Oncogene (2010) 29, 3173–3184 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 www.nature.com/onc REVIEW TWISTing an embryonic transcription factor into an oncoprotein S Ansieau1, A-P Morel4, G Hinkal1, J Bastid1,2 and A Puisieux1,2,3,4 1Inserm, U590, Lyon, France; 2Universite´ de Lyon, Lyon I, Institut des Sciences Pharmaceutiques et Biologiques, Lyon, France; 3Universite´ de Lyon, Lyon I, France and 4Centre Le´on Be´rard, Lyon, France Over the past decade, the reactivation of TWIST TWIST proteins are pleiotropic gene regulators embryonic transcription factors has been described as a frequent event and a marker of poor prognosis in an TWIST proteins belong to the large family of basic impressive array of human cancers. Growing evidence now Helix–Loop–Helix (bHLH) transcription factors origin- supports the premise that these cancers hijack TWIST’s ally found to modulate the expression of various embryonic functions, granting oncogenic and metastatic target genes through canonical E-box responsive ele- properties. In this review, we report on the history and ments (Lee et al., 1997; Yin et al., 1997). Well conserved recent breakthroughs in understanding TWIST protein during evolution, two TWIST genes exist in vertebrates, functions and the emerging role of the associated epithelial– TWIST1 and TWIST2 (formerly Dermo-1). The mesenchymal transition (EMT) in tumorigenesis. We then TWIST proteins display a high degree of sequence broaden the discussion to address the general contribution similarity in their C-terminal half, including the bHLH of reactivating embryonic programs in cancerogenesis. domain and a ‘TWIST-box’ protein interaction surface Oncogene (2010) 29, 3173–3184; doi:10.1038/onc.2010.92; and are more divergent in their N terminus (Wolf et al., published online 12 April 2010 1991; Li et al., 1995; Bialek et al., 2004). Among the bHLH proteins, TWIST transcription factors are unique Keywords: TWIST; embryogenesis; EMT; failsafe in their ability to form functional homo- and hetero- program escape; cell dissemination dimers. These two types of complexes were found to display distinct, even antagonistic, properties as exem- plified by the mesoderm specification and the subse- Introduction quent allocation of mesodermal cells to the somatic muscle fate by Twi-Twi (Twi is the Drosophila TWIST Originally shown to be central to embryonic develop- ortholog) homodimers whereas heterodimers between ment, aberrant functions of the TWIST transcription Twi and the E protein homolog Daughterless (Da) factor family have grown and spread into many facets of repress genes required for somatic myogenesis (Casta- tumor development and progression. The initial link non et al., 2001). The balance between TWIST1 homo- between these proteins and cancer identified TWIST1 as and heterodimers (TWIST1-TCF3/E2A (TE) or a pro-metastatic protein in breast carcinomas. Since TWIST1-Hand2 protein complexes) was also shown to then, the roles of this family (TWIST1 and TWIST2) in direct appropriate limb development as well as cranial various types of solid human cancers have expanded to suture patterning and fusion in mice (Firulli et al., 2005; include the development of primary tumor growth, cell Connerney et al., 2006, 2008), suggesting that this dissemination and chemoresistance. Recent progress in TWIST-regulatory mechanism is conserved across understanding TWIST physiological and oncogenic species. These different dimers further differ in stability functions highlights a number of parallels suggesting as the association of TWIST with TCF3/E2A proteins that reactivation of TWIST-associated embryonic pro- protect TWIST from the ubiquitin-dependent protea- grams de facto contribute to tumor progression. The aim some degradation pathway (Hayashi et al., 2007). of this review is to evaluate the status of the field by Multiple factors, including the relative amounts of the highlighting the most recent breakthroughs in under- binding partners, their phosphorylation status as well as standing TWIST functions and identifying future the expression of Id HLH proteins (that preferentially avenues of investigation. We will attempt to substantiate interact with E2A, titrating it and thereby favoring the parallel between the embryonic and oncogenic TWIST homodimers) determine the relative amounts of properties of TWIST proteins and, in particular, address both types of dimers and thereby TWIST activity (for a the role of the associated epithelial–mesenchymal recent review, see Firulli and Conway, 2008). transition in tumor progression and cell dissemination. TWIST proteins behave either as transcriptional repressors, recruiting histone deacetylases or inhibiting Correspondence: Dr S Ansieau, Inserm, U590, Centre Le´on Be´rard, 28 acetyl-transferases, or as transcriptional activators rue Laennec, Lyon, 69008, France. (Hamamori et al., 1999; Gong and Li 2002; Pan et al., E-mail: [email protected] or Professor A Puisieux, Inserm U590, 2009). These functions are independent of the dimer Centre Le´on Be´rard, 28 rue Laennec, 69008 Lyon, France. E-mail: [email protected] type and appear to be tissue-dependent in Drosophila. Received 30 June 2009; revised 9 November 2009; accepted 22 February This switchable behavior is linked to a particular 2010; published online 12 April 2010 conserved domain of Da, which inactivates both Twi TWIST embryonic and oncogenic properties S Ansieau et al 3174 and Da transactivation domains through an intra- pathway (Maestro et al., 1999; Sosic et al., 2003; molecular mechanism (Wong et al., 2008). As the dom- Puisieux et al., 2006; Doreau et al., 2009). Cell death ain is conserved across species, we anticipate that it might also result from inappropriate premature differ- similarly dictates TWIST tissue-specific effects in higher entiation as evidence strongly suggests a role of TWIST organisms. in maintaining myoblast progenitor cells during devel- In addition, TWIST proteins directly interact with opment (Hebrok et al., 1994). In vertebrates, TWIST1 is several other transcription factors, including MyoD, initially expressed throughout the somatic mesoderm MEF2, RUNX1, RUNX2, PGC1-a, p53 and NF-kB, and is subsequently silenced in the forming myotome and inhibit their activities (Spicer et al., 1996; Hama- (Hopwood et al., 1989; Wolf et al., 1991). Similarly, in mori et al., 1999; Sosic et al., 2003; Bialek et al., 2004; Drosophila, Twi is expressed in the somatic mesoderm at Sharabi et al., 2008; Shiota et al., 2008; Pan et al., 2009). high levels before differentiation and is required for As discussed in the following section, inactivation of differentiation of the somatic muscles (Baylies and Bate these transcription factors provides TWIST proteins 1996). Its expression is maintained in clusters of with the ability to inhibit an array of differentiation undifferentiated proliferating cells that will give rise to pathways while endowing survival properties. adult myoblasts and ultimately muscles throughout the larval and early pupal stages (Bate et al., 1991). TWIST1 inhibitory properties rely on their ability to interact and inactivate myogenic factors such as MyoD and MEF2 Physiological functions of Twist proteins (Spicer et al., 1996; Hamamori et al., 1997). The TWIST1 protein is also likely to be important for Embryonic development and tissue specification maintaining the immature phenotype of chondrocytes. Twi was originally identified in Drosophila as a zygotic By comparing chick caudal and cephalic sternal developmental gene crucial for the establishment of the chondrocytes that display distinct capabilities to term- ventral furrow, a prerequisite for the development of all inally differentiate, TWIST1 was found to inhibit mesoderm-derived internal organs (Thisse et al., 1987). chondrocyte differentiation by turning down RUNX2 In mice, Twist1 haploinsufficiency produces viable expression (Dong et al., 2007). Accordingly, in mice, offspring but leads to abnormal craniofacial structures Twist1 mRNA was found to be expressed in immature and polydactyly in the hind limb, a phenotype chrondrocytes whereas the transcript is undetectable in reminiscent to the dominantly inherited Saethre–Chot- more mature ones (Hinoi et al., 2006). Similarly, zen syndrome in humans where TWIST1 is mutated TWIST1 was found to inhibit osteoblastic differentia- (Bourgeois et al., 1998). Twist 1À/À mice, in contrast to tion. In mice, Twist1 is downregulated during endo- flies, undergo normal gastrulation but die at E10.5-11 chondral and intramembranous fetal bone marrow displaying severe defects in closure of the cephalic neural development (Alborzi et al., 1996). During calvaria tube, deficient cranial mesoderm, malformed branchial and suture development, Twist1 is expressed at sides of arches and facial primordium, and retarded limb bud immature bone cells and to less extent in more mature development. TWIST1 therefore acts after mesoderm ones (Rice et al., 2000; Connerney et al., 2006). specification in vertebrates, but remains essential in In addition, in mice and humans (Saethre–Chotzen mesoderm differentiation including muscle, cartilage syndrome), haploinsufficiency of TWIST1 has been and osteogenic cell lineages (Hebrok et al., 1994; Spicer shown to cause craniosynostosis, a premature closure et al., 1996; Lee et al., 1999). Mesoderm formation in of the cranial sutures and limb development abnormal- vertebrates is actually controlled by SNAI1 and SNAI2 ities (Howard et al., 1997; el Ghouzzy et al., 1997; transcription factors