Skeletal muscle formation in vertebrates constitutes an excellent system to study the signals and the molecular mechanism that govern cellular differentiation. During both embryonic development and adult myogenesis (e.g. muscle regeneration) a population of muscle progenitors, embryonic or fetal myoblasts and satellite cells, respectively, differentiate into mature myofibers. Recent studies have revealed striking similarities in the mechanisms and the mediators of skeletal myogenesis in the embryo and in adult life. For instance, the sequential activation of genes belonging to the same families (e.g., Pax3/7 and Myf5/MyoD), in response to the Wnt signalling pathway, promotes the transition towards a differentiated phenotype (Snider and Tapscott, 2003). Moreover, increasing evidence supports the notion that myogenic precursors can derive from non-muscle sources, such as bone marrow, endothelium, and other tissues. The epigenetic modifications that orchestrate the pattern of gene expression underlying the acquisition and the maintenance of the myogenic lineage, as well as the transition from muscle precursors to myofibers, have only recently become the object of investigation. Deciphering the network of signal that regulate these modifications holds the promise of revealing the targets for novel interventions toward boosting regeneration of skeletal muscles and, by analogy, possibly other tissues.
This review focuses on our current knowledge of the epigenetic changes regulating gene expression at the chromatin and DNA level, independently on the primary DNA sequence, to reprogram the nuclei of muscle precursors during developmental myogenesis and muscle regeneration. These epigenetic marks provide the blueprint by which the extra-cellular cues are interpreted at the nuclear level by the transcription machinery to select the repertoire of tissue-specific genes to be expressed. The reversibility of some of these changes necessarily reflects the dynamic nature of skeletal myogenesis, which entails the progression through two antagonistic processes—proliferation and differentiation. Other epigenetic modifications are instead associated to events conventionally considered as irreversible—e.g. maintenance of lineage commitment and terminal differentiation. However, recent results support the possibility that these events can be reversed, at least upon certain experimental conditions, thereby revealing a dynamic nature of many of the epigenetic modifications underlying skeletal myogenesis. The elucidation of the epigenetic network that regulates transcription during developmental myogenesis and muscle regeneration might provide the information instrumental to devise pharmacological interventions toward selective manipulation of gene expression to promote regeneration of skeletal muscles and possibly other tissue. J. Cell. Physiol. 207: 1–11, 2006. © 2005 Wiley-Liss, Inc.
DEVELOPMENTAL AND ADULT MYOGENESIS: COMMON MECHANISMS AND MEDIATORS
During development, myogenesis is first initiated from mesodermal precursors expressing the transcription factor Pax3 and located in the dermomyotome of the somites (epithelial spheres along the anterior–posterior axis of the embryo). In this early stage, the action of both positive (Wnt, Sonic hedgehog (Shh) and noggin) and negative (BMP4) signals emanated from adjacent structures in the embryo, controls the spatio-temporal expression pattern of two basic helix-loop-helix proteins belonging to the family of the muscle regulatory factors (MRFs), named Myf5 and MyoD (Cossu et al., 1996; Cossu and Borello, 1999; Buckingham et al., 2003). Both Myf5 and MyoD are necessary for the proper initiation of the skeletal muscle program, as deduced from the analysis of the single and double knock out mice (Arnold and Braun, 1996). Disruption in mice of the Myf5 locus, but not of MyoD, leads to a delayed and reduced myogenesis (Braun et al., 1994; Braun and Arnold, 1996). The presence of a very mild muscle phenotype in mice containing null mutations of MyoD is probably due to a compensatory effect of Myf5 (Rudnicki et al., 1992). The double mutant mice die soon after birth due to a lack of muscle structures, highlighting the importance of these two proteins in the establishment of the myogenic lineage (Rudnicki et al., 1993). MyoD and Myf5 are activated in distinct mesodermal sub-populations of muscle precursors (Braun and Arnold, 1996; Tajbakhsh et al., 1998). Once the mesodermal precursors are commited to the myogenic lineage in the dermomyotome, they downregulate Pax3 expression and delaminate from the dermomyotome to give rise to the myotome, where the differentiation proceeds with the expression of the late MRFs (myogenin and MRF4) and the MEF2 proteins (Buckingham et al., 2003). After these primitive myogenic precursors, other myoblasts proliferate and migrate to the forming muscles, where they align and fuse into multinucleated, terminally differentiated myotubes expressing structural and contractile proteins (Relaix et al., 2005).
Recent results have challenged this hierarchical model, showing that MRF4 is indeed transiently expressed in muscle precursors before MyoD, therefore suggesting a role for MRF4 in myogenic lineage determination (Kassar-Duchossoy et al., 2004). Muscles develop in embryos lacking MyoD and Myf5 in experimental conditions in which the MRF4 locus—which lies in the vicinity of Myf5 locus—is not disrupted (Kassar-Duchossoy et al., 2004). A re-examination of the expression pattern of the four MRFs during embryonic development also showed that MRF4 appears before or at the same time as Myf5 in somites (Kassar-Duchossoy et al., 2004); an early expression of MRF4 transcripts during somitogenesis was also observed by others (Hinterberger et al., 1991; Summerbell et al., 2002). However, the role of MRF4 at this early stage of development remains to be elucidated.
Muscle regeneration in the adult is an unfrequent event under normal circumstances, with very few fibers being replaced during the life of the organism. A regenerative response is observed either upon physiological injuries (e.g., mechanical stress or exercise) or after muscle degeneration occurs, as a consequence of the degeneration caused by genetic muscular diseases (e.g., dystrophies) (Morgan and Partridge, 2003). In these circumstances, regeneration is a remarkable feature of skeletal muscles, aimed to avoid massive loss of muscular mass. Muscle regeneration entails several processes, including necrosis of the damaged tissue, reactive inflammation and local release of cytokines, growth factors, and other soluble substances (reviewed in Charge and Rudnicki, 2004). The main source of myogenic precursors in the adult is provided by a population of quiescent mononucleated cells called satellite cells. Satellite cells are located within the basal lamina of individual myofibers. Several biomarkers of satellite cell identity and activation have been defined, including transcription factors such as Pax7, Msx1, or MyoD, cell surface molecules like the tyrosin kinase receptor c-Met and the calcium-dependent adhesion molecule m-cadherin, or structural proteins like desmin (Cornelison and Wold, 1997; Beauchamp et al., 2000; Cornelison et al., 2000; Seale et al., 2000; Charge and Rudnicki, 2004). Upon muscle injury, satellite cells first enter the cell cycle and undergo several rounds of proliferation. A small subset of the cells, however, exits the cycle and relocates in the basal lamina, thereby replenishing the satellite cells pool for future regeneration processes. Zammit et al., reported that these two populations within the satellite cell compartment are discriminated by their different patterns of gene expression (Zammit et al., 2004). Upon activation, Pax7 positive satellite cells upregulate Myf5 and MyoD. Most of the Pax7 positive/MyoD positive cells undergo several rounds of proliferation. Downregulation of Pax7 coincides with the activation of the differentiation program, which is initiated by MyoD, entails the coordinate expression of the late MRFs, of the cell cycle inhibitor p21 and the structural and contractile proteins, and culminates with the formation of new myofibers repairing the damaged muscles. A distinct population of satellite cells downregulate MyoD, while maintaining the expression of Pax7, and remain located in small clusters of cells beneath the basal lamina, being responsible for maintaining the pool of cells available for further regenerative responses (Zammit et al., 2004). The ability of the satellite cells to sustain muscle regeneration becomes impaired with aging due to a decline in satellite cells proliferation upon activation, rather than to an age-associated decrease in the number of progenitor cells (Conboy et al., 2003). This age-related impairment of the regenerative potential is the consequence of an insufficient Notch signalling in old muscles (Conboy and Rando, 2002, 2005).
The importance of Pax7 in satellite cell development is underscored by the analysis of the knock out mice. Pax7−/− mice show a decrease in skeletal muscle mass and, although it was originally described that they lack satellite cells (Seale et al., 2000), a recent report by Oustanina et al. showed that they contain a reduced, yet detectable, number of these myogenic precursors (Oustanina et al., 2004). Interestingly, these authors reported that despite the reduction in satellite cells in skeletal muscle mass, Pax 7 (−/−) mice contain enough satellite cells to sustain normal postnatal growth of myofibers. However, injury-activated muscle regeneration is impaired in the adult Pax7 knock out mice due to a defect in satellite cells renewal and propagation (Oustanina et al., 2004). According to an essential role of Pax7 in regeneration, Pax 7 is necessary for the expression of MyoD (Seale et al., 2004). And MyoD appears to have a unique role in promoting satellite-mediated regeneration (Megeney et al., 1996). Recently, Polesskaya et al., observed that upon muscle injury there is an up-regulation of several Wnt isoforms (Wnt 5a, 5b, 7a, and 7b) suggesting a role for these proteins in muscle regeneration (Polesskaya et al., 2003). Independent studies failed to detect increase of Wnt members (Wnt 1, 3a, 7a, and 11) in regenerating muscles (Zhao and Hoffman, 2004). The Wnt family of proteins regulates skeletal muscle development through a signalling network that is initiated by their binding to Frizzled receptors (Fzd) expressed by target cells. Binding of Wnts to Fzd leads to the inactivation of the glycogen synthase kinase-3 (GSK-3) and promotes beta-catenin nuclear traslocation and activity (Cossu and Borello, 1999; Logan and Nusse, 2004; Reya and Clevers, 2005). Polesskaya and co-workers demonstrated that Wnt signalling induces Pax7 expression in a subpopulation of muscle derived stem cells through a canonical beta-catenin pathway (Polesskaya et al., 2003). The Wnt/Pax7/MyoD signaling during muscle regeneration in the adult is reminiscent of the network that establishes the myogenic identity during embryonic myogenesis, and underscores the similarities between these two processes (Snider and Tapscott, 2003). Recent studies have identified CREB as a novel nuclear target of a non-canonical Wnt pathway (Chen et al., 2005). Mice deficient for either the activity or the expression of CREB show an impaired expression of Pax3, MyoD, and Myf5. Thus, the Wnt signalling pathway can establish the myogenic identity via distinct pathways.
MULTIPOTENT PRECURSORS AS A SOURCE OF MUSCLE CELLS
Despite being the main cellular source for muscle regeneration, satellite cells are not the only cells able to activate the myogenic program in the adult. An increasing number of studies during the past years indicate that bone marrow-derived cells (BMDC), adult stem cells—referred as muscle Side Population, mSP)—mesoangioblasts and other mesodermal-derived potential muscle progenitors can incorporate into the muscle of the recipient mice, albeit with a low efficiency (Goodell et al., 1996; Ferrari et al., 1998; Gussoni et al., 1999; Jackson et al., 1999; Minasi et al., 2002; Cossu and Bianco, 2003; Majka et al., 2003). The description of both “stemness” and cell identity surface markers, together with the analysis of the biochemical properties of the different cell populations, has allowed the isolation and characterization of non muscle precursor cells with myogenic potential. Of them, BMDC and in particular adult hematopoietic stem cells (HSC) sorted from the bone marrow have extensively been studied. Adult HSCs are multipotent progenitors, responsible for maintaining the hematopoietic system during the whole life of the organism (Weissman et al., 2001). They were identified and isolated in basis of several surface markers. HSC contain undetectable levels of lineage specific markers, and are therefore considered lineage negative (lin−). Mouse HSCs are positive for two surface antigens: the c-kit receptor tyrosin kinase, and the glycophosphatidyl inositol-linked immunoglobulin superfamily molecule, Sca-1, and present low levels of Thy-1.1, whereas human HSC can be isolated from umbilical cord blood as the lin-, c-kitlo, CD34+, Thy-1+, and CD38− population (revised in Weissman et al., 2001).
A subpopulation of cells isolated from blood and muscles and defined by the peculiar property of excluding organic dyes, such as rhodamine 123 or Hoescht 33342, has been named Side Population (SP), and contains high levels of progenitor cells, which retain the potential to differentiate into several lineages (Goodell et al., 1996; Majka et al., 2003). Bone marrow SP can contribute to muscle regeneration, while muscle SP (mSP) can reconstitute the hematopoietic system of lethally irradiated mice (Gussoni et al., 1999; Jackson et al., 1999). Importantly, muscle stem cells that differentiate into hematopoietic lineages retain their ability to differentiate into myotubes (Cao et al., 2003) Finally, the CD45+, Sca1+ muscle resident population is capable to contribute to myofiber repair, at least under some circumstances (discussed below).
Despite the low incorporation of non canonical muscle progenitors into the muscle of the recipient mice, the efficiency with which these multipotent cells incorporate into pre-existing fibers from the host increases in response to muscle lesions or in mice models of muscular dystrophies (Fukada et al., 2002; LaBarge and Blau, 2002; Camargo et al., 2003; Sampaolesi et al., 2003; Sherwood et al., 2004; Palermo et al., 2005). This suggests that activation of the muscle program in multipotent muscle precursors can be modulated by environmental cues. In accordance with this hypothesis, it has been shown that mSP cells activate the myogenic program if cultured in the presence of myoblasts, and fail to form myotubes when cultured alone (Asakura et al., 2002). Similarly, muscle resident CD45+, Sca1+ stem cells give place to myogenic cells only when co-cultured with myoblasts or in response to Wnt activation (Polesskaya et al., 2003). Moreover, intramuscular injection experiments showed that CD45+, Sca1+ cells sorted from normal muscle are not myogenic in vitro, and do not efficiently contribute to muscle repair in vivo (McKinney-Freeman et al., 2002), whereas following injury there is an expansion of the CD45+, Sca1+ population that can undergo myogenic differentiation both in vitro and in vivo (Polesskaya et al., 2003). Finally, a recently identified population of cells of endothelial derivation—the mesoangioblasts—has been reported to differentiate toward the myogenic lineage when co-cultured with muscle cells (Minasi et al., 2002). Collectively, these data suggest that environmental cues, possibly provided by muscle cells or other cell types adjacent to muscles, somehow instruct progenitor cells to enter the myogenic program. This notion highlights the importance of elucidating the intracellular signalling that transmits external cues to the nucleus to achieve the differentiation-associated reprogramming of the genome (Pomerantz and Blau, 2004).
In spite of the isolation and characterization of different cell sources that can be exploited to enhance muscle regeneration, the actual contribution of non-satellite, multipotent muscle precursors to myofiber repair in vivo is still under debate. In an effort to clarify the origin of adult muscle progenitor cells, Sherwood et al. (2004)) used a clonal assay to identify myofiber associated cells that were able to form myogenic colonies at high frequency in vitro and to efficiently regenerate muscle in vivo. They identified a subpopulation of muscle resident CD45−, Sca1−, CD34+ cells as highly myogenic both in vivo an in vitro, in contrast to previous work that placed the highest myogenic potential upon injury in the CD45+ fraction (Polesskaya et al., 2003). When the intrinsic myogenic potential of myofiber-associated cells, including BMDC, circulating cells, and hematopoietic stem cells (HSC), was evaluated in vivo, BMDC displayed limited myogenic activity, despite the fact that they localize in the same anatomical compartment as muscle satellite cells, and that a subpopulation of the cells express some myogenic markers. When sorted and tested in a colony-forming assay, muscle-resident BMDC cells lacked the myogenic potential, unless they were co-cultured with myofiber-associated cells previously isolated from an experimental model of muscle injury (Sherwood et al., 2004). On the other hand, lin-, c-kit+, Sca1+, Thy1.1lo HSC do not appear to have myogenic activity in vitro (Sherwood et al., 2004).
Altogether, these studies suggest that some, but not all, non-muscle derived progenitor cells can differentiate into muscle when exposed to the appropriate signals. However their myogenic potential is limited as compared to conventional muscle progenitors—the satellite cells. The above reported studies reveal that the main source of cells suitable for muscle regeneration is provided by satellite cells; yet, they emphasize the potential interest of non-satellite muscle cells as target for pharmacological intervention toward boosting muscle regeneration. Further in vivo studies will be necessary to unequivocally assess the actual role of these adult derived multipotent cells during muscle regeneration. Notably, it still remains unknown whether non-satellite muscle precursors transit through the satellite state during their differentiation.
The complex pattern of gene expression underlying the determination of satellite cell identity, their activation and differentiation in response to muscle injury, reflect a network of epigenetic modifications that are only partially known and will be the focus of the following sections.
CHROMATIN MODIFICATIONS DURING THE TRANSITION FROM MULTIPOTENT MUSCLE PRECURSORS TO DIFFERENTIATED MUSCLE CELLS
Coordinated activation and repression of different subsets of genes underlies the progression of multipotent muscle progenitors into differentiated muscle cells (Pomerantz and Blau, 2004). These changes allow the acquisition of the myogenic identity and restrict the expression of genes of unrelated lineages—a process known as lineage commitment. In order to complete the differentiation program, the nucleus of a multipotent cell must therefore be sequentially reprogrammed to acquire and maintain the new pattern of gene expression. This global genome reprogramming entails both the repression of genes associated to the acquisition of non-muscle lineages and the selective activation of genes involved in the establishment of myogenic differentiation. Undifferentiated muscle precursors maintain the myogenic lineage while proliferating, and finally enter the differentiation program to give place to terminally differentiated cells. Terminal differentiation is characterized by the sequential activation of different subset of muscle specific genes (Bergstrom et al., 2002), the maintenance of their expression, and the silencing of genes involved in cell cycle progression. A scheme showing the different stages of muscle differentiation is shown in Figure 1.
The epigenetic marks that allow the progression through these stages are described below.
Stage 1—acquisition of the myogenic lineage
At the molecular level, muscle differentiation is controlled by a family of muscle specific basic helix-loop-helix (bHLH) transcription factors called Muscle Regulatory Factors (MRFs) that cooperate with members of the MEF2 family of proteins (MEF2A-C) in the activation of muscle genes from previously silent loci (Molkentin et al., 1995; Olson et al., 1995; Puri and Sartorelli, 2000). Enforced activation of any of the four muscle bHLH factors (Myf5, MyoD, myogenin, or MRF4) converts 10T1/2 fibroblasts to a muscle phenotype (Davis et al., 1987; Braun et al., 1989; Rhodes and Konieczny, 1989; Wright et al., 1989; Miner and Wold, 1990), making them the ideal candidates to start the nuclear reprogramming necessary to specify the myogenic lineage. In cultured myoblasts induced to differentiate, the expression of the early muscle MRFs, MyoD and/or Myf5, preceeds the expression of myogenin and MRF4 and is therefore likely to be the first step in the commitment of multipotent precursors to muscle (Sassoon et al., 1989; Bober et al., 1991; Ott et al., 1991; Zhang et al., 1995). However, the mechanisms leading to MyoD (or Myf 5) induction in muscle precursors are still largely unknown. It has been recently described that the homeobox protein Msx1 plays a pivotal role in the temporal control of myogenesis (Lee et al., 2004). Msx-1 is implicated in inhibiting the differentiation of several mesenchymal lineages, including skeletal muscle, during embryogenesis (Bendall et al., 1999; Bendall and Abate-Shen, 2000). When looking for interacting partners for Msx-1 in C2C12 myoblasts Lee et al., found that it preferentially binds to the linker histone H1b (Lee et al., 2004). Linker histones are so called because of their role in keeping the nucleosomes juxtaposed, to form a highly compacted chromatin structure. Apart from their structural role, they participate in other regulatory functions, such as DNA repair (Downs et al., 2003), apoptosis (Konishi et al., 2003) and gene expression (Bouvet et al., 1994; Steinbach et al., 1997). Msx-1 targets H1b to the regulatory region of MyoD, and this recruitment is necessary for Msx-1 dependent repression of MyoD and inhibition of myogenesis (Lee et al., 2004). Of the several isoforms present in mice, the H1b subtype is expressed in undifferentiated cells, including muscle precursors, and its expression decreases as differentiation proceeds (Wang et al., 1997), consistent with a role in temporally restricting gene expression during development. It remains to be established whether induction of MyoD is achieved by simple de-repression or if an active mechanism contributes to stimulate MyoD expression in myogenic precursors.
Stage 2—maintainance of the myogenic lineage during proliferation and cytokinesis
In proliferating myoblasts, the ability of the MRFs to activate the differentiation program is countered by their association of muscle regulatory regions with histone deacetylases (HDACs) and co-repressor complexes, including YY1 and Polycomb proteins, which preclude premature muscle-gene expression by promoting histone modifications (hypoacetylation and hypermethylation of specific lysines) (Lu et al., 2000a,b; McKinsey et al., 2000a,b, 2002; Mal et al., 2001; Puri et al., 2001; Zhang et al., 2002; Caretti et al., 2004). Posttranslational modifications of the histone tails constitute a key step in the control of gene expression (Fischle et al., 2003). Acetylation of specific lysine residues in histones 3 and 4 by acetyltransferases results in a more relaxed chromatin structure due the lost of a negative charge in the lysine residue, thereby leading to gene activation. Conversely, histone hypoacetylation is often associated to gene repression. Three distinct families of HDACs (Thiagalingam et al., 2003) play an important role in keeping the inactive state of muscle regulatory regions in proliferating myoblasts. Of them, class I HDACs associate with MyoD in undifferentiated myoblasts, and this association is disrupted upon induction of differentiation (Mal et al., 2001; Puri et al., 2001). The interactions between class I HDACs and MyoD is regulated by cell cycle dependent changes in the phosphorylation status of pRb (Puri et al., 2001). Class II HDACs bind and repress MEF2 proteins (Lu et al., 2000a,b; McKinsey et al., 2000a,b, 2002; Zhang et al., 2002). The mechanisms that regulate class II HDACs dissociation from MEF in response to differentiation stimuli have been defined in detail (see below). A third class of histone deacetylase that negatively regulates the myogenic program is the NAD(+)-dependent histone deacetylase Sir2. Sir2 regulates skeletal muscle differentiation probably by sensing the redox status of the cell in response to exercise, food intake, or starvation (Fulco et al., 2003).
More recently, histone methylation has emerged as a new mechanism of epigenetic marking (Rice and Allis, 2001; Zhang and Reinberg, 2001). However, while histone acetylation almost always correlates with transcriptional activation, histone methylation has been associated to both transcriptional repression and activation (Zhang and Reinberg, 2001). Furthermore, while acetylation always targets specific lysines in histone tails, methylation by distinct methyltransferases can be directed towards either lysine or arginine, leading to different effects on transcription. For instance, arginine methylation in histones 3 and 4 and methylation of lysine 4 (K4) in histone 3 are linked to active transcription (Zhang and Reinberg, 2001). On the other hand, methylation of lysine 9 (K9) in histone 3 by the Suv39h1 methyltransferase represses gene expression due to the recruitment of the heterochromatin associated protein HP1 (Bannister et al., 2001; Nakayama et al., 2001; Nielsen et al., 2002), which further associates with several repressory complexes (Taddei et al., 2001; Zhang et al., 2002; Geiman et al., 2004). Moreover, the recruitment of HP1 helps to spread then epigenetic mark by recruiting Suv39h1 to the vicinity of the modified histones, constituting a mechanism for stably silencing gene expression. In myoblasts, H3 K9 methyltransferase activity is recruited to MEF2 target genes via class II HDACs. Zhang et al., showed that two members of the family, HDAC4 and 5, and the MEF2 repressor MITR (a splicing variant of HDAC9 that lacks of enzymatic activity), associate with HP1. Furthermore, the HP1–HDAC and HP1–MITR interactions are disrupted upon induction of differentiation, correlating with a decrease of K9 H3 methylation in the vicinity of the MEF2 binding sites (Zhang et al., 2002). Another family of histone methyltransferases that regulate muscle specific gene transcription is formed by the Ezh Polycomb proteins (Caretti et al., 2004). Polycomb proteins are transcription repressors that play an important role in the regulation of the Hox genes and therefore in the establishment of the anterior-posterior axis during embryonic development in vertebrates (Simon et al., 1992; Orlando, 2003). Amongst the members of the Polycomb family, Ezh1 and 2 contain a conserved SET domain that confers to them methylatransferase activity toward H3 K27 (Cao et al., 2002; Cao and Zhang, 2004). Recently, Caretti et al. (2004) showed that Ezh2 is downregulated in the dermomyotome of the somites during embryonic development, coincident with the activation of the myogenic lineage. Ezh2 expression also decreases during satellite cells activation. During myoblasts differentiation in culture, Ezh2 downregulation correlates with hypomethylation of H3-K9 at muscle regulatory regions. Moreover, increased Ezh expression inhibits muscle differentiation due to the methylation of muscle regulatory regions and the recruitment of the correpresor YY1 (Caretti et al., 2004). SET domains are found in several methyltransferases, whose activity toward distinct lysines is associated to either repression or activation of gene transcription (Xiao et al., 2003; Bottomley, 2004). Unlike histone acetylation, histone lysine methylation is commonly considered as a stable repressive modification, associated with heterochromatin formation and permanent silencing of gene expression, possibly via the recruitment of chromodomain-containing proteins. However, recent evidence has begun to challenge this notion. The observation that histone methylation is implied in regulating reversible gene expression suggests a broader role for this modification in the control of cellular processes (Sims et al., 2003). On the other hand, this “reversibility” implicates the need for a demethylation mechanism. There are currently several models to explain the methylation turnover from histones tails, including the replacement of the modified histone (Johnson et al., 2004) and the proteolytic processing of the histone tail (Allis et al., 1980). The most direct, and also the most suitable, mechanism to explain a dynamic regulation of histone methylation, is the rupture of the covalent bond between the methyl group and the lysine or arginine at the histone tail, reaction catalysed by a long searched histone demethylase. A recent report describing a demethylase that specifically removes the methyl group from dimethylated (but not trimethylated) lysine 4 in histone 3, via an oxidation reaction, strongly supports this model and has increased the efforts in identifying novel demethylases (Shi et al., 2004). However, it is necessary to note that, as methylation of K4-H3 is a mark of active chromatin, the newly described demethylase is actually a repressor of transcription. Up to date there are no activities identified involved in eliminating repressive methyl groups from histone tails. One alternative mechanism to erase repressive methylation is provided by histone variant exchange (Johnson et al., 2004; Sarma and Reinberg, 2005), although the significance of this event in controlling gene expression during myogenesis is unknown so far.
The presence of MyoD and/or Myf5 in the nucleus of myoblasts prior to the differentiation is induced raises an important question: how are these proteins able to establish and maintain the muscle identity in proliferating myoblasts without inducing terminal differentiation? It was originally assumed that MyoD was unable to bind their recognition sequences—Eboxes—on muscle specific regulatory regions in myoblasts. This assumption was based on the observation that in myoblasts high levels of the serum-inducible inhibitor of differentiation (Id) proteins sequestrate the heterodimeric partners of MyoD, the products of the E2A gene (E12 and E47), thereby preventing MyoD from binding the DNA (Benezra et al., 1990; Lassar et al., 1991). More recently, chromatin inmunoprecipitation (ChIP) studies showed an absence of MyoD binding to most muscle regulatory regions in myoblasts (Bergstrom et al., 2002; Caretti et al., 2004; Simone et al., 2004; Lluis et al., 2005), although in a report by Mal et al., the recruitment of MyoD to myogenin promoter was detected before any phenotypic differentiation was observed (Mal and Harter, 2003). The discrepancies between these studies may result from different culture conditions. Alternatively, they could reflect a limited sensitivity in the detection assay.
The introduction of the ChIP on chip technique, which allows the identification of transcription factors binding sites through the genome, has provided a powerful tool to address these questions. Using a combination of microarray analysis and chromatin inmunoprecipitation, Blais et al., have recently shown that MyoD is transiently bound to a series of regulatory regions in proliferating myoblasts, many of which are also target genes in differentiated myotubes (Blais et al., 2005). However, further studies are necessary to analyze if these genes are actually activated by MyoD at both stages. On the other hand, the report by Blais et al., points at the function of MyoD as an active determinant of the myogenic programme, whose role would be to mark specific chromatin regions for the subsequent differentiation-associated reprogramming. In this respect, many of the sequences occupied by MyoD in myoblasts belong to regulatory elements of transcription factors themselves, which could cooperate with MyoD and other MRFs to amplify the myogenic signal as differentiation proceeds. Thus, MyoD can generate a feed-forward circuit that temporally patterns gene expression during myoblast-to-myotube transition (Penn et al., 2004).
The observation that MyoD is bound to the chromatin of regulatory regions in proliferating myoblasts raises the question of how this transcription factor accesses to its binding sites in the context of repressive chromatin. A recent report by Berkes et al., provides a plausible model to explain the stable recruitment of MyoD to myogenin promoter in myoblasts (Berkes et al., 2004). According to this model, MyoD would be recruited to myogenin promoter through an interaction with the homeodomain protein Pbx, which is constitutively bound to the chromatin of myogenin promoter.
Stages 3 and 4—activation and maintainance of the differentiation program
Upon induction of differentiation, the HDACs and their associated co-repressors dissociate from the MRFs and MEF2 factors, allowing the productive recruitment to the chromatin of muscle genes of chromatin modifying complexes, including the acetyltransfersases (HATs) CBP/p300, PCAF, p/CIP, SRC1 and GRIP, the arginine methyltransferase CARM-1 and the ATP-dependent SWI/SNF chromatin remodelling complexes (Eckner et al., 1996; Yuan et al., 1996; Sartorelli et al., 1997; Puri et al., 1997a,b; Chen et al., 2000, 2002; de la Serna et al., 2001; Wu et al., 2005). Acetylation of the histone tails by p300 and PCAF results in a relaxed chromatin structure permissive for transcription, while acetylation of MyoD by the same acetyltransferases increases the affinity for its recognition site in the DNA (Puri et al., 1997b; Sartorelli et al., 1999; Polesskaya and Harel-Bellan, 2001; Polesskaya et al., 2000; Polesskaya et al., 2001a,b; Dilworth et al., 2004). Despite being two acetyltransferases present in the same complex and both capable to acetylate MyoD and histones, p300 and PCAF do not provide redundant enzymatic activities (Puri et al., 1997b). Recent in vitro experiments with purified recombinant proteins and “chromatinized” templates have demonstrated a preferential activity of p300 toward histones, with PCAF-dependent acetylation of MyoD on three evolutionary conserved lysines (K99-K102 and K104) providing a further step required for initiation of transcription (Dilworth et al., 2004) Furthermore, p300-mediated acetylation of multiple lysines is essential for MEF2 function (Ma et al., 2005). The importance of the acetyltransferases p300 and PCAF in the myogenic program is underscored by experiments of either functional or genetic inactivation of p300 or PCAF, which is sufficient to block skeletal myogenesis in cultured cells and in mouse embryos (Puri et al., 1997b; Polesskaya et al., 2001b; Roth et al., 2003). According to their early pattern of expression in muscle progenitors, MyoD and Myf5 have the unique ability to initiate the myogenic program by promoting chromatin remodelling at previously silent loci (Gerber et al., 1997). This ability is conferred by two conserved regions, a cystein–histidine rich region and the carboxy-terminal region. They differ from the corresponding regions of myogenin (Bergstrom and Tapscott, 2001), and are likely to provide the surface for the recruitment of SWI/SNF remodelling complexes to muscle regulatory regions (de la Serna et al., 2005). As hyperacetylation is one important signal to recruit bromodomain-containing proteins, like acetyltransferases and SWI/SNF complex members, lysine acetylation of histones, MRFs and MEF2 proteins is likely to provide the epigenetic mark that stabilizes the myogenic transcriptosome. However, pharmacological disruption of interactions between the components of this transcriptosome has revealed that local hyperacetylation alone is not sufficient to recruit the SWI/SNF complex to the chromatin of muscle-regulatory regions (Simone et al., 2004).
Simultaneous to the expression of muscle genes, terminal differentiation is accompanied by an exit from the cell cycle, as achieved by the induction of cyclin dependent kinase (CDKs) inhibitors such as p21/CIP or p27, and the inhibition of pro-mitogenic genes (Kitzmann and Fernandez, 2001). MyoD promotes the transcription of both p21/CIP (Guo et al., 1995; Halevy et al., 1995) and the tumor suppressor pRb (Martelli et al., 1994; Magenta et al., 2003). High levels of p21 allow the cell to exit the cycle by targeting and inactivating the CDKs, leading to pRb dephosphorylation (Kitzmann and Fernandez, 2001). Hypophosphorylated pRb represses the transcription of E2F dependent genes that are necessary for cell cycle progression by both binding the E2F family of transcription factors and impairing their transactivation potential (Flemington et al., 1993; Blais and Dynlacht, 2004), and by recruiting repressor complexes to E2F-regulated promoters (Zhang et al., 1999). Amongst these multiprotein complexes there are chromatin modifying enzymes such as the catalytic subunits of the SWI/SNF complex (Brg-1 and Brahma (Brm)), class I HDAC, histone methyltransferases like Suv39h1 and members of the Polycomb family of transcriptional repressors (Dunaief et al., 1994; Strober et al., 1996; Trouche et al., 1997; Zhang et al., 1999, 2000; Strobeck et al., 2000; Nielsen et al., 2001; Ogawa et al., 2002; Ait-Si-Ali et al., 2004). In quiescent and terminal differentiated cells, the combined activities of multi-protein complexes endows the Rb-associated transcriptosome with the enzymatic activity necessary to repress the genes that are necessary for cell cycle progression. Two recent studies indicate that the role of pRb is restricted to the establishment of the cell cycle withdrawal and mitogen resistance in differentiating myoblasts, and does not extend to the maintenance of the post-mitotic state in differentiated myotubes (Camarda et al., 2004; Huh et al., 2004). In these studies, acute pRb deletion in either cultured myotubes or myofibers of adult mice did not cause reactivation of DNA synthesis, despite the re-activation of E2F-dependent transcription of genes leading to G1–S phase progression. Since pRb-interaction with histone deacetylases and lysine methyltransferases is essential for the establishment of the post-mitotic state in myotubes, it is likely that pRb promotes epigenetic modifications at particular loci (e.g., cell cycle genes), such as histone hypoacetylation and methylation, leading to chromatin condensation and formation of heterochromatin, which eventually persist in the absence of pRb.
Furthermore, it is worthy to note that chromatin modifications to temporary silence cell cycle regulatory regions in quiescent cells must differ from those that control the permanent exit from the cell cycle during terminal differentiation. For instance it has been shown that cyclinA repression in quiescent mouse embryonic fibroblasts (MEFs) is associated to Brahma-dependent remodelling of its promoter, whereas terminal differentiation in muscle C2C12 cells correlates with Suvar39h1-dependent K9-H3 methylation and silencing of S phase genes (Ait-Si-Ali et al., 2004).
BEHIND THE CHROMATIN: DNA METHYLATION CHANGES DURING MYOGENESIS
DNA methylation is the major epigenetic modification in mammals, affecting around 70% of the CpGs in the genome. The addition of a methyl group into cytosines within the context of CpG dinucleotides, catalysed by a family of DNA methylatransferases (Dnmts), inhibits gene expression due to the recruitment of a family of proteins that specifically recognize methylated CpGs, the methyl binding domain proteins (MBDs) (Bird, 1992; Hendrich and Bird, 1998). MBDs proteins act by recruiting complexes containing histone deacetylase and methyltransferase activities to the chromatin of methylated regions (revised in (Jaenisch and Bird, 2003). More recently, an association between MBDs, Dnmts and members of the SWI/SNF chromatin remodellers has been observed, linking DNA methylation associated transcriptional silencing with local ATP-dependent remodelling (Geiman et al., 2004; Harikrishnan et al., 2005). Numerous studies have shown a correlation between tissue specific gene expression and demethylation of the associated regulatory region (Saluz et al., 1986; Sullivan and Grainger, 1987; Paroush et al., 1990; Bruniquel and Schwartz, 2003) The first link between DNA methylation and the activation of the myogenic program stems from the observation that treatment with a methyltransferase inhibitor, 5-azacytidine, converts 10T1/2 embryonic fibroblasts to muscle, and it was this observation what led to the cloning of MyoD (Taylor and Jones, 1982). The same myoblastic conversion also occurs after the transfection of 10T1/2 fibroblasts with an antisense RNA against the maintainance DNA methylatransferase, Dnmt-1 (Szyf et al., 1992). An analysis at the molecular level showed that treatment with azacytidine of 10T1/2 fibroblasts leads to demethylation of the CpG island surrounding MyoD promoter. However, this finding raised the question of whether demethylation of MyoD promoter was indeed the signal required to initiate the myogenic program during embryonic development, as CpG islands in general, and MyoD promoter in particular (Jones et al., 1990), are regions constitutively free of methylation in all the tissues of the organism (Bird, 1986). This problem was solved when Brunk et al. showed that another regulatory region in MyoD, the distal control element, was specifically demethylated during somitogenesis in mice (Brunk et al., 1996). Demethylation of MyoD distal enhancer is previous to gene activation. However, mutation of the CpGs found to be methylated in vivo did not lead to precocious activation of MyoD regulatory regions in transgenic mice (Brunk et al., 1996), suggesting that DNA demethylation is not sufficient for gene activation, but may be playing a role in imprinting the cells that will then activate the myogenic program. It is worthy to note, that the distal control element in MyoD enhancer plays a pivotal role in first activating the expression of human and mouse MyoD during embryonic development (Goldhamer et al., 1995; Kablar et al., 1999). The correlation between muscle differentiation and DNA methylation is further underscored by the finding that myogenin promoter becomes demethylated in culture at the onset of C2C12 muscle cells differentiation (Lucarelli et al., 2001). Moreover, a genome-wide demethylation has been observed in mouse myoblasts induced to differentiate, which reaches a maximum 2 days after the differentiation is induced (Jost et al., 2001). This global demethylation seems to be a two-step process that starts with the generation of a hemymethylated site, which in turn becomes the substrate of 5-methylcytosine DNA glycosilase. After 2 days, however, there is a partial remethylation of the genome, suggesting a broader role for this epigenetic modification during myogenesis (Jost et al., 2001). The role of DNA methylation in inactivating specific loci during differentiation is further supported by the discovery of a muscle specific Dnmt-1 (Aguirre-Arteta et al., 2000). This shorter isoform of the maintainance methyltransferase is not detected in myoblasts, but its expression is upregulated in differentiating myotubes. Further studies will be necessary to unequivocally assign a role to this protein during myogenesis.
In summary, demethylation of muscle regulatory regions at the beginning of the differentiation program seems to be a necessary step in the commitment of cells towards the muscle lineage. Other events, such as Msx1-dependent positioning of H1b histone (Lee et al., 2004) and Rho-mediated signalling (Carnac et al., 1998) must provide additional elements contributing to induce MyoD expression in muscle cells, although the relationship between these events remains unknown.
The role of DNA methylation in restricting the differentiation of multipotent precursors has also been demonstrated in different systems. Incomplete DNA demethylation leads to unsuccessful nuclear cloning in mammals (Humpherys et al., 2001; Bortvin et al., 2003; Dean et al., 2003), indicating that demethylation of the somatic nucleus is necessary for proper reprogramming. This hypothesis has recently proved to be correct in Xenopus with the work of Simonsson and Gurdon. Using nuclear and DNA transfer into Xenopus oocytes, these authors show that DNA demethylation is absolutely necessary for transcription of the stem cell marker Oct4 (Simonsson and Gurdon, 2004). Finally, global DNA hypomethylation obtained by genetic inactivation of the methyltransferases (Dnmt-1, 3a or 3b) blocks embryonic stem (ES) cells differentiation (Panning and Jaenisch, 1996; Jackson et al., 2004), whereas treatment of ES cells with 5-azacytidine reverses the differentiation process (Tsuji-Takayama et al., 2004), suggesting that this modification may be important in epigenetically marking pluripotent precursors. It will be interesting to analyze if similar mechanisms operate in the determination of more restricted multipotent precursors towards the myogenic lineage.
EXTRACELLULAR SIGNALS AND CYTOPLASMIC CASCADES THAT MARK SPECIFIC CELL FATES BY CHROMATIN REPROGRAMMING
The importance of external cues in imparting to muscle progenitors the information necessary to differentiate into myofibers is emphasized by the role of injury-derived soluble substances and cytokines in regulating the myogenic potential of BMDC and satellite cells (Langen et al., 2001, 2002, 2004; Coletti et al., 2002; Horsley et al., 2003; Broussard et al., 2004; Charge and Rudnicki, 2004; Sorci et al., 2004).
Several signalling pathways contribute to initiate myogenesis in response to different extra-cellular cues. The PI3K/Akt pathway, activated by growth factors, such as the insulin-like growth factor 1 (IGF-1), the stress activated MKK6/p38 MAPK pathway (Cuenda and Cohen, 1999; Zetser et al., 1999; Wu et al., 2000), and the calcium/calmodulin-dependent kinase (CaMK) pathway (McKinsey et al., 2000a,b) all promote muscle gene transcription. It is becoming increasingly clear that extracellular signals alter gene expression at least in part by targeting the different chromatin modifying activities. For instance, differentiation-induced CaMK I and IV phosphorylate class II histones deacetylases at two conserved serines, disrupting their association with MEF2 and thereby allowing MEF2 to become transcriptionally competent (McKinsey et al., 2000a,b). On the other hand, activation of the p38 pathway leads to phosphorylation of all four MEF2 members (Han et al., 1997; Ornatsky et al., 1999; Yang et al., 1999; Zhao et al., 1999; Cox et al., 2003), and this modification is necessary for MEF2 transcriptional activity (Wu et al., 2000). Differentiation-activated p38 promotes MyoD-dependent transcription by phosphorylating its heterodimerization partner E47 (Lluis et al., 2005) and by targeting components of the SWI/SNF remodelling complex to the muscle-gene regulatory regions (Simone et al., 2004). This recruitment is necessary to provide the transcriptional competence to the myogenic transcriptosome. Likewise, the role of the IGF-1/PI3K/AKT pathway during skeletal myogenesis is complex, with several targets described so far (Glass, 2003; Sandri et al., 2004; Stitt et al., 2004), including the acetyltransferase p300 (our unpublished results). Interestingly, in proliferating myoblasts IGF1-activated PI3K/AKT signalling contributes to stimulate proliferation, whereas upon serum withdraw the IGF-1/PI3K pathway is converted into a pro-myogenic pathway, possibly by differentiation activated p38 kinases (our unpublished results).
Further studies in the field will help to understand how the extra-cellular signals and the subsequently activated cytoplasmic cascades mark specific cell fates by targeting the chromatin of lineage-associated regulatory regions.
REVERSIBILITY OF THE DIFFERENTIATION PROGRAM
Although it was originally assumed that both lineage determination and terminal differentiation were irreversible processes in mammals, studies with cultured muscle cells challenged this idea. On one hand, ectopic expression of viral oncoproteins like the adenoviral E1A antigen or the SV40 large T antigen into myotubes blocks muscle gene expression and induces cell cycle reactivation, without completing mitosis (Ohkubo et al., 1994; Tiainen et al., 1996; Latella et al., 2000). A similar effect is observed in muscle cells derived from mice lacking the retinoblastoma protein, which re-enter the cell cycle when stimulated with mitogens (Schneider et al., 1994). The first direct evidence showing dedifferentiation of C2C12 myotubes in response to extracellular stimuli comes from the work of Rosania et al. (2000). These authors showed that when exposing C2C12 myotubes to myoseverine (a microtubule-binding molecule) they undergo cellular fission to give place to mononucleated cells that were in turn able to re-enter the cell cycle when exposed to mitogens. However, myoseverine did not affect the expression of the MRFs, suggesting that only terminal differentiation, but not commitment to the myogenic lineage, was affected. In a different study, Odelberg et al. showed that overexpression of the homeobox protein Msx1 in myotubes together with exposure of the cells to mitogens, lead to downregulation of the cell cycle inhibitor p21, myogenin, MRF4, and MyoD (Myf5 was not analyzed in this study) (Odelberg et al., 2000). Concomitant with this change in gene expression, a percentage of the C2C12 myotubes underwent cytokinesis to form mononucleated cells. More interestingly, clonal populations of the newly formed mononucleated cells were able to differentiate along the chrondogenic, adipogenic, osteogenic, and myogenic lineages when incubated with the appropriate culture conditions, indicating that both terminal differentiation and commitment to the myogenic lineage had been reversed. Although a plausible explanation for the inhibition of the myogenic program by Msx-1 in uncommitted precursors has recently arised from the work of Lee and colleagues (Lee et al., 2004), the dramatic nuclear reprogramming observed after Msx-1 overexpression in myotubes is more difficult to be interpreted. In proliferating precursors, Msx1 targeting of the linker histone H1b to MyoD regulatory regions leads to its transcriptional downregulation and inhibition of muscle differentiation. This model, however, is difficult to apply to the dedifferentiation observed after Msx1 overexpression in myotubes. First, the phenotypic reversion observed by Odelberg et al., seems to be sequential, with the late MRFs and the cell cycle inhibitor p21 being downregulated prior to MyoD in myotubes expressing Msx1, suggesting that Msx1 targets other muscle effectors apart from MyoD. Secondly, the observation that H1b is downregulated during normal muscle differentiation and in myotubes makes unlikely an Msx1 directed generation of repressive chromatin at muscle genes through this linker histone. While addressing the molecular mechanisms leading to this dramatic reprogramming will provide new insights into the factors conferring plasticity to differentiated nuclei, it is also important to verify these results in more relevant contexts, such as primary cells and in vivo studies.
Pier Lorenzo Puri is an Assistant Telethon Scientist, also supported by grants from Muscular Dystrophy Association and Parent Project Organization. We are grateful to Dr. Giulio Cossu and Dr. Valerio Orlando for critically reading the manuscripts and all the members of Puri's lab for stimulating discussions during the preparation of this review. We apologize to colleagues whose work could not be cited due to space limitations.