Smad4 induces the tumor suppressor E-cadherin and P-cadherin in colon

carcinoma cells

Nicole Muller1,4, Anke Reinacher-Schick1,4, Stephan Baldus2, Jolanda van Hengel3, Geert Berx3,Anke Baar1, Frans van Roy3, Wolff Schmiegel1 and Irmgard Schwarte-Waldhoff*,1

1Department of Internal Medicine, IMBL, University of Bochum, Bochum, Germany; 2Department of Pathology, University ofCologne, Germany; 3Department for Molecular Biomedical Research, Molecular Cell Biology Unit, VIB, Ghent University, Ghent,Belgium

Smad4 is an intracellular transmitter of TGF-b signals andits tumor suppressor function is presumed to reside in itscapacity to mediate TGF-b-induced growth inhibition.However, there is accumulating evidence that thishypothesis may be too simple. The roles of TGF-b incarcinogenesis are complex and also comprise tumorpromoting functions particularly in late stage carcinogen-esis. Importantly, functional inactivation of Smad4 incolon carcinomas frequently occurs at late stages whentumors acquire invasive and metastatic capabilities. Wehave previously reported that stable re-expression ofSmad4 in SW480 human colon carcinoma cells wasadequate to suppress tumor growth in nude mice. However,it did not affect cell growth in vitro nor did it restore TGF-b responsiveness. Here, we report that Smad4 transcrip-tionally induced classical cadherins including the invasionsuppressor E-cadherin, presumably re-establishing epithe-lial morphology. Smad4-induced cadherins were able torecruit catenins to the plasma membrane and werefunctionally active in cell – cell adhesion. These resultsindicate a novel pathway of Smad4-mediated tumorsuppression and suggest that Smad4 in colon cells maybe involved in the maintenance of epithelial traits.Oncogene (2002) 21, 6049 – 6058. doi:10.1038/sj.onc.1205766

Keywords: Smad4; tumor suppressor gene; TGF-b;E-cadherin; invasion

Introduction

Smad4, first identified as DPC4 (deleted in pancreaticcarcinoma, locus 4), is a tumor suppressor genefunctionally inactivated in one half of pancreatic

adenocarcinomas (Hahn et al., 1996), in one third ofmetastatic colorectal cancers (Miyaki et al., 1999), inevery fourth carcinoma of the small intestine (Blaker etal., 2002) and in smaller subsets of other tumor types(Hahn et al., 1998; Schutte et al., 1996). Smad4 belongsto the Smad gene family, which encodes intracellularsignaling mediators of the TGF-b superfamily ofcytokines. TGF-b cytokines signal through TGF-btype II and type I receptors, a family of transmem-brane serine/threonine kinase receptors, which uponactivation phosphorylate receptor interacting Smadproteins (R-Smads). Activated R-Smads form hetero-meric complexes with Smad4, the common partnerSmad (co-Smad), and translocate into the nucleuswhere they function as transcription factors.

As Smad4 is a key mediator of TGF-b responses, itscellular functions are commonly investigated as part ofthis signaling pathway. TGF-b cytokines have beenshown to regulate a wide range of cellular responsessuch as proliferation and differentiation, migration andapoptosis during embryogenesis and in the mainte-nance of tissue homeostasis in the adult (Kingsley,1994; Massague, 1998; Roberts and Sporn, 1990). Intumor biology most attention has been paid to TGF-b’s role as a prototype negative growth factor potentlyinhibiting the growth of normal epithelial cells. Tumorcells typically acquire TGF-b resistance, allowing themto escape normal growth constraints. Mutationalinactivation of the TGF-b type II receptor has beenidentified as the underlying mechanism for TGF-bresistance in microsatellite unstable colon cancers(Markowitz et al., 1995). When Smad4 was identifiedas a tumor suppressor, its functional inactivation wasconsidered an alternative mechanism to cause TGF-bresistance and Smad4’s ability to transmit TGF-binduced growth inhibitory signals was supposed tounderlie its tumor suppressor function. Experimentalproof for this hypothesis, however, is still missing:suppression of tumorigenicity in vivo through restora-tion of TGF-b growth inhibition by Smad4 re-expression has not yet been shown. Moreover, thereis accumulating evidence that the acquisition of TGF-bresistance and loss of Smad4 may be independentevents. For example, certain TGF-b transcriptionalresponses are retained in Smad4 deficient cells

Received 26 February 2002; revised 15 May 2002; accepted 14 June2002

*Correspondence: I Schwarte-Waldhoff, Immunologisch-Molekular-biologisches Labor (IMBL), Medizinische Universitatsklinik derRuhr-Universitat Bochum Knappschaftskrankenhaus, In der Schor-nau 23-25, D-44892 Bochum, Germany;E-mail: irmgard.schwarte-waldhoff@ruhr-uni-bochum.de4N Muller and A Reinacher-Schick contributed equally to this paperand should be considered as first authors

Oncogene (2002) 21, 6049 – 6058ª 2002 Nature Publishing Group All rights reserved 0950 – 9232/02 $25.00

www.nature.com/onc

(Hocevar et al., 1999; Sirard et al., 2000) and severalSmad4 deficient tumor cell lines are still growthinhibited by TGF-b (Dai et al., 1999; Fink et al., 2001).

The relationship of TGF-b and Smad4 is furthercomplicated by the fact that TGF-b has a dual andparadoxical role in carcinogenesis (Akhurst andBalmain, 1999; Derynck et al., 2001; Gold, 1999;Massague et al., 2000; Piek and Roberts, 2001; Reiss,1997, 1999). While acting as a tumor suppressor inearly stages of cancer development through growth-inhibitory and pro-apoptotic activities, TGF-b is oftenoverexpressed in late stage carcinogenesis, suggestingthat it promotes tumorigenesis. TGF-b’s tumorpromoting function may reside in part in its paracrineeffects. For example, TGF-b exerts local and systemicimmune suppression and can induce angiogenesis. Inaddition, TGF-b has been implicated in the epithelialto mesenchymal transition of tumor cells (Bhowmick etal., 2001; Cui et al., 1996; Ellenrieder et al., 2001;Kitagawa et al., 1996; Miettinen et al., 1994; Oft et al.,1998; Piek et al., 1999; Portella et al., 1998; Zavadil etal., 2001), which characterizes invasive and metastaticcarcinomas. The roles of the Smad proteins in theseprocesses remain to be elucidated.

Mediating TGF-b responses is only one of the manyputative functions of Smad4 as a signaling molecule.Smad4 is the only co-Smad in mammalian cells and isthus potentially involved in signal transmission of allmembers of the TGF-b superfamily, many of which aremultifunctional regulators in development and growthcontrol. Smad signaling is also interconnected withother signaling cascades through a variety of cross-talkmechanisms (for review see Attisano and Wrana, 2000;Massague et al., 2000; ten Dijke et al., 2000;Verschueren and Huylebroeck, 1999; Zhang andDerynck, 1999). Moreover, Smads functionally coop-erate with a number of transcription factors,transcriptional coactivators and repressors, and theseare in turn targets for regulation through othersignaling pathways. The loss of Smad4 in an incipienttumor cell, consequently, may not only affect onelinear signaling pathway, but may induce complexalterations in the signaling networks. Thus, decipheringthe functions of Smad4 as a tumor suppressor requiresmodels that are able to reflect the complexity of thismolecular and cellular context.

For this we have recently established stable Smad4transfectants derived from Smad4-deficient humancarcinoma cells. Re-expression of Smad4 in SW480colon and Hs766T pancreatic carcinoma cells wassufficient to suppress tumor formation in nude miceproviding definite proof for its functional activity(Schwarte-Waldhoff et al., 1999, 2000). Yet, Smad4re-expression did not exert growth inhibitory effectsnor did it restore sensitivity to TGF-b in vitroindicating that the acquisition of TGF-b resistanceand loss of Smad4 may be independent events(Schwarte-Waldhoff et al., 1999, 2000). RetainedTGF-b resistance in Smad4-positive cells may be dueto limited expression levels of the TGF-b type IIreceptor.

To dissect mechanisms of Smad4 mediated tumorsuppression we have analysed constitutive differencesbetween Smad4 re-expressing and Smad4-negativeSW480 cell clones. The previous observation, thatSmad4 re-expression induced a transition from aspindle cell shape to an epithelioid morphology withclose intercellular contacts, prompted us to investigatethe expression of cadherins, key molecules in theestablishment and maintenance of cell – cell adhesionand of epithelial differentiation (Bracke et al., 1996;Gumbiner, 1996). This study shows that stable re-expression of Smad4 in SW480 cells transcriptionallyinduces classical cadherins, influences subcellulardistribution of catenins and thus restores epithelialtype intercellular adhesion. These findings indicate anovel TGF-b independent way of Smad4 mediatedtumor suppression and suggest that the tumorsuppressor Smad4 may be involved in transcriptionalregulation of the tumor suppressor E-cadherin and thehighly related P-cadherin.

Results

Re-expression of Smad4 at physiological levels is notadequate to restore TGF-b responses in SW480 cells

SW480 human colon carcinoma cells are Smad4-deficient. They have lost one copy of chromosome18, where the Smad4 gene is located; the second Smad4allele harbors a splice site mutation and is notexpressed (Woodford-Richens et al., 2001). Throughstable restoration of Smad4 in these cells we havepreviously provided definite proof for Smad4’s func-tional activity as a tumor suppressor. Re-expression ofSmad4 at roughly physiological levels was adequate tosuppress tumor growth in vivo in nude mice (Schwarte-Waldhoff et al., 1999). Surprisingly, however, re-expression of Smad4 did not affect cell growth in vitroand was not sufficient to restore TGF-b antiprolifera-tive responses (Schwarte-Waldhoff et al., 1999).

To further elucidate the relationship of TGF-b andSmad4 we analysed TGF-b target genes in SW480transfectant clones. Interestingly, basal expression levelsof the cyclin dependent kinase inhibitor p21(Cip-1/WAF-1), a prototype antiproliferative TGF-b responsegene, were consistently increased in all Smad4-positiveSW480 cell clones, whereas levels of plasminogen-activator-inhibitor 1 (PAI-1), another ‘classical’ TGF-bresponse gene, were reduced (Figure 1a). However,neither in Smad4-negative control clones nor in Smad4re-expressing tumor suppressed revertants could weobserve any significant alterations of either p21(Cip-1/WAF-1) (Figure 1b) or PAI-1 RNA levels (data notshown) in response to incubation with high concentra-tions of recombinant active TGF-b. In contrast, twowell-known TGF-b responsive control cell lines, HaCaTand MDA-MB-231, yielded the expected results:p21(Cip-1/WAF-1) and PAI-1 RNA levels were rapidlyand strongly induced by incubation with TGF-b1(Figure 1c), confirming that performance of the assaywas adequate.

Tumor suppressor Smad4 induces cadherinsN Muller et al

6050

Oncogene

We have previously provided evidence that lowexpression levels of TGF-b receptors may limit TGF-b responses in SW480 cells (Schwarte-Waldhoff et al.,1999). Smad4 re-expressing SW480 cells displayedmoderate responses in transient transfections withp3TP-Lux, the most commonly used reporter tomeasure TGF-b responsiveness. When a plasmidencoding a constitutively active TGF-b receptor wascotransfected the responses were significantly increased.Smad4 negative SW480 cells still did not respond top3TP-Lux transfection, whether the receptor was co-expressed or not. Here, we investigated endogenousexpression levels of the classical TGF-b receptorsthrough sensitive ribonuclease protection analysis.Expression levels of the TGF-b type I receptor werehigh, however, RNAs encoding the type II receptorwere virtually undetectable in SW480 cells (Figure 1d).Thus, retained TGF-b resistance in Smad4-re-expres-sing SW480 clones may be due to low expression levelsof TGF-b RII.

Re-expression of Smad4 induces the expression ofE-cadherin and P-cadherin in SW480 cells

We have shown that restoration of TGF-b responses isdispensable for Smad4-mediated tumor suppression.To address TGF-b independent mechanisms thatunderlie Smad4-mediated tumor suppression in stableSW480 revertants, we further investigated constitutivedifferences between Smad4-negative and Smad4-re-expressing SW480 cells. We have previously reportedthat Smad4 restoration induced significant alterationsin cellular morphology in vitro (Schwarte-Waldhoff etal., 1999). SW480 parental cells as well as Smad4-negative vector transfectants displayed a spindle shapeor rounded morphology and a criss-cross growthpattern at confluency (Figure 2a). In contrast, Smad4revertant clones consistently showed a conversion intoan epithelioid morphology: cells were well-attached,grew in clusters of interconnected cells at low densitiesand formed epithelial-like layers when reachingconfluency (Figure 2a). As intercellular contacts inepithelial cells are primarily mediated by E-cadherin,this observation prompted us to analyse expression ofE-cadherin. In dense cultures of Smad4-negativeSW480 control clones, virtually no E-cadherin reactiv-ity was detected by indirect immunofluorescence(Figure 2b). In contrast, Smad4-positive SW480 cellsdisplayed a honeycomb-like pattern of E-cadherinexpression at sites of cell – cell contacts which is typicalfor epithelial cell layers.

E-cadherin protein levels were consistently increasedin all independent Smad4-positive SW480 clonesanalysed indicating that E-cadherin induction was a

Figure 1 Northern blot analysis of TGF-b response genes inSW480 and control cell-lines and RPA analysis of TGF-b recep-tor expression. (a) Constitutive expression levels of p21 andPAI-1 in Smad4 negative and Smad4-re-expressing SW480clones, (b) p21 expression in response to TGF-b treatment inSmad4 negative and Smad4 re-expressing SW480 clones. Allclones retain TGF-b resistance irrespective of their Smad4status. (c) p21 and PAI-1 expression in response to TGF-b inHaCaT and MDA-MB 231 control cell lines. (d) Expression ofTGF-b RI and RII by RPA analysis. Relative levels of RNAloading are shown by rehybridization with a GAPDH probeor through simultaneous GAPDH hybridization in riboquantRPA analysis

Figure 2 Morphology and E-cadherin expression in Smad4-nega-tive and Smad4-re-expressing SW480 cells. (a) Phase contrastmicrophotograph of a representative Smad4 negative controlclone and a Smad4-re-expressing clone at confluency. (b) Expres-sion of E-cadherin as analysed by indirect immunostaining andconfocal microscopy

Tumor suppressor Smad4 induces cadherinsN Muller et al

6051

Oncogene

direct result of Smad4 re-expression rather than aneffect of clonal selection (Figure 3a). Northernblotting revealed strongly increased steady-stateRNA levels E-cadherin in Smad4-positive SW480cells, suggesting that Smad4 operates at the level oftranscription (Figure 3b). This was confirmed inpromoter analyses using transient transfection analyseswith an E-cadherin promoter luciferase construct. E-cadherin promoter activity proved to be significantlyincreased in Smad4-positive SW480 cells, suggestingthat E-cadherin induction is due to increased tran-scription (Figure 3c).

Expression analysis of other classical cadherinsrevealed that P-cadherin was also strongly induced

through Smad4 re-expression (Figure 4). N-cadherin-specific transcripts were not detected, either in Smad4-negative or in Smad4-positive clones (data notshown).

Smad4-upregulated cadherins are functional

Cadherins are transmembrane proteins whose extra-cellular domains interact in a zipper-like fashion toconnect adjacent cells. Critical for cell – cell adhesion isthe interaction of the intracellular cadherin domainwith the catenins, a group of proteins which linkcadherins to the actin cytoskeleton. Thus, we analysedcatenin expression and localization through indirectimmunofluorescence. Smad4-negative SW480 cellsshowed strong staining for a- and b-catenin in thecytoplasm and the nucleus (Figure 5a). Smad4 re-expressing cells, in contrast, appeared depleted ofcytoplasmic and nuclear catenins and showed stronga- and b-catenin staining at the plasma membranecolocalizing with E-cadherin (Figure 5a). To assess thesubcellular distribution of catenins in a more quan-titative way, membrane and soluble cell fractionswere separated by concanavalin-A precipitation andanalysed by Western blotting (Figure 5b). E-cadherinwas used as a control for membrane association. Thisanalysis confirmed that a- and b-catenins wererecruited to the plasma membrane in Smad4-positivecells, presumably through interaction with the cadher-ins. Most of the cellular catenins, however, remained inthe soluble fraction (note that different portions ofmembrane and soluble fractions are loaded on the gel).

Figure 3 Smad4 impact on E-cadherin expression. (a) Westernblot analysis of E-cadherin expression. Relative levels of proteinloading are shown by b-tubulin staining of the lower parts ofthe membranes. (b) Northern blot analysis of E-cadherin expres-sion. Relative levels of RNA loading are shown by rehybridiza-tion with a GAPDH probe. (c) E-cadherin promoter activity inSW480 cells as analysed in transient transfections with an E-cad-herin promoter reporter construct. Luciferase activity was nor-malized versus b-galactosidase activity. Values were determinedin triplicate and bars show the mean values and the standard de-viation. Similar results were obtained in more than three experi-ments

Figure 4 Smad4 impact of P-cadherin expression. (a) Westernblot analysis of P-cadherin expression. (b) Northern blot analysis.Loading controls were performed as in Figure 3

Tumor suppressor Smad4 induces cadherinsN Muller et al

6052

Oncogene

Definite proof for functional competence of Smad4-induced cadherins was provided through cell aggrega-tion assays. Smad4-positive SW480 cells and Smad4-negative vector control cells, respectively, weredetached from cell culture dishes, replated as singlecell suspensions on bacterial plates and placed on agyratory shaker to keep cells in suspension. Whereasvector controls remained as single cells Smad4revertant cells formed multicellular aggregates within1 h (Figure 6a). Cellular aggregation of Smad4-positivecells was Ca2+-dependent (Figure 6a,b) stronglysupporting the argument that the observed effects aremediated by cadherins.

Smad4 induces E-cadherin in vivo

We have previously reported, that re-expression ofSmad4 in SW480 cells was adequate to suppress tumorgrowth after subcutaneous injection in nude mice.Smad4 negative SW480 cells produced progressivelygrowing tumors which reached a size of 1 cm indiameter after 5 – 8 weeks. No tumors derived fromSmad4 re-expressing cells could be detected after thesame period. Interestingly, we repeatedly detected verysmall tumors (1 – 4 mm in diameter) roughly 2 weeksafter injection of both Smad4 negative or Smad4positive SW480 cells. Tumors derived from Smad4positive cells completely disappeared 3 – 4 weeks afterinjection. As this behavior was highly reproducible, wesacrificed mice 2 weeks after injection of the cells tostudy these ‘transient’ tumors compared to tumorsderived from Smad4 negative cells.

Smad4 negative SW480 cells formed poorly differ-entiated tumors with a predominantly solid growthpattern (Figure 7a). On the other hand, tumors derived

from Smad4 positive cells displayed an increasedproportion of stromal constituents and showed atendency towards tubular differentiation (Figure 7a).Both exhibited abundant surrounding fibrous tissuecontaining mesenchymal as well as inflammatory cells.Expression of Smad4 and E-cadherin in vivo wasanalysed by immunohistochemistry. Smad4 re-expres-sing SW480 tumor cells displayed moderate expressionlevels of nuclear and cytoplasmic Smad4 (Figure 7b).Murine stromal and inflammatory cells were alsoSmad4 reactive, since the antibody also detects themurine Smad4 protein. In tumors derived from Smad4negative SW480 cells scattered Smad4 positive cells ofmurine origin were also detected, but the carcinomacells did not show any immunoreactivity (Figure 7b).Importantly, immunohistochemical staining for E-cadherin expression showed strong differences in E-cadherin expression levels in vivo: E-cadherin stainingwas barely seen in Smad4 negative tumor cells (Figure7c). In contrast, membrane-associated E-cadherin wasreadily detected in Smad4 positive carcinoma cells,confirming that Smad4-dependent induction of E-cadherin expression was functional in vivo (Figure 7c).

Discussion

Through stable re-expression of Smad4 we haveinvestigated the cellular and molecular mechanismsthat underlie Smad4 function as a tumor suppressor.Here we show that re-expression of Smad4 in SW480human colon carcinoma cells, adequate to mediatetumor suppression in vivo, is associated with theinduction of functionally competent cadherins, withrestoration of cell – cell adhesion and with morpholo-

Figure 5 Recruitment of a- and b-catenin to the plasma membrane. (a) Localization of a- and b-catenin and E-cadherin as ana-lysed by indirect immunostaining and confocal microscopy. (b) Subcellular distribution of the catenins as analysed by Western blot-ting of ConA-fractionated lysates. P: plasma membrane fraction; S: soluble fraction. In Smad4-positive cells, the membraneousportions of catenins are slightly increased as compared to Smad4-negative cells. Note, however, that the plasma membrane fractionsare overrepresented by threefold (a-catenin) and sixfold (b-catenin), respectively

Tumor suppressor Smad4 induces cadherinsN Muller et al

6053

Oncogene

gical reversion to an epithelioid phenotype. Thesefindings indicate a novel mechanism for Smad4-mediated tumor suppression and suggest that loss ofSmad4 in colon carcinogenesis may be associated withloss of epithelial traits and with the acquisition ofinvasive and metastatic capabilities of incipient tumorcells.

We have previously shown that stable re-expressionof Smad4 is adequate to suppress tumor growth innude mice whereas it does not affect cell proliferationin vitro (Schwarte-Waldhoff et al., 1999, 2000). To ruleout potential artefacts due to Smad4 overexpression wecarefully compared Smad4 protein levels in SW480 cellclones to endogenous Smad4 levels of various normaland transformed cell lines (Schwarte-Waldhoff et al.,

1999 and data not shown). Smad4 levels in recon-stituted SW480 cells were among the lowest detected.Moreover, in nude mouse tumors derived from Smad4re-expressing SW480 cells, Smad4 protein expressionwas similar in human epithelial cells and mousestroma, whereas in the normal mouse gut as in humansepithelial cells stain more intensely as compared to thestromal constituents (data not shown). Our findings arealso not due to a selection bias, as all Smad4dependent changes are consistently found in eachindependent SW480 clone analysed. Initial transfectionefficiencies using the Smad4 expression vector and theempty vector control were very similar, thus excludinga general selection bias.

The failure to restore TGF-b responsiveness throughstable re-expression of Smad4 suggests that TGF-bresistance of SW480 cells is not due to Smad4deficiency. SW480 human colon carcinoma cells haveaccumulated a number of genetic alterations character-istic for colon carcinomas, among them the mutationalinactivation of the tumor suppressor gene APC, anactivating mutation of the Ki-ras oncogene and amutated p53 gene, all of which had previously been

Figure 6 Analysis of cell aggregation. (a) Representative micro-photographs of Smad4-negative and Smad4-expressing SW480cells in aggregation assays, incubated in the absence or presenceof EGTA. Results were confirmed in at least three independentexperiments. (b) Quantification of aggregate formation. The num-ber of clusters (as defined by 55 cells per aggregate) per totalnumber of objects is plotted. Bars show mean values of four ran-domly chosen fields of Smad4-negative and Smad4-expressingSW480 cells

Figure 7 Growth pattern, Smad4 and E-cadherin expression insubcutaneous nude mouse tumors derived from Smad4-negativeand Smad4-re-expressing cells. (a) Smad4-negative cells formedsolid tumors, whereas Smad4-re-expressing SW480 cells weregrouped within the tumor tissue and showed features of beginningpolarization (H&E staining). (b) Smad4 immunoreactivity presentin the cytoplasm and nucleus of Smad4-re-expressing SW480 car-cinoma cells (closed arrows). Murine stromal cells (open arrows)as well as inflammatory cells (mostly lymphocytes, arrowheads)were also Smad4 reactive. (c) E-cadherin was virtually absentfrom Smad4 negative tumors, but showed a strong membraneousexpression on Smad4 re-expressing carcinoma cells in vivo. Mag-nification: 1006 and 2006 (inserts)

Tumor suppressor Smad4 induces cadherinsN Muller et al

6054

Oncogene

implicated in the control of TGF-b responsiveness. Wehave here provided evidence that expression levels ofthe TGF-b type II receptor may be limiting TGF-bresponsiveness. Interestingly, inactivation of both theSmad4 gene and a TGF-b receptor gene have beenreported to occur together both in pancreatic and incolorectal tumors (Kern, 1998). Also, nuclear localiza-tion of Smad4, indicating transcriptional activity, hasbeen reported in human carcinomas which lack afunctional TGF-b receptor (Montgomery et al., 2001),suggesting that TGF-b responsiveness is not necessaryfor Smad4 functions as transcriptional regulator.Smad4 was also localized nuclear in transient nudemouse tumors derived from Smad4 reconstitutedSW480 cells studied here, indicating that the Smad4protein is functionally active in these cells in the in vivoenvironment.

Focusing on Smad4 dependent changes in theabsence of exogenous administration of ligand, weshow here that Smad4 upregulates E-cadherin and P-cadherin. The role of E-cadherin in cancer developmentand progression has been extensively investigated andits status as a suppressor of invasion and metastasis isfirmly established based on a number of immunohis-tochemical, functional and genetic approaches(Behrens, 1999; Christofori and Semb, 1999; Guilford,1999; Nollet et al., 1999; Wijnhoven et al., 2000). Inhuman cancers, E-cadherin can be inactivated throughgenetic mechanisms including germline mutations ininherited diffuse-type gastric cancers and somaticmutations in lobular breast cancers (Berx et al.,1998). In addition, functional inactivation due topromoter hypermethylation has been described. Inmany cancer types, however, mechanisms for E-cadherin downregulation remain elusive. Here weprovide evidence for a role of the tumor suppressorSmad4 in positively regulating E-cadherin suggestingthat functional inactivation of Smad4 during carcino-genesis may represent a novel mechanism for E-cadherin reduction in invasive tumors.

How Smad4 mediates increased expression of E-cadherin and P-cadherin at the molecular level is notfully understood. We propose that Smad4 acts at thelevel of transcription because the activity of an E-cadherin promoter construct was increased in Smad4re-expressing cells. Smad4 may thus act on thecadherin promoter directly or indirectly either bycompetition for transcriptional coregulators or byinduction or reduction of other transcription factorsregulating cadherins. E-cadherin transcriptional regula-tion is complex and presumably involves the combinedactivities of positive and negative regulators binding todifferent regulatory sequence elements in the proximalpromoter region and in intron 1 (Bussemakers et al.,1994; Giroldi et al., 1997; Hennig et al., 1995, 1996).The E-boxes have been investigated in detail and havepreviously been implicated in silencing E-cadherinexpression in tumor cell lines and in fibroblasts(Hennig et al., 1995). A number of transcriptionfactors have been identified that may function inrepression of E-cadherin, among them Snail (Batlle et

al., 2000; Cano et al., 2000), SIP-1 (Smad interactingprotein-1) (Comijn et al., 2001; Verschueren et al.,1999) and SLUG (Hajra et al., 2002; Savagner et al.,1997), all of which bind to the E-box elements. Smad4might directly or indirectly impinge on (one of) thesefactors and thus relieve E-cadherin from repression.

Smad4-induced cadherins proved to be functionallycompetent in recruiting catenins to the plasmamembrane and in mediating cell – cell adhesion inaggregation assays. As loss of intercellular adhesion isa prerequisite for invasive tumor growth, these findingssuggest an association of Smad4 loss with theacquisition of invasiveness. This conclusion is alsoconsistent with recent results from detailed genetic andimmunohistochemical analyses of colorectal andpancreatic carcinomas which have assigned Smad4 lossto invasive and particularly metastatic stages (Luettgeset al., 2001; Maitra et al., 2000; Miyaki et al., 1999;Wilentz et al., 2000). In addition, a mouse knock-outmodel also suggests an association of Smad4 loss withthe acquisition of invasiveness. Whereas APC singleknock-out mice, a model for human familial adeno-matous polyposis (FAP), develop numerous benignpolyps in the small intestine and colon (Oshima et al.,1995), tumors in APC/DPC4 compound double knock-out mice frequently progressed into malignant lesionsinvading into the submucosa (Takaku et al., 1998).

Stable re-expression of Smad4 in SW480 cells, inaddition to increasing the levels of E-cadherin and P-cadherin, induced a complex reprogramming of geneexpression profiles (Schwarte-Waldhoff, unpublisheddata) and phenotypic properties. Cells are reverted toan epithelioid phenotype, showing features of polariza-tion (Reinacher-Schick et al., manuscript inpreparation). Thus, Smad4-re-expression may impingeon a number of cellular properties and several of thesemay contribute to Smad4’s tumor suppressor func-tion(s). As E-cadherin is a bona fide tumor suppressorwe propose that induction of cadherin expression playsa significant role in Smad4-mediated tumor suppres-sion. E-cadherin is supposed to exert tumor suppressoractivity through its role as a prime mediator ofadherens junctions in epithelia. Intercellular contactsprevent cells from detaching from the epithelial celllayer, an essential step in the acquisition of invasive-ness. Moreover, cell adhesion is intimately linked tosignaling cascades that regulate cell proliferation andsurvival in vitro and in vivo. E-cadherin tumorsuppressor activity could also result from its impacton b-catenin signaling activity through recruitment ofb-catenin to the plasma membrane and depletion of thesignaling competent cellular b-catenin pool. SW480cells, however, are known to express very high levels ofb-catenin presumably due to inactivation of the APCtumor suppressor. In Smad4-positive SW480 cells wedid show recruitment of a- and b-catenins to theplasma membrane but only a small portion of the totalcellular catenins was affected. In line with this, Smad4re-expression was not associated with decreased b-catenin/TCF signaling as determined by a reporterassay (data not shown).

Tumor suppressor Smad4 induces cadherinsN Muller et al

6055

Oncogene

In summary, our results indicate that Smad4’s tumorsuppressor functions are not necessarily confined to itsrole as a component of the linear TGF-b signalingpathway. We here provide evidence that Smad4 may beinvolved in transcription regulation of the tumor andinvasion suppressor E-cadherin and P-cadherin. Impor-tantly, Smad4-mediated induction of E-cadherin isretained in vivo in nude mouse tumors. Though themechanisms underlying suppression of tumor growth inSmad4 positive SW480 cells are not fully understood,the association of Smad4, a tumor suppressorfunctionally inactivated upon the acquisition ofinvasiveness, with regulation of the invasion suppressorE-cadherin is compelling and may stimulate furtherresearch to investigate the scope of these results.

Materials and methods

Cell cultures

SW480 and MDA-MB-231 cells were obtained from theAmerican Type Culture Collection (Rockville, MD, USA),HaCaT cells (Boukamp et al., 1988) were the kind gift of Drs PBoukamp and N Fusenig. Cells were maintained in DMEMsupplemented with antibiotics and 10% fetal calf serum(GIBCO). The full-length coding sequence of Smad4/DPC4was cloned into the pBK-CMV expression vector (Stratagene)and Smad4 re-expressing SW480 cell clones and negativecontrol transfectants were established by a standard calciumphosphate coprecipitation method as described (Schwarte-Waldhoff et al., 1999). Transfectants were maintained inmedium with geneticin (0.2 mg/ml; GIBCO). To analyseTGF-b responses serum-deprived cells were incubated withrecombinant TGF-b1 (5 ng/ml; R&D Systems) and RNA washarvested at the indicated time points.

Northern blot analysis and ribonuclease protection assay

RNA was isolated by acid phenol extraction or using acommercial kit (RNeasy; Quiagen). Northern blots andhybridizations were performed as described (Schwarte-Waldhoff et al., 1999). For loading control blots were strippedand reprobed for GAPDH. Ribonuclease protection assayswere performed with the ‘human cytokine receptor set hCR-4’(Pharmingen) according to the manufacturer’s protocol.Signals were quantified by PhosphorImager analysis (Packard).

Preparation of proteins, glycoprotein fractionation and Westernblot analysis

Cells were lysed in NP-40 lysis buffer (25 mM Tris HCl,pH 7.4, 0.5% NP-40, 100 mM NaCl, 1 mM EDTA) contain-ing a protease inhibitor cocktail (Roche) and 1 mM PMSF.Purification of glycoprotein bound and unbound fractions ofa- and b-catenin or E-cadherin was performed as described(Fagotto et al., 1996). In brief, 300 mg of total protein wereincubated overnight at 48C with 50 ml (*300 mg concanava-lin) concanavalin A (conA)-Sepharose (Sigma Aldrich). Totalprotein lysates or ConA-fractionated proteins were subjectedto standard SDS–PAGE and immunoblot analysis whichwas performed as described (Schwarte-Waldhoff et al., 1999).In case of a-catenin or b-catenin detection a threefold orsixfold larger proportion of the conA-bound fractioncompared to unbound fraction, respectively, were loaded onthe gel. The blots were incubated with monoclonal antibodies

against Smad4 (anti-Smad4 B8; dilution 1 : 500, Santa Cruz),E-cadherin (HECD-1, dilution 1: 1000, Zymed) and P-cadherin (NCC-CAD-299, dilution 1 : 500, Zymed) or withpolyclonal rabbit antiserum (anti-b-catenin or anti-a-catenin,diluted 1 : 5000 and 1 : 2000, respectively). As loading controlfor Western analysis, blots were co-stained for b-tubulin(TUV 2.1, dilution 1 : 1000, Sigma).

Confocal immunofluorescence microscopy

Cells were grown to confluency on glass coverslips, rinsedwith PBS and fixed with ice-cold 100% methanol for 5 min at7208C. The fixed cells were incubated with primaryantibodies HECD-1 (anti human E-cadherin, Zymed, dilution1 : 100 in PBS+1% BSA) and rabbit antisera (anti human b-catenin or anti human aE-catenin; dilution 1 : 200 inPBS+1% BSA) for 1 h followed by FITC or Cy3-conjugatedsecondary antibodies (FITC-coupled anti-mouse Ig, dilution1 : 200, Dianova; Cy3-coupled anti-rabbit Ig, dilution 1 : 1000,Dianova). Samples were mounted with Vectashield (VectorLaboratories, Burlinghame, CA, USA) and examined with aZeiss LSM 410 confocal laser-scanning immunofluorescencemicroscope (Carl Zeiss, Jena, Germany).

Aggregation assay

Cells were grown to 70 – 80% confluency on 10-cm cellculture dishes, washed and detached by incubation in PBSwithout Ca2+ and Mg2+ for 15 – 20 min. Cells werecentrifuged for 5 min at 800 r.p.m. and resuspended as singlecell suspensions by repeated pipetting. Cells at a concentra-tion of 56105 cells/ml in standard growth medium wereplated on bacterial plates and incubated at 378C on ahorizontal shaker at 100 r.p.m. for 1 h. Then, photographs ofrandomly chosen areas were taken with a digital camera. Thenumber of single cells, cell aggregates (as defined by morethan five cells per cluster) and the mean area of aggregateswere determined using ImagePro software. For determinationof calcium dependency of aggregate formation, cells wereincubated in HBSS medium containing 2% FCS plus either2 mM EGTA (pH 8.0) or 2 mM CaCl2.

E-cadherin reporter gene assay

SW480 cells (36105 cells per 10 cm2 well) with either no orreconstituted SMAD4 expression were cotransfected with500 ng human E-cadherin reporter plasmid (Comijn et al.,2001) and 500 ng internal standard, i.e. b-galactosidaseexpression plasmid pUT651 (Eurogentec). Three days aftertransfection, cells were harvested and cell extracts wereassayed for luciferase and b-galactosidase activity with theGalacto-Star kit (Tropix). E-cadherin promoter activity wasnormalized versus b-galactosidase activity.

Immunohistochemistry

Tumorigenicity assays in nude mice were performed asdescribed (Schwarte-Waldhoff et al., 1999). Subcutaneoustumor tissues were fixed in neutral-buffered formaldehyde(4%) and embedded in paraffin according to routineprocedures. Serial sections (4 mm thick) were stained withH&E and immunostained with mabs directed against Smad4(clone B8, dilution 1 : 100; Santa Cruz Biotechnology, SantaCruz, CA, USA) and E-cadherin (clone 5H9, dilution 1 : 40;Progen Biotechnik, Heidelberg, Germany). Antigen retrievalwas achieved by microwave treatment (665 min for Smad4;465 min for E-cadherin) and staining was performed as

Tumor suppressor Smad4 induces cadherinsN Muller et al

6056

Oncogene

described earlier (Baldus et al., 1998). Briefly, after blockingof endogenous peroxidase by 1% H2O2/methanol, normalswine serum (X0901, dilution 1 : 20; Dako, Hamburg,Germany) was added for 30 min at 378C. Primary mABswere incubated overnight at 48C, followed by biotinylatedrabbit-anti-mouse immunoglobulin (E354, dilution 1 : 400;Dako) for 30 min at room temperature (RT) and streptavi-din-peroxidase conjugate (P397; dilution 1:400, Dako) for30 min, RT. The reaction was visualized by 200 mg/ml 3-amino-9-ethyl-carbazol (Sigma, Munich, Germany) in 50 mM

sodium acetate buffer containing 5% dimethylformamide and0.1% H2O2 for 30 min, RT. Slides were counterstained withhematoxylin and mounted in glycerol jelly.

Abbreviations

conA, concanavalin A; DMEM, Dulbecco’s modified Eaglemedium; EDTA, ethylenediaminetetraacetic acid; EGTA,ethyleneglycoltetraacetic acid; FITC, fluorescein isothiocya-nate; HBSS, Hank’s balanced salt solution; IHC,immunohistochemistry; PMSF, phenylmethylsulphonyl fluor-

ide; RPA, ribonuclease protection assay; TGF-b,transforming growth factor beta

Acknowledgements

The authors thank S Landsberg for IHC, C Theiss forassistance with confocal microscopy, D Thomas for RPAanalysis and C Niessen for critically reading the manu-script. The polyclonal rabbit antisera against a- and b-catenin were a kind gift of BM Gumbiner (MemorialSloan-Kettering Cancer Center, NY, USA). This work wassupported by grants from the Deutsche Krebshilfe – DrMildred Scheel-Stiftung, from the Wilhelm Sander-Stiftungand from the Ruhr-Universitat Bochum, FoRUM program(to I Schwarte-Waldhoff, A Reinacher-Schick and WSchmiegel) and by grants from the FWO, the Geconcer-teerd Onderzoeksacties of Ghent University, and the FortisBank-Verzekeringen, Brussels, Belsium (to F van Roy). Jvan Hengel and G Berx are postdoctoral fellows with theFWOP (Fund for Scientific Research, Flanders).

References

Akhurst RJ and Balmain A. (1999). J. Pathol., 187, 82 – 90.Attisano L and Wrana JL. (2000). Curr. Opin. Cell Biol., 12,235 – 243.

Baldus SE, Zirbes TK, Engel S, Hanisch FG, Monig SP,Lorenzen J, Glossmann J, Fromm S, Thiele J, PichlmaierH and Dienes HP. (1998). Int. J. Cancer, 79, 133 – 138.

Batlle E, Sancho E, Franci C, Dominguez D, Monfar M,Baulida J and Garcia DHA. (2000). Nat. Cell Biol., 2, 84 –89.

Behrens J. (1999). Cancer Metast. Rev., 18, 15 – 30.Berx G, Becker K, Hofler H and van Roy F. (1998). Hum.Mutat., 12, 226 – 237.

Bhowmick NA, Ghiassi M, Bakin A, Aakre M, LundquistCA, Engel ME, Arteaga CL and Moses HL. (2001). Mol.Biol. Cell, 12, 27 – 36.

Blaker H, von Herbay A, Penzel R, Gross S and Otto HF.(2002). Oncogene, 21, 158 – 164.

Boukamp P, Petrussevska RT, Breitkreutz D, Hornung J,Markham A and Fusenig NE. (1998). J. Cell Biol., 106,761 – 771.

Bracke M, van Roy F and Mareel M. (1996). Curr. Top.Microbiol. Immunol., 213, 123 – 161.

Bussemakers M, Giroldi L, van Bokhoven A and Schalken J.(1994). Biochem. Biophys. Res. Commun., 203, 1284 – 1290.

Cano A, Perez-Moreno M, Rodrigo I, Locascio A, BlancoMJ, del Barrio MG, Portillo F and Nieto MA. (2000). Nat.Cell Biol., 2, 76 – 83.

Christofori G and Semb H. (1999). Trends Biochem. Sci., 24,73 – 76.

Comijn J, Berx G, Vermassen P, Verschueren K, vanGrunsven L, Bruyneel E, Mareel M, Huylebroeck D andvan Roy F. (2001). Mol. Cell, 7, 1267 – 1278.

Cui W, Fowlis DJ, Bryson S, Duffie E, Ireland H, Balmain Aand Akhurst RJ. (1996). Cell, 86, 531 – 542.

Dai JL, Schutte M, Bansal RK, Wilentz RE, Sugar AY andKern SE. (1999). Mol. Carcinog., 26, 37 – 43.

Derynck R, Akhurst RJ and Balmain A. (2001). Nat. Genet.,29, 117 – 129.

Ellenrieder V, Hendler SF, Boeck W, Seufferlein T, MenkeA, Ruhland C, Adler G and Gress TM. (2001). CancerRes., 61, 4222 – 4228.

Fagotto F, Funayama N, Gluck U and Gumbiner BM.(1996). J. Cell Biol., 132, 1105 – 1114.

Fink SP, Swinler SE, Lutterbaugh JD, Massague J,Thiagalingam S, Kinzler KW, Vogelstein B, Willson JKand Markowitz S. (2001). Cancer Res., 61, 256 – 260.

Giroldi LA, Bringuier PP, de Weijert M, Jansen C, vanBokhoven A and Schalken JA. (1997). Biochem. Biophys.Res. Commun., 241, 453 – 458.

Gold LI. (1999). Crit. Rev. Oncog., 10, 303 – 360.Guilford P. (1999). Mol. Med. Today, 5, 172 – 177.Gumbiner B. (1996). Cell, 84, 345 – 357.Hahn SA, Bartsch D, Schroers A, Galehdari H, Becker M,Ramaswamy A, Schwarte-Waldhoff I, Maschek H andSchmiegel W. (1998). Cancer Res., 58, 1124 – 1126.

Hahn SA, Schutte M, Hoque ATM, Moskaluk CA, da CostaLT, Rozenblum E, Weinstein CL, Fischer A, Yeo CJ,Hruban RH and Kern SE. (1996). Science, 271, 350 – 353.

Hajra KM, Chen DY and Fearon ER. (2002). Cancer Res.,62, 1613 – 1618.

Hennig G, Behrens J, Truss M, Frisch S, Reichmann E andBirchmeier W. (1995). Oncogene, 11, 475 – 484.

Hennig G, Lowrick O, Birchmeier W and Behrens J. (1996).J. Biol. Chem., 271, 595 – 602.

Hocevar BA, Brown TL and Howe PH. (1999). EMBO J., 18,1345 – 1356.

Kern SE. (1998). Curr. Opin. Oncol., 10, 74 – 80.Kingsley DM. (1994). Genes Dev., 8, 133 – 146.Kitagawa K, Murata A, Matsuura N, Tohya K, Takaichi S,MondenM and Inoue M. (1996). Int. J. Cancer, 66, 91 – 97.

Luettges J, Galehdari H, Brocker V, Schwarte-Waldhoff I,Henne-Bruns D, Kloppel G, Schmiegel W and Hahn SA.(2001). Am. J. Pathol., 158, 1677 – 1683.

Maitra A, Molberg K, Albores-Saavedra J and Lindberg G.(2000). Am. J. Pathol., 157, 1105 – 1111.

Markowitz S, Wang J, Myeroff L, Parsons R, Sun L,Lutterbaugh J, Fan RS, Zborowska E, Kinzler KW,,Vo-gelstein B, Brattain M and Willson JKV. (1995). Science,268, 1336 – 1338.

Massague J. (1998). Ann. Rev. Biochem., 67, 753 – 791.Massague J, Blain S and Lo R. (2000). Cell, 103, 295 – 309.

Tumor suppressor Smad4 induces cadherinsN Muller et al

6057

Oncogene

Miettinen PJ, Ebner R, Lopez AR and Derynck R. (1994). J.Cell Biol., 127, 2021 – 2036.

Miyaki M, Iijima T, Konishi M, Sakai K, Ishii A, Yasuno M,Hishima T, Koike M, Shitara N, Iwama T, Utsunomiya J,Kuroki T and Mori T. (1999). Oncogene, 18, 3098 – 3103.

Montgomery E, Goggins M, Zhou S, Argani P, Wilentz R,Kaushal M, Booker S, Romans K, Bhargava P, Hruban Rand Kern S. (2001). Am. J. Pathol., 158, 537 – 542.

Nollet F, Berx G and van Roy F. (1999). Mol. Cell Biol. Res.Commun., 2, 77 – 85.

Oft M, Heider KH and Beug H. (1998). Curr. Biol., 8, 1243 –1252.

Oshima M, Oshima H, Kitagawa K, Kobayashi M, ItakuraC and Taketo M. (1995). Proc. Natl. Acad. Sci. USA, 92,4482 – 4486.

Piek E, Moustakas A, Kurisaki A, Heldin CH and ten DijkeP. (1999). J. Cell Sci., 112, 4557 – 4568.

Piek E and Roberts AB. (2001). Adv. Cancer Res., 83, 1 – 54.Portella G, Cumming SA, Liddell J, Cui W, Ireland H,Akhurst RJ and Balmain A. (1998). Cell Growth Differ., 9,393 – 404.

Reiss M. (1997). Oncol. Res., 9, 447 – 457.Reiss M. (1999). Microbes. Infect., 1, 1327 – 1347.Roberts AB and Sporn MB. (1990). Peptide growth factorsand their receptors. Sporn MB and Roberts AB (eds).Heidelberg: Springer-Verlag, pp 419 – 472.

Savagner P, Yamada KM and Theiry JP. (1997). J. Cell Biol.,137, 1403 – 1419.

Schutte M, Hruban RH, Hedrick L, Cho KR, Molnar G,Weinstein G, Bova S, Isaacs WB, Cairns P, Nawroz H,Sidransky D, Casero B, Meltzer PS, Hahn SA and KernSE. (1996). Cancer Res., 56, 2527 – 2530.

Schwarte-Waldhoff I, Klein S, Blass-Kampmann S, Hintel-mann A, Eilert C, Dreschers S, Kalthoff H, Hahn SA andSchmiegel W. (1999). Oncogene, 18, 3152 – 3158.

Schwarte-Waldhoff I, Volpert O, Bouck N, Sipos B, Hahn S,Klein-Scory S, Luttges J, Kloppel G, Graeven U, Eilert-Micus C, Hintelmann A and Schmiegel W. (2000). Proc.Natl. Acad. Sci. USA, 97, 9624 – 9629.

Sirard C, Kim S, Mirtsos C, Tadich P, Hoodless PA, Itie A,Maxson R, Wrana JL andMak TW. (2000). J. Biol. Chem.,275, 2063 – 2070.

Takaku K, OshimaM,Miyoshi H, Matsui M, Seldin MF andTaketo MM. (1998). Cell, 92, 645 – 656.

ten Dijke P, Miyazono K and Heldin C. (2000). TrendsBiochem. Sci., 25, 64 – 70.

Verschueren K and Huylebroeck D. (1999). Cytokine GrowthFactor Rev., 10, 187 – 199.

Verschueren K, Remacle J, Collart C, Kraft H, Baker B,Tylzanowski P, Nelles L, Wuytens G, Su M, Bodmer R,Smith J and Huylebroeck D. (1999). J. Biol. Chem., 274,20489 – 20498.

Wijnhoven B, Dinjens W and Pignatelli M. (2000). Br. J.Surg., 87, 992 – 1005.

Wilentz RE, Iacobuzio-Donahue CA, Argani P, McCarthyDM, Parsons JL, Yeo CJ, Kern SE and Hruban RH.(2000). Cancer Res., 60, 2002 – 2006.

Woodford-Richens KL, Rowan AJ, Gorman P, Halford S,Bicknell DC, Wasan HS, Roylance RR, Bodmer WF andTomlinson IP. (2001). Proc. Natl. Acad. Sci. USA, 98,

9719 – 9723.Zavadil J, Bitzer M, Liang D, Yang YC, Massimi A, KneitzS, Piek E and Bottinger EP. (2001). Proc. Natl. Acad. Sci.USA, 98, 6686 – 6691.

Zhang Y and Derynck R. (1999). Trends Cell. Biol., 9, 274 –279.

Tumor suppressor Smad4 induces cadherinsN Muller et al

6058

Oncogene