Mutations in MECP2 (methyl-CpG-binding protein 2) are linked to the severe postnatal neurodevelopmental disorder RTT (Rett syndrome). MeCP2 was originally characterized as a transcriptional repressor that preferentially bound methylated DNA; however, recent results indicate MeCP2 is a multifunctional protein. MeCP2 binding is now associated with certain expressed genes and involved in nuclear organization as well, indicating that its gene regulatory function is context-dependent. In addition, MeCP2 is proposed to regulate mRNA splicing and a mouse model for RTT shows aberrant mRNA splicing. To further understand MeCP2 and potential roles in RTT pathogenesis, we have employed a biochemical approach to identify the MeCP2 protein complexes present in the mammalian brain. We show that MeCP2 exists in at least four biochemically distinct pools in the brain and characterize one novel brain-derived MeCP2 complex that contains the splicing factor Prpf3 (pre-mRNA processing factor 3). MeCP2 directly interacts with Prpf3 in vitro and in vivo and many MECP2 RTT truncations disrupt the MeCP2–Prpf3 complex. In addition, MeCP2 and Prpf3 associate in vivo with mRNAs from genes known to be expressed when their promoters are associated with MeCP2. These results support a role for MeCP2 in mRNA biogenesis and suggest an additional mechanism for RTT pathophysiology.
MeCP2 (methyl-CpG-binding protein 2) was originally identified by its ability to preferentially bind double-stranded DNA containing symmetrically methylated CpG dinucleotides and is the founding member of the MBD (methyl-CpG-binding domain) family of proteins [1,2]. The first biological role for MeCP2 was illustrated by showing that the protein interacts with methylated DNA in vivo and could repress transcription by association with a transcriptional co-repressor complex containing Sin3A and HDAC (histone deacetylase) [3–5]. In 1999, a genetic analysis identified mutations in MECP2 as causal for RTT (Rett syndrome), providing the first direct link between an epigenetic regulator and a human disease . RTT is a severe postnatal neurodevelopmental disorder and one of the most common causes of mental retardation in females . First described in 1966 by Andreas Rett , RTT is characterized by a period of apparently normal development from birth to 6–18 months followed by a regression of obtained language and motor skills . RTT patients usually exhibit a deceleration of head growth, respiratory dysfunction, scoliosis, cognitive impairment, seizures and social withdrawal [8,9]. In addition to RTT, numerous MECP2 mutations have now been linked to a variety of additional disorders, including autism, Angelman syndrome, learning disabilities and mental retardation syndromes [7,10–14].
MeCP2 has been reported to associate with myriad protein partners, including Sin3A [3,5], c-REST and Suv39h1 , c-Ski and N-CoR (nuclear receptor co-repressor) , Brm  and HP1 , all supporting a model of MeCP2 interacting with or being a stable component of transcriptional co-repressor complexes, resulting in targeted transcriptional repression of methylated DNA through modification of the chromatin state or chromatin-associated proteins. However, the biological relevance and implications towards RTT for these numerous documented MeCP2 interactions is not clear due in part to the particular methods utilized and non-neuronal choices for initial cellular protein sources. In fact, contradicting these numerous studies, it has been proposed that endogenous MeCP2 does not form any stable protein–protein interactions in vivo . Compounding the issue, recent work has expanded MePC2's proposed gene regulatory role beyond mere transcriptional repression; MeCP2 is implicated in transcriptional activation, genome-wide transcriptional silencing, mediating chromatin and nuclear architecture, and regulating pre-mRNA splicing as well [20–23]. Thus the in vivo protein–protein interaction profile of endogenous MeCP2, particularly in the brain, is still an open question and increasingly more important to understand as new functions for MeCP2 are emerging.
Genetic studies in mice suggest that expression of functional MeCP2 in neurons is essential for normal synapse formation and neuronal function during postnatal development and re-expression of MeCP2 in differentiated neurons alone rescues an RTT mouse model [24–29]. However, this idea is being challenged by a recent study that indicates the lack of MeCP2 specifically in glial cells contributes to RTT phenotypic neurons by an unknown secreted glial factor . This discrepancy illustrates the need for more unbiased approaches in determining the molecular roles of MeCP2 in both normal and RTT brain; thus, intact mammalian brain tissue would be the ideal source to study endogenous MeCP2 protein function. In the present study, we use the power of biochemistry to characterize MeCP2 in the mammalian brain and show that native MeCP2 protein purified from adult rat brain exists in multiple biochemically distinct pools/complexes, consistent with MeCP2 working as a multi-functional protein. We further characterize one brain-derived MeCP2 complex that contains Prpf3 (pre-mRNA processing factor 3), a known spliceosome-associated protein , as well as the Sdccag1 (serologically defined colon cancer antigen 1) , a mediator of nuclear export . MeCP2 shows specific, direct interactions with Prpf3 and Sdccag1, and these interactions are disrupted by certain RTT mutations. In addition, we show that MeCP2 and Prpf3 co-associate in vivo with mRNAs from genes activated by MeCP2, further supporting the previously identified regulatory role of MeCP2 in mRNA biogenesis  and providing another potential mechanism disrupted during RTT pathogenesis.
MATERIALS AND METHODS
MeCP2 protein purification
Rat brain nuclei were isolated generally as described . Adult rat brains (n=200 per extract preparation; Pel-Freez Biologicals) were thawed on ice, homogenized and nuclei were collected. Nuclei were suspended in buffer A [20 mM Hepes (pH 7.5), 1.5 mM MgCl2, 1 mM EGTA, 10% glycerol, 0.5 mM DTT (dithiothreitol), 1 μg/ml leupeptin, 1 μg/ml pepstatin and 1 μg/ml aprotinin] supplemented with 350 mM NaCl (A-350) and extracted for 30 min at 4°C. Insoluble material was removed by centrifugation at 200000 g for 20 min at 4°C. The supernatant was diluted with buffer A to <200 mM NaCl and used as a soluble protein source for chromatography.
Chromatography was preformed using an AKTA-FPLC apparatus (GE Healthcare) at 4°C in buffer A with indicated concentrations of NaCl. Soluble protein (130 mg per purification) was fractionated over a MonoQ10/10 column with bound protein eluted by a 20 CV (column volume) linear salt gradient from 100 to 1000 mM NaCl, collecting 0.5 CV fractions. The MeCP2-containing fractions peaking at 230 mM NaCl (QB1/2) were pooled, diluted with buffer A and fractionated over MonoS5/5 with bound protein eluted by a 20 CV linear salt gradient from 250 to 1000 mM NaCl, collecting 0.5 CV fractions. The MeCP2 containing fractions peaking at 550 mM NaCl (QB2) were combined and loaded on to a 1 ml Heparin FastFlow column and eluted with a 10 CV linear gradient from 450 to 1000 mM, collecting 0.5 CV fractions. Pooled heparin column fractions containing MeCP2 were applied to a 110 ml Superose6 column and fractionated in buffer A-150 with 0.1% Triton X-100. Subsequently, 0.5 ml of each fraction was trichloroacetic acid-precipitated, subjected to SDS/PAGE, and used for Western blotting or silver staining. Silver-stained polypeptide bands precisely co-fractionating with MeCP2 (by Western blotting) were excised and analysed by MS.
For the nuclease treatment experiments, MonoQ-eluted MeCP2 was dialysed against buffer A-100 containing 1 mM CaCl2, 10 mM MgCl2 and without EGTA then either treated with or without 500 units of benzonase nuclease (Sigma) for 30 min at 37°C, and re-tested for ability to bind the MonoQ resin.
Antibodies and Western blot analysis
Protein samples were separated by SDS/PAGE and transferred to ECL nitrocellulose membrane (GE Healthcare) for Western blotting by standard methods. For each experiment, the Western blotting images presented are from the same exposure on the same piece of film, linearly adjusted for brightness in Adobe Photoshop. The anti-MeCP2 3998 antibody is a rabbit polyclonal antibody derived from bacterially expressed recombinant protein encoding the full-length human MeCP2e2 isoform. The anti-(MeCP2 7–18) antibody is a rabbit polyclonal derived from bacterially expressed recombinant protein encoding amino acid residues 310–388 of the human MeCP2e2 isoform. Antibodies were used at the following concentrations: anti-(MeCP2 7–18), 1:2000; anti-MeCP2 (Upstate 07–013) 1:1000; anti-Prpf3 (MBL D171–3) 1:2000; and anti-HA (haemagglutinin) high affinity (clone 3F10) at 1:1000.
MS was carried out at the Protein Sciences Facility at the University of Illinois. FPLC-purified fractions were separated by SDS/PAGE, visualized by MS-compatible silver staining. Polypeptide bands were excised, destained (50% acetonitrile and 25 mM ammonium bicarbonate) and digested in 25 μl of Sequencing Grade Trypsin (12.5 ng/μl in 25 mM ammonium bicarbonate, G-Biosciences St Louis, MO, U.S.A.) using a CEM Discover Microwave Digestor for 15 min at 55°C (60W). Digested peptides were extracted using 50% acetonitrile with 5% formic acid, dried in a Savant SpeedVac apparatus and suspended in 13 μl of 5% acetonitrile containing 0.1% formic acid with 10 μl of sample used for MS analysis. The mass spectrometer used was a Waters quadrupole time-of-flight (Q-ToF) mass spectrometer connected to a Waters nano-Acquity UPLC. The column used was Waters Atlantis C18 (0.075 mm ×150 mm) with a flow rate of 250 nl/min. Peptides were eluted using a linear gradient of water/acetonitrile containing 0.1% formic acid (0–60% B) in 60 min. The mass spectrometer was set for result-dependent acquisition; MS/MS (tandem MS) was performed on the most abundant four peaks at any given time. Data analysis was performed using Waters Protein Lynx Global Server 2.2.5, Mascot (Matrix Sciences) and BLAST against NCBI NR database.
Generation of plasmids
RNA was purified from rat brain tissue or HeLa cells using TRIzol® Reagent (Invitrogen) as per the manufacturer's instructions. All cDNAs were generated using SuperScript III one-step RT (reverse transcription)–PCR with Platinum Taq (Invitrogen). All PCRs were performed with Phusion polymerase (New England Biolabs) and cloned into pGEM-T easy (Promega) for sequencing before subcloning into pGEX5-X1 (GE Life sciences), pCDNA 3.1 (Invitrogen) or the pCDNA 3.1 HA vector . All primers are listed in Supplementary Table S1 (at http://www.bioscirep.org/bsr/031/bsr0310333add.htm). All human MeCP2 constructs were generated from the human MECP2E2 cDNA (NM_004992). To generate constructs for in vitro synthesized proteins, cDNAs were PCR-amplified using primers listed in Supplementary Table S1 and subcloned into the specified restriction sites of pCDNA 3.1. Rat cDNAs were subcloned between NotI and XhoI; the human Prpf3 cDNA clone was subcloned between EcoRI and XhoI; the human full-length MeCP2 was amplified from full-length cDNA and subcloned between NotI and XhoI. Constructs for bacterially generated GST (glutathione transferase) fusion proteins were PCR-amplified from full-length cDNAs, subcloned into the specified restriction sites of pGEX5-X1 using primers listed in Supplementary Table S1. Rat GST–Prpf3 was subcloned between BamHI and XhoI. The rat and human GST–MeCP2 full-length, GST–MeCP2 deletion and human GST–MeCP2 RTT cDNAs were subcloned between EcoRI and XhoI. For the pCDNA3PHA-MECP2 vector, full-length human MeCP2 was subcloned between NotI and XhoI restriction sites of the pCDNA3.1 HA vector and subsequent digestion with NdeI and XhoI for cloning of the HA–MeCP2 fragment into the pCDNA3.1P puromycin vector.
HT-22 cells were transfected with pCDNA 3.1P HA-MeCP2 using Fugene HD transfection reagent (Roche). Stable integrants were selected as pools in puromycin (1 μg/ml), and maintained in DMEM (Dulbecco's modified Eagle's medium; Biowhittaker) supplemented with 10% fetal bovine serum, glutamine, antibiotics and puromycin.
GST pull-down assay
Assays were preformed essentially as previously described , using recombinant GST fusion proteins and in vitro transcribed/translated proteins radiolabelled with [35S]methionine using the T7 TnT Quick Coupled Transcription/Translation System (Promega) with or without 250 units of Benzonase nuclease.
Co-IPs from fractionated brain extracts were carried out as follows: MonoQ-fractionated rat brain nuclear extracts (QB2) were diluted to 100 mM NaCl with buffer A (0). Protein A Dynabeads (Invitrogen) were blocked with BSA and added to the extracts with 10 μl of either anti-MeCP2 3998 or normal rabbit serum IgG for 2 h at 4°C with rotation. Co-immunoprecipitates were washed three times with buffer A225, eluted by boiling in Laemmli buffer, separated by SDS/PAGE and subjected to Western-blot analysis. Co-IPs from cell culture were carried out as previously described  with modifications. Confluent 10-cm plates of HT-22 cells stably expressing HA–MeCP2 were lysed in 1 ml of IPH buffer [50 mM Tris/HCl (pH 8.0), 150 mM NaCl, 5 mM EGTA, 0.5 % Igepal and 1 mM PMSF] and debris removed by centrifugation. Lysates were pre-cleared with Protein A agarose, then incubated for 4 h with anti-HA (Sigma, E6779) or irrelevant IgG with Protein A agarose beads (Santa Cruz Biotechnologies) at 4°C with rotation. Nuclease treatment was preformed by suspension of protein-bound beads with 125 units of benzonase in the manufacturer's recommended buffer for 10 min at 37°C. IPs were washed three times with IPH buffer, eluted with Laemmli buffer, separated by SDS/PAGE and subjected to Western-blot analysis.
RIP (RNA IP) and Re-RIP
Experiments were performed as previously described with slight modifications . HT-22 cells (108) stably expressing HA–MeCP2 were collected, washed and suspended in 4 ml of PBS. Cross-linking buffer (50 mM Hepes, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA and 11 % formaldehyde) was added (400 μl), incubated at room temperature (23°C) for 30 min with rocking and quenched for 5 min at room temperature. Cell pellets were washed with PBS and lysed in 1 ml of FA buffer [50 mM Hepes/KOH (pH 7.5), 140 mM NaCl, 1 mM EDTA, 1 % Triton X-100, 0.1 % sodium deoxycholate and protease inhibitors] by sonication. The lysed cells were treated with 200 units of DNase I (Promega) in 25 mM MgCl2, 5 mM CaCl2 and 6 μl of RNasin (Promega) for 30 min at 37°C, cleared by centrifugation at 4°C, diluted 1:10 with ChIP (chromatin immunoprecipitation) dilution buffer [50 mM Hepes (pH 7.5), 140 mM NaCl, 1 mM EDTA, 10 % glycerol and 0.5 % Nonidet P40] and incubated with either anti-HA antibody or non-specific IgG with 4 μl of RNasin, for 12 h, rotating at 4°C. Protein A or G Dynabeads were added for 1 h rotating at 4°C and washed three times for 10 min each with wash buffer [50 mM Tris/HCl (pH 7.4), 500 mM NaCl, 1 % Triton X-100 and 0.1 % SDS]. For Re-RIP experiments, beads were washed and bound immune-RNA complexes were released in 20 mM DTT for 30 min at 37°C and resuspended in 1 vol. of ChIP dilution buffer for Re-RIP with antibody for Prpf3 and washed as described above. Beads were brought up in 200 μl of elution buffer [50 mM Tris/HCl (pH 7.4), 200 mM NaCl and 20 μg of proteinase K] for 1 h at 42°C, and cross-links reversed at 65°C. Samples were extracted with acid-equilibrated (pH 4.8) phenol/chloroform (5:1, v/v) and ethanol-precipitated. Precipitated material was suspended in 50 μl of DEPC (diethyl pyrocarbonate)-treated water and used for RT–PCR analysis with SuperScript III One-Step RT–PCR with Platinum Taq (Invitrogen) for mouse Cdk10 (cyclin-dependent kinase 10) and Frg1. Reaction mixtures were analysed by electrophoresis on a 2% agarose gel.
MeCP2 exists in at least four distinct protein pools in rat brain nuclei
A large-scale biochemical purification of endogenous MeCP2 protein from rat brain nuclear extract was performed to characterize the native MeCP2 in the mammalian brain. Whole rat brains were homogenized under non-denaturing conditions, the intact nuclei were purified by centrifugation through a sucrose cushion, and nuclear proteins were extracted under mild ionic conditions. Proteins were initially fractionated by strong anion exchange (MonoQ) column chromatography with the MeCP2 protein being tracked by Western blotting with multiple MeCP2-specific antibodies (Figure 1A and Supplementary Figure S1 at http://www.bioscirep.org/bsr/031/bsr0310333add.htm; and results not shown), all providing virtually identical profiles. The majority of soluble MeCP2 protein (~90%) did not associate with the MonoQ resin agreeing with a previous report ; however, a significant fraction (~10%) of this brain-derived MeCP2 consistently bound to MonoQ resin and was eluted in two distinct peaks along the linear salt gradient (Figure 1A), indicating multiple biochemically distinct pools of MeCP2.
Brain-derived nuclear MeCP2 exists in multiple biochemically distinct pools
The MonoQ bound (QB) MeCP2 elution profile showed an initial broad peak (QB1/2) concentrated at 230 mM NaCl and tailing to 370 mM, suggesting multiple MeCP2 complexes, with a minor yet distinct peak (QB3) centred at 450 mM NaCl (Figure 1A). Since MeCP2 is a DNA-binding protein and also known to interact with RNA, the potential of nucleic acids mediating the anionic association of MeCP2 with the cationic resin was addressed. The QB fractions were treated with benzonase nuclease to remove both DNA and RNA and then MonoQ chromatography was repeated. We found that all of the nuclease-treated QB MeCP2 still bound the MonoQ resin and was released by step elution, indicating the interaction was both RNA and DNA independent (Figures 1B and 1C). Finally, the presence of MeCP2 in the MonoQ flow through (QF) and each elution peak (QB1/2 and QB3) was subsequently confirmed by MS (Supplementary Figure S2 at http://www.bioscirep.org/bsr/031/bsr0310333add.htm).
The early eluting MonoQ-bound pool of MeCP2 (QB1/2) was characterized further for complexity and content (Figure 1A). The QB1/2 peak fractions were pooled and fractionated over the MonoS strong cation exchange resin, resolving into two peaks of MeCP2 protein (Figure 2D, top panel), the first being eluted at 400 mM NaCl (QB1) and the second being eluted at 550 mM NaCl (QB2). These results indicate at least three biochemically distinct pools of MeCP2 exist in the initial QB fraction and four MeCP2 pools exist overall in rat brain nuclear extract (QF and QB1, QB2 and QB3).
The QB2 pool of MeCP2 co-purifies with six candidate proteins including Prpf3
Brain-derived MeCP2 exists in a complex with the splicing factor Prpf3
To identify potential MeCP2-interacting proteins, the more abundant MonoS-bound pool of MeCP2 (QB2) was purified further by fractionation using heparin affinity chromatography and gel filtration through Superose6 (Figure 2A). Western blotting and silver-stain analysis of the final fractionation showed MeCP2 peaking with an apparent molecular mass of 600–700 kDa and precisely co-purifying with six additional polypeptides (Figures 2B and 3). Since monomeric MeCP2 has previously been shown to exhibit an unusual Superose6 gel filtration profile migrating equivalent with an apparent mass of approx. 450 kDa , the apparent mass for QB2 MeCP2 is consistent with it being associated with additional proteins. Therefore all six MeCP2 co-purifying polypeptides were identified by MS with significant coverage (Figure 3 and Supplementary Figure S3 at http://www.bioscirep.org/bsr/031/bsr0310333add.htm) as Prpf3, Sdccag1, ATP-binding cassette 50 and three components of a translation initiation complex (Eif2 subunits 1, 2 and 3). Since a commercial antibody was available against Prpf3, Western-blot analysis of the size exclusion chromatography fractions was carried out and showed that MeCP2 and Prpf3 precisely co-fractionate (Figure 2B, bottom) confirming the silver-stain analysis (Figure 2B, top).
Identification of candidate MeCP2 complex proteins
With Prpf3 identified as a putative MeCP2-interacting protein, the MeCP2 fractionation scheme (Figure 2A) was analysed for Prpf3 by Western blotting to determine whether Prpf3 was present in all MeCP2 pools or specific to QB2 and similarly to determine whether all Prpf3 in brain extracts associated with MeCP2. Not surprisingly, only a fraction of Prpf3 in brain extracts co-fractionated with MeCP2 when assaying the MonoQ separation profile (Figure 2C); however, all of the detectable Prpf3 overlapping with MeCP2 specifically co-fractionated with the QB2 pool of MeCP2 from the MonoS fractionation and not with the QB1 MeCP2 pool (Figure 2D). Thus Prpf3 distinguished QB2 from other MeCP2 protein pools. Because Prpf3 is a known component of the spliceosome  and MeCP2 has been shown to interact with another spliceosome-associated protein, YB-1 , the QB2 fractionation was screened for YB-1. However, YB-1 was not found by Western blotting and YB-1 was absent from the MS analysis indicating YB-1 did not co-purify with QB2 MeCP2 under these conditions (Figure 3, and results not shown).
To confirm that native MeCP2 and Prpf3 are in a complex in vivo, co-IP experiments were performed (Figure 4). In brain extracts, as determined by the biochemical fractionations described above, the vast majority of MeCP2 is not associated with Prpf3 and likewise the majority of Prpf3 in brain is not associated with MeCP2 making it difficult to visualize any interaction by co-IP directly from crude nuclear extracts. Thus two approaches were used: co-IPs from fractionated brain nuclear extract (Figure 4A) and co-IPs from crude extracts using tissue culture cells overexpressing an epitope-tagged version of MeCP2 (Figure 4B). Rat brain nuclear extract was prepared and fractionated over MonoQ as described above. The QB1/2 MeCP2 fractions were diluted to 100 mM NaCl in buffer A (0) and used for co-IP experiments assayed by Western blotting. Using the anti-MeCP2 3998 antibody to IP, Prpf3 was specifically co-immunoprecipitated from these fractions and the interaction between MeCP2 and Prpf3 was not dependent on nucleic acids (Figure 4A, lanes 3 and 4). Alternatively, to show co-IP interaction between MeCP2 and Prpf3 from crude cell extracts, the murine hippocampal HT-22 cell line was used to generate a pool of cells stably expressing HA epitope-tagged human MeCP2. Immunofluorescence showed that HA–MeCP2 in these cell lines predominantly concentrated to DAPI (4′,6-diamidino-2-phenylindole)-rich heterochromatic foci as expected for endogenous MeCP2 (supplemetnary Figure S4 at http://www.bioscirep.org/bsr/031/bsr0310333add.htm). Anti-HA antibodies were then used to specifically co-IP HA–MeCP2 and MeCP2-associated proteins from these cell lines. Western blotting showed that HA–MeCP2 specifically co-immunoprecipitated a fraction of the endogenous mouse Prpf3 (Figure 4B, lane 3), indicating that Prpf3 and MeCP2 exist in a stable complex within these neuronal cells, and their interaction was not dependent on nucleic acids (Figure 4B, lane 4). We conclude that MeCP2 and Prpf3 associate in vivo.
Co-IPs confirmed MeCP2 associates with Prpf3
MeCP2 interacts directly with Prpf3 and Sdccag1, independent of nucleic acids
To investigate which of the co-purifying candidate proteins directly interacted with MeCP2, the cDNAs for each identified protein was cloned and the recombinant proteins were tested for the ability to interact with MeCP2 in vitro (Figure 5 and Supplementary Figure S4). GST pull-down assays using a GST–MeCP2 fusion protein showed that rat Prpf3 specifically interacted with full-length rat MeCP2, but not GST alone (Figure 5A, left panel), and reciprocally rat GST–Prpf3 interacted with rat MeCP2 (Figure 5B). Furthermore, treatment of the GST–MeCP2/Prpf3 reaction with benzonase nuclease showed the interaction was not dependent on nucleic acids (Figure 5A and 5B). Similarly, rat Sdccag1 specifically interacted with rat GST–MeCP2 in a nucleic acid-independent manner (Figure 5A, middle panel). Interestingly, neither rat GST–MeCP2 nor rat GST–Prpf3 interacted with any of the other four identified polypeptides supporting the specificity of the observed interactions of MeCP2, Prpf3 and Sdccag1 (Figure 5A, right panel, and results not shown).
MeCP2 directly interacts with Prpf3 and Sdccag1
Mapping the Prpf3 and Sdccag1 interaction domains of MeCP2
The regions of MeCP2 that directly interact with Prpf3 and Sdccag1 were mapped by GST pull-down analysis. A series of bacterially generated human GST–MeCP2 deletion mutants were tested for their ability to interact with in vitro synthesized human Prpf3 (Figures 5C and 5D). Prpf3 retained the ability to interact with MeCP2 N-terminal deletions up to and including amino acid residue 195 as well as a C-terminal deletion lacking amino acid residues 309–486. However, Prpf3 did not interact with an MeCP2 deletion lacking amino acid residues 1–308, indicating that the TRD (transcription repression domain) region is required for interaction (Figure 5C, upper panel). Therefore a fragment of MeCP2 that contains only amino acid residues 207–308, corresponding to the TRD domain, was tested and found to interact with Prpf3 with similar apparent affinity as full-length MeCP2. Similarly, Sdccag1 was synthesized in vitro and subjected to the same series of GST–MeCP2 deletions as Prpf3 (Figure 5C, lower panel). Sdccag1 was capable of interacting with all of the N-terminal MeCP2 deletion proteins tested, but failed to interact with a MeCP2 deletion lacking the region C-terminal to the TRD. Thus the interaction domain with Sdccag1 resides between amino acid residues 309 and 486, adjacent with the Prpf3 interaction domain (Figure 5E).
Considering most known RTT mutations reside in MECP2 and many are nonsense mutations resulting in a truncated MeCP2, it is likely that some of these mutations would also disrupt the interactions of Prpf3 and/or Sdccag1 with MeCP2. A series of bacterially generated human GST–MeCP2 RTT nonsense mutants were produced and tested for their ability to interact with in vitro synthesized human Prpf3 and rat Sdccag1 (Figure 5D). The series of RTT mutations truncate MeCP2 ranging from amino acids 49 to 204. All of these RTT nonsense mutants lack the C-terminal region of MeCP2 and therefore disrupted its interaction with Sdccag1 as expected (Figure 5D, bottom panel). The Prpf3 interaction was disrupted by RTT truncations at amino acids S49X, S68X and W104X as well. However, the Prpf3 interaction with MeCP2 was maintained with the RTT truncations Y141X, R168X and S204X. This indicates there are two MeCP2 regions capable of interacting with Prpf3 flanking the inter-domain region between the MBD and TRD of MeCP2. Interestingly, this same inter-domain region of MeCP2 has been previously characterized as the RG domain and is required for MeCP2's RNA-binding activity . We conclude that MeCP2 contains two domains sufficient for interaction with Prpf3, one in the MBD between amino acids 104 and 141, and the second in the TRD between amino acids 207 and 294 and one domain for MeCP2's interaction with Sdccag1, residing between amino acids 311 and 486 (Figure 5E). Therefore any RTT mutation truncating MeCP2 at or before amino acid residue 104 will abolish the Prpf3 and Sdccag1 interaction, whereas any RTT truncation at or before amino acid residue 297 would disrupt the Sdccag1 interaction with MeCP2, all of which could affect MeCP2's role in RNA biogenesis.
MeCP2 interacts with mRNA
Several lines of evidence suggest that MeCP2 is associated with mRNA biogenesis; MeCP2 binds RNA in vitro , is part of an RNP complex with YB-1 in cell culture , MeCP2 knockout mice show mis-spliced transcripts in the brain , it has been identified as a transcriptional activator at the majority of gene promoters it has been shown to regulate , and in the present study MeCP2 was shown to interact with the splicing factor Prpf3. To determine whether MeCP2 interacts with the RNA transcripts expressed from the genes it regulates in vivo, RIPs were performed. Using anti-HA antibodies on HT-22 (HA–MeCP2) cell lysates, mRNAs for the MeCP2-regulated genes Cdk10 and Frg1 were capable of being specifically RNA immunoprecipitated (Figure 6A). Assaying the IPs by RT–PCRs using oligonucleotide primers designed to amplify across exon junctions of Cdk10 and Frg1 show that MeCP2 is associated with the spliced form of the genes, whereas MeCP2 is not associated with the Casc3 gene transcript (Figure 6A), a gene not regulated by MeCP2 , supporting the specificity of the MeCP2–mRNA interaction. The RIPs were RNase sensitive confirming that RNA, and not DNA, was specifically immunoprecipitated. Interestingly, RT–PCR for RIPs using primers for Cdk10 additionally show what appears to be pre-mRNA by size and RNase sensitivity (Figure 6A, top). Although the Cdk10 pre-mRNA product was not observed in 100% of the RT–PCR analyses for Cdk10, it was observed repeatedly and suggests that MeCP2 associates with the Cdk10 RNA at a step before the completion of splicing and remains bound after RNA processing is completed.
The MeCP2–Prpf3 complex interacts with mRNA
To investigate whether Prpf3 was part of the HA–MeCP2–mRNA complex, an RIP and Re-RIP approach was implemented. Anti-HA antibodies were used to RNA immunoprecipitate HA–MeCP2–mRNA complexes from HT-22 (HA–MeCP2) cell lysates, the bound complexes were eluted from the HA antibodies intact using DTT to disrupt the IgG structure and then anti-Prpf3 antibodies were used to Re-RIP. Therefore any RNAs present in the Re-RIP must have been associated with both MeCP2 and Prpf3. RT–PCR analysis showed that RIP for HA–MeCP2 and subsequent Re-RIP for Prpf3 protein immunoprecipitated mRNA for Cdk10, indicating that both proteins are in fact associated with this target mRNA (Figure 6B), further supporting their being in a complex in vivo.
MeCP2 has been characterized as a multifunctional protein using a variety of techniques and from numerous cellular contexts; however, only the biochemical characteristics of the endogenous MeCP2 protein in the mammalian brain are relevant to RTT. MeCP2 mutations identified from RTT patients are generally point mutations resulting in a single missense or non-sense amino acid change, and have been identified throughout all protein domains of MECP2 (http://www.rettsyndrome.org/), suggesting that all domains are critical for MeCP2 function . Although all these mutations produce clinical RTT pathology, certain mutations are more strongly associated with particular symptoms and disease severity suggesting that all RTT mutations in MECP2 are not equal, with some potentially being more disruptive towards MeCP2's many functions . The underlying mechanisms of how these myriad mutations lead to RTT pathophysiology remain unclear due to a lack of understanding towards the complete scope of MeCP2's normal function in the brain. In the present study, we identified four distinct brain-derived MeCP2 protein pools, agreeing with the functional results proposing multiple roles for MeCP2 . Characterizing one complex as containing MeCP2, Pprf3, Sdccag1 and mRNA strongly complements previously published results that MeCP2 is an RNA-binding protein involved in mRNA splicing .
MeCP2's association with Prpf3, a major component of the spliceosome, supports MeCP2 as having a role in modulating mRNA splicing ; however, it is not clear what exactly that role might be. MeCP2 is present at the promoter of many actively transcribed genes , and activates a majority of genes it regulates in the mouse hippocampus . Interestingly, a recent study suggests that differential gene body methylation may play a role in RNA transcript splicing . Genome-wide methylation profiling in multiple human cell types found that exons were more highly methylated than introns, and that there were sharp transitions of DNA methylation at exon–intron boundaries . Experimentally, it is known that MeCP2 binds RNA in vitro with an affinity similar to that of methylated DNA, and that the two activities are mutually exclusive . Thus MeCP2 at an active promoter or within the body of a transcribed gene would be well positioned spatially on its release to interact with the transcripts of the genes it activates and influence splice site selection. Therefore it is reasonable to propose that MeCP2-activated genes would also be targets for splicing regulation by an MeCP2 containing complex. Supporting this model, RNA transcripts from Cdk10, a gene positively regulated by MeCP2 binding at its promoter  and abnormally spliced in Mecp2308Y mice , were associated with both MeCP2 and Prpf3 in vivo (Figure 6).
Based on what is known about MeCP2 and Prpf3 independently, it is possible that the direct interaction of MeCP2 with Prpf3 could function in splice site selection. Prpf3 is one of multiple associated proteins of the U4/U6.U5 tri-snRNP (tri-small nuclear ribonucleoprotein) complex that is recruited to the splice site to form and stabilize the functional spliceosome, and is essential for pre-mRNA splicing [31,41]. Importantly, MeCP2 can bind RNA directly  and has been implicated as a regulator of alternative splicing in the HeLa and Neuro2A cell lines through an RNA-dependent interaction with YB-1 , a protein known to participate in splicing of mRNAs . Although we were unable to detect YB-1 as a component of the MeCP2–Prpf3–Sdccag1 complex in brain, it is inefficient to inhibit the highly active and stable RNases that would abolish any YB-1–RNA–MeCP2 interaction during purification, hence we cannot rule out that YB-1 could be part of a larger RNA-dependent alternative splicing complex. Nevertheless, a direct interaction of MeCP2 and Prpf3 further supports a role for MeCP2 in mRNA splicing regulation. Notably, Mecp2308/Y mice, which produce a truncated form of MeCP2 and reproduce many of the classical features of RTT , have been shown to have multiple genes that are abnormally spliced in the brain . This suggests the C-terminal portion of MeCP2, which we have identified as the putative Sdccag1 interaction domain, plays a critical role in regulating alternative splicing.
Sdccag1 was originally identified from colon cancer patients by a serological analysis of recombinant cDNA expression libraries (SEREX)  and later identified as a tumour suppressor by its ability to cause cell cycle arrest in a non-small-cell lung cancer cell line . The significance of finding Sdccag1 as part of the MeCP2–Prpf3 complex can only be implied due to the limited information on its function in vertebrates and Drosophila. Bioinformatically the Sdccag1 protein contains a predicted RNA-binding domain homologous with a eukaryotic small nuclear RNP , suggesting the capacity to function in mRNA splicing. Functionally, the Drosophila Sdccag1 homologue Caliban has been shown to interact with and mediate the nuclear export of the Prospero homeodomain transcription factor (Prox in mammals) and this interaction and function is conserved in mammalian cells . This raises the possibility that the MeCP2–Prpf3–Sdccag1 complex may not only be involved in splicing mRNA but also transporting mRNAs. Consistent with this model, MeCP2 has been shown to have both a nuclear and cytoplasmic localization in neuronal cell lines .
Considering the range of mutations throughout MECP2 in RTT, it is not surprising that many RTT mutations are within, or predicted to affect both the Prpf3- and Sdccag1-binding domains, disrupting the MeCP2–Prpf3–Sdccag1 complex and presumably MeCP2-mediated splicing regulation and mRNA transport. With MeCP2 emerging as a multifunctional protein and its biological role in RTT unclear, identifying biochemically distinct pools of MeCP2 in the brain containing novel MeCP2-interacting proteins is a valuable tool towards an understanding MeCP2 function and its dysfunction in RTT. The present study adds to the mounting evidence indicating that one such critical function of MeCP2 in the brain involves RNA biogenesis.
cyclin-dependent kinase 10
methyl-CpG-binding protein 2
pre-mRNA processing factor 3
serologically defined colon cancer antigen 1
transcription repression domain
Steven Long performed the biochemistry, IPs, co-IPs, RIPs, transfections and wrote the paper; Jenny Ooi initiated the project, performed the biochemistry and the initial MeCP2 fractionation into multiple complexes; Peter Yau performed the MS and identifications; Peter Jones conceived the project, funded the project, performed the biochemistry and wrote the paper.
We greatly appreciate the gift of HT-22 cells from Dr Stephanie Ceman. We thank Dr Ann Cheever and Dr Brian Imai for technical support, and Dr Takako I. Jones for help with the manuscript.
This work was supported by the National Institutes of Health, National Institute of Child Health and Human Development [grant number 5K22HD001338 (to P.L.J.)], and start-up funds from the School of Molecular and Cellular Biology and Department of Cell and Developmental Biology, University of Illinois at Urbana-Champaign (to P.L.J.).