N E W S A N D V I E W S

NATURE CELL BIOLOGY VOLUME 6 | NUMBER 10 | OCTOBER 2004 913

GSK-3β sets Snail’s paceKarni Schlessinger and Alan Hall

Snail, a transcription factor that promotes epithelial-to-mesenchymal transitions during development, has also beenimplicated in human cancer. A new mechanism in which GSK-3β phosphorylation regulates the rate of Snail proteindegradation may provide insight into invasion and the metastatic progression of tumours.

Epithelial cells, which typically are polarized,spatially organized and tightly packed, canundergo dynamic changes in morphologyand motility termed epithelial-mesenchymaltransitions (EMTs)1. Although EMT is a keyprocess in development and wound healing, ithas also been implicated in neoplastic trans-formation, particularly invasion and metasta-sis. Gene expression studies have identifiednumerous changes in cells undergoing EMTand transcriptional repression of E-cadherinseems to be a key event, resulting in disassem-bly of cell–cell adherens junctions2. E-cad-herin is lost in many metastatic carcinomasand this is often due to transcriptional down-regulation rather than gene loss or mutation3.

The Snail family of zinc-finger transcriptionfactors — originally identified in Drosophilaand later in other organisms, includinghumans — interacts with the E-cadherin pro-moter to repress transcription during EMT indevelopmental processes such as mesodermspecification4. In addition, Snail activates thetranscription of genes associated with mes-enchymal differentiation, such as vimentin andfibronectin5–7. RT–PCR analysis has demon-strated an accumulation of Snail transcripts inmany metastatic human tumours, but, surpris-ingly, concomitant increases in the concentra-tion of Snail protein are not always observed. Apossible explanation for this discrepancyemerges on page 931 of this issue, where Zhouet al. describe a new signal transduction path-way that regulates the stability of Snail proteinthrough GSK-3β-mediated phosphorylation8.

Zhou et al. screened 26 cancer cell lineswhere snail mRNA is expressed, but failed todetect any Snail protein by immunoblotanalysis8. As several transcription factors areregulated by proteasome-mediated degrada-tion, they examined whether Snail might alsobe regulated in this way. Treatment with theproteasome inhibitor MG132, however,resulted in only modest increases in Snail pro-tein, and even that was in just a few of the celllines. Undeterred, the authors proceeded tocombine MG132 with lithium, a GSK-3βinhibitor, and found a synergistic interaction,resulting in marked increases in Snail proteinin ten cell lines tested.

Sequence analysis predicts the presence oftwo GSK-3β consensus sites in Snail (Fig. 1):Site I overlaps with a destruction box recog-nized by β-TRCP — an F-Box protein of theSCF E3 ubiquitin-ligase complex — whereasSite II lies next to a nuclear export signal(NES)9. Using both in vivo and in vitro assays,Zhou et al. find that both sites can function assubstrates for GSK-3β and, further, thatphosphorylation most probably occurs at two

serines in Site I and four serines in Site II. Toanalyse the role of these two sites, wild-typeSnail and mutant variants were expressed inthe human breast cancer cell line, MCF-7,which lacks detectable Snail protein ormRNA. Wild-type Snail localized mainly inthe nucleus, but treatment with MG132resulted in increased cytoplasmic levels.Substitution of the two serine residues at SiteI (2SA) resulted in increased protein concen-trations detectable in both the nucleus andthe cytoplasm, and no effect following treat-ment with MG132. Substitution of the fourserine residues at Site II caused retention inthe nucleus and no cytoplasmic Snail couldbe seen, even after MG132 treatment.Similarly, and as first reported by Dominguezet al., when all six serine residues were substi-tuted (6SA), Snail was found exclusively inthe nucleus8,9. Together with additional bio-chemical data presented in this paper, andwith the previous report showing that phos-phorylation of Snail controls CRM1-depend-ent nuclear export9, this leads to a model forSnail regulation in epithelial cells: a nuclear

DSGXXS SXXXSXXXSXXXS

Site I Site II

β-TRCPdegradation

Nuclear export

N C

Figure 1 GSK-3β phosphorylation consensus sites in Snail. Zhou et al. identify two consensussites for GSK-3β, containing six serine residues that undergo phosphorylation. Site I containstwo serine residues that overlap with a destruction box recognized by β-Trcp. Site II contains fourtarget serine residues and lies next to a nuclear export sequence.

Karni Schlessinger and Alan Hall are at theMRC Laboratory for Molecular Cell Biology &Cell Biology Unit, University College London,Gower Street, London WC1E 6BT, UK.e-mail: alan.hall@ucl.ac.uk andk.schlessinger@ucl.ac.uk

Oct N&V final 16/9/04 4:55 PM Page 913

© 2004 Nature Publishing Group

© 2004 Nature Publishing Group

N E W S A N D V I E W S

914 NATURE CELL BIOLOGY VOLUME 6 | NUMBER 10 | OCTOBER 2004

pool of GSK-3β phosphorylates Site II andpromotes export of Snail from the nucleus,whereas a cytoplasmic pool of GSK-3β phos-phorylates Site I and induces recognition byβ-TRCP and subsequent ubiquitination anddegradation (Fig. 2; green arrows). BecauseGSK-3β is active in its basal state, this pro-vides an explanation as to why Snail proteinis often undetectable in epithelial cells.

To strengthen the physiological significanceof this GSK-3β signalling pathway, the authorsanalysed the stability of wild-type Snailexpressed in MCF-7 cells in response to epi-dermal growth factor (EGF) and insulin-likegrowth factor 1 (IGF-1), two growth factorsknown to inactivate GSK-3β through phos-phorylation of Ser-9 by the phosphatidylinos-itol-3-OH kinase (PI(3)K)/Akt pathway. Theyfind that Snail is stabilized following growth-factor treatment and that this is PI(3)K- andERK-dependent. The correlation foundbetween the inactivation of GSK-3β and thestabilization and nuclear accumulation ofSnail, suggests a model for mesenchymal tran-sitions (Fig. 2; red line). Unfortunately, the

authors do not show whether IGF-1 or EGFinduces EMT in these cells, as would be pre-dicted. The same growth factors would also beexpected to induce EMT in the ten cancer celllines, which they initially found responded tothe MG132/lithium treatment; however, thiswas not tested. These observations will cer-tainly generate a great deal of interest in thispathway, as many human cancers express highlevels of the EGF receptor. This could result ininhibition of GSK-3β and so promote Snailstabilization, tumour progression and metas-tasis. Many other growth factors, for exampleHGF and FGF, induce EMT and activatePI(3)K, and so their ability to affect Snail pro-tein levels should now be examined10. Finally,the canonical Wnt pathway, which leads to aPI(3)K-independent inhibition of GSK-3β,has been implicated in both cancer and EMT.It will be of interest to see whether Wnt caninduce Snail stabilization, as well as the moreclassic effect of β-catenin stabilization11.

To provide further evidence for theirmodel, Zhou et al. used the Snail mutants thatlack the putative phosphorylation sites. Most

notably, expression of the Snail-6SA variant inMCF-7 cells induced EMT, with a loss ofE-cadherin mRNA expression, upregulationof vimentin and fibronectin, and a markedincrease in cell motility. They conclude thatSnail, which cannot be exported from thenucleus and cannot be degraded, is able topromote EMT. Interestingly, both wild-typeSnail and the variant with Site-II substitutions(4SA) are also found substantially in thenucleus when overexpressed in the MCF-7cells; however, the wild-type protein at least,does not induce EMT. One possibility is thateven though the protein reaches high steadystate levels in the nucleus, it is phosphorylatedand this might block its ability to function asa transcriptional regulator.

In contrast to this study, others havereported that overexpression of wild-typeSnail can result in EMT in some epithelial celllines, raising the old problem of differences incell type7. Obviously, variations in the activi-ty of the endogenous GSK-3β nuclear poolwould be expected to make a significant con-tribution, but little is known about how the

Nucleus

Cytoplasm

snail vimentin, fibronectin, SMA, N-cadherin

E-cadherin, occludin, mucin, desmoplakin E-cadherin, occludin, mucin, desmoplakin

β-TRCP

Snail

Snail

MesenchymalEpithelial

Snail

GSK-3βCyt

GSK-3βNuc

Snail

P

TGF-β, FGF, BMPIntegrins

EGF, IGFHGF/Wnt/wounding?

Snail

Snail SnailP P

P

Figure 2 Multiple levels of Snail regulation. Green arrows highlight thepathway responsible for Snail degradation in epithelial cells. Aftertranslation, Snail shuttles into the nucleus by an unknown mechanismand is phosphorylated at Site II by a nuclear pool of active GSK-3β topromote CRM1-dependent nuclear export. Snail is then phosphorylatedat Site I by a cytoplasmic pool of active GSK-3β (phosphorylation ofSite I may also occur in the nucleus) and is then recognized by β-TRCP,leading to ubiquitination and proteasomal degradation. Red arrows

highlight the pathways promoting EMT. Snail transcription can beinduced by growth factors or integrins, but for EMT to occur GSK-3βactivity must also be inhibited. EGF or IGF, and perhaps other growthfactors that activate PI(3)K, such as HGF, promote phosphorylation andinactivation of GSK-3β. Other extracellular stimuli, such as Wnt orwounding of a monolayer, may inhibit GSK-3β in other ways. Theaccumulation of non-phosphorylated nuclear Snail functions as both atranscriptional repressor and activator to promote EMT.

Oct N&V final 16/9/04 4:55 PM Page 914

© 2004 Nature Publishing Group

© 2004 Nature Publishing Group

N E W S A N D V I E W S

NATURE CELL BIOLOGY VOLUME 6 | NUMBER 10 | OCTOBER 2004 915

subcellular distribution of this kinase is con-trolled or how its activity in the nucleus isregulated. In addition, other cooperating fac-tors may be required. Twist, for example, isanother transcription factor recently shownto induce EMT and metastasis12.Interestingly, Twist induces expression ofSnail during Drosophila mesoderm inductionand, although snail mRNA levels did notchange after introducing Twist into humanmammary epithelial cells, Snail protein levelswere not examined12. Finally, other Snail-family members, such as Slug and Scratch,may also contribute to EMT induction5.

Snail is a major contributor to EMTs, but it isnot the only one and its relationship with Twistand other Snail family members is going to be amajor focus of interest. Numerous studies

suggest that loss of E-cadherin is necessary,although not sufficient, to induce EMT andmetastasis. In agreement with this, Zhou et al.find that re-expression of E-cadherin in theSnail-6SA-expressing MCF-7 cells blocks theincreased cell motility8. Others, however, havereported that re-expression of E-cadherin doesnot revert the EMT phenotype12,13. It is likely,therefore, that combinations of signalling path-ways are required to induce EMT and metasta-sis, and that other factors essential for EMTremain to be identified14. The work by Zhou etal. adds another level of complexity to thisprocess, but it also raises some new and excitingideas about the regulation of EMT by growthfactors, with the hope that this might lead to abetter understanding of the transition to ametastatic phenotype during tumorigenesis.

1. Thiery, J. P. Nature Rev. Cancer 2, 442–454 (2002).2. Jechlinger, M. et al. Oncogene 22, 7155–7169

(2003).3. Cavallaro, U. & Christofori, G. Nature Rev. Cancer 4,

118–132 (2004).4. Alberga, A. et al. Development 111, 983–992

(1991).5. Nieto, M. A. Nature Rev. Mol. Cell Biol. 3, 155–166

(2002).6. Cano, A. et al. Nature Cell Biol. 2, 76–83 (2000).7. Batlle, E. et al. Nature Cell Biol. 2, 84–89 (2000).8. Zhou, B. D. et al. Nature Cell Biol. 6, 931–940

(2004).9. Dominguez, D. et al. Mol. Cell. Biol. 23, 5078–5089

(2003).10. Comoglio, P. M. & Boccaccio, C. Semin. Cancer Biol.

11, 153–165 (2001).11. Taki, M. et al. Cancer Sci. 94, 593–597 (2003).12. Yang, J. et al. Cell 117, 927–939 (2004).13. Ohkubo, T. & Ozawa, M. J. Cell Sci. 117, 1675–1685

(2004).14. Pagliarini, R. A. & Xu, T. Science 302, 1227–1231

(2003).

Crossing the tracksClassically, we have learnt that kinesin and dynein move alongmicrotubules, whereas myosin binds to actin. But in more recenttimes, the need for crosstalk between the two cytoskeletons inmany cellular processes has become increasingly apparent, and thefinding now that Myo10 may bind to microtubules as well as actinoffers yet another mechanism by which this crosstalk can occur.

Weber et al., in a recent issue of Nature (431, 325–329; 2004),first reasoned that Myo10 may be a microtubule–F-actin linker onthe basis of its primary structure: in its carboxyl terminus there isa MyTH4 domain, which mediates interactions with microtubulesin other systems. This suggested that Myo10 might directly asso-ciate not only with actin, but also with microtubules. Indeed,Myo10 co-sedimented with microtubules in Xenopus egg extracts,and colocalized with microtubules independently of F-actin.Furthermore, it colocalized with meiotic spindle microtubulesspecifically at the interface between the spindle and the cortex.Deletion analysis confirmed that the MyTH4–FERM domain ofMyo10 was essential for the interaction with microtubules.

When Weber et al. expressed the Myo10 tail domain, whichfunctions as a dominant negative, they observed displacement ofthe oocyte nucleus from its characteristic asymmetric localizationin the animal hemisphere to the cortex. This phenomenon wasnot the result of microtubule depolymerization and could bereproduced by the addition of anti-Myo10 antibodies. Thus,Myo10 is required for microtubule-dependent asymmetricanchoring of the oocyte nucleus.

Next, the authors tried to understand the basis of the pheno-types they observed in the oocyte — rotation failure, abnormalspindle structure and multiple microtubule organizing centres(MTOCs). These effects are all similar to those that occur afteractin depolymerization during meiotic maturation; however, noactin disassembly was observed. Instead, there was a concentra-tion of F-actin in aggregates on or near to spindles or abortive

MTOCs, indicating that Myo10 is essential for proper F-actin–meiotic-spindle interactions. Together, these data suggestthat this actin-based motor functions to link the microtubule andF-actin cytoskeletons.

As Weber et al. highlight, Myo10 could function in nuclearanchoring by binding to phosphoinositides through its PHdomain and by binding to actin and microtubules; alternative-ly, the motor activity of Myo10 could function in the transportof microtubule-associated spindle components. In either case,this highlights a previously unappreciated function for amyosin in microtubule-based spindle function. Furthermore,as similar MyTH4–FERM domains are present in othermyosins, this may be a more general function of other uncon-ventional myosins.

JON REYNOLDS

Meiotic spindles viewed from inside of the cell, slightly to the side.Blue, actin; green, myosin; red, microtubules.

Oct N&V final 16/9/04 4:55 PM Page 915

© 2004 Nature Publishing Group

© 2004 Nature Publishing Group