Research Article 2135

IntroductionStudies of inherited and sporadic colorectal cancer (CRC) revealedthe Wnt/b-catenin pathway (Kinzler and Vogelstein, 1996) as theprimary mutation target, and mutations that lead to thehyperactivation of genes regulated by b-catenin–T-cell-factor (TCF)are among the first genetic changes and also a major cause forCRC development and progression (Clevers, 2006; Conacci-Sorrellet al., 2002a; Polakis, 2007). An important question in b-catenin-mediated oncogenesis is the identification and mode of action ofgenes targeted by b-catenin–TCF that are induced during invasiveCRC formation. In addition to the b-catenin target genes thatconfer a growth advantage in cancer cells – such as cyclin D1(Shtutman et al., 1999; Tetsu and McCormick, 1999) and MYC (Heet al., 1998) – the accumulation of b-catenin in the nuclei of cancercells continues with disease progression and is most evident at theinvasive edge of CRC tissue (Brabletz et al., 1998; Brabletz et al.,2001). This indicates that b-catenin signaling also activates genesthat are involved in later stages of CRC development. The changesin molecular and cellular properties that occur at the invasive frontof CRC tissue cells are reminiscent of epithelial-to-mesenchymaltransition (EMT) (Brabletz et al., 2005; Conacci-Sorrell et al.,2003; Gavert and Ben-Ze’ev, 2008; Thiery, 2002; Thiery andSleeman, 2006), but the mechanisms driving these changes remainlargely unknown. In recent studies, we identified genes encodingmembers of the neuronal L1-CAM family of immunoglobulin-likecell-adhesion receptors (the neural cell-adhesion moleculeL1CAM – hereafter referred to as L1, and the neuron-glia-relatedcell-adhesion molecule NRCAM – hereafter referred to as Nr-CAM) as targets of b-catenin–TCF signaling in CRC cells

(Conacci-Sorrell et al., 2002b; Gavert et al., 2005). We detected L1in a small population of CRC cells at the invasive tumor front thatdisplayed nuclear b-catenin (Gavert et al., 2005). Furthermore,forced expression of L1 in human CRC cells promotes their motilityin vitro and metastatic capacity to the liver when injected into thespleen of nude mice (Gavert et al., 2007). However, the mechanismsby which L1 confers these dramatic changes in the behavior ofCRC cells remain unclear.

The nuclear factor kB (NF-kB), known for its capacity toactivate genes involved in immune response and inflammation butalso in controlling oncogenesis (Bassères and Baldwin, 2006; Karinet al., 2002; Naugler and Karin, 2008), was recently implicated inboth EMT and metastasis (Huber et al., 2004; Julien et al., 2007;Min et al., 2008; Solanas et al., 2008). In this study, we askedwhether signaling by NF-kB is involved in eliciting the cellularchanges induced by L1 that lead to invasive CRC development.We found that L1 induces NF-kB signaling in CRC cells and showthat, by inhibiting NF-kB signaling in L1-expressing CRC cells, itis possible to block the metastatic capacity of these cells. We alsofound that this process requires the cytoskeletal crosslinking proteinezrin. Ezrin is a member of the ERM (ezrin, radixin, moesin)family of actin-associated proteins that have a key role in theformation of microvilli and other membrane protrusions (Bretscheret al., 2002). Several studies implicated the involvement of ezrinin the metastatic spread of various neoplasms (Khanna et al., 2004;Yu et al., 2004). We found that ezrin forms a complex with L1 andIkB in the juxtamembrane region, and that suppression of ezrinblocks the metastatic capacity of CRC cells. We demonstrate thepresence of ezrin, L1 and the more active phosphorylated subunit

Nuclear factor-kB signaling and ezrin are essential forL1-mediated metastasis of colon cancer cellsNancy Gavert1,*, Amir Ben-Shmuel1,*, Vance Lemmon,2 Thomas Brabletz3 and Avri Ben-Ze’ev1,‡

1Department of Molecular Cell Biology, Weizmann Institute of Science, Hertzel str. 1, Rehovot, 76100, Israel2University of Miami Miller School of Medicine, 1095 NW 14th Terrace, Miami, FL 33136, USA3Department of Visceral Surgery, Hugstetterstr. 55, University of Freiburg, Freiburg, 79095, Germany*These authors contributed equally to this work‡Author for correspondence (avri.ben-zeev@weizmann.ac.il)

Accepted 26 March 2010Journal of Cell Science 123, 2135-2143 © 2010. Published by The Company of Biologists Ltddoi:10.1242/jcs.069542

SummaryHyperactivation of b-catenin–T-cell-factor (TCF)-regulated gene transcription is a hallmark of colorectal cancer (CRC). The cell-neuraladhesion molecule L1CAM (hereafter referred to as L1) is a target of b-catenin–TCF, exclusively expressed at the CRC invasive frontin humans. L1 overexpression in CRC cells increases cell growth and motility, and promotes liver metastasis. Genes induced by L1are also expressed in human CRC tissue but the mechanisms by which L1 confers metastasis are still unknown. We found thatsignaling by the nuclear factor kB (NF-kB) is essential, because inhibition of signaling by the inhibitor of kB super repressor (IkB-SR) blocked L1-mediated metastasis. Overexpression of the NF-kB p65 subunit was sufficient to increase CRC cell proliferation,motility and metastasis. Binding of the L1 cytodomain to ezrin – a cytoskeleton-crosslinking protein – is necessary for metastasisbecause when binding to L1 was interrupted or ezrin gene expression was suppressed with specific shRNA, metastasis did not occur.L1 and ezrin bound to and mediated the phosphorylation of IkB. We also observed a complex containing IkB, L1 and ezrin in thejuxtamembrane region of CRC cells. Furthermore, we found that L1, ezrin and phosphorylated p65 are co-expressed at the invasivefront in human CRC tissue, indicating that L1-mediated activation of NF-kB signaling involving ezrin is a major route of CRCprogression.

Key words: Colon cancer, L1, Metastasis, NF-kB, IkB, Ezrin

Jour

nal o

f Cel

l Sci

ence

p65 (p65-P) of NF-kB in human CRC tissue cells at the invasivefront of tumors, and suggest that induction of NF-kB signaling andezrin are key components in the mechanism(s) by which L1promotes invasive metastatic colon cancer development.

ResultsExpression of L1 in CRC cells induces NF-kB signaling byenhancing IkB phosphorylationTo determine whether L1 expression in CRC cells enhancessignaling by NF-kB, a NF-kB-responsive reporter plasmidcontaining three copies of the NF-kB DNA-binding sequencelinked to the luciferase gene (3xkB.luc) was introduced into Ls174Tcontrol cells and Ls174T cells stably transfected with L1. Levelsof luciferase activity in these cells were then measured. Twoindependently derived L1-expressing cell clones (Gavert et al.,2007) displayed higher levels of NF-kB transactivation comparedwith control Ls174T cells that lacked L1 (Fig. 1A). L1 also induceda dramatic increase in NF-kB activity in HEK293T cells (Fig. 1B),indicating that its capacity to induce NF-kB signaling is notrestricted to a single cell type. We also compared NF-kB signalingin SW620 CRC cells (Fig. 1C), which express L1endogenously,before and after suppressing L1 levels with specific small hairpinRNA (shRNA; Fig. 1D) and found that NF-kB signaling wasreduced when L1 levels were suppressed (Fig. 1C). SW620 cellswhose L1 levels were suppressed also displayed decreasedtumorigenic capacity when injected subcutaneously into nude mice(supplementary material Fig. S1).

A key step in the activation of NF-kB is the phosphorylation ofthe inhibitor of kB (IkB) leading to the release of NF-kB for

nuclear translocation, and the polyubiquitylation and degradationof pIkB by the proteasome (Bassères and Baldwin, 2006; Karinand Ben-Neriah, 2008). In accordance with this mechanism, wedetected in L1-expressing cells an increase in IkB phosphorylation(Fig. 1E; IkB-P, lane 4). Naturally, pIkB could only be detectedwhen its degradation by the proteasome was inhibited by MG132(Fig. 1E, lane 4). Employing immunofluorescence microscopy, wealso observed the nuclear accumulation of p65-P, a phosphorylated(more active form) of the p65 NF-kB subunit (Perkins, 2006), inL1 expressing CRC cells (Fig. 1F, arrowheads), but not in controlcells displaying diffuse cytoplasmic p65-P distribution. Theseresults suggest that L1 expression induced nuclear translocationand increased activity of NF-kB signaling by a mechanisminvolving the classic pathway of IkB phosphorylation.

IkB-SR binds to and colocalizes with L1, thereby inhibitingCRC cell proliferation, motility and L1-mediatedmetastasisL1 could be affecting NF-kB signaling by binding to IkB andreleasing NF-kB for signaling in the nucleus. We wished to determinewhether L1 binds to IkB thereby affecting NF-kB signaling. Weisolated clones of CRC cells that stably expressed both L1 and theIkB super repressor (IkB-SR) – a stable point mutant of IkB (Fig.2A, lanes 4,5) – and found that in such cells the L1-mediatedinduction of NF-kB signaling was suppressed (Fig. 2B). We thenexamined the localization of L1 and IkB-SR in CRC cells that stablyexpressed both proteins. Whereas IkB-SR was only localized incytoplasmic aggregates in cells that expressed IkB-SR alone (Fig.2C, left panel), cells that expressed both L1 and IkB-SR, showed

2136 Journal of Cell Science 123 (12)

Fig. 1. L1 induces NF-kB signaling and thephosphorylation of IkB in colon cancer cells. (A) TheNF-kB-responsive reporter plasmid 3xkB.luc wastransfected together with pSV b-galactosidase controlvector (for transfection efficiency normalization) intohuman Ls174T CRC cells that do not express L1, andinto two individually selected clones stably transfectedto express L1 (L1 Cl1 and Cl2) and a control clone(pcDNA3). Fold NF-kB activation was determined afterdividing luciferase activity by the values obtained withan empty reporter plasmid. (B) 293T cells weretransfected as in A with either an empty pcDNA3plasmid or with the same plasmid containing L1.Luciferase activity was then determined. (C) HumanSW620 CRC cells that express endogenous L1, twoindividually selected clones stably transfected withshRNA targeting L1 (shRNA L1 Cl1 and Cl2) and acontrol clone (control) were transfected as in A, andluciferase activity was determined. (D) The cellsdescribed in C were analyzed for L1 RNA and proteinlevels by PCR and western blotting (bottom and topblots, respectively. (E) Ls174T cells stably transfectedwith the empty pcDNA3 vector or vector expressing L1were either left untreated (lanes 1,3) or were treatedwith the proteasome inhibitor MG132 for 3 hours beforeharvesting (lanes 2,4). Levels of IkB total (IkB ) andphosphorylated IkB (IkB-P) were determined bywestern blotting. Tubulin served as loading control. (F) Ls174T cells stably transfected with either pcDNA3or vector expressing L1 were immunostained with anti-p65-P and anti-L1 antibodies (red and green staining,respectively). Nuclei were stained with DAPI (blue).Scale bar: 20 mm. pIkB is IkB-P and pp65 is p65-P.

Jour

nal o

f Cel

l Sci

ence

IkB-SR localization also in the juxtamembrane region together withL1 (Fig. 2C, right panel and merged image). In addition, in co-immunoprecipitation experiments (using anti-L1 antibody), we

identified a complex containing both IkB-SR and L1 (Fig. 2D, lane8), and also wild-type (wt) IkB in complex with L1 (Fig. 4E),suggesting that L1 can interact with IkB and thereby affects NF-kBsignaling. We determined the consequences of blocking NF-kBsignaling in L1-expressing CRC cells (by IkB-SR) on several cellularproperties that are altered by L1 (Gavert et al., 2005; Gavert et al.,2007). We found that proliferation of cells grown in 0.5% serum

2137NF-kB and ezrin in L1-induced CRC metastasis

Fig. 2. IkB-SR binds to and colocalizes with L1, inhibiting CRC cellproliferation, motility and L1-mediated metastasis of CRC cells to theliver. (A) pcDNA3 empty plasmid or L1-transfected Ls174T cell clones, werefurther transfected with the puror plasmid or the puror plasmid containing IkB-SR. Western blotting of these cell lines identified two cell clones thatexpressed both L1 and IkB-SR (lanes 4,5). (B) NF-kB activation in the variouscell clones was determined as described for Fig. 1A. (C) The subcellularlocalization of cells stably expressing L1, IkB-SR or L1 and IkB-SR togetherwas determined by double immunofluorescence microscopy with antibodiesagainst L1 (green, aL1) and IkB (red, aIkB-SR). Scale bar: 20 mm.(D) Lysates of 293T cells transfected with L1, IkB-SR or both wereimmunoprecipitated (IP) with antibodies against L1. The immunoprecipitatedproteins were analyzed by western blotting with antibodies against L1 andIkB. (E) The proliferation of Ls174T CRC cell clones (described in A) in thepresence of 0.5% serum was determined in quadruplicate during 6 days. (F) An artificial wound was introduced into confluent monolayers of Ls174Tcell clones (described in A), and the extent of wound closure in four differentwounds for each cell clone was determined after 18 hours. (G) The CRC cellclones (described in A) were injected into the spleen of nude mice, and tumorgrowth at the site of injection (spleen) and formation of metastases (liver) weredetermined. Arrows point to tumors formed at the site of injection in the spleenand arrowheads to large macrometastases in the liver of mice.

Fig. 3. Overexpression of the NF-kB p65 subunit promotes cellproliferation and motility, and confers metastasis in human CRC cells.(A) Two cell clones that stably expressed p65 (p65 Cl1 and Cl2) or pcDNA3(western blot, left panel) and also L1-transferred cells (Fig. 2A, lane 3) weretransiently transfected with 3xkB.luc and b-galactosidase together. Activationof NF-kB was determined (bar graph, right panel) as described for Fig. 1A.(B) Proliferation (in 0.5% of serum) of pcDNA3-transfected cells and the twop65-transfected cell clones was compared. (C) The capacity of a p65-transfected cell clone to close an artificial wound was compared with that ofcontrol Ls174T cells after 24 hours. The same areas were photographedimmediately after wounding (0 hours) and 24 hours later. (D) The motility oftwo p65 overexpressing cell clones was compared with that of control Ls174Tcells. Wound closure was determined simultaneously in four wounds for eachcell line. (E) The metastatic capacity of Ls174T cells stably transfected withpcDNA3, L1 or p65 was determined as described for Fig. 2G. (F) Sections oflivers stained with hematoxylin and eosin show tumor metastases in mice thathad been injected with either L1- or p65-expressing CRC cells. Scale bar:250 mm.

Jour

nal o

f Cel

l Sci

ence

(Fig. 2E) and cell motility (Fig. 2F) (n=18, P<0.0001) are bothinhibited in L1-expressing cells when NF-kB signaling was blockedby IkB-SR.

To examine whether NF-kB signaling is required for theinduction of metastasis conferred by L1 in CRC cells, we injectedinto the spleen of nude mice control cells, Ls174T cells thatexpressed IkB-SR alone, or L1 either alone or in combination withIkB-SR, and determined their ability to grow at the site of injectionand to form metastases in the liver. Whereas the injection of L1-expressing cells resulted in the formation of large liver metastasesin all the injected animals (100%, n=9, P<0.05), co-expression ofIkB-SR in two independently derived L1-expressing cell clonescompletely blocked their metastasis (100%; n=13, P<0.001) (Fig.2G). Cells expressing IkB-SR alone (Fig. 2A, lane 2) did not formmetastases (Fig. 2G), similar to control Ls174T cells. By contrast,all cell clones formed tumors to varying extents at the site ofinjection in the spleen, irrespective of their metastatic potential; afinding that has previously been reported (Gavert et al., 2007). Weconclude that signaling by NF-kB is essential for the ability of L1to confer a variety of cellular properties associated with CRC cellinvasion and metastasis.

Overexpression of the p65 NF-kB subunit in CRC cellsconfers enhanced proliferation, increased motility andinduces liver metastasisTo more closely link the tumorigenic and metastatic capacitiesconferred by L1 in CRC cells to NF-kB signaling, we askedwhether this downstream activation of the NF-kB pathway (in theabsence of L1) is sufficient to induce the cellular characteristicsseen when L1 is expressed. We stably transfected the NF-kB p65subunit in Ls174T cells (that lack L1) (Gavert et al., 2005; Gavertet al., 2007) and isolated two independent cell clones thatoverexpressed the p65 subunit (Fig. 3A, left panel). These cellsdisplayed strong NF-kB signaling (Fig. 3A, right panel) . Similarto L1-transfected cells (Fig. 2E), the p65-overexpressing cells grewbetter in 0.5% serum than control cells (Fig. 3B). Expression ofp65 in these cells dramatically enhanced their capacity to close anartificial wound introduced in a confluent monolayer; cells werealso more motile than control Ls174T cells (Fig. 3C,D) (n=9,P<0.0001). This increase was similar to that observed after L1transfection (Fig. 2F). The injection of p65-overexpressing Ls174Tcells into the spleen of nude mice resulted in the formation ofmetastases (Fig. 3E; 50%, n=15, P<0.08), but these were markedlysmaller than those seen in L1-transfected cells. This might havebeen the result of insufficient levels of the p52 subunit, which isneeded for the optimal activity of the p65-p52 NF-kB heterodimer.Histological sections showed that the gross morphology of themetastatic tissue formed in p65-transfected cells wasindistinguishable from that formed in L1-overexpressing cells (Fig.3F).

A point mutation at Tyr1151 within the L1 cytodomainabolished its capacity to induce NF-kB signaling, cellproliferation, tumorigenesis and metastasisTo begin to unravel the molecular pathways through which L1mediates the activation of NF-kB signaling, we employed severaldeletion and point mutations in the L1 cytodomain sequence, whichis known to interact with various partners (Fig. 4A) (Cheng et al.,2005). The capacity of these L1 mutants to induce NF-kB activationwas determined in 293T cells (Fig. 4B). The results showed thatlarge deletions starting at amino acids (aa) 1176 or 1180 in the L1cytodomain (L1-1176 or L1-1180, respectively) (Fig. 4A) reducedthe capacity of L1 to activate NF-kB. However, four pointmutations of lysine, lysine, tyrosine and valine to alanine within

2138 Journal of Cell Science 123 (12)

Fig. 4. The juxtamembrane L1 cytodomain contains a tyrosine residue atposition 1151 that is necessary for L1-mediated cancer progression.(A) Diagram of the L1 cytoplasmic domain in wt and mutant molecules. (B) The NF-kB-responsive reporter plasmid 3xkB.luc was transfected into293T cells together with an empty pcDNA3 plasmid (control), or with thesame plasmid expressing wt L1 or mutant L1 in the cytodomain. ConstructsL1-1176 and L1-1180 express truncated versions of L1, whereas constructsL1-4A and L1-Y1151A express L1 point mutants. (C) Cells stably transfectedwith either pcDNA3 or L1-Y1151A were immunostained with anti-L1antibody, and their levels of expression were determined by western blottingwith antibodies against L1 and IkB (D). (E) Lysates of 293T cells transfectedto express L1 or L1-Y1151A together with IkB-SR were immunoprecipitated(IP) with antibodies against L1 and analyzed by western blotting withantibodies against L1 and IkB. (F) Ls174T cells stably transfected with emptypcDNA3 vector, or plasmids expressing L1, L1-Y1151A or the p65 NF-kBsubunit were either left untreated (lanes 1,3,5,7) or were treated with theproteasome inhibitor MG132 for 3 hours before cell harvesting (lanes 2,4,6,8),and levels of phosphorylated IkB (IkB-P) were determined. (G) Proliferationof CRC cell clones described in C in the presence of 10% or 0% serum (H)was determined during 7 days in quadruplicate. (I) Cell clones described in Cwere injected into the flanks of nude mice – on one side L1-expressing cellsand on the other side of the same mouse L1-Y1151A-expressing cells or cellsexpressing the empty pcDNA3 vector (control). Tumor size was determined 14days after injection. (J) The metastatic capacity of the CRC cell clonesdescribed in C was determined as described for Fig. 2G. IkB is IkB-P.

Jour

nal o

f Cel

l Sci

ence

the sequence KGGKYSV in the juxtamembrane L1 domain (L1-4A) were sufficient to completely block NF-kB activation (Fig.4B, L1-4A). Moreover, one single point mutation that ofTyr1151Ala (L1-Y1151A), suppressed the ability of L1 to activateNF-kB signaling in 293T cells (Fig. 4B). We, therefore, isolatedCRC cells that stably expressed the L1-Y1151A mutant (Fig. 4C,D)and found that the L1-Y1151A mutant protein is expressed in thecell membrane at levels that are similarly to those in wt L1 (Fig.4C). Like wtL1, L1L1-Y1151 still formed a complex with wt IkB,(Fig. 4E), but unlike wt L1 and p65, L1-Y1151A did not inducethe phosphorylation of IkB (Fig. 4F, lane 8), a finding that isconsistent with the inability of L1-Y1151A to activate NF-kBsignaling (Fig. 4B). Moreover, unlike cells that express wt L1,CRC cells that express the mutant L1-Y1151A were unable toproliferate without serum (Fig. 4H), had a low tumorigenic capacity(similar to control Ls174T cells; Fig. 4I; n=23, P=0.0417) andwere unable to form liver metastases (Fig. 4J; n=20, P<0.001).

These results suggested that Tyr1151 in the L1 cytodomain isessential for NF-kB activation and to confer growth, tumorigenicand metastatic capacities in CRC cells.

The interaction of ezrin with L1 requires Tyr1151 and isnecessary for the L1-mediated increase in cell motility andthe induction of liver metastasisPrevious studies pointed to the importance of Tyr1151 in the L1cytodomain (Cheng et al., 2005; Sakurai et al., 2008) and a closehomologue of L1 (Schlatter et al., 2008) for the interaction with theactin-cytoskeleton adaptor protein ezrin (Bretscher et al., 2002;Fievet et al., 2008). We wished to determine whether ezrin forms amolecular complex with L1 and whether the L1-Y1151A mutant isdefective in this capacity. In 293T cells transfected with wt and theL1-Y1151A mutant we found that wt L1 forms a complex that co-immunoprecipitates with ezrin (Fig. 5A, lane 7), whereas the L1-Y1151A mutant has a reduced capacity to bind ezrin (Fig. 5A, lane

2139NF-kB and ezrin in L1-induced CRC metastasis

Fig. 5. Ezrin is required for L1-mediated metastasis, and L1, IkB and ezrin form a complex in CRC cells. (A) Lysates of 293T cells transfected with differentcombinations of ezrin, L1 and L1-Y1151A, were immunoprecipitated (IP) with antibodies against ezrin. Samples of cell lysates (Total) and immunoprecipitatedproteins (IP:Ezrin) were analyzed by western blotting with antibodies against L1 and ezrin. (B) L1-expressing Ls174T cell clones were transfected with shRNAtargeting ezrin or a scrambled siRNA sequence (L1+siRNA-Sc). In two independently isolated cell clones (lanes 4,5) the expression of ezrin was suppressed. (C) Theproliferation of CRC cell clones (described in B) in the presence of 0.5% serum was determined during 5 days. (D,E) An artificial wound was introduced into confluentmonolayers of the cell clones described in B, and the extent of wound closure in four different wounds for each cell clone was determined after 24 hours (an image ofone representative wound is shown in E). (F) The metastatic capacity of the CRC cell clones described in B was determined as described for Fig. 2G. (G) Cells stablytransfected with either the empty pcDNA3 vector, the construct expressing L1 alone or together with shRNA targeting ezrin, or with the construct expressing the p65subunit of NF-kB were either left untreated (–) or were treated with the proteasome inhibitor MG132 (+) for 3 hours before cell harvesting. Protein levels of L1 andIkB-P were determined. Ponceau staining shows the protein gel loading. (H) 293T cells were transfected with different combinations of L1, L1-Y1151A, ezrin andIkB-SR, and cell lysates were immunoprecipitated (IP) with antibodies against ezrin. Immunoprecipitated proteins were analyzed by western blotting with antibodiesagainst L1, IkB and ezrin. Ezrin co-precipitates with IkB and wt L1 (lanes 1,3), but to a much lesser extent with L1-Y1151A and IkB. pIkB is IkB-P.

Jour

nal o

f Cel

l Sci

ence

8). Next, we wished to determine whether this interaction of L1with ezrin is an essential part of the L1-mediated response we seein CRC cells. We isolated individual clones of CRC cells expressingL1, in which the level of ezrin was suppressed by shRNA (Fig. 5B,lanes 4,5). Cells that expressed shRNA targeting ezrin lost theincreased capacity to grow in the absence of serum conferred by L1(Fig. 5C), and their ability to close a wound in a monolayer wasreduced to that of control Ls174T cells lacking L1 (Fig. 5D,E)(n=24, P<0.0001). Moreover, the metastatic capacity of L1-transfected cells with reduced ezrin levels was blocked (Fig. 5F;n=20, P<0.001), indicating that ezrin is essential for L1-mediatedmetastasis. The induction of NF-kB signaling through an increasein IkB phosphorylation displayed by L1- or p65-overexpressingCRC cells (Fig. 5G, lanes 6,8) was also inhibited when ezrin levelswere suppressed (Fig. 5G, lane 10). This suggested that NF-kBactivation and the metastatic ability conferred by L1 require ezrin.Similar results were obtained by the expression of dominant-negativeezrin (which lacks the actin-binding C-terminus; supplementarymaterial Fig. S2), indicating that the actin-binding function of ezrinis required for the properties conferred to CRC cells by L1.

Since our results showed that L1 interacts with both IkB (Fig.4E) and ezrin (Fig. 5A), we wished to determine whether a ternarycomplex containing L1, IkB and ezrin is formed. The formation ofsuch ternary complex between L1, ezrin and IkB-SR was examinedin 293T cells transfected with cDNAs encoding these molecules,and compared with the capacity of the L1-Y1151A mutant to formsuch a complex. Wt L1 can be found in a ternary complexcontaining both IkB and ezrin (Fig. 5H, lane 3), but much lessIkB-SR was precipitated with ezrin when L1-Y1151A wasemployed (Fig. 5H, lane 4). We conclude that ezrin acts as ascaffold molecule that promotes the assembly of a multi-molecularcomplex containing L1 and IkB, and that enhances IkBphosphorylation and NF-kB signaling.

We then wished to determine the subcellular location of the L1-ezrin complex and found that, normally, endogenous ezrin is mostlydetected in filopodia of Ls174T cells when they are lacking L1(Fig. 6E,F). However, when L1 was expressed in such cells, ezrinwas recruited to membrane areas of cell adhesion containing L1(Fig. 6B,C), in a manner similar to that of sequestering IkB-SR toL1-containing cell adhesions in CRC cells following L1 expression(Fig. 2C). We examined the localization of endogenous ezrin andL1 in SW620 CRC cells, before and after suppressing L1 levelswith L1-specific shRNA. As in Ls174T cells transfected with L1,endogenous ezrin and L1 colocalized at cell-cell contact sites inSW620 cells (Fig. 6M-O). Suppression of L1 (Fig. 1D) led to theredistribution of ezrin, mainly to filopodia but also to the cytoplasm(Fig. 6K,L, insert). Furthermore, when we looked at the localizationof ezrin in cells that expressed the L1-Y1151A mutant, we foundthat more ezrin was found in filopodia and less in L1-containingcell-cell contacts (Fig. 6G-I) when compared with wt L1 (Fig. 6A-C). This is in accordance with the co-immmunoprecipitation results(Fig. 5A, lanes 7,8). Together with the results of Fig. 2C,D and Fig.5H, these findings indicate that L1 recruits ezrin and IkB to asubmembranal complex that promotes IkB phosphorylation andNF-kB signaling.

Ezrin, activated NF-kB and L1 colocalize at the invasivefront of human CRC tissueWe have previously reported on the specific localization of L1 atthe invading edge of CRC tumors (Gavert et al., 2005). We nowwished to determine whether ezrin is enriched in cells of human

CRC tissue samples and whether NF-kB is activated in the nucleiof these cells. Analysis of human CRC tissue samples byimmunohistochemistry on serial sections using anti-L1 and anti-p65-P antibodies, revealed that 75% (n=20) of the samplesexpressed L1 in invading tumor cells (Fig. 7A). In 80% (n=15) ofL1-expressing samples of tumor tissue, the immunostaining ofserial sections revealed coexpression of L1 in the membrane ofinvasive CRC cells (Fig. 7A,B) with activated NF-kB p65-Pstaining in their nuclei (Fig. 7D,E). The presence of p65-P wasalso evident in some cells localized in the central more differentiatedareas of the tumor, but these were primarily mitotic cells (Fig. 7F,arrows) that did not express L1 (Fig. 7C). Sequential staining ofCRC tissue for ezrin (Fig. 7J,K) and L1 (Fig. 7G,H) indicatedstrong expression and colocalization of ezrin and L1 at the invasivetumor front, and only weak ezrin staining in differentiated areas ofthe tumor (Fig. 7L). Normal colonic epithelial tissue neitherexpressed L1 (Fig. 7M), p65-P (Fig. 7N) nor ezrin (Fig. 7O) at

2140 Journal of Cell Science 123 (12)

Fig. 6. Endogenous wt L1 recruits ezrin from filopodia to cell-cell contactareas. The localization of L1 and ezrin in Ls174T cells stably transfected witheither empty pcDNA3 vector, or the constructs expressing L1 wt or L1-Y1151A mutant, was determined by double immunofluorescence microscopywith antibodies against L1 (red, A,D,G) and ezrin (green, B,E,H). Mergedimages are shown in (C,F,I). The expression of endogenous L1 was stablysuppressed in SW620 CRC cells with shRNA targeting L1, and thelocalization of endogenous L1 and ezrin was determined using antibodiesagainst L1 (red, J,M) and ezrin (green, K,N). (L,O) Merged images. Scale bar:20 mm for all panels.

Jour

nal o

f Cel

l Sci

ence

detectable levels, and only weak occasional staining for thesemolecules was observed in the stroma. These results stronglysuggest that activation of NF-kB signaling by L1 and theinvolvement of ezrin at the invasive CRC tumor front is animportant route for the development of invasive colon cancer.

DiscussionHere, we demonstrated that NF-kB signaling and ezrin are essentialfor the capacity of L1 to induce metastasis in CRC cells becauseby inhibiting NF-kB transactivation or by suppressing ezrin levels,or by using an L1 mutant with an altered ezrin-binding domain, the

ability of L1 to induce liver metastasis by CRC cells could beblocked. We further showed that L1 recruits both IkB and ezrininto a juxtamembrane complex in CRC cells, and that ezrin isfound in a ternary complex together with L1 and IkB in cells.Ezrin is required for the elevated IkB phosphorylation that leadsto NF-kB signaling. We also found that L1, ezrin and thephosphorylated p65 NF-kB subunit are present in the samepopulation of invading human CRC tissue cells. Together, theseresults strongly suggest a major route by which L1 confers aninvasive metastatic potential in human CRC cells, by activating theNF-kB signaling pathway through a mechanism involving thecytoskeletal protein ezrin as indicated in the proposed model (Fig.8). To our knowledge, this is the first indication of the involvementof this signaling complex in human CRC metastasis.

The essential function of ezrin in metastasis of CRC cell that wehave detected is in agreement with studies on the involvement ofezrin in the metastasis of pediatric cancers (osteosarcoma andrhabdomyosarcoma) (Khanna et al., 2004; Yu et al., 2004). Arecent study also reported on elevated ezrin expression in humanCRC tissue (Wang et al., 2009). Our findings on the presence ofp65-P in invasive CRC tissue cells in L1-expressing cells aresupported by a recent study that reported on higher p65immunoreactivity in CRC tissue compared with normal intestinalmucosa, and an increase in p65 that correlated with histologicaltumor progression (Aranha et al., 2007). In a CRC cell-culturesystem that mimicks the invasive versus the more differentiatedareas of the tumor by varying the density of cells in culture(Brabletz et al., 2001; Conacci-Sorrell et al., 2003), we foundmuch higher levels of L1 in less dense cell cultures, which alsodisplayed higher nuclear b-catenin signaling – reminiscent ofinvasive CRC cells at the tumor edge (Gavert et al., 2005). Suchsparse cultures also displayed a more substantial nuclear localizationof NF-kB (Aranha et al., 2007). Together, these studies supports alink between the activation of b-catenin–TCF-regulatedtranscription (which induces L1) with the induction of NF-kBsignaling leading to increased cell motility and the acquisition ofinvasive capacity by CRC cells.

These changes in the cellular phenotype of CRC also includethe downregulation of E-cadherin (Brabletz et al., 2001; Conacci-Sorrell et al., 2003) through induction of the major EMT regulatorsSlug and ZEB1 (Spaderna et al., 2008), in both the cell-culturemodels of CRC cells and at the invasive front of CRC tissue,suggesting that L1 is a key mediator of an EMT-like phenotypicshift. In a recent study that used MCF7 breast cancer cells, theintroduction of L1 into these cells resulted in the downregulationof E-cadherin levels and an increase in the invasive motility ofthese cells (Shtutman et al., 2006). EMT and NF-kB activationwere also found to be essential for metastasis of breast cancer cells(Huber et al., 2004) and in the process of mesenchymal geneinduction during EMT (Chua et al., 2007; Julien et al., 2007;Solanas et al., 2008). It is noteworthy that the expression of thehomeoprotein Six1 that expands the stem cell pool in mousemammary tumors and induces EMT and tumorigenesis (McCoy etal., 2009) also activates the promoter of the ezrin gene and promotesmetastasis of rhabdomyosarcoma (Yu et al., 2006).

By which mechanism(s) does L1 activate the NF-kB signalingpathway? Previous studies indicated that L1 interacts with tyrosinekinase growth-factor receptors, thereby modulating the activation ofMAPK, AKT and other signaling pathways (Gavert et al., 2008).These pathways were reported to activate the IkB kinase (IKK)complex upstream of the IkB-NF-kB complex, thereby inducing

2141NF-kB and ezrin in L1-induced CRC metastasis

Fig. 7. NF-kB, L1 and ezrin are preferentially co-expressed at the invadingedge of human CRC tissue. Immunohistochemical staining for L1 (red; A-C,G-I), p65-P (D-F) and ezrin (J-L) was conducted in sequential paraffinsections of human colorectal adenocarcinoma tissue with differentiated (diff)central regions indicated by tubular structures and undifferentiated invadingtumor cells (inv). Boxed areas in the first column of images indicate respectiveareas at higher magnifications that are shown in the middle and right column.Note that L1-expressing invading tumor cells (inv) contain nuclear p65-P(B,E); these invading tumor cells also coexpress ezrin (H,K). By contrast,tumor cells in central tumor regions lack L1 (diff), and p65-P is only detectedin some mitotic cells (diff, arrows) (D,F). Ezrin staining in more central regionsof the tumor is weak (J,L). (M-O) Staining of normal colon tissue. Scale bar:400mm (A,D,G,J), 75mm (all other panels). In the magnification boxes seen inpanels (H,I,K,L) the bar represent 40mm. pp65 is p65-P.

Jour

nal o

f Cel

l Sci

ence

NF-kB signaling (Bassères and Baldwin, 2006). Another possibility,on the basis of the results of this study, is the formation of asubmembrane complex between IkB, ezrin and L1 that leads toincreased phosphorylation and p65-P degradation, and then to therelease of NF-kB for transactivation in the nucleus. This notion issupported by a recent study demonstrating the presence of asubmembrane complex between E-cadherin and the IkB-NF-kBcomplex, whose disruption during EMT (driven by thedownregulation of E-cadherin) led to the release of NF-kB from thissubmembrane pool and its activation of mesenchymal genes (Solanaset al., 2008).

The most obvious mechanism by which NF-kB signalingcontributes to cancer development is through suppression ofapoptosis, by inducing the transcription of Bcl2 family members(Naugler and Karin, 2008; Van Antwerp et al., 1996); this wasdemonstrated in breast cancer cells (Wang et al., 2007) and othercancer cell types (Bassères and Baldwin, 2006). Our microarrayanalysis of gene expression in L1-transfected CRC cells, however,did not reveal the induction of a Bcl2 gene expression signature orof other anti-apoptotic genes (Gavert et al., 2007). Moreover, arecent study of apoptosis and Bcl2 expression in CRC tissue revealeda higher level of Bcl2 in normal mucosa compared with cancertissue and Bcl2 protein levels inversely correlated with apoptosis –with more extensive apoptosis observed in the tumor tissue thatincreased from the adenoma to carcinoma stage, concluding thatNF-kB does not exert an anti-apoptotic function through Bcl2activation in CRC tissue (Aranha et al., 2007). Another means bywhich NF-kB often contributes to cancer progression is the inductionof angiogenesis through the coordinated activation of inflammatorycytokines (Bassères and Baldwin, 2006; Naugler and Karin, 2008).Our gene-array studies of L1-induced genes in CRC cells did notsupport this notion either (Gavert et al., 2007), because we observeda coordinated downregulation of inflammatory cytokine RNAs inL1-transfected CRC cells (data not shown). Future studies shouldfocus on the molecular details of the link between L1 and the IkB–NF-kB complex, and on the analysis of gene-expression patterns ofCRC cells that overexpress L1 genes whose levels are reduced byIkB or ezrin shRNA (in L1-overexpressing cells). Such studies willhopefully narrow down the number and provide information on thegenes whose activation by NF-kB is essential for conferring invasivemetastatic capacities by L1 in CRC.

Materials and MethodsCell culture, transfection and transactivation293T, Ls174T and SW620 cells were grown as described (Gavert et al., 2005; Gavertet al., 2007). For cell growth, 5�103 cells per well were plated into 24-well dishesand their number was determined in triplicate every 24 hours for 5-7 days. Transienttransfection of 293T cells was performed using the calcium-phosphate method.Ls174T and SW620 cells were transfected using LipofectamineTM 2000 (Invitrogen,Carlsbad, CA). In transactivation assays, 0.25 mg of b-galactosidase was co-

transfected with 1 mg of a plasmid containing three copies of an NF-kB-responsivepromoter sequence linked to luciferase (3xkB.luc) and 1 mg of the empty pGL3plasmid. Cells were plated in triplicate, lysed after 48 hours, and luciferase and b-galactosidase levels determined by using enzyme assay kits from Promega (Madison,WI). Co-transfection with human IkB super repressor (IkB-SR; with mutations inSer32Ala and Ser36Ala) stabilized against degradation (Zhang et al., 2000) wasfollowed by selection with puromycin (10 mg/ml). SW620 cells stably expressingL1shRNA were co-transfected with four plasmids containing different shRNAs,followed by selection with puromycin. Ls174T cells stably expressing ezrin shRNAwere established by co-transfecting cells with four plasmids containing differentezrin shRNA, followed by selection with zeocine (500 mg/ml).

PlasmidsThe L1 cDNA and L1 mutants have been described previously (Hlavin and Lemmon,1991). The IkB-SR cDNA, the NF-kB responsive reporter plasmid (3xkB.luc) andthe cDNA for human p65 NF-kB subunit were from David Wallach and AndrewKovalenko (Weizmann Institute of Science, Rehovot, Israel) (Zhang et al., 2000).shRNA against ezrin and unspecific control shRNA were from Monique Arpin(Curie Institute, Paris). The ezrin target sequences were: 5�-GATT GGCTTT -CCTTGGAGTGA-3�, 5�-GAATCCTTAGCGATGAGATCT-3�. The scrambled ezrinsequence was: 5�-GATATGTGCGTACCTAGCAT-3�. L1 shRNA was prepared inpSuper according to the manufacturer’s instructions (pSuper RNAi System,OligoEngine, Seattle, WA) using the target sequences 5�-GGATGGTGTCCAC -TTCAAA-3�, 5�-GAGAAGGGTGGGCTTCCC-3�, 5�-AGAC CAGAAGTACCGG -ATA-3� and the scrambled sequence 5�-AGC GC GC TTTGTAGGATTC-3�.

RT-PCRRNA was isolated from cells using EZ-RNA kit (Biological Industries, Israel). PCRwas performed using the sequences for L1: forward 5�-ACGGGC AACAAC -AGCAAC-3�, reverse 5�-CGGCTTCCTGCTAATCATG-3�.

Immunofluorescence and immunoblottingCells cultured on glass coverslips were permeabilized with 0.5% Triton X-100 andfixed with 3% PFA. Staining for p65-P was conducted with cells fixed in methanolfor 5 minutes at 4°C. Polyclonal and monoclonal antibodies (pAb and mAb,respectively) against L1, recognizing both the extracellular and intracellular domains,have been described previously (Cheng and Lemmon, 2004). Other antibodies were,mAb IkBa/MAD-3 (BD Biosciences, Franklin Lakes, NJ), goat Ab against NF-kBp65 (sc-109, Santa Cruz Biotechnology, Inc. Santa Cruz, CA), rAb phosphospecificpS529 NF-kB p65 (Rockland, Gilbertsville, PA), and mAb ezrin (Sigma, St Louis,MI). The secondary antibodies were Alexa-488-conjugated goat anti-mouse, anti-rabbit IgG (Invitrogen, Carlsbad, CA) and Cy3-labeled goat anti-mouse or anti-rabbit IgG (Jackson ImmunoResearch Laboratories, West Grove, PA). Images wereacquired by using Eclipse E1000, Nikon objectives 60�/1.4 NA with a camera(ORCA-ER; Hamamatsu) and Volocity acquisition software (Improvision). Westernblots were developed by using the ECL method (Amersham Biosciences, UK) withthe antibodies described above and mouse anti-tubulin (Sigma-Israel).

ImmunohistochemistryImmunohistochemistry was carried out on 25 paraffin-embedded human colorectaladenocarcinomas as described (Brabletz et al., 1999). For L1 and p65-P detection,polyclonal rabbit anti-L1 antibody (described above; diluted 1:1000), rabbit antiserumagainst p65-P (no. 3037, Cell Signaling, Frankfurt, Germany; diluted 1:50), andmouse anti-ezrin antibody (described above) were used. The streptomycin/AB systemwas used for detection of Ab binding according to the manufacturer’s protocol(Dako, Hamburg, Germany). Sections were counterstained with hemalaun (Merck,Darmstadt, Germany).

Artificial wound healingA round wound was introduced into a confluent cell monolayer with a micropipettetip by using suction to remove the cells. Fresh culture medium with 40 mg/mlmitomycin C was added to inhibit cell proliferation. Cells were placed in a 37°C

2142 Journal of Cell Science 123 (12)

Fig. 8. Scheme for L1-mediated activation of NF-kB that involves ezrin. (A) NF-kB is inactive when bound to IkB. Phosphorylation of IkB releases NF-kB,and allows its translocation to the nucleus and the subsequent activation of target genes. L1 can form a complex with IkB and can promote the phosphorylation ofIkB by a process involving the binding of ezrin to L1 at the juxtamembrane domain. (B) When Tyr1151 in L1is mutated to alanine (L1-Y1151A) in thejuxtamembrane domain, ezrin can no longer bind to L1 and, although L1 can still form a complex with IkB, the L1-1Y1151A mutant no longer promotes IkBphosphorylation. pIkB is IkB-P.

Jour

nal o

f Cel

l Sci

ence

chamber above the microscope, the wounds were photographed in quadruplicate for18-24 hours and percentage wound closure was calculated using the Photoshop CS3analyzer to measure the open wound area compared with the wound area at thebeginning of the experiment.

Metastasis assaysGroups of five mice were injected with 106 cells into the distal tip of the spleen; after4-6 weeks their spleen and livers were removed as described (Gavert et al., 2007).Tissue histology was performed on 3-mm serial sections of formalin-fixed, paraffin-embedded livers that were stained with hematoxylin-eosin. Tumor growth wasinduced by injecting 2�106 cells into the flanks of nude mice (five mice per group).Control cells were injected into the opposite flank of the same mouse and tumorswere removed and compared after 2 weeks.

StatisticsStatistical significance was determined by Fisher’s exact test (Langsrud et al., 2007)for mouse metastasis experiments. Tumor mass was compared and significancedetermined by ANOVA. In wound closure, significance was determined using non-paired Student’s t-test. A P value of <0.05 was considered significant.

These studies were supported by grants from Israel ScienceFoundation, Israel Cancer Research Fund, Yad Abraham Center forCancer Diagnosis and Therapy, the YY fellowship from UICC (to A.B.-Z.). Studies in the laboratory of T.B. were supported by theDeutsche Forschungsgemeinschaft (no. BR 1399/4-3 and 5-1), theEuropean Union (MCSCS contract no. 037297) and the Walter G.Ross Foundation.

Supplementary material available online athttp://jcs.biologists.org/cgi/content/full/135/12/2135/DC1

ReferencesAranha, M. M., Borralho, P. M., Ravasco, P., Moreira da Silva, I. B., Correia, L.,

Fernandes, A., Rodrigues, C. M. and Camilo, M. (2007). NF-kB and apoptosis incolorectal tumourigenesis. Eur. J. Clin. Invest. 37, 416-424.

Bassères, D. S. and Baldwin, A. S. (2006). Nuclear factor-kB and inhibitor of kB kinasepathways in oncogenic initiation and progression. Oncogene 25, 6817-6830.

Brabletz, T., Jung, A., Hermann, K., Günther, K., Hohenberger, W. and Kirchner, T.(1998). Nuclear overexpression of the oncoprotein b-catenin in colorectal cancer islocalized predominantly at the invasion front. Pathol. Res. Pract. 194, 701-704.

Brabletz, T., Jung, A., Dag, S., Hlubek, F. and Kirchner, T. (1999). b-catenin regulates theexpression of the matrix metalloproteinase-7 in human colorectal cancer. Am. J. Pathol.155, 1033-1038.

Brabletz, T., Jung, A., Reu, S., Porzner, M., Hlubek, F., Kunz-Schughart, L. A., Knuechel,R. and Kirchner, T. (2001). Variable b-catenin expression in colorectal cancers indicatestumor progression driven by the tumor environment. Proc. Natl. Acad. Sci. USA 98,10356-10361.

Brabletz, T., Hlubek, F., Spaderna, S., Schmalhofer, O., Hiendlmeyer, E., Jung, A. andKirchner, T. (2005). Invasion and metastasis in colorectal cancer: epithelial-mesenchymaltransition, mesenchymal-epithelial transition, stem cells and b-catenin. Cells TissuesOrgans 179, 56-65.

Bretscher, A., Edwards, K. and Fehon, R. (2002). ERM proteins and merlin: integrators atthe cell cortex. Nat. Rev. Mol. Cell Biol. 3, 586-599.

Cheng, L. and Lemmon, V. (2004). Pathological missense mutations of neural cell adhesionmolecule L1 affect neurite outgrowth and branching on an L1 substrate. Mol. CellNeurosci. 27, 522-530.

Cheng, L., Itoh, K. and Lemmon, V. (2005). L1-mediated branching is regulated by twoezrin-radixin-moesin (ERM)-binding sites, the RSLE region and a novel juxtamembraneERM-binding region. J. Neurosci. 25, 395-403.

Chua, H. L., Bhat-Nakshatri, P., Clare, S. E., Morimiya, A., Badve, S. and Nakshatri,H. (2007). NF-kB represses E-cadherin expression and enhances epithelial to mesenchymaltransition of mammary epithelial cells: potential involvement of ZEB-1 and ZEB-2.Oncogene 26, 711-724.

Clevers, H. (2006). Wnt/b-catenin signaling in development and disease. Cell 3, 469-480.Conacci-Sorrell, M., Ben-Yedidia, T., Shtutman, M., Feinstein, E., Einat, P. and Ben-

Ze’ev, A. (2002a). Nr-CAM is a target gene of the b-catenin/LEF-1 pathway in melanomaand colon cancer and its expression enhances motility and confers tumorigenesis. GenesDev. 16, 2058-2072.

Conacci-Sorrell, M., Zhurinsky, J. and Ben-Ze’ev, A. (2002b). The cadherin-cateninadhesion system in signaling and cancer. J. Clin. Invest. 109, 987-991.

Conacci-Sorrell, M., Simcha, I., Ben-Yedidia, T., Blechman, J., Savagner, P. and Ben-Ze’ev, A. (2003). Autoregulation of E-cadherin expression by cadherin-cadherininteractions: the roles of b-catenin signaling, Slug, and MAPK. J. Cell Biol. 163, 847-857.

Fievet, B., Louvard, D. and Arpin, M. (2007). ERM proteins in epithelial cell organizationand functions. Biochim. Biophys. Acta 1773, 653-660.

Gavert, N. and Ben-Ze’ev, A. (2008). Epithelial-mesenchymal transition and the invasivepotential of tumors. Trends Mol. Med. 14, 199-209.

Gavert, N., Conacci-Sorrell, M., Gast, D., Schneider, A., Altevogt, P., Brabletz, T. andBen-Ze’ev, A. (2005). L1, a novel target of b-catenin signaling, transforms cells and isexpressed at the invasive front of colon cancers. J. Cell Biol. 168, 633-642.

Gavert, N., Sheffer, M., Raveh, S., Spaderna, S., Shtutman, M., Brabletz, T., Barany, F.,Paty, P., Notterman, D., Domany, E. et al. (2007). Expression of L1-CAM and ADAM10in human colon cancer cells induces metastasis. Cancer Res. 67, 7703-7712.

Gavert, N., Ben-Shmuel, A., Raveh, S. and Ben-Ze’ev, A. (2008). L1-CAM in canceroustissues. Exp. Opin. Biol. Ther. 8, 1749-1757.

He, T. C., Sparks, A. B., Rago, C., Hermeking, H., Zawel, L., da Costa, L. T., Morin, P.J., Vogelstein, B. and Kinzler, K. W. (1998). Identification of c-MYC as a target of theAPC pathway. Science 281, 1509-1512.

Hlavin, M. L. and Lemmon, V. (1991). Molecular structure and functional testing of humanL1CAM: an interspecies comparison. Genomics 11, 416-423.

Huber, M. A., Azoitei, N., Baumann, B., Grünert, S., Sommer, A., Pehamberger, H.,Kraut, N., Beug, H. and Wirth, T. (2004). NF-kB is essential for epithelial-mesenchymaltransition and metastasis in a model of breast cancer progression. J. Clin. Invest. 114, 569-581.

Julien, S., Puig, I., Caretti, E., Bonaventure, J., Nelles, L., van Roy, F., Dargemont, C.,de Herreros, A. G., Bellacosa, A. and Larue, L. (2007). Activation of NF-kB by Aktupregulates Snail expression and induces epithelium mesenchyme transition. Oncogene26, 7445-7456.

Karin, M. and Ben-Neriah, Y. (2008). Phosphorylation meets ubiquitination: the control ofNF-kB activity. Annu. Rev. Immunol. 18, 621-663.

Karin, M., Cao, Y., Greten, F. R. and Li, Z. W. (2002). NF-kB in cancer: from innocentbystander to major culprit. Nat. Rev. Cancer 2, 301-310.

Khanna, C., Wan, X., Bose, S., Cassaday, R., Olomu, O., Mendoza, A., Yeung, C.,Gorlick, R., Hewitt, S. M. and Helman, L. J. (2004). The membrane-cytoskeleton linkerezrin is necessary for osteosarcoma metastasis. Nature Med. 10, 182-186.

Kinzler, K. and Vogelstein, B. (1996). Lessons from hereditary colon cancer. Cell 87, 159-170.

Langsrud, Ø., Jørgensen, K., Ragni Ofstad, R. and Næs, T. (2007). Analyzing designedexperiments with multiple responses. J. Appl. Stat. 34, 1275-1296.

McCoy, E., Iwanaga, R., Jedlicka, P., Abbey, N.-S., Chodosh, L., Heichman, K., Welm,A. L. and Ford, H. L. (2009). Six1 expands the mouse mammary epithelial stem/progenitorcell pool and induces mammary tumors that undergo epithelial-mesenchymal transition. J.Clin. Invest. 119, 2663-2677.

Min, C., Eddy, S. F., Sherr, D. H. and Sonenshein, G. E. (2008). NF-kB and epithelial tomesenchymal transition of cancer. J. Cell Biochem. 104, 733-744.

Naugler, W. E. and Karin, M. (2008). NF-kB and cancer-identifying targets and mechanisms.Curr. Opin. Genet. Dev. 18, 19-26.

Perkins, N. D. (2006). Post-translational modifications regulating the activity and functionof the nuclear factor kB pathway. Oncogene 25, 6717-6730.

Polakis, P. (2007). The many ways of Wnt in cancer. Curr. Opin. Genet. Dev. 17, 45-51.Sakurai, T., Gil, O. D., Whittard, J. D., Gazdoiu, M., Joseph, T., Wu, J., Waksman, A.,

Benson, D. L., Salton, S. R. and Felsenfeld, D. P. (2008). Interactions between the L1cell adhesion molecule and ezrin support traction-force generation and can be regulatedby tyrosine phosphorylation. J. Neurosci. Res. 86, 2602-2614.

Schlatter, M., Buhusi, M., Wright, A. and Maness, P. (2008). CHL1 promotes Sema3A-induced growth cone collapse and neurite elaboration through a motif required forrecruitment of ERM proteins to the plasma membrane. J. Neurochem. 104, 731-744.

Shtutman, M., Zhurinsky, J., Simcha, I., Albanese, C., D’Amico, M., Pestell, R. and Ben-Ze’ev, A. (1999). The cyclin D1 gene is a target of the b-catenin/LEF-1 pathway. Proc.Natl. Acad. Sci. USA 96, 5522-5527.

Shtutman, M., Levina, E., Ohouo, P., Baig, M. and Roninson, I. B. (2006). Cell adhesionmolecule L1 disrupts E-cadherin-containing adherens junctions and increases scatteringand motility of MCF7 breast carcinoma cells. Cancer Res. 66, 11370-11380.

Solanas, G., Porta-de-la-Riva, M., Agustí, C., Casagolda, D., Sánchez-Aguilera, F.,Larriba, M. J., Pons, F., Peiró, S., Escrivà, M., Muñoz, A. et al. (2008). E-cadherincontrols b-catenin and NF-kB transcriptional activity in mesenchymal gene expression. J.Cell Sci. 121, 2224-2234.

Spaderna, S., Schmalhofer, O., Wahlbuhl, M., Dimmler, A., Bauer, K., Sultan, A.,Hlubek, F., Jung, A., Strand, D., Eger, A. et al. (2008). The transcriptional repressorZEB1 promotes metastasis and loss of cell polarity in cancer. Cancer Res. 68, 537-544.

Tetsu, O. and McCormick, F. (1999). b-Catenin regulates expression of cyclin D1 in coloncarcinoma cells. Nature 398, 422-426.

Thiery, J. P. (2002). Epithelial-mesenchymal transitions in tumour progression. Nat. Rev.Cancer 2, 442-454.

Thiery, J. P. and Sleeman, J. P. (2006). Complex networks orchestrate epithelial-mesenchymal transitions. Nat. Rev. Mol. Cell Biol. 7, 131-142.

Van Antwerp, D. J., Martin, S. J., Kafri, T., Green, D. R. and Verma, I. M. (1996).Suppression of TNF-a-induced apoptosis by NF-kB. Science 274, 787-789.

Wang, H.-J., Zhu, J.-S., Zhang, Q., Sun, Q. and Guo, H. (2009). High levels of ezrinexpression in colorectal cancer tissue is closely related to tumor malignancy. World J.Gastroenterol. 15, 2016-2019.

Wang, X., Belguise, K., Kersual, N., Kirsch, K. H., Mineva, N. D., Galtier, F., Galtier, F.,Chalbos, D. and Sonenshein, G. E. (2007). Oestrogen signalling inhibits invasivephenotype by repressing RelB and its target BCL2. Nat. Cell Biol. 9, 470-478.

Yu, Y., Khan, J., Khanna, C., Helman, L., Meltzer, P. and Merlino, G. (2004). Expressionprofiling identifies the cytoskeletal organizer ezrin and the developmental homeoproteinSix-1 as key metastatic regulators. Nature Med. 10, 175-181.

Yu, Y., Davicioni, E., Triche, T. and Merlino, G. (2006). The homeoprotein Six1transcriptionally activates multiple protumorigenic genes but requires ezrin to promotemetastasis. Cancer Res. 66, 1982-1989.

Zhang, S. Q., Kovalenko, A., Cantarella, G. and Wallach, D. (2000). Recruitment ofthe IKK signalosome to the p55 TNF receptor: RIP and A20 bind to NEMO (IKKg)upon receptor stimulation. Immunity 12, 301-311.

2143NF-kB and ezrin in L1-induced CRC metastasis

Jour

nal o

f Cel

l Sci

ence