Oncogene (2010) 29, 3490–3500 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 www.nature.com/onc ORIGINAL ARTICLE ZEB1 represses E-cadherin and induces an EMT by recruiting the SWI/SNF chromatin-remodeling BRG1

ESa´nchez-Tillo´1,ALa´zaro2, R Torrent3, M Cuatrecasas4, EC Vaquero5, A Castells5, P Engel2 and A Postigo1,6

1Group of Transcriptional Regulation, Department of Oncology and Hematology, IDIBAPS, Barcelona, Spain; 2Immunology Unit, Department of Cellular Biology and Pathology, University of Barcelona Medical School, Barcelona, Spain; 3Master Program in Biotechnology, University Pompeu Fabra, Barcelona, Spain; 4Department of Pathology, Hospital Clinic, Barcelona, Spain; 5Department of Gastroenterology, Hospital Clı´nic, CIBEREHD, IDIBAPS, Barcelona, Spain and 6ICREA, Barcelona, Spain

Loss of E-cadherin is a key initial step in the transdiffer- event in the EMT is the downregulation/loss of the entiation of epithelial cells to a mesenchymal phenotype, E-cadherin adhesion protein. Dissecting the molecular which occurs when tumor epithelial cells invade into mechanisms that regulate E-cadherin expression has surrounding tissues. Expression of the nuclear factor therefore become pivotal for understanding tumor ZEB1 induces an epithelial-to-mesenchymal transition invasiveness and metastasis. and confers a metastatic phenotype on carcinomas by E-cadherin expression is regulated at multiple levels repressing the E-cadherin at the transcriptional level. having been described carcinomas with genetic, epi- In this study, we show that ZEB1 interacts with the SWI/ genetic, transcriptional and post-translational altera- SNF chromatin-remodeling protein BRG1 to regulate tions of the E-cadherin (Peinado et al., 2007). At the E-cadherin independently of CtBP, its traditional co- transcriptional level E-cadherin is repressed by a number repressor. Blocking the interaction between ZEB1 and of nuclear factors (E-cadherin transcriptional repres- BRG1 induces expression of E-cadherin and downregula- sors, EcTRs) such as ZEB1 (also known as dEF1, tion of the mesenchymal marker vimentin. ZEB1 and zfhx1a), ZEB2 (SIP1, zfhx1b), SNAI1 (Snail), SNAI2 BRG1 colocalize in E-cadherin-negative cells from cancer (Slug), E12/E47 and Twist1/2. Expression of these lines and in the stroma of normal colon. Colocalization of EcTRs in tumor epithelial cells is inversely correlated ZEB1 and BRG1 in epithelial cells is only found in those with E-cadherin and associates with increased invasive- de-differentiated cells characterized by nuclear b-catenin ness/metastasis and poorer clinical prognosis (Moreno- staining at the invasive edge of the tumor. Our results Bueno et al., 2008). identify ZEB1/BRG1 as a new transcriptional mechanism EMT may also be linked with the generation of cancer regulating E-cadherin expression and epithelial-to-me- stem cells in epithelial tumors (Rosen and Jordan, 2009; senchymal transdifferentiation that may be involved Peter, 2010). Stem cells from epithelial organs and during the initial stages of tumor invasion. carcinomas show a mesenchymal phenotype characteri- Oncogene (2010) 29, 3490–3500; doi:10.1038/onc.2010.102; zed by high levels of EcTRs. In turn, overexpression of published online 26 April 2010 EcTRs not only induces a stem cell antigenic phenotype but also increases the number of cells with the capacity Keywords: BRG1; E-cadherin; epithelial-mesenchymal of seeding new tumors (Mani et al., 2008). transition (EMT); ZEB1 ZEB1 is a factor that controls key regulatory during embryonic development and cell differentiation (Vandewalle et al., 2009). Among them, ZEB1 represses E-cadherin transcription by binding to Introduction two E box sequences in its promoter region (Grooteclaes and Frisch, 2000; Eger et al., 2005; Shirakihara et al., Invasion of carcinomas into surrounding tissues and 2007). Expression of ZEB1 in epithelial cells induces an their eventual metastasis requires that epithelial EMT and promotes tumor invasiveness in in vitro and cells undergo a de-differentiation process known as in vivo models (Aigner et al., 2007; Spaderna et al., epithelial-mesenchymal transition (EMT) (Polyak and 2008). Furthermore, mice with a targeted deletion of the Weinberg, 2009). During the EMT, epithelial cells lose ZEB1 gene show ectopic expression of E-cadherin and their polarity and intercellular adhesion and acquire a loss of mesenchymal genes such as vimentin (Liu et al., mesenchymal phenotype with higher motility. A hallmark 2008). Expression of ZEB1 is upregulated by mechanisms known to activate EMT such as transforming growth Correspondence: Dr A Postigo, Group of Transcriptional Regulation, factor-b, nuclear factor-kB, hypoxia and SNAI1 (re- IDIBAPS, Barcelona 08036, Spain. E-mail: [email protected] viewed in Peinado et al., 2007), and repressed by non- Received 8 November 2009; revised 9 February 2010; accepted 25 coding RNAs of the miR-200 and miR-205 families, February 2010; published online 26 April 2010 known to induce epithelial differentiation (Gregory Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3491 et al., 2008; Park et al., 2008). In turn, ZEB1 In this report, we show that ZEB1 interacts with transcriptionally represses miR-200 family members BRG1 to repress the E-cadherin promoter. ZEB1 and thus creating an EMT regulatory loop in ZEB1- BRG1 colocalize in vivo in E-cadherin-negative cells expressing tumors (Burk et al., 2008). Repression of from established lines, and stroma of normal colon as miR200c, miR203 and miR183 by ZEB1 is also required well as in de-differentiated epithelial cells at the invasion for maintenance of a stem cell phenotype further front of colorectal carcinomas. Our results show that supporting a link, mediated by ZEB1/miR, between BRG1 is required for the regulation E-cadherin and the EMT and cancer stem cells (Wellner et al., 2009; Peter, induction of EMT by ZEB1. 2010). ZEB1 could also contribute to overcome onco- genic addiction in tumors—ZEB1 is expressed in K-ras- independent cancer lines and its elimination induces K-ras dependency (Singh et al., 2009). Results ZEB1 represses transcription by recruiting the CtBP corepressor to its CtBP-interacting domain (CID) (see Differential pattern of E-cadherin repression by ZEB1 scheme in Figure 2a; Furusawa et al., 1999; Postigo across cell lines and Dean, 1999a; Chinnadurai, 2007) and CtBP As indicated above, it has been suggested that has been involved in ZEB1-mediated repression of ZEB1 could repress E-cadherin via CtBP-independent E-cadherin (Grooteclaes and Frisch 2000; Shi et al., mechanism(s). To elucidate such alternative mechani- 2003). Nevertheless, it has been reported that, at least in sms, we tested the ability of ZEB1 and a CID-mutant 293T cells, ZEB1 could repress E-cadherin through form of ZEB1 (ZEB1CIDm, which is unable to bind CtBP-independent mechanisms yet to be determined CtBP) to repress transcription of the E-cadherin (van Grunsven et al., 2003) and addressed in the present promoter in a panel of carcinoma cell lines. We found paper. that in MCF7, SW480, 293T, HCT116 and Hs578T Although loss of BRG1 and/or BRM—the two (Figure 1a and data not shown) mutation of the CID mutually exclusive ATPase subunits of the SWI/SNF region resulted in only a partial relief of ZEB1-mediated chromatin-remodeling complex—is found in a number repression thus supporting the existence of alternative of cancer lines and primary tumors (Reisman et al., repressor mechanisms. However, when similar experi- 2009), expression of BRG1, but not of BRM, has also ments were carried out in SW13 and C33a cells, not only been correlated with higher invasive/metastatic behavior did wild-type ZEB1 showed a lower repressor activity in carcinomas (Sentani et al., 2001; Sun et al.,2007). than in the other cell lines but mutation of the CID region BRG1 associates with DNA-binding transcription factors was sufficient to abrogate ZEB1-mediated repression and histone-modifying enzymes to activate/repress trans- (Figure 1b). cription as well as to regulate DNA replication, repair The repressor effect of ZEB1CIDm in MCF7 and and recombination (Trotter and Archer, 2008). SW480 but not in SW13 and C33a cells prompted us to

CtBP BRG1 CtBP

ZEB1 ZEB1 E Cad-Luc E Cad-Luc 120 120

80 80 RLU RLU 40 40

0 0 ZEB1 - + --+- ZEB1 - + --+-

ZEB1CIDm - - +--+ZEB1CIDm - - +--+ MCF7 SW480 SW13 C33a Figure 1 Differential pattern of E-cadherin repression by ZEB1 across cell lines. MCF7 and SW480 (a) or SW13 and C33a (b) were cotransfected with 0.15 mg of a firefly luciferase E-cadherin promoter reporter along with equal molar amounts (0.57–0.68 mg) of either ZEB1 or ZEB1CIDm. In this Figure, as well as in Figures 4b–d and Figures 5a–c, equal molar amounts of the corresponding empty vectors were transfected as controls. To normalize transfection efficiency, 0.1 mg of pCMV-b-gal was cotransfected with E-cadherin-luc. Luciferase and b-gal activities were determined as described in Materials and methods. Relative luciferase activity (RLU) values throughout the paper are expressed as the mean of duplicates and are representative of at least four different experiments.

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3492 IP endogenous N-term C-term Region (NtR) Central Region (CR) Region (CtR) inputZEB1 Contl BRG1 WB: BRG1 CtBP ZEB1 WB: ZEB1 Zinc CID Zinc Fingers Fingers WB: Control

FLAG-tagged: BRG1 FLAG-tagged: BRG1

Myc-tagged: vector Gem ZEB1 -tagged: NtR CR CtR

Direct WB WB: 9E10 Myc Direct WB WB: 9E10 Myc

WB: 9E10 Myc WB: 9E10 Myc IP FLAG IP FLAG WB: FLAG WB: FLAG

Figure 2 ZEB1 interacts with BRG1. (a) Schematic representation of ZEB1. The central region (CR) in between both zinc-finger clusters contains the CtBP-interacting domain (CID). (b) 293T cells were transfected with FLAG-BRG1 and either myc-Geminin or myc-ZEB1. After 48 h, cell lysates were immunoprecipitated (IP) for FLAG-BRG1 with FLAG M2 Ab. Pulled-down Geminin and ZEB1 were detected by western blotting (WB) with 9E10 myc Ab. (c) Endogenous interaction between ZEB1 and BRG1. 293T cell lysates were immunoprecipitated with ZEB1 (H102), BRG1 (H88) or control (SIP1 H260) Abs. Endogenous complexes were detected by WB with the indicated Abs. (d) BRG1 binds to the N-terminal domain of ZEB1. 293T were cotransfected with FLAG-BRG1 and myc-tagged versions of NtR, CR or CtR regions of ZEB1. After 48 hrs, cell lysates were IP with FLAG M2 Ab and the coimmunoprecipitated ZEB1 fragments were detected by WB with 9E10 myc Ab. See Supplementary Figures A and B for full scans of WB data. In (b–d), input/direct WB lanes represent 2.5% of total lysate.

hypothesize that the latter cell lines may lack some co- DNA as well as with a number of such as CtBP, repressor factor(s), different from CtBP but important Smad2/3 and p300 (Postigo and Dean 1997, 1999a, b; for ZEB1 regulation of E-cadherin. Interestingly, SW13 Furusawa et al., 1999; Postigo 2003) (Figure 2a). We and C33 cells—but not the other lines tested above— therefore sought to define the region within ZEB1 that have been described as deficient in the two ATPases of mediates the interaction with BRG1. To this end, we the SWI/SNF-remodeling complex, BRG1 and BRM generated three tagged fragments of ZEB1 encompass- (Zhang et al., 2000). This observation led us to analyze ing its C-terminal (CtR), central (CR) and N-terminal whether these two proteins may be involved in the (NtR) regions and co-expressed them in 293T along with repression of E-cadherin by ZEB1. full-length BRG1. As shown in Figure 2d (and Supplementary Figure B), the NtR of ZEB1, but not the other two regions, bound to BRG1. As CID is ZEB1 interacts with BRG1 located in the CR, BRG1 and CtBP appear to interact We therefore started testing whether ZEB1 interacts through different regions within ZEB1. with BRG1 by coexpressing tagged versions of both proteins in 293T cells followed by their co-immunopre- cipitation. The cell-cycle protein Geminin, which is ZEB1 and BRG1 colocalize in vivo at the invasion front known to bind BRG1 and BRM (Seo et al., 2005), was of carcinomas used as a positive control. This experiment revealed We next wondered whether ZEB1 and BRG1 colocalize that ZEB1 was able to bind BRG1 (Figure 2b and in living cells. Indeed, confocal microscopy and spectral Supplementary Figure A). analysis in E-cadherin-negative SW480 cells, revealed To confirm that the interaction between ZEB1 and the colocalization of both proteins with identical BRG1 also occurs without overexpression, we carried matching patterns of nuclear condensation (Supplemen- out two-way co-immunoprecipitation experiments in tary Figure C). Colocalization of ZEB1 and BRG1 was cell lysates from 293T that express ZEB1 and BRG1 also detected in E-cadherin-negative cells located in the endogenously. Immunoprecipitation with an anti-ZEB1 stromal compartment of normal colon tissue (Figure 3a antibody specifically pulled down endogenous BRG1 and Supplementary Figure D). and, conversely, an anti-BRG1 antibody co-immuno- At the membrane of normal epithelial cells E-cadherin precipitated endogenous ZEB1 (Figure 2c). These results associates with the oncoprotein b-catenin (Peinado indicate that ZEB1 and BRG1 are indeed capable of et al., 2007). Although b-catenin maintains its membra- interacting to form an endogenous complex. nous/cytoplasmic expression at the center of a carcinoma, ZEB1 is a highly modular protein and separate just as it does in normal epithelium, it translocates to the domains have been identified for its interaction with nucleus in isolated and scattered E-cadherin-negative

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3493 a BRG1 E-cadherin ZEB1 Merged DAPI

b BRG1 -catenin ZEB1 Merged DAPI

Figure 3 Endogenous ZEB1 and BRG1 colocalize in E-cadherin-negative cells. (a) BRG1 (green) and ZEB1 (red) colocalize (yellow in merged) in cells negative for E-cadherin (blue) in normal colon tissue. E-cadherin was originally detected with DyLight649 (far red) but converted here to blue for better representation. A triple merge of BRG1, E-cadherin and ZEB1 is shown. Size bars here and in (b) correspond to 76 mm. (b) BRG1 (green) and ZEB1 (red) colocalize (white/pink in merged) in epithelial cells with nuclear b-catenin (blue) at the tumor front of colorectal carcinomas. b-catenin was originally detected with Dylight-649 (far red) but converted to blue for representation. A triple merge of BRG1, b-catenin and ZEB1 is shown. 4,6-diamidino-2-phenylindole (DAPI) staining (blue) for all panels is also shown but not used in the merged panel.

dedifferentiated epithelial cells present at the tumor regulate its transcription. To this end, we expressed the front (Brabletz et al., 1998, 2002). As cells at the invasive E-cadherin promoter along both proteins in MCF7 cells, edge of colorectal carcinomas are undergoing active which are defective for ZEB1 but positive for BRG1. EMT, we decided to examine the colocalization of ZEB1 Overexpression of BRG1 (or highly related BRM) alone and BRG1 at the tumor front (Figure 3b and did not have any significant effect on the basal activity Supplementary Figure D). ZEB1 is absent in differ- of the E-cadherin promoter (Figure 4b and Supplementary entiated epithelial cells but expressed in those scattered Figure E). As expected, ZEB1 repressed the E-cadherin cells at the tumor front. Importantly, ZEB1 and BRG1 promoter but its repressor effect was enhanced the only colocalize in these E-cadherin-negative and inva- overexpression of BRG1 but not by that of BRM. By sive undifferentiated epithelial cells expressing nuclear contrast, knock down of endogenous BRG1 with specific b-catenin (note the white/pink staining in the merged siRNA alleviated ZEB1-mediated repression (Figure 4b). panel of Figure 3b and Supplementary Figure D). To elucidate whether BRG1 co-repressor activity is dependent on CtBP binding to ZEB1, we carried out experiments similar as those in Figure 4b but using BRG1 acts as a corepressor of ZEB1 to regulate ZEB1CIDm instead of wild-type ZEB1. Although ZEB1- E-cadherin transcription CIDm is not as efficient repressor of E-cadherin its effect We then analyzed whether an endogenous ZEB1/BRG1 was heightened by the overexpression of BRG1 complex (as shown in Figure 2c) could be assembled (Figure 4c, left panel), while elimination of endogenous onto the E-cadherin promoter. In chromatin immuno- BRG1 with siRNA almost totally rescued its repressor precipitation assays, we found that ZEB1 and BRG1 activity (Figure 4c, right panel). Taken together, these bind to a fragment of the E-cadherin promoter contain- results indicate that BRG1 not only synergizes with ing the ZEB1-binding sites (Figure 4a, left panel). ZEB1 in the repression of the E-cadherin promoter, but Knocking down of ZEB1 with specific small interfering also, that its co-repressor role could take place in the RNA (siRNA) not only blocked binding of ZEB1 absence of CtBP. Furthermore, the fact that knock to E-cadherin but also that of BRG1 (Figure 4a, right down of BRG1 rendered ZEB1CIDm into a non-repressor panel), indicating that the recruitment of the latter to construct suggests that CtBP and BRG1 are the main, E-cadherin takes place through a ZEB1-dependent indeed perhaps the only, co-repressor mechanisms by mechanism. which ZEB1 regulates the E-cadherin promoter. Next, we analyzed whether the formation of a ZEB1/ The above experiments prompted us to further BRG1 complex onto the E-cadherin promoter could examine the role of ZEB/BRG1 complexes by using a

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3494 BRG1 CtBP ZEB1 siRNA Control siRNA ZEB1 120 E Cad-Luc

80 input input ZEB1 BRG1 IgG ZEB1 BRG1 IgG No DNA MW 300 200 RLU 40 100 0 ZEB1 - - ++ZEB1- + +

BRG1 - + - + siCtl - + - siBRG1- - +

CtBP CtBP BRG1 CtBP BRG1

BRG1 ZEB1 ZEB1CIDm E Cad-Luc E Cad-Luc ZEB1CIDm E Cad-Luc 120 120 80 80 RLU

RLU 40 40

0 0 ZEB1 - ++ - - ZEB1CIDm - + + ZEB1CIDm - + + BRG1 - - + - + BRG1- - + siCtl - + - ZEB1 siBRG1- - + CIDm - --++ Figure 4 BRG1 acts as a co-repressor of ZEB1 in the repression of E-cadherin promoter. (a) BRG1 is recruited to the E-cadherin promoter in a ZEB1-dependent manner. Chromatin immunoprecipitation (ChIP) assays were carried out in 293T cells transfected with either a specific siRNA for ZEB1 or a control siRNA. Chromatin was immunoprecipitated with antibodies against ZEB1, BRG1 or a control rabbit IgG serum. The input lane represents the starting chromatin used for immunoprecipitation. As negative control for the PCR reaction a lane without chromatin is also included (No DNA). (b) As in Figure 1a, MCF7 cells were co-transfected with E-cadherin-luc along with ZEB1 and/or 0.38 mg of BRG1 (left panel). Where indicated siRNA BRG1 or control was also added. (c)As in (b), but ZEB1CIDm was used instead of ZEB1 (d)Asin(b)or(c) but using C33a cells.

cell system where BRG1 was missing. Exogenous expres- however, fully replicated in transient transfection sion of BRG1 in C33a cells enhanced the ability of ZEB1 experiments. Previous studies have found that stable and ZEB1CIDm to inhibit the E-cadherin promoter integration of a promoter reporter construct into (Figure 4d). chromatin enhances the capacity of BRG1 to activate/ As the NtR of ZEB1 mediates interaction with BRG1, repress transcription (Fryer and Archer, 1998; Zhang we generated a form of ZEB1 in which this region et al., 2000). Therefore, we generated stable clones with was deleted (ZEB1DNTR) and examined its capacity the E-cadherin-luc reporter stably integrated into C33a to repress the E-cadherin promoter. Like ZEB1CIDm with cells. Although—as in transient experiments illustrated respect to CtBP, elimination of the BRG1-binding site in Figure 4d—ZEB1 showed limited repressor activity of reduced the repressor capacity of ZEB1DNTR (Figure 5a). the integrated E-cadherin reporter in C33a; co-expres- ZEB1 only represses transcription when directly sion of BRG1 resulted in a much more efficient recruited to DNA. We therefore reasoned that over- repression than in transiently transfected E-cadherin expression of the NtR of ZEB1 should bind and titrate (Figure 5c). out endogenous BRG1, thus preventing its interaction with ZEB1 bound to the E-cadherin promoter. Indeed, we found that overexpression of increasing amounts of ZEB1 and BRG1 regulate the expression NtR—but not an empty vector—efficiently blocked of the endogenous E-cadherin gene repression of E-cadherin by DNA-bound ZEB1 thus In preceding sections, we have established BRG1 as a working as a dominant negative (Figure 5b). ZEB1 co-represssor in the regulation of the E-cadherin BRG1 regulates transcription by altering the nucleo- promoter. Next, we wondered whether ZEB1/BRG1 some structure around genes, a configuration that is not, complexes also regulate endogenous E-cadherin protein.

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3495 BRG1 NtR-ZEB1 BRG1 CtBP ZEB1 ZEB1 E Cad-Luc E Cad-Luc

120 120

80 80 RLU RLU 40 40

0 0 ZEB1- + - - ZEB1 - + + ++ -

ZEB1ΔNTR - - + - Gal4-NtR-ZEB1- - +

ZEB1CIDm - --+

BRG1 ZEB1 E Cad-Luc Genomic DNA

120

80 RLU 40

0 ZEB1 -++ BRG1 --+

C33a clone #4 Figure 5 (a) The NtR of ZEB1 is required for repression of the E-cadherin-luc. MCF7 cells were co-transfected with E-cadherin-luc along equal molar amounts of either ZEB1DNTR or ZEB1CIDm. Control vectors and transcriptional assays were included and assessed as referred in Figure 1. (b) The NtR of ZEB1 acts as a dominant negative of E-cadherin repression by ZEB1. MCF7 cells were co- transfected with E-cadherin-luc and ZEB1 as referred in Figure 4b. Where indicated increasing molar amounts of Gal4-NtR-ZEB1 (0.3, 0.6 and 1.8 mg)—unable to bind the E-cadherin promoter—were also co-transfected. The maximum amount of Gal4-NtR-ZEB1 did not have by itself any significant effect on the E-cadherin promoter (last bar on the right). (c) Enhanced corepressor activity of BRG1 over a stably integrated E-cadherin reporter. C33a clones with E-cadherin-luc stably integrated into the chromatin were transiently co-transfected with ZEB1 and/or BRG1 as in Figure 4d. Results are representative of four independent C33a clonal lines. Control vectors and transcriptional assays were included and assessed as in Figure 1.

To this end, we examined the effect of eliminating induce the de novo expression of the E-cadherin protein endogenous ZEB1 or BRG1 in SW480 colorectal cells, and the downregulation of endogenous vimentin lacking basal E-cadherin expression and expressing high (Figure 6a). levels of ZEB1. In a reverse set of experiments in MCF7 cells— If overexpression of ZEB1 induces an EMT, its showing opposing patterns of E-cadherin and ZEB1 elimination triggers a mesenchymal-epithelial transition, expression than SW480—we examined whether ZEB1 characterized by loss of mesenchymal characteristics requires BRG1 in the regulation of endogenous in favor of an epithelial phenotype (Spaderna et al., E-cadherin and vimentin (Figure 6b and Supplementary 2008). We found that knocking down of BRG1 Figure F). Indeed, elimination of endogenous BRG1 in SW480 cells using specific siRNAs was able to through siRNA diminished the ability of ZEB1 to

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3496 regulate both genes—partially reverting the effect of mentary Figure F). In summary, these results indicate ZEB1 in the repression of E-cadherin and the induction that BRG1 is an important cofactor of ZEB1 in its of a mesenchymal phenotype (Figure 6b and Supple- regulation of E-cadherin and the EMT process.

siCtl siZEB1 siBRG1

WB: E-cadherin

WB: Vimentin

WB: ZEB1

WB: BRG1

WB: α-tubulin

Vector + siCtlVector + siBRG1 ZEB1 + siCtl ZEB1 + siBrg1

E-cadherin

Vimentin

ZEB1

BRG1

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3497 Discussion Our results indicate that BRG1 synergizes with ZEB1 in repressing the E-cadherin and in initiating an EMT The loss of E-cadherin in carcinomas is associated with process. Loss-of-function brg1 mutations have been increasing invasiveness and hence poorer clinical prog- described in a number of tumors; in addition, gene nosis. Although a number of EcTRs (SNAI1, SNAI2, deletion studies in mice have also shown a role of ZEB1, ZEB2, E12/E47, Twist1/2) have been identified BRG1—but not BRM—in tumor development (Bult- the molecular mechanism(s) by which some of them man et al., 2000; Kadam and Emerson, 2003; Glaros operate remains yet to be elucidated. In this paper, we et al., 2008; Medina and Sanchez-Cespedes, 2008). reported that BRG1 binds to the NtR of ZEB1 acting as However, it is possible that BRG1 could have different, its co-repressor in the regulation of the E-cadherin or even opposing, roles depending on the tissue, the promoter. BRG1-mediated repression by ZEB1 is factors (oncogenic or anti-oncogenic) with which BRG1 enhanced when the promoter is stably integrated into interacts and/or the functional activity of various genomic DNA suggesting that ZEB1/BRG1 complexes proteins within the SWI/SNF complex. For instance, may repress E-cadherin by remodeling the local while BRM expression is frequently lost in tumors structure of chromatin around the gene. We have affecting the alimentary tract (in particular gastric provided evidence that ZEB1 and BRG1 colocalize in carcinomas), BRG1 is usually retained (Yamamichi E-cadherin-negative cells in the stromal compartment of et al., 2007). In fact, BRG1 may act in an oncogenic normal colon as well as in de-differentiated epithelial manner in some tumors (Wang et al., 2009). Further- cells, characterized by nuclear b-catenin staining, at the more, increased expression of BRG1 has been associated invasive front of colorectal carcinomas. Impairing the with tumor growth and invasiveness in prostate tumors binding of BRG1 to ZEB1 diminishes the latter’s (Sun et al., 2007) as well as higher metastatic behavior in capacity to repress endogenous E-cadherin and induce gastric carcinomas (Sentani et al., 2001). Although brg1 vimentin in epithelial cells. heterozygotes have higher incidence of tumors, it is A wealth of reports has inversely correlated expres- worth noting that these are mostly differentiated sion of E-cadherin with that of ZEB1 and other EcTRs epithelial tumors (Bultman et al., 2000). in primary tumors and cell lines. Nevertheless, the BRG1- and BRM-containing complexes regulate specific role of each EcTR in E-cadherin expression and different set of genes during cellular proliferation and EMT regulation remains unclear. EcTRs seem to be differentiation (Kadam and Emerson, 2003). The fact using distinct co-repressor mechanisms and it has been that ZEB1 cooperates with the former but not with the suggested that the availability of these co-repressors latter could shed light onto the multiple roles of ZEB1 in could dictate which EcTR is involved in a given tumor gene regulation. BRG1 interacts with DNA-binding (Vandewalle et al., 2009). However, the situation is proteins (activators, repressors and general transcription likely to be more complex as binding between EcTRs machinery factors) to be recruited to specific gene and their co-repressors could be regulated by post- promoters (Trotter and Archer, 2008). In this study, translational modifications (Postigo et al., 2003). The we described BRG1 as a co-repressor of ZEB1. finding that ZEB1 and ZEB2 are regulated by SNAI1 Although BRG1 is better known for its role in also indicates the existence of hierarchical relations promoting transcriptional activation—via nuclear re- among EcTRs (Guaita et al., 2002; Beltran et al., 2008). ceptors, MEF2D, Smad3, STAT1/2—it also participates Furthermore, loss of E-cadherin could also induce ZEB1 in gene silencing through its direct interaction with and Twist in a self-reinforcing loop to maintain a transcriptional repressors (Zhang et al., 2000; Burkhart mesenchymal state (Onder et al., 2008). et al., 2005). ZEB1 expression is not only inversely correlated to It is worth noting that all EcTRs co-repressors E-cadherin but also to the K-ras dependency of identified until now (for example, BRG1, CtBP, epithelial tumors and the ability of K-ras to maintain PRC2, Sin3a, NuRD) are part of chromatin-remodeling cell viability—elimination of K-ras in K-ras-indepen- complexes potentially linking epigenetic and transcrip- dent cancer cells only induces cell death when ZEB1 is tional mechanisms of E-cadherin regulation (BRG1 in depleted (Singh et al., 2009). The same study links this paper, Postigo and Dean, 1999a, b, 2000; Peinado epithelial differentiation to a K-ras-dependent gene et al., 2004, 2007; Herranz et al., 2008; Verstappen et al., signature for pharmacological targets thus positioning 2008). SWI/SNF complexes have been associated with EMT regulators as potential therapeutic targets (Singh demethylation of E-cadherin and CD44 genes (Banine et al., 2009). et al., 2005) although interactions of SWI/SNF with

Figure 6 BRG1 is required for the regulation of endogenous E-cadherin and EMT by ZEB1. (a) Knock down of ZEB1 or BRG1 induce endogenous E-cadherin protein. SW480 were transfected with control siRNA or specific siRNAs for ZEB1 or BRG1. At 48 h, cell lysates were loaded into polyacrylamide gels and protein levels were assessed by WB with the corresponding Abs as described in Supplementary Information. (b) BRG1 is required for efficient repression of endogenous E-cadherin and induction of EMT by ZEB1. MCF7 cells were transfected with an empty vector or ZEB1 along with pBABE-puro and either control siRNA or a siRNA specific to knock down BRG1. Upon puromycin selection, cells were analyzed for E-cadherin and vimentin expression by immunofluorescence—imposed over DAPI nuclear staining—in confocal microscopy. Expression of ZEB1 and BRG1 is shown here as controls while their corresponding nuclear DAPI stainings are in Supplementary Figure F. Size bar corresponds to 38 mm.

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3498 methylated DNA could also induce transcriptional Cells, cell cultures and human tissues repression (Harikrishnan et al., 2005). In our own The origins of the cell lines used in this work are described knock down studies, elimination of BRG1 induced under Supplementary Information. Cell lines were cultured in E-cadherin protein expression—most likely by relieving Dulbecco’s modified Eagle’s medium (Lonza, Basel, Switzerland) the repressor effect exerted by endogenous ZEB1—in an containing 10% fetal calf serum (Sigma, St Louis, MO, USA). Use of normal and tumor human samples was approved and otherwise negative cell line. Future studies should followed the guidelines of the local clinical research ethics determine whether BRG1 contributes to the repression committee (Hospital Clinic, Barcelona, Spain). of other ZEB1 target genes regulated by methylation. In this study, we have defined the N-terminal region of ZEB1 as the BRG1-interacting domain, which differs Cell transfection and transcriptional assays Cells were transfected with expression or reporter vectors using from the CID region where CtBP binds. Although Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) or by using BRG1 and CtBP act as ZEB1 co-repressors indepen- the calcium phosphate method (Postigo et al., 1997, 2003). dently of each other both are required for the efficient siRNA oligonucleotides were transfected at 50 nM using Lipo- repression of E-cadherin. Given their additive effect in fectamine RNAiMAX (Invitrogen). After 36 and 48 h, cells were the downregulation of E-cadherin, it is tempting to processed for transcriptional or immunoprecipitation assays. In speculate that CtBP and BRG1 contribute to ZEB1 the former, the firefly luciferase activity of the E-cadherin targeting different transcriptional activators within the promoter was assessed using a Luciferase Assay kit (Promega E-cadherin promoter. In that regard, it would be Corp., Madison, WI, USA). As an internal control of transfection advantageous for ZEB1 to have different repressor efficiency cells were cotransfected with pCMV-b-gal (Clontech, mechanisms to regulate the E-cadherin (or other ZEB1 Mountain View, CA, USA) and values were determined with a b-gal detection kit (Clontech). Clones of C33a stably carrying target genes) at different tissues, or differentiation E-cadherin-luc were generated by cotransfecting the puromycin stages. resistant vector pBABE-puro. After 15 days of puromycin selection Mounting evidence seems to indicate that expression (1 mg/ml) (Sigma), clones were isolated and tested. of EcTRs may also vary within the tumor. Thus, in an orthotopic mouse model of pancreatic cancer SNAI1 is Immunoprecipitation and western blot assays present in the center of the tumor while SNAI2 is Immunoprecipitation and western blot assays were performed expressed at the invasive front (Hotz et al., 2007). as previously described (Postigo et al., 1999c). Briefly, cells Meantime, in adrenal tumors both SNAI1 and Twist are were lysed in radioimmunoprecipitation assay buffer (150 mM present in the periphery of the tumor (Waldmann et al., NaCl, 1% NP40, 0.5% sodium dodecyl sulfate, 50 mM Tris 2009). Although well-differentiated tumors might still pH8, 2 mM EDTA plus protease inhibitors). Wherein indicated express E-cadherin, its loss is an essential condition for lysates were immunoprecipitated with the corresponding ulterior invasion into surrounding tissues and metastasis primary Abs and protein A, and were then loaded onto 8% (Polyak and Weinberg, 2009). Over 80% of colorectal poly-acrylamide gels. Around 2.5% of the lysate was also carcinomas involve mutations of the APC tumor directly loaded onto gels as input control for direct western suppressor gene, which results in a lack of degradation blotting. Gels were then transferred to a PVDF membrane (ImmobilonP, Millipore, Bedford, MA, USA). After blocking and nuclear accumulation of b-catenin. At the tumor (4% non-fat milk), membranes were incubated with the center, however, b-catenin shows a predominantly corresponding primary and horseradish peroxidase-conjugated membranous/cytoplasmic distribution and is only found secondary Abs before the chemiluminescence reaction was nuclear at the tumor invasive front in isolated developed, using a Pierce ECL kit (Termo Fisher Scientific, de-differentiated E-cadherin-negative epithelial cells Waltham, MA, USA). (Brabletz et al., 1998, 2002), whose presence has been associated with poorer clinical prognosis (Baldus et al., Chromatin immunoprecipitation assays 2004). During tumor invasion, these nuclear b-catenin Chromatin immunoprecipitation assays were performed using cells seem to be the first to undergo EMT and lose the EpiQuick ChIP kit (Epigentek, Inc., Brooklyn, NY, USA) E-cadherin. ZEB1 is present neither in the epithelial cells as per manufacturer’s instructions. Rabbit polyclonal antiser- of normal colonic mucosa nor the tumor center, but um for ZEB1 and BRG1 and control rabbit IgG were used for rather in those E-cadherin-negative de-differentiated immunoprecipitation (Supplementary Information). Promoter cells containing nuclear b-catenin at the point in which occupancy was assessed by 28 cycles PCR amplification of the it colocalizes with BRG1. À86/ þ 60 sequence of human E-cadherin (containing binding sites In conclusion, the results presented here have for ZEB1) using primers described in Vandewalle et al. (2005). identified ZEB1/BRG1 as a new transcriptional mechanism in the repression of E-cadherin during Cell and tissue staining, immunofluorescence and confocal EMT and tumor progression. analysis SW480 cells and samples of normal and tumor colon tissues were fixed for 30 min with 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA), and then permea- Materials and methods bilized with phosphate-buffered saline-0.5% Triton X100 and 0.1% sodium borohydride (Sigma) for background reduction. Reagents, antibodies, plasmids and sequences After 1 h blocking, (5% primary antibody-matching normal The various sources of plasmids, antibodies and sequences for serum plus 4% bovine serum albumin in phosphate-buffered primers and siRNAs are described in the Supplementary saline, 0.5% Tween) cells were incubated overnight at 4 1C Information section. with the corresponding primary and secondary antibodies.

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3499 Slides were then mounted in Vectashield-DAPI medium We also thank M Morales (Department Physiological (Vector Labs, Burlingame, CA, USA) and visualized with a Sciences) and staff from the Technical Science Services at the TCS SP5 Spectral confocal microscope (Leica, Mannheim, University of Barcelona for their help in confocal analysis, Germany). Images were later analyzed using ImageJ software M Rickman (IDIBAPS) and the Pathology Department at the (Rasband, 1997–2004) applying a median filter of 0.5-1 pixels. Hospital Clinic of Barcelona for technical advice and M Parrizas (IDIBAPS) for helpful recommendations on ChIP assays. We are also grateful to R Gasa (IDIBAPS) for critical Conflict of interest reading of the paper and J Moore for help with English editing. We apologize to those researchers whose work was The authors declare no conflict of interest. cited indirectly through reviews because of space limitations. Initial work in this study was supported by a grant from La Caixa Foundation (BM-06570) to AP. At later stages, this Acknowledgements project was funded by the Spanish Ministry of Science and Innovation (MICINN) (BFU2007-60302) and Olga Torres We are indebted to all of the researchers who kindly provided Foundation (FOT-0902) to AP. EST’s salary was partly us with cell lines and plasmids (see Supplementary Information). funded by Grant BM-06570.

References

Aigner K, Dampier B, Descovich L, Mikula M, Sultan A, Schreiber M Gregory PA, Bert AG, Paterson EL, Barry SC, Tsykin A, Farshid G et al. (2007). The ZEB1 (deltaEF1) promotes et al. (2008). The miR-200 family and miR-205 regulate epithelial to tumour cell dedifferentiation by repressing master regulators of mesenchymal transition by targeting ZEB1 and SIP1. Nat Cell Biol epithelial polarity. Oncogene 26: 6979–6988. 10: 593–601. Baldus SE, Mo¨nig SP, Huxel S, Landsberg S, Hanisch FG, Engelmann Grooteclaes ML, Frisch SM. (2000). Evidence for a function of CtBP K et al. (2004). MUC1 and nuclear b-catenin are coexpressed at the in epithelial gene regulation and anoikis. Oncogene 19: 3823–3828. invasion front of colorectal carcinomas and are both correlated with Guaita S, Puig I, Franci C, Garrido M, Dominguez D, Batlle E et al. tumor prognosis. Clin Cancer Res 10: 2790–2796. (2002). Snail induction of epithelial to mesenchymal transition in Banine F, Bartlett C, Gunawardena R, Muchardt C, Yaniv M, tumor cells is accompanied by MUC1 repression and ZEB1 Knudsen ES et al. (2005). SWI/SNF chromatin-remodeling factors expression. J Biol Chem 277: 39209–39216. induce changes in DNA methylation to promote transcriptional Harikrishnan KN, Chow MZ, Baker EK, Pal S, Bassal S, Brasacchio activation. Cancer Res 65: 3542–3547. D et al. (2005). Brahma links the SWI/SNF chromatin-remodeling Beltran M, Puig I, Pen˜a C, Garcı´a JM, Alvarez AB, Pen˜aRet al. complex with MeCP2-dependent transcriptional silencing. Nat (2008). A natural antisense transcript regulates Zeb2/Sip1 gene Genet 37: 254–264. expression during Snail1-induced epithelial-mesenchymal transition. Herranz N, Pasini D, Dı´az VM, Francı´C, Gutierrez A, Dave N et al. Genes Dev 22: 756–769. (2008). Polycomb complex 2 is required for E-cadherin repression Brabletz T, Jung A, Hermann K, Gu¨nther K, Hohenberger W, by the SNAI11 transcription factor. Mol Cell Biol 28: 4772–4781. Kirchner T. (1998). Nuclear overexpression of the oncoprotein beta- Hotz B, Arndt M, Dullat S, Bhargava S, Buhr HJ, Hotz HG. (2007). catenin in colorectal cancer is localized predominantly at the Epithelial to mesenchymal transition: expression of the regulators invasion front. Pathol Res Pract 194: 701–704. snail, slug, and twist in pancreatic cancer. Clin Cancer Res 13: Brabletz T, Jung A, Kirchner T. (2002). Beta-catenin and the 4769–4776. morphogenesis of colorectal cancer. Virchows Arch 441: 1–11. Kadam S, Emerson BM. (2003). Transcriptional specificity of human Burkhart BA, Hebbar KB, Trotter KW, Archer TK. (2005). SWI/SNF BRG1 and BRM chromatin remodeling complexes. Mol Chromatin-dependent E1A activity modulates NFkB RelA- Cell 11: 377–389. mediated repression of glucocorticoid -dependent transcrip- Liu Y, El-Naggar S, Darling DS, Higashi Y, Dean DC. (2008). Zeb1 tion. J Biol Chem 280: 6349–6358. links epithelial-mesenchymal transition and cellular senescence. Bultman SJ, Gebuhr T, Yee D, La Mantia C, Nicholson J, Gilliam A Development 135: 579–588. et al. (2000). A Brg1 null mutation in the mouse reveals functional Mani SA, Guo W, Liao MJ, Eaton EN, Ayyanan A, Zhou AY et al. differences among SWI/SNF complexes. Mol Cell 6: 1287–1295. (2008). The epithelial-mesenchymal transition generates cells with Burk U, Schubert J, Wellner U, Schmalhofer O, Vincan E, Spaderna S properties of stem cells. Cell 133: 704–715. et al. (2008). A reciprocal repression between members of the Medina PP, Sanchez-Cespedes M. (2008). Involvement of the miR-200 family and ZEB1 promotes EMT and invasion in cancer chromatin-remodeling factor BRG1/SMARCA4 in human cancer. cells. EMBO Rep 9: 582–589. Epigenetics 3: 64–68. Chinnadurai G. (2007). Transcriptional regulation by C-terminal Moreno-Bueno G, Portillo F, Cano A. (2008). Transcriptional binding proteins. Int J Biochem Cell Biol 39: 1593–1607. regulation of cell polarity in EMT and cancer. Oncogene 27: Eger A, Aigner K, Sonderegger S, Dampier B, Oehler S, Schreiber M 6958–6969. et al. (2005). dEF1 is a transcriptional repressor of E-cadherin and Onder TT, Gupta PB, Mani SA, Yang J, Lander ES, Weinberg RA. regulates epithelial plasticity in breast cancer cells. Oncogene 24: (2008). Loss of E-cadherin promotes metastasis via multiple 2375–2385. downstream transcriptional pathways. Cancer Res 68: 3645–3654. Fryer CJ, Archer TK. (1998). Chromatin remodelling by the Park SM, Gaur AB, Lengyel E, Peter ME. (2008). The miR-200 family requires the BRG1 complex. Nature 393: determines the epithelial phenotype of cancer cells by targeting the 88–91. E-cadherin repressors ZEB1 and ZEB2. Genes Dev 22: 894–907. Furusawa T, Moribe H, Kondoh H, Higashi Y. (1999). Identification Peinado H, Ballestar E, Esteller M, Cano A. (2004). SNAI1 mediates of CtBP1 and CtBP2 as corepressors of zinc finger-homeodomain E-cadherin repression by the recruitment of the Sin3A/histone factor deltaEF1. Mol Cell Biol 19: 8581–8590. deacetylase 1 (HDAC1)/HDAC2 complex. MolCellBiol24: 306–319. Glaros S, Cirrincione GM, Palanca A, Metzger D, Reisman D. (2008). Peinado H, Olmeda D, Cano A. (2007). SNAI1, Zeb and bHLH Targeted knockout of BRG1 potentiates development. factors in tumour progression: an alliance against the epithelial Cancer Res 68: 3689–3696. phenotype? Nat Rev Cancer 7: 415–428.

Oncogene Repression of E-cadherin by ZEB1/BRG1 ESa´nchez-Tillo´ et al 3500 Peter ME. (2010). Regulating cancer stem cells the miR way. Cell Stem addiction’ reveals regulators of EMT and tumor cell survival. Cell 6: 4–6. Cancer Cell 15: 489–500. Polyak K, Weinberg R. (2009). Transitions between epithelial and Spaderna S, Schmalhofer O, Wahlbuhl M, Dimmler A, Bauer K, mesenchymal states: acquisition of malignant and stem cell traits. Sultan A et al. (2008). The transcriptional repressor ZEB1 promotes Nature Rev Cancer 9: 265–273. metastasis and loss of cell polarity in cancer. Cancer Res 68: Postigo AA. (2003). Opposing functions of ZEB proteins in the regulation 537–554.1 of the TGFb/BMP signaling pathway. EMBO J 22: 2443–2452. Sun A, Tawfik O, Gayed B, Thrasher JB, Hoestje S, Li C et al. (2007). Postigo AA, Dean DC. (1997). ZEB, a vertebrate homolog of Aberrant expression of SWI/SNF catalytic subunits. BRG1/BRM Is Drosophila Zfh-1, is a negative regulator of muscle differentiation. associated with tumor development and increased invasiveness in EMBO J 16: 3935–3943. prostate cancers. Prostate 67: 203–213. Postigo AA, Dean DC. (1999a). ZEB represses transcription through Trotter KW, Archer TK. (2008). The BRG1 transcriptional coregu- interaction with the corepressor CtBP. Proc Natl Acad Sci USA 96: lator. Nuc Rec Sign 6: e004. 6683–6688. van Grunsven LA, Michiels C, Van de Putte T, Nelles L, Wuytens G, Postigo AA, Dean DC. (1999b). Independent repressor domains in Verschueren K et al. (2003). Interaction between Smad-interacting ZEB regulate muscle and T-cell differentiation. Mol Cell Biol 19: protein-1 and the corepressor C-terminal binding protein is 7961–7971. dispensable for transcriptional repression of E-cadherin. J Biol Postigo AA, Dean DC. (2000). Differential expression and function of Chem 278: 26135–26145. the zfh-1 family of zinc finger/homeodomain repressors. Proc Natl Vandewalle C, Comijn J, De Craene B, Vermassen P, Bruyneel E, Acad Sci USA 97: 6391–6397. Andersen H et al. (2005). SIP1/ZEB2 induces EMT by repressing Postigo AA, Depp JL, Taylor JJ, Kroll KL. (2003). Regulation of genes of different epithelial cell-cell junctions. Nucleic Acids Res 33: Smad signaling through a differential recruitment of coactivators 6566–6578. and corepressors by ZEB proteins. EMBO J 22: 2453–2462. Vandewalle C, Van Roy F, Berx G. (2009). The role of the ZEB family Postigo AA, Sheppard AM, Mucenski L, Dean DC. (1997). c-Myb and of transcription factors in development and disease. Cell Mol Life Ets proteins synergize to overcome transcriptional repression by Sci 66: 773–787. ZEB. EMBO J 13: 3924–3934. Verstappen G, van Grunsven LA, Michiels C, Van de Putte T, Postigo AA, Ward E, Skeath JB, Dean DC. (1999c). Zfh-1, the Souopgui J, Van Damme J et al. (2008). Atypical Mowat-Wilson Drosophila homologue of ZEB, is a transcriptional repressor that patient confirms the importance of the novel association between regulates somatic myogenesis. Mol Cell Biol 19: 7255–7263. ZFHX1B/SIP1 and NuRD corepressor complex. Hum Mol Genet Rasband WS. (1997–2004). ImageJ. National Institutes of Health: 17: 1175–1183. Bethesda, MD, USA, http://rsb.info.nih.gov/ij/ . Waldmann J, Slater EP, Langer P, Buchholz M, Ramaswamy A, Walz Reisman D, Glaros S, Thompson EA. (2009). The SWI/SNF complex MK et al. (2009). Expression of the transcription factor snail and its and cancer. Oncogene 28: 1653–1668. target gene twist are associated with malignancy in pheochromocy- Rosen JM, Jordan CT. (2009). The increasing complexity of the cancer tomas. Ann Surg Oncol 16: 1997–2005. stem cell paradigm. Science 324: 1670–1673. Wang X, Sansam CG, Thom CS, Metzger D, Evans JA, Nguyen PT Sentani K, Oue N, Kondo H, Kuraoka K, Motoshita J, Ito R et al. et al. (2009). Oncogenesis caused by loss of the SNF5 tumor (2001). Increased expression but not genetic alteration of BRG1, a suppressor is dependent on the activity of BRG1, the ATPase of the component of the SWI/SNF complex is associated with the advanced SWI/SNF chromatin remodeling complex. Cancer Res 69: 8094– stage of human gastric carcinomas. Pathobiology 69: 315–320. 8101. Seo S, Herr A, Lim JW, Richardson GA, Richardson H, Kroll KL. Wellner U, Schubert J, Burk UC, Schmalhofer O, Zhu F, Sonntag A (2005). Geminin regulates neuronal differentiation by antagonizing et al. (2009). The EMT-activator ZEB1 promotes tumorigenicity BRG1 activity. Genes Dev 19: 1723–1734. by repressing stemness-inhibiting microRNAs. Nat Cell Biol 11: Shi Y, Sawada JI, Sui G, Affar EB, Whetsline JR, Lan F et al. (2003). 1487–1495. Coordinated histone modifications mediated by a CtBP co-repressor Yamamichi N, Inada K, Ichinose M, Yamamichi-Nishina M, complex. Nature 422: 735–738. Mizutani T, Watanabe H et al. (2007). Frequent loss of Brm Shirakihara T, Saitoh M, Miyazono K. (2007). Differential regulation expression in gastric cancer correlates with histological features and of epithelial and mesenchymal markers by deltaEF1 proteins in differentiation state. Cancer Res 67: 10727–10735. epithelial mesenchymal transition induced by TGF-beta. Mol Biol Zhang HS, Gavin M, Dahiya A, Postigo A, Ma D, Luo R et al. (2000). Cell 18: 3533–3544. Exit from G1 and S phase of the cell cycle is regulated by repressor Singh A, Greninger P, Rhodes D, Koopman L, Violette S, Bardeesy S complexes containing HDAC-Rb-hSWI/SNF and Rb-hSWI/SNF. et al. (2009). A signature associated with ‘K-Ras Cell 101: 79–89.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene