of splice variant 3 of vimentin, which was shown to be upregulated in kidney tumors (Fig. 3B) 55 . The EMT transcription factor Snail1 and the CSC marker CD44 were also downregulated by stable transfection with either wild-type PTEN or the G129E mutant. To determine whether PTEN-induced restoration of an epithelial phenotype reduces the CSC population in BT549 cells, we evaluated CSC activity of sphere-forming cells developed from stem cell-like clones by mammosphere formation 6,56 . Stable transfection of BT549 cells with wild-type PTEN and G129E led to a fivefold reduction in mammosphere formation, whereas transfection with C124S had no effect (Fig. 3C). This result suggests that the protein phosphatase activity of PTEN inhibits the self-renewal of CSCs. To date, most of the tumor suppressor functions of PTEN are thought to be mediated by its lipid phosphatase activity, which converts PIP3 to PIP2 and thus antagonizes the PI3K-Akt pathway 57 . To further examine if the lipid phosphatase activity of PTEN is involved in EMT suppression, we treated BT549 cells with the PI3K inhibitor wortmannin or the Akt1/2 inhibitor for 24 h. As expected, these inhibitors effectively reduced Akt phosphorylation at S473, but failed to induce E-cadherin or suppress vimentin expression (Fig. 3D). In fact, inhibition of the PI3K-Akt pathway further downregulated E-cadherin. Inhibition of PI3K or Akt also failed to induce morphological changes in BT549 cells (Fig. 3E). Altogether, these results suggest that the protein phosphatase activity of PTEN is responsible for the suppression of the EMT and CSC activity in breast cancer cells.

Abi1 is a PTEN substrate and upregulated after PTEN loss in breast cancer cells. Using PTEN-
null embryoid bodies reconstituted with PTEN G129E or C124S 43 , we performed phosphotyrosine immunopre- www.nature.com/scientificreports/ cipitation followed by mass spectrometric analysis and identified the WRC subunit Abi1 as a new substrate of PTEN 44 . PTEN dephosphorylates Abi1 at Y213 and S216, and causes Abi1 degradation via the calpain pathway.
To examine whether Abi is expressed in breast cancer cells and regulated by PTEN, we analyzed BT474 and BT549 cells by immunoblotting. The expression level of Abi1 was low in PTEN-positive BT474 cells and significantly upregulated in PTEN-negative BT549 cells together with the WRC component WAVE2 (Fig. 4A). The increased expression of Abi1 in BT549 cells is unlikely to occur at the transcriptional level since the Abi1 mRNA level is lower than that seen in BT474 cells (Fig. 4B). The Abi1 protein level was also higher in MDA-MB-468 cells deficient in PTEN (Fig. 4C). On the contrary, Abi1 was barely detectable in MCF-10A normal mammary epithelial cells (Fig. 4D). Stable transfection of BT549 cells with PTEN decreased Abi1 and its phosphorylation at Y213 and S216, which confirmed that PTEN also dephosphorylates and negatively regulates the Abi1 protein in breast cancer cells (Fig. 4E). In contrast to the Abi1 protein, Abi1 mRNA was elevated upon PTEN restoration, likely due to feedback regulation by reduction of the Abi1 protein (Fig. 4F). To further explore the relationship between PTEN and Abi1 in vivo, we generated whole body Pten knockout mice by crossing Pten fl/fl mice with UBC-Cre/ERT2 transgenic mice in which a Cre-ERT2 fusion protein is expressed under the control of the ubiquitin C promoter 58,59 . Intraperitoneal injection of tamoxifen into Pten fl/fl ;UBC-Cre/ERT2 mice induces the deletion of the Pten gene. Two weeks later, the PTEN protein was significantly reduced in knockout mammary tissues (Fig. 4G). As a consequence, levels of phospho-Abi1 S216, Abi1, and WAVE2 were increased. However, there was no significant difference in Abi1 mRNA between control and Pten knockout mammary tissues (Fig. 4H). Taken together, these results suggest that PTEN dephosphorylates and downregulates Abi1 in breast cancer cells.

Overexpression of Abi1 in mammary epithelial cells induces the EMT and enriches stem cell
activity. Abi1 is a core adaptor protein essential for the stabilization of the WRC, which control actin dynamics by stimulating Arp2/3-mediated actin nucleation 60 . To determine if elevated Abi1 levels contribute to EMT and CSC activity induced by PTEN loss, we stably transfected MCF10A cells, the most widely used normal mammary epithelial cell model derived from benign human breast tissue, with human Abi1. We chose MCF10A instead of BT474 cells because BT474 cells are very resistant to transfection. Overexpression of Abi1 transformed MCF10A cells from an epithelial to a fibroblast-like shape with prominent lamellipodial formation (Fig. 5A). Immunoblot analysis revealed that Abi1 overexpression increased vimentin, c-Myc, and decreased E-cadherin. The WRC component WAVE2 was also elevated, likely due to reduced degradation. Induction of the EMT as a result of Abi1 overexpression was supported by the upregulation of the EMT-inducing transcription factors Snail1, Twist1 and 2, and ZEB1 ( Fig. 5B-D). Moreover, MCF10A cell transfected with the vector alone formed www.nature.com/scientificreports/ smooth cell aggregates on Matrigel and developed into epithelial cysts similar to mammary acini after 8 days in culture (Fig. 5E). In contrast, the Abi1-overexpressing cells formed cords on day 1 and became scattered in a spindle shape in Matrigel by day 8. Abi1 overexpression also promoted mammosphere formation and switched the expression pattern of cell surface markers from CD44 low /CD24 high to CD44 high /CD24 low (Fig. 5B-D,F). In addition, the Matrigel invasion assay demonstrated that overexpression of Abi1 increased invasiveness of MCF10A cells more than 50-fold in comparison with the cells transfected with vector alone, which rarely invaded into Matrigel (Fig. 5G,H). Taken together, these results suggest that overexpression of Abi1 in mammary epithelial cells induces the EMT and enriches stem cell activity. Elevation of Abi1 after PTEN loss may be responsible for the EMT and increased CSC activity in breast cancer. www.nature.com/scientificreports/

Depletion of Abi1 in breast cancer cells suppresses the EMT and CSC activity. If Abi1 elevation
in PTEN-deficient breast cancer causes the EMT and enriches CSC activity, depletion of Abi1 should suppress these processes. To test this hypothesis, we knocked down Abi1 in BT549 cells by stable expression of shRNAs targeting human Abi1. Knockdown of Abi1 partially converted BT549 cells from a mesenchymal to an epithelial morphology (Fig. 6A). At the molecular level, Abi1 depletion induced the expression of E-cadherin and CD24 and reduced the expression of vimentin, Snail1, and CD44 (Fig. 6B). The self-renewal capability of the Abi1 knockdown cells was also significantly decreased as demonstrated by reduced mammosphere formation (Fig. 6C). In addition, their invasion into Matrigel was attenuated (Fig. 6D,E). Combined with the overexpression studies, these results suggest that elevation of Abi1, induced by PTEN loss, contributes to the EMT and increased CSC activity in breast cancer.

Discussion
A recent meta-analysis of 27 published studies involving 10,231 breast cancer patients revealed that loss of PTEN is associated with large tumor size, less differentiation, metastasis, and poor prognosis 61 . The EMT and CSCs are thought to be key contributors to aggressive behavior and recurrence of breast cancer. However, the molecular mechanisms whereby PTEN regulates EMT and CSCs are incompletely understood. The involvement of PTEN's protein phosphatase activity in CSCs has not been reported. In the present study, we demonstrate that loss of PTEN promotes the EMT and enriches CSC activity in breast cancer cells. This effect is mediated by the protein phosphatase activity of PTEN rather than the lipid phosphatase activity. Furthermore, PTEN dephosphorylates and downregulates Abi1. Gain-and loss-of-function analysis revealed that downregulation of Abi1 contributes to the tumor suppressor effect of PTEN on the EMT and CSCs. These results suggest a novel mechanism whereby PTEN regulates the mammary epithelial phenotype and stem/progenitor cell activity. www.nature.com/scientificreports/ The EMT was initially described as a developmental process observed in the formation of the primitive streak from the epiblast epithelium during chick gastrulation 62 . The EMT is widely adopted during development in many species to generate mesenchymal tissues from the epithelium 62-64 . The EMT also participates in tissue repair as well as many pathological processes such as fibrosis, tumor invasion and metastasis 65 . PTEN has been implicated in the regulation of the EMT in both development and cancer. In the chick embryo, overexpression of PTEN inhibited the EMT of the epiblast epithelium in the primitive streak 10 . This is unlikely to be mediated by the lipid phosphatase activity of PTEN because the G129E mutant, which is unable to convert PIP3 to PIP2, showed the same inhibitory effect. Similarly, in a three dimensional mammary epithelial culture model, shRNA-mediated depletion of PTEN disrupted mammary acinar formation 66 . Rescue experiments with the PTEN mutants that lack either protein or lipid phosphatase activity or both suggest that the protein and lipid phosphatase activity are both required for the formation of the acinar epithelial structure. These studies suggest that the protein phosphatase activity of PTEN may be involved in the maintenance of an epithelial phenotype. In the present study, we provide striking evidence that restoration of PTEN expression converted PTEN-deficient breast cancer cells from a mesenchymal phenotype to an epithelial phenotype and suppressed the EMT. This was assessed by the change of cell morphology, induction and/or increase of epithelial marker proteins, reduced expression of mesenchymal markers and EMT-inducing transcription factors, and decreased invasion into Matrigel. PTEN reconstitution also suppressed CSC activity, which is associated with and likely results from the EMT 6 . Rescue with the PTEN mutants that can distinguish between the protein and the lipid phosphatase activity suggests that PTEN's protein phosphatase activity is required for the suppression of the EMT and CSCs. In addition, pharmacological inhibition of PI3K or Akt failed to reverse the EMT. In fact, it further downregulated the epithelial marker E-cadherin. These results suggest that the lipid phosphatase activity of PTEN is not important for EMT suppression under these conditions.
We have identified Abi1 as a new substrate of PTEN during epiblast epithelial differentiation. After PTENmediated dephosphorylation, Abi1 is degraded through the calpain pathway 44 . Abi1 contains multiple PEST sequences [rich in proline (P), glutamic acid (E), serine (S), and threonine (T)]. The PEST sequence is often found in proteins with short half-lives and acts as a signal for degradation 67 . S216 is in the first PEST sequence and predicted to be a calpain cleavage site. Calpains are a family of intracellular cysteine proteases involved in many cellular activities, including cytoskeletal remodeling, cell differentiation and cell survival 68 . The ubiquitously expressed µ-and m-calpain consist of the large catalytic subunit calpain 1 (CAPN1) or 2 (CAPN2) dimerized with the common small regulatory subunit 1 (CAPNS1). CAPNS1 is required for the stability and activity of both calpain large subunits. Treatment of PTEN-dephosphorylated Abi1 with calpain led to Abi1 degradation. In addition, genetic ablation of Capn2 or Calpn1 blocked Abi1 degradation 44 . Ablation of Abi1 has been shown to destabilize the WRC and induce its degradation 69 . The results of the present study suggest this new PTEN-Abi1 pathway is conserved in human mammary epithelial cells and breast cancer cells.
The WRC controls actin cytoskeletal dynamics by stimulating Arp2/3-mediated actin nucleation 60 . It consists of 5 protein subunits, WAVE2 (or WAVE1 or WAVE3), Abi1 (or Abi2), Nck-associated protein 1 (Nckap1or Nck-ap1l), specifically Rac1-associated protein 1 (Sra-1, also termed as cytoplasmic FMR1-interacting protein, Cyfip1 or Cyfip2), and Brick1 (Brk1). Rac1-GTP binding to Sra-1, acidic phospholipids, especially PIP 3 , and kinasemediated phosphorylation of WRC subunits can activate WAVE2 and stimulate Arp2/3 complex-dependent actin nucleation 60,70,71 . In the WRC, Abi1 acts as a core adaptor to mediate membrane recruitment and stabilize other subunits 69,72,73 . Abi1 has also been shown to enter into a complex with the SH3-containing proteins Eps8 (epidermal growth factor receptor pathway substrate 8) and Sos1, and thus activating Rho GTPase Rac1 74,75 . Furthermore, Abi1 binds to Abl kinases to regulate kinase activity 73,[76][77][78][79] . In breast cancer, high levels of Abi1 expression correlates with early recurrence and poor prognosis 80 . In concordance with this finding, less invasive breast cancer cell lines express lower levels of Abi1, whereas highly invasive breast cancer cell lines express high levels of Abi1 81 . shRNA-mediated depletion of Abi1 decreased the formation of lamellipodia and invadopodia and the degradation of the extracellular matrix 82 . In cultured breast cancer cells, Abi1 formed a complex with IRS53 to activate Rac1, which drives lamellipodial formation 83 . How Abi1 regulation is linked to CSC activity remain unknown. A single-cell analysis showed that inhibition of actin polymerization blocked epidermal stem cell differentiation 84 . Stabilization of F-actin promoted differentiation. This effect is mediated by the change of G-actin that acts on serum response factor to regulate gene transcription. In mouse mammary carcinoma cells, an integrin-linked kinase-and mDia2-induced integrin-actin cascade regulated tumor initiation 85 . More recently, the WRC component WAVE3 was shown to promote CSC activity in breast cancers 86 . We speculate that Abi1 upregulation may enrich CSCs via interaction with mDia2 and WAVE3 87 .
In summary, we demonstrated that Abi1 is significantly upregulated in PTEN-deficient breast cancer cells. PTEN dephosphorylates Abi1 and causes its downregulation. Overexpression of Abi1 in non-tumorigenic mammary epithelial cells induces the EMT and increases stem/progenitor cell activity. On the other hand, depletion of Abi1 in breast cancer cells suppresses the EMT and inhibits CSC activity, mimicking the effect of PTEN reconstitution. These findings support a notion that PTEN suppresses the EMT by dephosphorylating and downregulating Abi1. The importance of this new PTEN pathway in suppression of breast cancer formation and metastasis in vivo warrants further investigation.

Stable transfection of mammary epithelial and breast cancer cells. For stable expression of wild-
type PTEN, the G129E and C124S mutants in BT549 cells and Abi1 in MCF10A cells, the cells were transfected with the corresponding vectors using Lipofectamine 3000 reagent (Invitrogen). Stable cell clones were selected with 500 µg/ml G418. GFP positive colonies were cloned and expanded. For knockdown of Abi1 in BT549 cells, specific shRNAs and the scrambled controls in the pRFP-C-RS vector were introduced into the cells by transfection using Lipofectamine 3000 reagent. The cells were selected with 1 µg/ml puromycin and RFP-positive clones were expanded and grown in the medium containing puromycin. Reconstitution of PTEN and Abi1 overexpression and knockdown were confirmed by immunoblotting. and UBC-Cre/ERT2 (B6.Cg-Ndor1Tg(UBC-cre/ERT2)1Ejb /2 J, RRID:IMSR_JAX:008085) transgenic mice were purchased from The Jackson Laboratory. Pten fl/fl mice were crossed with UBC-Cre/ERT2 to generate Pten fl/+ ; UBC-Cre/ERT2 mice, which were further crossed with Pten fl/fl mice to generate Pten fl/fl ; UBC-Cre/ERT2 mice. At 8 weeks of age, Pten fl/fl ; UBC-Cre/ERT2 mice were injected peritoneally with tamoxifen (150 mg/kg/day) for 5 days. Tamoxifeninjected Pten fl/fl mice served as controls. Mice were sacrificed at 2 weeks after injection and mammary tissues were harvested for immunoblot analysis. The animal studies were conducted in accordance with regulations of the Institutional Animal Care and Use Committee (IACUC) at Rutgers University, New Brunswick, New Jersey. The experimental protocol for this study was approved by the Rutgers IACUC.

Generation of
Phase contrast and fluorescence microscopy. Live cells were observed with a Nikon inverted fluorescence microscope (Eclipse TE2000) and digital images were acquired with a Hamamatsu CCD camera controlled by IP Lab software (Scanalytics).

Matrigel invasion assay.
Corning Costar transwell cell culture inserts (8 µm pore size, 24 mm diameter) were coated with 0.5 ml Matrigel (2 mg/ml) that gelled at 37 °C for 2 h. 2.5 × 10 5 cells were seeded in the upper chamber in RPMI-1640 (BT549) or DMEM/F12 medium (MCF-10A) without fetal bovine serum or growth factors. Complete growth media were added to the lower chamber. After 24 h of incubation in a tissue culture incubator, cells and Matrigel in the upper chamber were removed with a cotton swab. The cells that migrated through Matrigel to the bottom of the transwell were visualized with Giemsa stain and photographed. Cells in 24 fields (6 fields per well using a 20 × objective, quadruplicates) were counted for each group.
Statistical analysis. Results are presented as mean ± SD. Statistical differences between multiple groups were evaluated by one-way ANOVA. Statistical analysis between two groups was performed using unpaired Student's t test.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information files).