1 Inserm, U590, Lyon, F-69008, France;2 Université de Lyon, Lyon I, ISPB, Lyon, F-69008, France;3 Centre Léon Bérard, Lyon, F-69008, France
*Correspondence: ansieau@lyon.fnclcc.fr, puisieux@lyon.fnclcc.fr Tel: 0478785160, Fax: 0478782720
December 2009
Abstract
The TWIST proteins are embryonic transcription factors that play key roles in embryonic development. While they remain largely undetectable in healthy adult tissues, both TWIST1 and TWIST2 genes are frequently reactivated in a wide array of human cancers, where they invariably correlate with more aggressive, invasive and metastatic lesions. In the last decade, the role of TWIST proteins in cancer has been deeply investigated, now offering a general overview how these transcription factors might promote tumor progression. In this review, we aim to summarize the current state of knowledge on TWIST oncogenic functions and to describe how hijacking embryonic processes by tumor cells is to their advantage.
TWIST protein features
The TWIST1 and TWIST2 (formerly Dermo-1) proteins belong to the huge bHLH transcription factor family. While they differ in their N-terminus, their C-terminal halves are sequentially very close, encompassing a conserved bHLH (basic Helix Loop Helix) motif as well as an interaction domain named "TWIST box". Through their bHLH motif, the TWIST proteins are able to recognize E-box responsive elements (CANNTG) and behave either as transcription repressors or activators, depending on the cellular context (Hamamori et al., 1999; Gong and Li, 2002; Pan et al., 2009). TWIST proteins are known to directly interact with a large set of transcription factors and to modulate their activity (Table 1). TWIST proteins additionally present the particular ability to either form homo- or heterodimers that display distinct, sometimes even antagonistic activities (Castanon et al., 2001; Firulli et al., 2005; Connerney et al., 2006; Connerney et al., 2008) (Table 1).
Table 1: Listing of the referred Twist partners and associated functions.
TWIST functions during embryogenesis
In Drosophila, the ancestral Twist gene twi was found as essential for proper gastrulation as well as for the generation of neural crest cells (Thisse et al., 1987; Leptin, 1991). The twi knockout is embryonic lethal during gastrulation. These embryos lack mesoderm-derived tissues leading to a torsion of the abdomen, giving rise to the TWIST name (Thisse et al., 1987). Twist1+/- heterozygote mice (Bourgeois et al., 1998) displays limb and craniofacial abnormalities reminiscent to the Saethre-Chotzen syndrome, a human craniosynostosis disorder associated with Twist1 haploinsufficiency (Pantke et al., 1975; Gripp et al., 2000). While TWIST1 is dispensable for gastrulation in mammalians, Twist1 -/- homozygous mice die at E10.5, displaying multiple defects including failure of neural tube closure, abnormal limb buds and increased apoptosis in the somites (Chen and Behringer, 1995). After birth, the expression of the murine Twist1 gene is mainly restricted to some cellular precursors of mesodermal origin (Wolf et al., 1991). While molecularly distinct, Twist1 and Twist2 expression patterns overlap (Li et al., 1995). Twist2 expression in mice increases in somites and limb buds during embryogenesis and restricts, with time, to the perichondrium, which becomes the connective tissue bordering cartilage and bones (Li et al., 1995). On the other hand, its expression is progressively increased in the dermis until birth and it is maintained in neonates but downregulated in adult tissues (Li et al., 1995). Twist2 knockout mice develop normally but die shortly after birth due to an overexpression of proinflammatory cytokines resulting in severe cachexia (Sosic et al., 2003). Thus, despite their sequence homology, TWIST1 and TWIST2 display distinct embryonic functions. However, Twist1 +/- Twist2 +/- and Twist2 -/- mouse phenotypes share similarities, suggesting some functional redundancies (Sosic et al., 2003).
TWIST functions in adult tissues
TWIST proteins in adult humans are mainly expressed in precursor cells including the myogenic, osteoblastic, chondroblastic, odontoblastic and myelomonocytic lineages, maintaining their undifferentiated state. TWIST1 protein was also found to be expressed in brown fat where it behaves as a key regulator of adaptative thermogenesis (a process that consists in dissipating energy as heat) (Pan et al., 2009). More recently, TWIST proteins were additionally found to perform important roles in lymphocyte function and maturation. In collaboration with the group of N. Bonnefoy-Berard, we demonstrated that TWIST1 and TWIST2 are key regulators of B cell activation in an inflammatory environment such as autoimmune disease (Doreau et al., 2009). TWIST1 expression is induced following B cell stimulation with IL-17 and BAFF cytokines, and is essential for naive or memory B cell survival and proliferation as well as their differentiation into immunoglobulin producing cells. Interestingly, this might explain the persistence and activation of auto-reactive B cells in lupus and thus the presence of autoimmune complex deposits that damage the organs of these patients. TWIST1 also functions downstream NF-κB in T helper 1 lymphocytes and is induced following repeated T cell Receptor binding. In these cells, TWIST1 inhibits the production of IFN-γ, IL-2 and INF-α, thus preventing their pro-inflammatory action (Niesner et al., 2008). Noticeably, in both B and T lymphocytes, TWIST1 behaves as an early response gene, suggesting common regulation mechanisms (Doreau et al., 2009; Niesner et al., 2008). Similarly, in macrophages, IFN-induced TWIST1 expression was shown to prevent TNF-α production (Sharif et al., 2006), contributing in down-modulating the inflammatory response. In adult mice, Twist2 expression is restricted to the dermis where it remains barely detectable (Li et al., 1995). Its expression in adult human tissues has not been directly investigated. It is worth mentioning that TWIST2 transcripts are undetectable in normal colon, esophagus, lung, kidney and diseased tissues of melanocytic controls (Ansieau et al., 2008). TWIST2 is transcriptionally active in a handful of mesenchymal tissues including myelomonocytic progenitors, it inhibits their proliferation and differentiation, and in mature myeloid cells, where it inhibits the production of various cytokines including IL-4, IL-10 and IL-12 (Sharabi et al., 2008).
TWIST genes are re-induced in multiple human cancer types
As previously mentioned, while these transcription factors are essential for embryonic development, their expression turns down after birth and is mainly restricted to precursor cells in adult tissues (Puisieux et al., 2006). Invariably, TWIST proteins are undetectable in epithelial cells. Conversely, TWIST1 and/or TWIST2 gene expression was found to be active in multiple carcinomas (breast, bladder, lung, kidney, colon, gastric, liver, pancreas, ovarian, prostate, head and neck and esophageal squamous cell carcinomas) and also frequently expressed in melanomas and sarcomas (Puisieux et al., 2006; Ansieau et al., 2008). In all cancer types, their expression is associated with poor prognosis, high grade, invasive and metastatic lesions (Puisieux et al., 2006). The mechanisms leading to the reactivation of both TWIST genes in human cancers have been deeply investigated. Likely resulting from a combination of additive effects rather than a single event, their reactivation relies on the deregulation of pathways that normally subtly regulate their expression during development, highlighting a striking relationship between embryogenesis and tumorigenesis (Bastid et al., 2008). A significant example is provided by stress conditions such as hypoxia or mechanical constrains that are known to turn on TWIST1 expression in both physiological and pathological conditions. Additionally, the constitutive activation of diverse signaling pathways (Wnt, IGF, EGF, TNF-α, IL-17) positively impact on TWIST expression and/or protein stability. Downstream effectors of these pathways include transcription factors such as NF-κB, STAT3 and c-MYC (Sosic et al., 2003; Rodrigues et al., 2008; Cheng et al., 2008). For example, TWIST1 was identified as a c-MYC target gene, essential for vasculature development in Xenopus and zebrafish (Bellmeyer et al., 2003; Rodrigues et al., 2008), and is presumably induced by N-MYC in neuroblastomas (Kawagoe et al., 2007). TWIST protein activity is also tightly regulated by post-translational modifications, offering an alternative to quickly adapt its activity to the cellular context. TWIST proteins are subjected to the ubiquitin-proteasome degradation pathway and their stabilization is driven by the phosphorylation of PKA sites that favors its dimerization with partners such as the E12/E47 and Hand bHLH transcription factors (Firulli and Conway, 2008). TWIST proteins also form labile homodimers that modulate the expression of distinct downstream target genes, providing them with specific functions (Firulli et al., 2005). The balance between the two complex types is highly regulated during development and is also likely to evolve during tumor progression. In particular, the reported modulation of Id expression in response to cell environment and mechanical constrains might be determinant in regulating the balance between these two complexes (Firulli and Conway, 2008; Castanon et al., 2001). Epigenetic events might also contribute to alter TWIST expression as their promoter sequences are subjected to methylation, although this still remains a matter of discussion (Martin et al., 2005; Gort et al., 2008; Fackler et al., 2009). At last, TWIST1 and TWIST2 might be subjected to chromosomal rearrangements in cancer cells as both genes are located in unstable chromosomal regions. The TWIST1 containing 7p21 locus is amplified in colon (Aragane et al., 2001), cervical (Choi et al., 2007) and gastric (Wu et al., 2002) cancers as well as pediatric osteosarcomas (Entz-Werle et al., 2005). Similarly, the TWIST2 locus containing 2q37 is involved in prostate and pancreatic tumors (Pierce et al., 2007; Loukopoulos et al., 2007).
Prometastatic properties of TWIST proteins
Comparison of the tumorigenic and metastatic potentials of murine isogenic breast cancer cell lines first identified TWIST1 as a prometastatic factor (Yang et al., 2004). A growing body of evidence now supports this conclusion, highlighting a preponderant role of TWIST proteins in promoting invasion and metastatic dissemination in multiple cancer types including breast (Mironchik et al., 2005; Yang et al., 2004; Cheng et al., 2007), prostate (Kwok et al., 2005), gastric (Yang et al., 2007), liver (Lee et al., 2006; Matsuo et al., 2009), head and neck (Yang et al., 2008), nasopharyngeal (Horikawa et al., 2007) cancers and gliomas (Elias et al., 2005). TWIST pro-metastatic potential relies on their ability to induce an epithelial to mesenchymal transition (EMT), a process that converts joined and polarized epithelial cells into isolated and motile mesenchymal ones, able to bypass the basement membrane and infiltrate into the surrounding extracellular matrix (Thiery and Sleeman, 2006; Yang and Weinberg, 2008; Nakaya and Sheng, 2008). This programmed embryonic transdifferentiation process involves the loss of cellular junctions (tight and adherent junctions as well as desmosomes) and a profound reorganization of the cytoskeleton. It is determinant in promoting multiple morphogenetic movements during the embryonic development, including the mesoderm formation during gastrulation, the neural crest delamination, the palate fusion and the formation of the heart cushions (Thiery et al., 2009; Yang and Weinberg, 2008). EMT contributes to several human pathologies, including renal, hepatic and lung fibrosis, by turning epithelial cells into collagen-producing mesenchymal ones (Iwano et al., 2002; Boutet et al., 2006; Zeisberg et al., 2007; Kim et al., 2006). More recently, it turns out to be a driving force of cancer cell dissemination during tumor development (Thiery et al., 2009). As epithelial cells that have undergone an EMT are phenotypically indistinguishable from fibroblasts, a role of EMT in cancers was controversial. Nevertheless, the physiological relevance of EMT in cancers in vivo has now unquestionably been experimentally confirmed (Spaderna et al., 2006; Ao et al., 2006; Rees et al., 2006; Fendrich et al., 2007; Trimboli et al., 2008). TWIST proteins promote EMT by turning-down the expression of epithelial specific proteins, such as the E-cadherin and by upregulating the expression of mesenchymal markers such as the N-cadherin, the vimentin and the smooth-muscle actin. Particular attention has been devoted on E-cadherin regulation as it behaves as the guardian of the epithelial phenotype, its knock-down being sufficient to promote an EMT (Onder et al., 2008). Inhibition of E-cadherin by TWIST proteins implies either direct or indirect (through induction of the SNAI1 zinc finger protein) effects on the E-cadherin promoter (Vesuna et al., 2008; Smit et al., 2009).
TWIST proteins display oncogenic properties
Beyond their prometastatic properties, TWIST proteins were found by our own group to display oncogenic properties by preventing senescence and apoptosis induction in response to oncogenic insults (Ansieau et al., 2008). These two oncosuppressive mechanisms are known to function in premalignant lesions (Gorgoulis et al., 2005; Bartkova et al., 2005; Collado et al., 2005; Braig et al., 2005; Chen et al., 2005; Bartkova et al., 2006; Sarkisian et al., 2007; Swarbrick et al., 2008; Michaloglou et al., 2005; Michaloglou et al., 2008; Ansieau et al., 2008; Dhomen et al., 2009) and their inhibition is required for malignant conversion. TWIST1 anti-apoptotic properties were originally highlighted in embryonic development as Twist1 -/- mice display a significant increase of apoptosis in their somites (Chen and Behringer, 1995). This observation was next supported by the detection of apoptotic osteoblasts in Saethre-Chotzen syndrome patients, a human syndrome of cranial malformation due to reduced Twist1 function (Yousfi et al., 2002). TWIST proteins were identified in a functional screen as proteins able to counteract c-MYC induced apoptosis, thus depicting them as potential survival factors promoting cell transformation (Maestro et al., 1999). In line with this observation, we have shown that TWIST1 overexpression is a prevalent p53 inactivation mechanism in neuroblastomas, a tumor where the p53 gene is usually not mutated, promoting tumor growth by alleviating N-MYC induced apoptosis (Valsesia-Wittmann et al., 2004). Recently, we further demonstrated that TWIST proteins were able to prevent oncogene-induced senescence, by neutralizing both Rb- and p53- dependent pathways, and to cooperate with oncogenes such as RASV12 or ERBB2 in murine primary fibroblast transformation. Inactivation of p53 and Rb-dependent pathways mainly relies on the ability of TWIST1 proteins to prevent the induction of cyclin-kinase inhibitors p21CIP1, p16INK4A and ARF encoding genes. Additionally, TWIST proteins directly interact with p53, prevent its activation by posttranslational modifications and titrate its co-activator CBP/p300 (Stasinopoulos et al., 2005; Maestro et al., 1999; Hamamori et al., 1999). By overriding failsafe programs, TWIST proteins are therefore likely to promote malignant conversion, playing a de facto role at the early steps of tumor progression. Since this discovery, TWIST1 expression was shown to be induced during the malignant conversion of bladder (Zhang et al., 2007), liver (Li et al., 2006), ovary (Yoshida et al., 2009), prostate (Kwok et al., 2005; Yuen et al., 2007), pancreatic (Ohuchida et al., 2007) tumors as well as melanomas (Ansieau et al., 2008) and pheochromocytomas (Waldmann et al., 2009). By inhibiting Rb and p53 tumor suppressors, TWIST proteins are suggested to provide cells with proliferative and survival advantages at the primary tumor sites.
Intimate crosstalk between failsafe program escape and cancer cell dissemination
Growing evidence suggest that proteins such as TWIST, instead of being constitutive EMT inducers, behave as sensors, promoting EMT in response to appropriate micro-environmental signals. Although these signals might cooperate with EMT inducers through the induction of independent and complementary signaling pathways, they are also believed to directly activate EMT-inducer activity. This is exemplified by the stabilization of SNAI1 in response to TNFα, providing an explanation of how immune cells might promote EMT and thereby foster metastatic dissemination (Wu et al., 2009). By performing oncogenic cooperation assays in epithelial cells, we also demonstrated that TWIST proteins override ERBB2 or RAS-induced senescence and cooperate with the mitogenic oncoproteins in promoting an EMT (Ansieau et al., 2008). Although RAS activation is known to strongly impact different components of the cytoskeleton, and could thereby synergize with TWIST proteins, this mitogenic protein is likely to directly modulate TWIST's stabilization and/or dimerization and, as a consequence, its activity. Importantly, this cooperative effect raises the possibility that failsafe program escape, when associated with the reactivation of embryonic genes such as TWIST, might promote malignant conversion, and concomitantly be determinant in fostering cancer cell dissemination (Figure 1). The initiation of the metastatic process thus would not be restricted to late stages of tumor progression, as commonly claimed. Instead, at least in some cancers, dissemination may be initiated much earlier, giving rise to disseminated single cells evolving independently of the primary tumor. Consistent with this idea, in a murine mammary tumor progression model, precancerous cell dissemination was found to initiate from atypical ductal hyperplasia, an early stage in tumorigenesis, in association with TWIST1 reactivation (Husemann et al., 2008). Some mechanisms associated with failsafe program escape might therefore lead to the pre-selection of cells particularly prone to disseminate in permissive conditions, giving a rationale to the established metastatic predictive signatures.
Figure 1: Reactivation of TWIST1 or TWIST2 expression in a benign tumor favors malignant conversion (by inhibiting apoptosis and senescence). In the presence of a abnormal mitogenic activation and in response to a permissive environment, tumor cells concomitantly acquire a mesenchymal and motile phenotype (through EMT) favoring invasion and dissemination, acquisition of cancer stem cell properties, a widespread chemoresistance to therapeutics and a global genomic instability. Furthermore, it also induces angiogenesis in hypoxic conditions, thus setting the table for the primary tumor growth as well as the formation of distant metastasis.
The pleiotroic oncogenic properties of TWIST proteins
By promoting an EMT, TWIST proteins provide cells with motility, thereby fostering cell dissemination. Recently, we and the Weinberg laboratory demonstrated in vitro that EMT also provides mammary epithelial cells with stem-like properties including a CD44high CD24low antigenic phenotype, the ability to form mammospheres in low adherent conditions and a high tumorigenic potential (Mani et al., 2008; Morel et al., 2008). This observation raises the possibility that "cancer-stem" cells, or "cancer-initiating" cells, may arise not only from the alteration of normal stem (or progenitor) cells but also from the dedifferentiation of differentiated cells. Knowing that EMT is highly sensitive to the microenvironment and is a reversible process, the balance between cancer-stem cells and differentiated cells in tumors is presumed to be highly dynamic (Gupta et al., 2009). Growing evidence strengthens the relevance of this observation in vivo. The CD44high CD24low cell population (an antigenic phenotype allotted to mammary stem cells (Al-Hajj et al., 2003)), isolated from mammoplasty, displays mesenchymal features (Mani et al., 2008). Consistently, various EMT inducers, including TWIST1 and SNAI2, take part of the cancer-stem cell signature that characterizes the most aggressive breast tumors (Hennessy et al., 2009). Additionally, the majority of disseminated cells arising from metastatic breast cancers, display mesenchymal and stem cell-like features (Aktas et al., 2009). Collectively, these data highlight an intimate crosstalk between embryonic transcription factors, EMT and the generation of cancer-stem cells. As both TWIST proteins display anti-apoptotic properties, their expression in tumor cells also induces a widespread chemoresistance to therapeutics drugs. TWIST1 or TWIST2 expression promotes resistance to paclitaxel (Cheng et al., 2007; Wang et al., 2004; Kwok et al., 2005; Kajiyama et al., 2007; Zhang et al., 2007), daunorubicine (Pham et al., 2007), etoposide (Pham et al., 2007), doxorubicine (Li et al., 2009) and cisplatine (Pham et al., 2007) in various cancer types. As a consequence, TWIST transcription factors are associated with recurrence and their expression is increased after radio- or chemotherapeutic treatments (Li et al., 2009; Yang et al., 2009; Hosono et al., 2007). Resistance is also provided by additional embryonic transcription factors, particularly members of the Snail family. Snail and TWIST proteins share the ability to down-regulate the p53 pathway and to promote EMT (Ansieau et al., 2008; Lee et al., 2009; Peinado et al., 2007). Although yet not clarified, both properties (potentially linked) might contribute to provide cells with higher resistance. Additional functions undoubtedly contribute to TWIST oncogenic properties. Induced in hypoxic conditions in response to HIF1α and HIF2α activation (Yang et al., 2008), TWIST1 promotes the production of VEGF and thereby induces neoangiogenesis. By turning down p53 activity, TWIST proteins also promote chromosomal instability (Mironchik et al., 2005; Vesuna et al., 2006), favoring appearance of secondary events and contributing thereby to tumor progression (Sieber et al., 2003). Compelling evidence supports a preponderant and unusual role of TWIST proteins (and probably most embryonic EMT-inducers) in tumor progression. They are capable of simultaneously promoting malignant conversion, cancer cell dissemination, acquisition of self-renewal capabilities, chromosomal instability and neoangiogenesis. These pleiotropic effects raise important questions regarding their prognostic and therapeutic interests. The proof of concept for TWIST as a prognostic marker in cancer has already been demonstrated in a variety of epithelial cancers (Zhang et al., 2007; Fondrevelle et al., 2009; Martin et al., 2005; Vesuna et al., 2008; Kyo et al., 2006; Yang et al., 2008; Niu et al., 2007; Yang et al., 2009; Song et al., 2006; Horikawa et al., 2007; Kajiyama et al., 2006; Hosono et al., 2007; Xie et al., 2009) as well as in melanomas (Hoek et al., 2004) and sarcomas (Man et al., 2005), at both mRNA and protein levels. The importance of post-translational modifications that affect the balance between homo- and heterodimers may well complicate the analysis. The interest of TWIST proteins as future therapeutic targets is revealed by the reactivation of apoptosis and senescence programs following the inhibition of gene expression, through RNA interference, in human beast, neuroblastoma and melanoma cell lines (Ansieau et al., 2008; Valsesia-Wittmann et al., 2004) . Presumably, this effect would be further enhanced when combined with standard anticancer chemotherapeutics such as paclitaxel, doxorubicine and cisplatine (Kwok et al., 2005; Zhang et al., 2007; Li et al., 2009; Zhuo et al., 2008). TWIST1 depletion also restores an epithelial phenotype, thereby reducing invasive and metastatic abilities of epithelial cancer cells (Kwok et al., 2005; Yang et al., 2007; Luo et al., 2008; Yang et al., 2004; Qin et al., 2009; Terauchi et al., 2007). In vivo, its inhibition abrogates the transformation, tumorigenic and metastatic potentials of tumor cell lines (Smit et al., 2009) and restores their sensitivity to conventional therapies (Li et al., 2009). We could therefore reasonably expect that further investigation on these EMT-inducers might open novel routes to eradicate tumors
Atlas of Genetics and Cytogenetics in Oncology and Haematology
Role of TWIST proteins in cancer progression
Online version: http://atlasgeneticsoncology.org/deep-insight/20081/role-of-twist-proteins-in-cancer-progression