The existence of eukaryotic cells is governed by the coordinated expression of the genome in response to environmental cues. These signals, starting from transmembrane receptors, are propagated toward the nucleus by enzymatic cascades that finally affect gene transcription.
Genomic DNA is packaged into nucleosomes, the basic unit of chromatin structure formed by the wrapping of 147 bp of DNA around an octamer of four different histones (H2A, H2B, H3, and H4). Nucleosomes are then separated by a short linker DNA to which histone H1 binds. The packaging of DNA into chromatin has a repressing effect on gene expression and, therefore, reconfiguration of chromatin is mandatory for transcriptional initiation. Increasing evidence indicates that gene expression is regulated by means of facilitating access of DNA-binding factors to DNA via chromatin-remodeling and covalent histone modifications, which are considered the biochemical hallmarks of chromatin configuration (Emerson, 2002; Narlikar et al., 2002). The repressive state is marked by hypoacetylation and methylation of particular residues catalyzed by histone deacetylases and methyltransferases. By contrast, the permissive configuration results from the concerted activity of histone acetyltransferases, kinases, and additional methyltransferases, which promote local phospho-hyperacetylation together with activating methylation. These site-specific modifications, defined as the “histone code,” might influence chromatin function through two mechanisms: first, by affecting nucleosome-nucleosome or histone-DNA interactions that control the folding of nucleosomal arrays; second, by promoting or disrupting the chromatin binding of non-histone proteins (Strahl and Allis, 2000; Jenuwein and Allis, 2001; Agalioti et al., 2002; Kouzarides, 2002; Jaskelioff and Peterson, 2003).
The histone-modifying enzymes exert their activity in concert with another class of nuclear factors, termed chromatin-remodeling complexes, which use the free energy derived from ATP hydrolysis to actively alter nucleosomal structure. These factors peel up to 50 bp of DNA from the edge of the nucleosome, slide the histone octamer beyond the DNA ends, and bind the spooled out DNA end to the exposed surface of the octamer in an intramolecular manner to form an accessible entry/exit-site DNA loop. This stable altered nucleosome conformation is achieved in less then one second, hydrolyzing fewer than ten ATPs per event (Kassabov et al., 2003). Different chromatin-remodeling complexes were characterized, which are defined by a unique subunit composition and the presence of a distinct ATPase. Mammalian SWI/SNF is a multiprotein chromatin-remodeling complex composed by at least 10 elements. Two distinct SWI/SNF complexes were described, each characterized by the presence of a unique subunit: either BAF250 or BAF180, defining BAF and PBAF, respectively. In addiction, BAF can contain either BRG1 or BRM as the core motor subunit, whereas PBAF only contains BRG1. All subunits of SWI/SNF are well conserved from yeast to humans, and structural analysis of their protein domains suggests specific functional properties (Fig. 1). The central core subunits BRG1 and BRM contain an ATPase domain and a bromodomain, a recognition motif for acetylated lysines in histone tails or in other proteins. Further structural domains involved in protein–protein interaction are present in BAF250, BAF180, BAF170, BAF155, BAF60 (a, b, c1/2), BAF53 (a, b), BAF47, and G-actin, whereas BAF250, BAF180, BAF170, BAF155, and BAF57 contain different sequence-dependent and sequence-independent DNA binding domains (for review see Mohrmann and Verrijzer, 2005).
This review will focus on the role of mammalian SWI/SNF as the crossroads surface integrating the execution of diverse or even opposite biological programs involving the coordinated expression of the genome.
SWI/SNF ROLE IN TISSUE-SPECIFIC TRANSCRIPTION
Tissue-specific transcription is conferred by the promoter binding of tissue-restricted transcription factors and chromatin-modifying enzymes in response to ligand-dependent activation of signaling cascades. These pathways impart covalent modifications on nuclear proteins to recruit transcriptional complexes on the chromatin of subsets of differentiation-specific loci to reprogram gene expression.
Differentiation-activated MKK6/p38 cascade targets SWI/SNF to muscle-specific loci
At the onset of muscle differentiation, tissue-specific transcription is sustained by IGF1/PI3K/Akt and MKK6/p38 signaling pathways, and remains active during myotube formation and post-mitotic growth with the additional support of MEK1/ERK (Wu et al., 2000). These pathways stimulate the expression of muscle-specific genes through the combinatorial activity of the muscle regulatory factors MyoD and MEF2 family members, which bind DNA consensus sequences within the regulatory elements recruiting chromatin-modifying enzymes. The major chromatin modifications consist in the phospho-acetylation of histone tails, and the remodeling of chromatin (for review see Puri and Sartorelli, 2000; Sartorelli and Puri, 2001).
Upon MKK6-dependent enzymatic activation, p38 localizes on the chromatin of muscle-gene regulatory elements promoting the recruitment of SWI/SNF possibly via BAF60 phosphorylation (Simone et al., 2004). Notably, inhibition of p38 enables the assembly of MyoD and MEF2 with p300 and PCAF acetyltransferases, leading to local hyperacetylation, but specifically prevents the engagement of BRG1 and Brm, and chromatin remodeling at these elements. Activation of p38 by constitutively active MKK6, but not by inflammatory cytokines, promotes unscheduled BRG1 recruitment on muscle promoters in myoblasts; BRG1 inactivation, by antibody microinjection or genetic deficiency, abrogates MKK6-dependent induction of muscle gene expression. Moreover, the MKK6/p38 pathway promotes SWI/SNF binding to both MyoD and MEF2 multiprotein complexes, suggesting that the phosphorylation of the BAF60 subunit, a surface for protein–protein interaction, could represent the biological event leading to the transcription factor-dependent recruitment of SWI/SNF on the chromatin of muscle-specific loci. The essential role of BAF60c in skeletal muscle and heart development has been underscored by genetic inactivation studies (Lickert et al., 2004), indicating this subunit as the muscle-specific interface in mammalian SWI/SNF.
SWI/SNF tissue-specific surfaces for protein–protein interaction: Other targets for signaling cascades?
The model we described in muscle differentiation indicates that, in addition to targeting transcription factors, kinases can alter gene expression by targeting chromatin-modifying enzymes (see Gillespie and Rudnicki, 2004). The identification of this mechanism suggests that all of the other SWI/SNF subunits, especially those involved in protein–protein interaction with tissue-specific factors, might represent target substrates for signaling-dependent kinase cascades.
During development, (β-globin genes are differentially expressed by the combinatorial activity of EKLF and GATA-1 zinc finger transcription factors, both directly interacting with BRG1, BAF170, and BAF155 (Kadam et al., 2000). Moreover, SWI/SNF components (BAF170, BAF60a, BAF57, and BAF47) participate in the PYR complex's binding specifically to an intergenic pyrimidine-rich DNA sequence to facilitate fetal-to-adult β-globin gene switching (O'Neill et al., 1999). Analogously, SWI/SNF contributes to CD4/CD8 lineage bifurcation of T-cells via Brg1 enzymatic activity and BAF57 DNA binding ability (Chi et al., 2002).
During neuronal differentiation the basic-helix-loop-helix (bHLH) transcription factors NeuroD and Ngnr recruit SWI/SNF via the physical association with BRG1 (Seo et al., 2005). In adipocytes, the nuclear hormone receptor PPAR( engages the remodeling activity of SWI/SNF via the direct protein–protein interaction with the BAF60c subunit (Lemon et al., 2001; Debril et al., 2004). In both human and mouse cells, two BAF60c isoforms are formed, termed c1 and c2, with a differential pattern of expression: BAF60c1 is more abundant in neural cells, while BAF60c2 in heart, skeletal muscle, and adipocytes. Furthermore, other two BAF60 forms are encoded by additional genes, BAF60a is preferentially expressed in the adrenal gland, kidney, and liver, whereas BAF60b in the pancreas, placenta, and spleen (Gene Expression Atlas, http://expression.gnf.org). Interestingly, BAF60a interacts with and coactivates the c-jun/c-fos heterodimer in regulating cellular growth, differentiation, and development via the AP-1 DNA binding sites (Ito et al., 2001). The tissue-restricted expression of different BAF60 forms sustains the model that tissue-specific subunits are responsible for the transcription factor-mediated recruitment of SWI/SNF to remodel chromatin of tissue-specific subsets of genes.
SWI/SNF HORMONE-RESPONSIVE SUBUNITS
Steroid hormones activate receptors by inducing conformational changes that lead to nuclear translocation, dimerization, and binding to steroid response elements of target genes. Activated receptor bound to target DNA recruits coactivators that are essential to modulate chromatin structure and activate target gene expression. In addition, subpopulations of steroid receptors can associate in a hormone-dependent manner with cytoplasmic or cell membrane molecules, activating discrete signaling pathways, such as PI3K, p38, ERK, PKC, PKA, and others. This additional mechanism might influence gene transcription by phosphorylating different targets: the hormone-receptor itself or an interacting coactivator, otherwise a transcription factor that cooperates with the receptor or that mediates activation of genes lacking steroid response elements (for review see Edwards, 2005). SWI/SNF is an essential coactivator of nuclear hormone receptors and its subunits might represent target surfaces for these signaling cascades. In fact, the glucocorticoid receptor (GR) recruits remodeling activity by interacting with a portion of the SWI/SNF complex composed by BAF250 (Nie et al., 2000), BAF60a, and BAF57 (Hsiao et al., 2003), while the vitamin D receptor (VDR) heterodimers complex binds only BAF60a (Koszewski et al., 2003); estrogen (ER) and androgen receptors (AR) interact with BAF57 (Belandia et al., 2002; Link et al., 2005).
SWI/SNF CONTROL OF CELLULAR PROLIFERATION
Bi-directional communication between cyclin E and SWI/SNF regulates cell-cycle progression
The pivotal mechanism regulating cell proliferation is considered to be the control of the progression through the G1/S phase boundary of the cell-cycle, which is ensured by the precisely timed accumulation and degradation of cyclin E. At the mid G1 phase, cyclin E expression is induced by the cyclin D/cdk4 complex via pRb phosphorylation. This event releases the transcriptional repression of cyclin E promoter exerted by a multiprotein complex containing E2F, pRb, histone deacetylase, and SWI/SNF (Zhang et al., 2000). At this point, cyclin E expression is maintained by an autonomous mechanism enhancing E2F-mediated transcription via cyclin E/cdk2-mediated pRb inactivation (Moroy and Geisen, 2004). Recently, generation of cyclin E-deficient mice revealed that cyclin E functions are dispensable for proliferation of normal cells, while are necessary for endoreplication, cell cycle re-entry from the quiescent state, and oncogenic transformation (Geng et al., 2003). Furthermore, the notion that pRb-deficient cells still require cyclin E to proliferate (Lukas et al., 1997) opens the question about pRb-independent cyclin E functions.
In human cells, cyclin E associates with SWI/SNF via a pRb-independent interaction with phosphorylated BAF155 and BRG1 (Shanahan et al., 1999). Moreover, cyclin E/cdk2 complex phosphorylates in vitro both SWI/SNF subunits, and coprecipitates with BAF47 and Brm as well. SWI/SNF-dependent remodeling of chromatin structure regulates DNA methylation, replication, recombination and repair as well as gene expression. As these functions would predict, inactivating modification of the proteins that remodel chromatin causes cell transformation and tumor progression. In fact, mutations of BAF47 and BRG1, and reduced expression of BRM, have been identified in primary tumors and tumor-derived cell lines (Klochendler-Yeivin et al., 2002). To date, BRG1 and BRM were linked to adrenal, cervical, lung, breast, and prostate cancer, while BAF47 with pediatric tumors and cancer-prone families (Cho et al., 2004). According to tumor suppressor function, reintroduction of BRG1 into deficient tumor cell lines induces G1-arrest by downregulation of genes promoting cell growth, as for cyclin E, and upregulation of genes inhibiting proliferation, as for the cdk inhibitors (CKIs) p21 and p15 (Hendricks et al., 2004). Intriguingly, concomitant expression of cyclin E is able to abrogate this effect (Shanahan et al., 1999), establishing a functional bi-directional communication between SWI/SNF and the cyclin.
Moreover, to efficiently promote transcription, SWI/SNF completes and stabilizes pre-initiation complex formation and promotes transcription elongation by the hyperphosphorylated form of RNA polymerase II (Salma et al., 2004). Recently, two independent studies identified cyclin E/cdk2 as a kinase complex able to interact and phosphorylate RNA polymerase II in vivo (Deng et al., 2002; Nekhai et al., 2002), suggesting another functional link between cyclin E and SWI/SNF, and defining a new role for cyclin E/cdk2 at the chromatin level in the transcription process.
Collectively, this evidence suggests a scenario where cyclin E/cdk2 localize at the chromatin as a part of a multiprotein complex involving SWI/SNF and RNA polymerase II. At the G1/S phase boundary, the covalent modifications imparted to BAF155, BRG1, and the polymerase by cyclin E/cdk2 could represent the cellular signal switching gene expression toward S phase-specific loci. This mechanism could lead to the displacement of SWI/SNF from the regulatory elements where BRG1 prevents gene expression, such as those of cyclin A and cdk1 (Zhang et al., 2000), and the redistribution to chromatin loci requiring remodeling and activation.
Cooperation of SWI/SNF and Lkb1 pathway in the control of G1 progression
LKB1 is a serine-threonine kinase mutated in an autosomal dominant inherited cancer disorder termed Peutz-Jeghers Syndrome (PJS) and in different sporadic tumors (Resta et al., 1998; Boudeau et al., 2003). Homozygous LKB1-knockout mice die at early embryonic stages, indicating that LKB1 plays an important role during development (Ylikorkala et al., 2001). In human cells, LKB1 forms a hetero-trimeric complex with STRAD and MO25, both required for full kinase activity, and contributes to different signaling pathways involved in metabolism, cellular polarity, and proliferation. In these cascades, some downstream targets are described: LKB1 phosphorylates and actives the AMP-activated protein kinase (AMPK), a regulator of cellular energy charge, and 11 other AMPK-related kinases in an energy stress-independent manner (for review see Hardie, 2004). Furthermore, LKB1 associates with and modulates the functions of BRG1, by kinase-dependent and-independent mechanisms (Marignani et al., 2001). In vitro, both the wild type and a kinase-deficient mutant are able to stimulate BRG1 ATPase activity, but the full kinase activity of LKB1 is required for BRG1-dependent growth arrest of cancer cells. The concurrent functional role in the control of G1 progression is underscored by the fact that LKB1, as well as BRG1, regulates p21 expression, and that reintroduction of the wild-type kinase in LKB1-deficient tumor cells causes G1 arrest (Tiainen et al., 1999, 2002).
Phosphorylation-dependent inactivation of SWI/SNF in mitosis
During mitosis, chromatin is condensed into mitotic chromosomes ensuring the equal distribution of packaged DNA between daughter cells, and contributing to the general inhibition of transcription. Since the nucleosome remodeling activity of SWI/SNF plays a direct role in activating gene expression by increasing the chromatin accessibility to the transcriptional machinery, thus the inactivation of this complex might be one mechanism of inhibiting gene expression during mitosis. Moreover, it might be important to allow the tight compaction of DNA to reach the condensation of chromatin structure.
Upon cells' entry into mitosis, BRG1, Brm, and BAF155 undergo phosphorylation, and although the level of Brg1 remains constant, Brm appears to be degraded. These molecular events result in the reversible inactivation of SWI/SNF activity and the displacement of the remodeling complexes from the condensed chromatin (Muchardt et al., 1996; Reyes et al., 1997; Sif et al., 1998). To date, it is not clear which kinases are responsible for the phosphorylation of SWI/SNF subunits in vivo, and which phosphatases are required for reactivating dephosphorylation at the beginning of the G1 phase of the cell cycle. In fact, although cdk1 is the essential kinase for G2-M transition and phosphorylates a wide variety of proteins, neither cyclin A/cdk1 nor cyclin B/cdk1 complexes are able to phosphorylate SWI/SNF subunits in vitro. Intriguingly, MEK1-activated ERK1 inhibits SWI/SNF activity by direct phosphorylation of BRG1, Brm, and BAF155, while PP2A phosphatase reactivates the chromatin remodeling function of the complex (Sif et al., 1998). This model supports the concept that the assembly of BRG1/Brm with BAF155 might form the specific substrate for cell cycle-regulated kinase cascade signaling to SWI/SNF to dictate the configuration of the chromatin.
Proliferation-specific surfaces for protein–protein interaction
In addition to BAF155 and BRG1, also BAF53 and BAF47 subunits are essential for the control of cell cycle progression and, as those involved in tissue-specific transcription, might represent additional target surfaces for signaling cascades. These subunits directly interact with the bHLH/leucine zipper transcription factor c-Myc (Cheng et al., 1999; Park et al., 2002), ensuring the recruitment of SWI/SNF on target genes such as E2F, cdk4, and cyclin E. c-Myc is involved in at least three distinct pathways controlling the progression through the G1 phase of the cell cycle: the suppression of p27 function, the induction of the transcriptional activity of E2F family members, and the promotion of cell growth and increase in cell mass (Beier et al., 2000). Furthermore, BAF47 interacts with p53, mediating the recruitment of SWI/SNF on the chromatin of genes implicated in growth suppression and apoptosis in response to several cell stress-dependent pathways (Lee et al., 2002).
As described above, BRG1 cooperates with the retinoblastoma protein to control the G1/S phase transition via a direct interaction with the A/B pocket domain (Dunaief et al., 1994; Strobeck et al., 2000). Also Brm behaves as BRG1, and both ATPase are able to bind the other two pRB family members p107 and p130 (Strober et al., 1996; Trouche et al., 1997). Intriguingly, the ability of Brm to induce G1 arrest is negatively regulated by acetylation of specific lysines on the C-terminus possibly mediated by PCAF and p300 acetyltransferases (Bourachot et al., 2003). The acetylation of this domain seems to destabilize Brm interaction with pRb family members, indicating that covalent modifications other than phosphorylation might affect SWI/SNF activity in response to signal-dependent pathways.
CONCLUSIONS AND PERSPECTIVES
The cellular response to an environmental cue requires switching transcription from certain subsets of genes to others for the completion of the new cell request. This action necessitates that different nuclear complexes modify the chromatin to reach the permissive or the repressive configuration. SWI/SNF is a component of these multiprotein complexes able to remodel chromatin toward cellular necessity. In addition to the transcriptional activator functions, we have herein described that SWI/SNF can repress transcription in concert with pRb and HDAC, and possibly with the SUV39H1/HP1 complex (Nielsen et al., 2001), as for cyclin E, cyclin A, and cdk1 genes during cell cycle regulation. Moreover, BAF57 interacts with MeCP2, the methyl-CpG binding protein mutated in Rett Syndrome (Harikrishnan et al., 2005), and the transcriptional repressor Ets-2 (Baker et al., 2003), evidencing that SWI/SNF is also involved in the maintenance of the inactive state of the chromatin.
Recent studies of higher-order chromatin arrangements and their dynamic interactions with other nuclear components suggest a topological model for gene regulation (for review see Cremer and Cremer, 2001). The mammalian cell nucleus is organized into chromosome territories (CTs) and an interchromatin compartment (IC) that contains macromolecular complexes required for replication, transcription, splicing and repair. Long-term changes in gene expression require a higher-order chromatin remodeling that reflects the repositioning of genes in open or closed high-order chromatin compartments. Long-term active genes that need to be turned on and off rapidly are exposed at the chromatin-domain surfaces that line the IC, while long-term silenced genes are located in the interior of the chromatin domain far from the IC. During differentiation, cells establish a tissue-specific pattern of gene locations with respect to certain nuclear compartments, such as heterochromatin, the IC, or the nuclear lamina.
This model is compatible with the hypothesis that active polymerases are immobilized and clustered in discrete nuclear foci (for review see Cook, 2002). Several lines of evidence suggest that polymerases bind to genomic transcription units, and then active transcription complex aggregate, forming the DNA into surrounding loops. Once the stable attachments between an engaged polymerase, its transcript, and template are lost on transcriptional termination, a new, larger loop is generated and the polymerase becomes free to exchange with others in solution. The release transcription unit is still near a polymerase cluster, favoring reengagement. Active transcription units are likely to be associated with a cluster; units located adjacent to each other in the primary DNA sequence tend to attach to the same cluster; and groups of active units separated by long stretches of inactive DNA in between will be out in a loop.
The ATP-dependent activity that “slides” nucleosomes along the “naked” DNA filament seems to indicate that SWI/SNF is the motor core unit of macromolecular complexes, located in the IC, that translocate genes within CTs to reach active chromatin domains near clusters of polymerases. This view of the nuclear organization is congruent with the finding that components of the human SWI/SNF complex are enriched in active chromatin and are associated with the nuclear matrix (Reyes et al., 1997). As such, it will be interesting to define the role of G-actin in the SWI/SNF complex, and to identify the molecules forming the “nuclear-skeleton” that sustains the nuclear architecture.
To conclude, we observed that during cellular differentiation, the mRNA and protein levels of SWI/SNF subunits remain constant from myoblasts to myotubes (Simone et al., 2004), indicating that cellular SWI/SNF availability is limited. This evidence suggests that the pharmacological manipulation of those pathways signaling to SWI/SNF could represent the ideal approach to switch gene expression by targeting the chromatin remodeling complexes to loci of specific therapeutical interest, potentially for tissue-specific regeneration, inhibition of cell transformation and cancer progression or prevention and treatment of genetic diseases.
Thanks to Lorenzo Puri and the members of his laboratory for the past 4 years of intense, fulfilling and fruitful work done together. Thanks to Dr. Giordano, Dr. Puri, and Dr. Sartorelli for helpful discussion and Marie Basso for editorial assistance.