|Home | About | Journals | Submit | Contact Us | Français|
Mediator occupies a central role in RNA polymerase II transcription as a sensor, integrator, and processor of regulatory signals that converge on protein-coding gene promoters. Compared to its role in gene activation, little is known regarding the molecular mechanisms and biological implications of Mediator as a transducer of repressive signals. Here, we describe a protein interaction network required for extra-neuronal gene silencing comprising Mediator, G9a histone methyltransferase, and the RE1 silencing transcription factor (REST; also known as neuron restrictive silencing factor, NRSF). We show that the MED12 interface in Mediator links REST with G9a-dependent histone H3K9 di-methylation to suppress neuronal genes in non-neuronal cells. Notably, missense mutations in MED12 causing the X-linked mental retardation (XLMR) disorders FG syndrome and Lujan syndrome disrupt its REST corepressor function. These findings implicate Mediator in epigenetic restriction of neuronal gene expression to the nervous system and suggest a pathologic basis for MED12-associated XLMR involving impaired REST-dependent neuronal gene regulation.
The specification and maintenance of cell fate in multicellular organisms is critically dependent upon the precise spatiotemporal control of RNA polymerase II transcription in response to a determinative set of cell-intrinsic and –extrinsic signals. Accordingly, genetic or environmental factors that perturb physiologic transcription controls can alter cell fate decisions leading to a variety of pathologic conditions including developmental defects and cancer. Because of its central importance in organismal biology, metazoans have therefore evolved an elaborate protein machinery to ensure proper transcription control. Work over the last decade has identified Mediator as a critical component of this regulatory apparatus. Mediator is a conserved multiprotein interface between gene-specific transcription factors and RNA polymerase II (Kornberg, 2005). In this capacity, Mediator serves to channel regulatory signals from activator and repressor proteins to affect changes in gene expression programs that control diverse physiological processes, including cell growth and homeostasis, development, and differentiation (Conaway et al., 2005; Malik and Roeder, 2005).
The role of Mediator in transcriptional activation has been characterized extensively in recent years. Mediator has been shown to promote the assembly, activation, and regeneration of transcription complexes on core promoters during the initiation and re-initiation phases of activator-dependent transcription (Kornberg, 2005; Malik and Roeder, 2005). Comparably little is known regarding the role of Mediator in transcriptional repression, although its inhibitory properties have nonetheless been linked to a dissociable module comprised of MED12/MED13/CDK8/CycC [hereafter referred to as the MED12 module, since we and others have shown that MED12 is required for stable incorporation of CDK8/CyC into Mediator (Kim et al., 2006; Myer and Young, 1998)]. Several lines of evidence support a role for MED12 module components in transcriptional repression. First, orthologous modules in both Saccharomyces cerevisiae and Schizosaccharomyces pombe Mediator have been implicated in repression of yeast transcription (Bjorklund and Gustafsson, 2005). Second, mammalian Mediator preparations enriched for the MED12 module inhibit reconstituted transcription reactions in vitro (Boyer et al., 1999; Gu et al., 1999; Sun et al., 1998; Taatjes et al., 2002). Finally, transcriptionally repressive or signal-activated forms of the mammalian transcription factors C/EBPβ and retinoic acid receptor have been shown to recruit, respectively, Mediator with or without CDK8 to target gene promoters (Mo et al., 2004; Pavri et al., 2005). Collectively, these studies suggest that the repressive function of Mediator derives principally from components within its MED12 module.
Within the MED12 module, CDK8/CycC have been ascribed direct repressive activity through inhibitory phosphorylation of activators and general transcription factors (Akoulitchev et al., 2000; Fryer et al., 2004). The functions of MED12 and MED13 appear to be more complex, and previous studies support overlapping negative (and positive) roles for these two subunits in signal-dependent developmental gene regulation. Thus, in Caenorhabditis elegans, MED12 and MED13 suppress Wnt- and Ras-target genes involved in vulval cell fate specification and asymmetric cell division, while in Drosophila melanogaster, MED12 and MED13 suppress Wnt-and Hedgehog-target genes, respectively, critical for differentiation in the eye and antennal discs (Janody et al., 2003; Moghal and Sternberg, 2003; Treisman, 2001; Yoda et al., 2005; Zhang and Emmons, 2000). More recently, human MED12 has been shown to function as a direct suppressor of Gli3-dependent Sonic hedgehog signalling (Zhou et al., 2006). Thus, considerable evidence suggests that MED12 can function as a signal- and context-dependent negative regulator of transcription. Although the mechanistic basis by which MED12 functions to repress transcription remains unknown, the recent genetic identification of MED12 in D. melanogaster as a novel Polycomb group (PcG) gene required for HOX gene repression suggests a potentially intriguing link to chromatin (Gaytan de Ayala Alonso et al., 2007). PcG proteins mediate epigenetic inheritance of silent chromatin states throughout development in a manner dependent upon posttranslational histone modifications, including methylation (Schuettengruber et al., 2007). Nonetheless, beyond the aforementioned genetic association, it has not heretofore been established whether and how MED12/Mediator physically and functionally interacts with histone modifying activities to repress transcription, nor how such collaborations might impact specific gene expression programs that dictate fundamental cellular processes.
Recent genetic studies in vertebrates have begun to illuminate the developmental and biological processes regulated by MED12/Mediator in higher metazoans. These studies have revealed an important role for MED12 in the specification of cell fate and function in the nervous system. In the zebrafish Danio rerio, MED12 has been shown to be required for proper development of the brain and neural crest, among other organs, where it plays an important role in the production of monoaminergic neurons and cranial sensory ganglia through selective regulation of neuronal gene expression (Hong et al., 2005a; Hong et al., 2005b; Rau et al., 2006; Wang et al., 2006). In humans, MED12 polymorphisms have been linked with neuropsychiatric illness, including schizpophrenia and pyschosis, and MED12 mutations have recently been shown to be causative for two X-linked mental retardation (XLMR) disorders, FG (also known as Opitz-Kaveggia) syndrome and Lujan (or Lujan-Fryns) syndrome (Philibert and Madan, 2007; Risheg et al., 2007; Schwartz et al., 2007). Nonetheless, the molecular bases by which MED12 controls neuronal gene expression and the means by which genetic variation in MED12 elicits behavioural and cognitive dysfunction remain to be fully established.
Herein, we identify an epigenetic basis to link the function of MED12/Mediator in transcriptional repression with its roles in neuronal gene control and XLMR. First, we identify and characterize a network of functional interactions involving MED12/Mediator, G9a histone methyltransferase (HMTase), and the RE1 silencing transcription factor (REST; also known as neuron restrictive silencing factor, NRSF) required for extra-neuronal gene silencing. We show that the MED12 interface in Mediator links REST with enzymatically active G9a to silence REST-target genes in non-neuronal cells through the imposition of transcriptionally repressive histone H3K9 di-methylation. Second, we show that missense mutations in MED12 causing the XLMR disorders FG syndrome and Lujan syndrome disrupt its REST-specific corepressor function. Together, these findings establish a direct link between MED12/Mediator and epigenetic gene silencing, reveal an essential biological function for MED12/Mediator in lineage-specific restriction of neunronal gene expression, and suggest a pathologic basis for MED12-associated XLMR involving impaired REST-dependent neuronal gene regulation.
To clarify the function of MED12 in transcription control, we processed the MED12 carboxyl terminus (MED12C; aa 1715 to 2177) through a yeast two-hybrid screen using a human fetal brain cDNA library (Zhou et al., 2006), and recovered a partial cDNA encoding G9a HMTase. Together with GLP (G9a-like protein; also Eu-HMTase1), G9a forms a stoichiometric heteromeric complex responsible for transcriptionally repressive H3K9 mono- and dimethylation within euchromatin (Peters et al., 2003; Rice et al., 2003; Tachibana et al., 2005).
To validate the physical interaction between MED12 and G9a and to map their reciprocal binding domains, we tested the ability of GST-MED12C or GST-full-length G9a (aa 1–1001) to bind, respectively, to a panel of radiolabeled G9a or MED12 truncation derivatives produced by transcription and translation in vitro. This analysis identified the principal reciprocal interaction surfaces on each protein to encompass a PQL domain on MED12 (aa 1616–2050) and an Ankyrin-repeat domain on G9 (aa 486–689), although a Cys-rich domain on G9a (aa 250–332) also bound weakly to MED12 (Figure 1A; Figure S1 and Figure S2). We confirmed a direct interaction between MED12 and G9a in vitro using purified baculovirus-expressed proteins and co-immunprecipitation analyses (Figure S3). To determine whether G9a and MED12 interact in vivo, we examined the ability of the two proteins to reciprocally coimmunoprecipitate with one another following their ectopic expression in HeLa human cervical carcinoma cells. This analysis revealed specific and efficient precipitation of HA-G9a by FLAG-specific antibodies only in the presence, but not in the absence, of FLAG-MED12. Reciprocally, HA-specific antibodies precipitated FLAG-MED12 only in the presence, but not in the absence, of HA-G9a (Figure 1B). These results confirm that MED12 and G9a associate in mammalian cells.
To determine if G9a interacts with Mediator, we immunoprecipitated Mediator from HeLa nuclear extracts using antibodies specific for either of three different Mediator subunits (MED4, MED30, or CDK8) and examined the immunoprecipitates by western blot assay for the presence of G9a. We readily detected G9a in Mediator immunoprecipitates washed with 300 mM salt and 0.2% NP-40, thus revealing a specific and relatively stringent association of G9a with Mediator in vivo (Figure 2A; Figure S4 and Figure S5). Further analyses revealed specific, albeit reduced, association of G9a with Mediator immunoprecipitates subjected to a variety of stringent wash conditions, including 500 mM KCl, 1% NP-40, and 0.5% DOC (Figure S5). Notably, the association between G9a and Mediator is likely to be conserved in a broad range of human cell types since this interaction was confirmed in two additional human cell lines, including HEK293 human embryonic kidney and BG-1 human ovarian adenocarcinoma cell lines (Figure S6).
To further validate the in vivo association between G9a and Mediator, we monitored the subnuclear localization of DsRed-G9a and FLAG-tagged Mediator subunits MED12, CDK8, MED30, and MED8 following their ectopic expression in COS-7 monkey kidney fibroblast cells. G9a and each of the Mediator subunits were found in the euchromatic regions of the nucleus (excluded from the nucleolus and heterochromatic regions) where they colocalized extensively in multiple foci (Figure 2B). Together, these results provide strong support for an in vivo association between G9a and Mediator.
G9a belongs to the SUV39 family of SET domain-containing HMTases, members of which specifically methylate H3K9 (Patnaik et al., 2004; Tachibana et al., 2005). Like G9a, Mediator immunoprecipitated from HeLa nuclear extract specifically methylated H3 among core histones in vitro (Figure 2C), and exhibited a preference for unmethylated and mono-methylated H3K9 as substrates (Figure 2D and Figure S7). We confirmed G9a as the Mediator-associated HMTase based on the finding that RNAi-mediated G9a depletion in HeLa cells prior to immunoprecipitation analysis concordantly reduced the amount of G9a and the level of H3 HMTase activity in Mediator immunoprecipitates (Figure 2E). Notably, RNAi-mediated MED12 knockdown also reduced the amount of G9a and the level of H3 HMTase activity associated with Mediator, thus revealing MED12 to be a likely G9a interface in Mediator (Figure 2E). Because MED12 is an essential determinant of CDK8/Cyc association with Mediator (Figure 2E) (Kim et al., 2006), we also examined the influence of CDK8 knockdown on G9a association with Mediator. RNAi-mediated CDK8 depletion destabilized CycC, but not MED12, in Mediator and had no influence on the level of Mediator-associated G9a (Figure 2F). Thus, G9a dissociation from Mediator accompanying MED12 knockdown does not derive from indirect loss of CDK8/CyC. Furthermore, RNAi-mediated knockdown of MED23, a different subunit in Mediator, had no influence on the level of Mediator-associated G9a (Figure 2F). Taken together, these results identify MED12 to be a likely interface in Mediator for physical and functional interaction with G9a.
To explore the biological consequence of the Mediator/G9a interaction, we first examined its requirement for global levels of di-methylated H3K9 (H3K9me2), an epigenetic mark dominantly regulated by G9a in mammalian cells (Tachibana et al., 2005). Disruption of the Mediator/G9a interaction by MED12 knockdown did not alter the global level of H3K9me2 (Figure S8), suggesting that the Mediator/G9a interaction might be functionally important at restricted genetic loci. In this regard, G9a-directed H3K9me2 has previously been implicated in gene-specific silencing via its targeted recruitment by transcriptional repressors including REST, a key non-neuronal lineage restrictor that suppresses the non-specific expression of neuronal genes in terminally differentiated non-neuronal cells (Chong et al., 1995; Gyory et al., 2004; Roopra et al., 2004; Schoenherr and Anderson, 1995). We therefore examined the possibility that Mediator and G9a collaborate to support transcriptional repression directed by REST.
First, we investigated whether REST physically associates with both Mediator and G9a in mammalian cells. Having shown that Mediator and G9a form an immunoprecipitable complex in vivo (Figure 2A and Figure S4–Figure S6), we tested whether REST could also be coimmunoprecipitated with Mediator. Myc-tagged REST ectopically expressed in HEK293 cells was observed to be specifically and efficiently precipitated by MED30-specific antibodies, confirming an interaction between REST and Mediator in vivo (Figure 3A). More importantly, analysis of Mediator immunoprecipitates bearing endogenous G9a also revealed the presence of endogenous REST, suggesting a three-way interaction between Mediator, G9a, and REST in three different cell lines of non-neuronal origin - HeLa, HEK293, and BG-1 (Figure 3B). We verified trimeric complex formation between Mediator, G9a, and REST by means of their reciprocal sequential coimmunopreciptation from nuclear extracts of MED31/HeLa cells that stably express a FLAG-tagged MED31 Mediator subunit. Thus, serial immunoprecipitations from MED31/HeLa nuclear extracts with FLAG- followed by G9a-specific antibodies yielded terminal immunoprecipitates containing Mediator, G9a, and REST, thereby revealing that all three species engage in trimeric complex formation in mammalian cells (Figure 3C)
Thereafter, we examined the requirement for MED12/Mediator in REST-directed repression of an episomal target gene. In BG-1 cells, endogenous REST repressed the expression of a luciferase reporter gene bearing an RE1 silencing element compared to an otherwise identical reporter lacking an RE1 element (Figure 3D). REST-directed reporter gene repression in this assay was confirmed by knockdown of REST, which alleviated repression from the RE1 element (Figure 3D). Notably, G9a depletion also relieved REST-directed repression from the RE1 element, confirming a role for G9a as a REST corepressor (Roopra et al., 2004) (Figure 3D). Importantly, knockdown of MED12, but not CDK8 or MED23, in Mediator also abrogated reporter gene repression, thus implicating MED12/Mediator in REST-directed, G9a-dependent repression from an RE1 silencing element (Figure 3D). The requirement for MED12 in REST-directed repression is specific, since MED12 knockdown did not influence the repressive activity of CoREST tethered to promoter DNA through a Gal4 DNA-binding domain (Figure S9). Therefore, all repression domains do not universally require MED12.
Next, we examined the MED12/Mediator requirement for REST-directed extra-neuronal silencing of neuronal genes in their natural chromosomal loci. Previous bioinformatics-based studies have identified almost 1000 human genes, more than 40% neuronally expressed, that harbour RE1 silencing elements within 10 kb of their transcribed regions (Bruce et al., 2004). Among seven neuronally-expressed genes from this list, RT-qPCR analysis revealed three (M4, SNAP25, and Synapsin1) to be de-repressed at least four-fold while four (SCN2A2, nMDAR1, BDNF, and TUBB3) were de-repressed less than two-fold following REST knockdown in HeLa cells (Figure 3E). The refractory nature of some RE1-containing genes to REST knockdown likely reflects intrinsic differences in their respective affinities for REST (Bruce et al., 2004; Otto et al., 2007; Sun et al., 2005) and/or the involvement of additional REST-independent repressive mechanisms (Otto et al., 2007). Nonetheless, this analysis identified the M4, SNAP25, and Synapsin1 genes to be measurable targets of REST-directed extra-neuronal silencing. Derepression of these three genes was also elicited by depletion of G9a (Figure 3E), providing additional evidence to support an important role for G9a in REST-directed extra-neuronal silencing of neuronal-specific genes (Roopra et al., 2004). Significantly, knockdown of MED12, but not CDK8 or MED23, also resulted in significant de-repression of these three REST-target genes, thus implicating MED12/Mediator in REST-directed, G9a-dependent extra-neuronal gene silencing (Figure 3E). We also confirmed a MED12 requirement for repression of REST-target genes in BG-1 cells, indicating a broad role for MED12/Mediator in REST-directed extra-neuronal gene silencing (Figure S10). The fact that CDK8 knockdown had no effect on REST-target gene expression (Figure 3E) is significant, since MED12 knockdown destabilizes CDK8/CycC, whereas CDK8 knockdown destabilizes CycC, but not MED12 (Figure 3F). Therefore, the influence of MED12 knockdown on REST-target gene expression does not reflect the indirect loss of CDK8/CycC from Mediator. Furthermore, MED12 knockdown did not appreciably influence the expression level of REST nor any of its established corepressors examined, including CoREST, G9a, HDAC1, HDAC2, mSin3A, MeCP2, LSD1, HP1α, HP1β, or HP1γ(Figure 3F and data not shown). Together these findings support a direct role for MED12/Mediator in REST-directed G9a-dependent extra-neuronal silencing of neuronal-specific gene expression.
REST-directed neuronal gene silencing through targeted recruitment of G9a has been shown to involve the generation of highly localized domains of H3K9me2 around RE1 sites within repressed genes (Roopra et al., 2004). These G9a-dependent epigenetic marks function to promote transcriptionally repressive higher order chromatin structure. We used chromatin immunoprecipitation (ChIP) to monitor the transcription factor binding and histone methylation profiles of RE1 silencing elements as well as upstream gene sequences within the REST-targeted M4, SNAP25, and Synapsin1 genes in HeLa cells. These analyses revealed specific occupancy, while re-ChIP experiments confirmed co-occupancy, of RE1 elements by REST, enzymatically active G9a (revealed by the presence of G9a and H3K9me2), and, notably, Mediator (Figure 4A and B).
Because Mediator and G9a co-occupy RE1 silencing elements within REST-target genes and also physically associate and functionally participate in REST-directed neuronal gene repression, we sought to investigate whether and how Mediator might contribute to G9a-dependent H3K9me2 around RE1 sites within REST-repressed genes. To this end, we examined the transcription factor binding and histone methylation profiles of the M4, SNAP25, and Synapsin1 RE1 silencing elements as a function of MED12 using RNAi and quantitative ChIP. First, we confirmed the importance of REST and G9a in the establishment of a repressive chromatin environment around RE1 elements by coupling independent knockdown of REST and G9a with quantitative ChIP analyses. As expected, depletion of REST reduced occupancy of RE1 elements by REST itself as well as G9a and H3K9me2 (Figure 4C), confirming previous findings that REST recruits enzymatically active G9a to its repressed target genes in non-neuronal cells (Roopra et al., 2004). Notably, REST knockdown also reduced occupancy of RE1 elements by Mediator, thus revealing a role for REST in Mediator recruitment to its repressed target genes (Figure 4C). Interestingly, REST knockdown did not influence RE1 site occupancy by its corepressor CoREST (Figure 4C), possibly due to the ability of CoREST to associate with nucleosomal DNA in the absence of REST (Lee et al., 2005; Yang et al., 2006). Depletion of G9a reduced the level of H3K9me2, as expected, without influencing the association of REST or Mediator with RE1 elements (Figure 4C). Strikingly, depletion of MED12 significantly reduced occupancy of RE1 elements by G9a and decreased the level of H3K9me2 without influencing RE1 site occupancy by REST (Figure 4C). These observations implicate MED12/Mediator in REST-directed recruitment of enzymatically active G9a. In control experiments, neither MED23 nor CDK8 depletion reduced the association of G9a with RE1 elements, confirming that reduced REST-dependent recruitment of G9a accompanying MED12 knockdown is specific and does not derive from indirect loss of CDK8/CycC from Mediator (Figure 4D). Together, these results suggest that MED12/Mediator links REST with G9a HMTase in epigenetic silencing of neuronal gene expression.
To clarify the physical interaction dynamics among REST, Mediator, and G9a, we subjected purified proteins (Figure 5A and B) to pair-wise binding assays in vitro. Direct interactions were noted between each of the three species; thus, REST bound directly to both Mediator and G9a (Figure 5C and D), while Mediator and G9a also interacted directly (Figure 5E). Direct interactions between all three components likely contribute to the assembly of multimeric repressive complexes on RE1 elements, since we observed that MED12/Mediator is required for REST-directed G9a recruitment in vivo (Figure 4C). Delineation of the Mediator and G9a-binding domains on REST identified a common internal REST fragment (amino acids 141–600) that includes the DNA-binding domain and a lysine-rich domain (Figure 5C and D).
Previously, REST has been characterized as a bipartite transcriptional repressor encompassing distinct N- and C-terminal repression domains that function to recruit, respectively, the mammalian Sin3A/HDAC and CoREST/LSD1 corepressor complexes (Andres et al., 1999; Grimes et al., 2000; Huang et al., 1999; Lee et al., 2005; Naruse et al., 1999; Roopra et al., 2000; Shi et al., 2005). We validated independent association of discrete REST domains with distinct corepressors present in HeLa nuclear extracts using a GST pull-down approach. Thus, GST-REST fragments corresponding to amino acids 1–140 (including the N-terminal repression domain), 141–600, and 601–1098 (including the C-terminal repression domain) independently bound components of the Sin3A/HDAC, Mediator/G9a, and CoREST/LSD1 corepressor complexes, respectively (Figure 6A). Because REST amino acids 141–600 independently bound to both Mediator and G9a, we asked whether this domain harbours autonomous repression activity. When tethered to the heterologous Gal4 DNA-binding domain, REST amino acids 141–600 repressed transcription comparably to the established N-and C-terminal REST repression domains (Figure 6B and C). Furthermore, and consistent with their differential corepressor binding profiles, we confirmed distinct corepressor requirements for repression directed by REST 141–600 compared to the N- and C-terminal REST repression domains. Thus, the repressive activity of REST 141–600 requires MED12/Mediator and G9a (Figure 6D and Figure S11) but not HDACs (Figure 6E), while the opposite is true for the REST N- and C-terminal repression domains (Figure 6D and E; Figure S11). Taken together, these findings identify a MED12/Mediator- and G9a-dependent internal repression domain in REST, and establish a molecular basis to explain the requirement for Mediator and G9a in REST-directed neuronal gene silencing.
Recently, two human XLMR disorders, FG syndrome and Lujan syndrome, were discovered to be allelic and caused by two different MED12 missense mutations, R961W and N1007S, respectively (Risheg et al., 2007; Schwartz et al., 2007). We examined the possibility that these mutations in MED12 disrupt its REST-specific corepressor function. Missense mutations R961W (FG) and N1007S (Lujan) were independently introduced into a siRNA-resistant MED12 derivative (MED12r) and compared with wild-type MED12r (MED12r WT) for their respective abilities to rescue REST-dependent repression in cells compromised for this activity due to siRNA-mediated suppression of endogenous MED12. Strikingly, both the FG/R961W and Lujan/N1007S mutations severely compromised the ability of MED12r to rescue REST 141–600-dependent reporter gene repression, indicating that pathological mutations in MED12 associated with XLMR disrupt its REST-specific corepressor function (Figure 7A and C). By contrast, neither mutation influenced the ability of MED12 to support β-catenin transactivation (Kim et al., 2006), thus revealing a selective defect in the corepressor as opposed to the coactivator function of MED12 arising from the FG/R961W and Lujan/N1007S mutations (Figure 7B). As a control for these experiments, deletion of the PQL domain within MED12 critical for its interaction with both G9a and β-catenin abolished the ability of MED12 to rescue both REST-dependent repression and β-catenin-dependent activation in cells depleted of endogenous MED12 (Figure 7A and B). These results confirm the importance of this surface for MED12 coregulator function (Kim et al., 2006; Zhou et al., 2006; Zhou et al., 2002), and further reveal the REST-specific corepressor function of MED12 to be dependent upon its direct interaction with G9a.
To examine the impact of the FG and Lujan mutations in MED12 on the epigenetic profiles and expression levels of REST target genes in vivo, we compared siRNA-resistant WT MED12 (MED12r) and its corresponding FG/R961W and Lujan/N1007S mutant derivatives for their respective abilities to restore in MED12 knockdown cells WT levels of transcriptionally repressive G9a-dependent H3K9me2 on RE1 elements within the Synapsin1 and M4 genes. Both mutations compromised the ability of MED12 to mediate REST-directed recruitment of G9a and the imposition of transcriptionally repressive H3K9me2 (Figure 7D–F), providing additional evidence to link MED12/Mediator with REST-dependent neuronal gene silencing and XLMR. Surprisingly, both mutations in MED12 also impaired recruitment of Mediator to RE1 elements (Figure 7E and F). As neither mutation significantly influenced the interaction of MED12 with G9a (Figure S12), impaired Mediator recruitment to RE1 elements likely explains how these mutations disrupt REST-imposed epigenetic restrictions on neuronal gene expression, since MED12/Mediator is essential to link RE1-bound REST with G9a-dependent H3K9me2.
Transcriptional repression in eukaryotes is achieved through the interaction of sequence-specific DNA-binding transcriptional repressors with corepressors – functional intermediaries that directly or indirectly serve to inhibit transcription preinitiation complex (PIC) assembly on core promoters. Corepressors have been grouped broadly into those that interface directly with RNA polymerase II and its general transcription factors to inhibit PIC assembly and those that modify chromatin and therefore facilitate nucleosome-mediated promoter occlusion. Mediator is generally considered representative of the former corepressor class, and prior studies have attributed its chromatin-independent repressive function to its resident CDK8/CycC moiety. Thus, CDK8/CycC have been shown to repress transcription through diverse effects on the basal transcription machinery, including phosphorylation and consequent inhibition of TFIIH kinase activity as well as through direct occlusion of RNA polymerase II binding (Akoulitchev et al., 2000; Elmlund et al., 2006). Nonetheless, whether and how Mediator physically and functionally collaborates with chromatin modifying activities to repress transcription has not previously been established. Our finding that the MED12 interface in Mediator links REST with G9a-dependent H3K9me2 establishes a direct link between Mediator and chromatin modification leading to transcriptional repression. These studies thus implicate Mediator directly in epigenetic gene silencing and further reveal a hitherto unknown role for Mediator in facilitating the imposition of repressor-driven transcriptionally restrictive higher order chromatin structure.
Our studies further serve to clarify what has heretofore remained an obscure and complex function for MED12 in transcriptional repression. Thus, while prior genetic studies in model metazoans have implicated MED12 in negative regulation of gene expression (Janody et al., 2003; Moghal and Sternberg, 2003; Zhang and Emmons, 2000), it has not previously been clear whether MED12 plays a direct or indirect role in this process. Our finding that MED12 links REST with G9a in neuronal gene silencing provides a clear mechanistic basis to explain a direct repressive function for MED12 in gene control, and also provides new insight concerning the role and requirement of MED12 module components in negative regulation of gene expression. Heretofore, repression exerted by the MED12 module has been attributed to its CDK8/CycC moiety, while MED12/MED13 have been ascribed regulatory or architectural roles critical to ensure anchored kinase activity. Our delineation of a MED12-dependent repressive function for Mediator independent of CDK8/CycC is consistent with recent genetic analyses in D. melanogaster revealing functional diversification among MED12 module components and distinct roles for MED12 in developmental gene regulation independent of CycC-regulated CDK8 activity (Loncle et al., 2007).
The identification herein of MED12 as a corepressor of REST has important implications for both the biological function of Mediator and the mechanism of REST-dependent neuronal gene silencing. REST occupies a central role in non-neuronal lineage restriction by virtue of its ability to silence neuronal-specific gene expression in terminally differentiated non-neuronal cells (Ooi and Wood, 2007). Our discovery that MED12/Mediator collaborates with REST and G9a in extra-neuronal gene silencing suggests an important and heretofore unrecognized function for Mediator in targeting neuronal gene expression to the nervous system.
Mechanistically, REST-directed neuronal gene silencing is attributed to its targeted recruitment, via distinct N- and C-terminal domains, of multiple enzymatically diverse corepressors that co-ordinately function to fine-tune the histone acetylation and methylation dynamics as well as the overall epigenetic signature associated REST-repressed genes (Ooi and Wood, 2007). Our delineation of a MED12/Mediator- and G9a-dependent internal repression domain in REST encompassing amino acids 141–600 suggests that REST itself is more complex than originally thought, and considerably expands the network of functional interactions through which REST represses neuronal gene expression. Notably, this region in REST has previously been ascribed transcriptional repressive function, although the mechanistic basis for such activity has not been established (Roopra et al., 2000). Our finding that Mediator and G9a physically and functionally interact with REST 141–600 not only establishes a molecular basis for the repressive activity of this domain, but also extends the notion of REST as an integrative hub for the recruitment and directed deployment of functionally diverse chromatin modifiers. Further studies will be required to establish the fundamental interaction dynamics through which individual components of the broader REST corepressor network, including MED12/Mediator, coordinately function and possibly crosstalk to impose epigenetic restrictions on REST-target gene expression.
In addition to its role in modulating chromatin structure, REST has been reported to repress neuronal gene expression through direct effects on RNA polymerase II and its accessory transcription factors. For example, REST binds directly to the TATA box-binding protein to inhibit transcription preinitiation complex (PIC) formation (Murai et al., 2004), and RNA polymerase II small CTD phosphatases to inhibit polymerase activity (Yeo et al., 2005). Mediator, by virtue of its ability to interact directly with both REST and RNA polymerase II, could conceivably function to assist REST in PIC inhibition, or possibly even coordinate this activity with REST-directed chromatin remodelling. Future studies will be required to distinguish among these possibilities.
Among the 30 subunits comprising human Mediator, MED12 is the only one in which genetic variation has been linked to neuropsychiatric illness and cognitive dysfunction. Our discovery that the FG/R961W and Lujan/N1007S missense mutations in MED12 disrupt its REST-specific corepressor function represents the first description of a functional defect associated with these mutations. Notably, we found that both mutations impaired recruitment of Mediator to RE1 elements with little impact on the MED12/G9a interaction, providing a plausible mechanistic basis to explain how these mutations disrupt REST-imposed epigenetic restrictions on neuronal gene expression. Thus, impaired recruitment of Mediator, identified herein to be essential link between RE1-bound REST and G9a, results in reduced G9a recruitment, diminished levels of G9a-dependent H3K9me2, and de-repression of REST-target gene expression. The underlying basis by which the FG and Lujan mutations in MED12 impair Mediator recruitment to RE1 elements remains to be clarified, but could involve a conformational change in MED12/Mediator and resultant masking of a REST target subunit(s) in Mediator other than MED12, which we have thus far been unable to identify as a direct interface for REST. Further studies will be required to validate this prediction and establish the definitive basis by which the FG and Lujan mutations in MED12 impair Mediator recruitment to RE1 elements.
Our discovery that the FG/R961W and Lujan/N1007S missense mutations in MED12 disrupt REST-imposed restrictions on neuronal gene expression offers a possible epigenetic perspective to explain the role of MED12 in XLMR. Because REST and MED12 have both been implicated in neuronal development (Ballas et al., 2005; Hong et al., 2005b; Kuwabara et al., 2004; Rau et al., 2006; Wang et al., 2006), misregulation of REST target genes arising as a consequence of pathological mutations in MED12 could conceivably affect neuronal differentiation and possibly contribute to XLMR. In this regard, we note that additional members of the REST interactome have been linked to mental retardation (MR) (Kleefstra et al., 2006; Tahiliani et al., 2007). Haploinsufficiency arising from mutations/deletions in the gene encoding the H3K9 HMTase GLP/Eu-HMTase1 (a stoichimetric binding partner of G9a) have been shown to be causative for the 9q34 subtelomeric deletion syndrome characterized by severe MR, while mutations in the H3K4 histone demethylase SMCX/JARID1C have been linked to XLMR (Kleefstra et al., 2006; Tahiliani et al., 2007). However, the impact of pathological mutations in either protein on REST-dependent repressive activity is unknown. Our findings linking altered REST repressor function with XLMR-associated mutations in MED12 could provide a paradigm for how pathological defects in a broader REST corepressor network contribute to MR.
In conclusion, our work reveals that pathological mutations in MED12 associated with XLMR disrupt its newly identified role as mediator of REST-directed G9a-dependent extra-neuronal gene silencing. These findings establish a new function for Mediator in epigenetic restriction of neuronal gene expression to the nervous system, and shed new light on the mechanism of Mediator in transcriptional repression as well as the etiology of MED12-associated XLMR syndromes.
Processing of the MED12 carboxyl terminus (MED12C; aa 1715 to 2177) through a yeast two-hybrid screen using a human fetal brain cDNA library has been described previously (Zhou et al., 2006). From among ~2 × 106 independent clones screened, we identified 60 hits representing 21 different cDNAs, 3 of which corresponded to G9a HMTase.
pRC/CMV-hMED12, p3XFLAG-CMV-hMED12 and pcDNA3.1-HA-hG9a mammalian expression plasmids (Gyory et al., 2004; Kim et al., 2006), TATA-Luc and RE1-TATA-Luc reporter plasmids (Kuwabara et al., 2004), and MED12 truncation derivatives used in GST pull-down assays (Kim et al., 2006) have all been described. siRNA-resistant MED12 WT, FG, Lujan and ΔPQL mutants were generated by site-directed mutagenesis of p3XFLAG-hMED12 using the QuickChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA). PCR-based strategies were used to sublcone G9a truncation derivatives into pCS2+, G9a (aa 1–1001) and MED12C into pGEX-4T-2, and REST truncation derivatives into either pGEX4T-2, pM3 or pBind vectors as indicated.
Conditions and procedures for cell culture, transfections, RNA interference, and RT-qPCR analyses are described in Supplemental Experimental Procedures.
Conditions and procedures for co-IP, GST pull-down, immunofluorescence, and HMTase analyses are described in Supplemental Experimental Procedures.
We thank Y. Shinkai, D. Anderson, J. Conaway, R. Conaway, M. Garabedian, M. Horwitz, T. Kuwabara, G. Mandel, A. Roopra, and K. Wright for providing reagents, N. Buckley for discussions concerning the genomic locations of RE1 sites, and M. Carey, Y. Shinkai, P.R. Yew, and Boyer laboratory members for advice and comments. This work was supported by Public Health Service grant CA-0908301 from the National Cancer Institute (TGB), and by U.S. Army Department of Defense BCRP grants DAMD17-03-1-0272 and DAMD17-02-1-0584 (TGB).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.