Decitabine alters the expression of Mecp2 isoforms via dynamic DNA methylation at the Mecp2 regulatory elements in neural stem cells

Background Aberrant MeCP2 expression in brain is associated with neurodevelopmental disorders including autism. In the brain of stressed mouse and autistic human patients, reduced MeCP2 expression is correlated with Mecp2/MECP2 promoter hypermethylation. Altered expression of MeCP2 isoforms (MeCP2E1 and MeCP2E2) is associated with neurological disorders, highlighting the importance of proper regulation of both isoforms. While known regulatory elements (REs) within the MECP2/Mecp2 promoter and intron 1 are involved in MECP2/Mecp2 regulation, Mecp2 isoform-specific regulatory mechanisms are unknown. We hypothesized that DNA methylation at these REs may impact the expression of Mecp2 isoforms. Methods We used a previously characterized in vitro differentiating neural stem cell (NSC) system to investigate the interplay between Mecp2 isoform-specific expression and DNA methylation at the Mecp2 REs. We studied altered expression of Mecp2 isoforms, affected by global DNA demethylation and remethylation, induced by exposure and withdrawal of decitabine (5-Aza-2′-deoxycytidine). Further, we performed correlation analysis between DNA methylation at the Mecp2 REs and the expression of Mecp2 isoforms after decitabine exposure and withdrawal. Results At different stages of NSC differentiation, Mecp2 isoforms showed reciprocal expression patterns associated with minor, but significant changes in DNA methylation at the Mecp2 REs. Decitabine treatment induced Mecp2e1/MeCP2E1 (but not Mecp2e2) expression at day (D) 2, associated with DNA demethylation at the Mecp2 REs. In contrast, decitabine withdrawal downregulated both Mecp2 isoforms to different extents at D8, without affecting DNA methylation at the Mecp2 REs. NSC cell fate commitment was minimally affected by decitabine under tested conditions. Expression of both isoforms negatively correlated with methylation at specific regions of the Mecp2 promoter, both at D2 and D8. The correlation between intron 1 methylation and Mecp2e1 (but not Mecp2e2) varied depending on the stage of NSC differentiation (D2: negative; D8: positive). Conclusions Our results show the correlation between the expression of Mecp2 isoforms and DNA methylation in differentiating NSC, providing insights on the potential role of DNA methylation at the Mecp2 REs in Mecp2 isoform-specific expression. The ability of decitabine to induce Mecp2e1/MeCP2E1, but not Mecp2e2 suggests differential sensitivity of Mecp2 isoforms to decitabine and is important for future drug therapies for autism.


Background
Methyl CpG Binding Protein 2 (MeCP2) is a key transcriptional regulator in the brain [1]. MECP2 mutations and expression deficits result in a broad range of neurodevelopmental disorders, including Rett syndrome (RTT) and autism spectrum disorders [2,3]. In mice (Mecp2) and humans (MECP2), alternative splicing of a single gene leads to the generation of two protein isoforms MeCP2E1 and MeCP2E2 (mature transcripts for Mecp2e1 and Mecp2e2 are shown in Figure 1A) [4,5]. We and others have shown differential expression of the two Mecp2/ MeCP2 isoforms in mouse brain [5][6][7]. Recent studies suggest that MeCP2E1 is the most relevant isoform for RTT pathology [8,9]. Moreover, overexpression of Mecp2e2, but not Mecp2e1, promotes neuronal cell death [10], implicating the importance of proper regulation of both Mecp2 isoforms in the brain.
In RTT mouse models, transgenic expression of either Mecp2 isoform can rescue RTT phenotypes to different extents [14,15]. However, gene therapy delivery of MECP2 into the affected cells or drug therapies to induce MECP2 expression has to be carried out with caution, as even mild overexpression of MeCP2 can lead to progressive neurological disorders [16,17]. Currently, limited knowledge exists on MECP2/Mecp2 regulation, with no specific knowledge on possible differential MECP2/Mecp2 isoform-specific regulatory mechanisms.
MECP2/Mecp2 gene expression is known to be regulated by regulatory elements (REs) within the promoter and a silencer element within the Mecp2 intron 1 [12,13,18] (Figure 1B). Implying the role of DNA methylation in MECP2 regulation, reduced MECP2 expression in the brains of male autistic patients correlates with human MECP2 promoter hypermethylation [2,19]. Moreover, reduced Mecp2 expression in the postnatal murine brain in response to early maternal separation and stress is associated with hypermethylation of the mouse Mecp2 promoter [20]. However, possible differential impact of DNA methylation on MECP2/Mecp2 isoforms is currently unknown. DNA methylation is a major epigenetic modification that controls gene expression without affecting the underlying DNA sequences (reviewed in [21,22]). DNA methylation at the cytosine residues (5methylcytosine (5mC)) of the CpG dinucleotides is carried out by DNA methyltransferases (DNMT) and is generally considered to be a repressive epigenetic modification [1,23]. Conversely, 5-hydroxymethylcytosine (5hmC), which is generated by oxidation of 5mC by TET proteins  [4,11]). Exons are denoted as Ex. (B) Regulatory elements of the MECP2/Mecp2 gene. The MECP2/ Mecp2 gene is reported to be regulated by negative and positive REs within the promoter and a silencer element within the intron 1 (information extracted from [12,13]). For our studies we selected a 500-bp region in the promoter upstream of the exon 1 and a 1-kb region in the intron 1 upstream of the exon 2. Each sequence was divided into three regions, R1 to R3 in the Mecp2 promoter and R4 to R6 in the intron 1. Note that there are no CpG dinucleotides in the mouse genomic sequence between R5 and R6.
is generally considered to be an active epigenetic mark [24,25]. Promoter methylation is mostly associated with gene silencing [26], while DNA methylation at both intronic and exonic regions are shown to correlate with isoform-specific transcription by alternative splicing or by utilizing alternate promoters [27,28].
Treatment with DNA demethylating agents or DNMT/ Dnmt inhibitors such as decitabine (also called 5-Aza-2′deoxycytidine) is a commonly used method to study the role of DNA methylation in gene expression [29,30]. While exposure to decitabine results in DNA demethylation, its subsequent withdrawal causes remethylation or methylation reprogramming [29], providing an excellent platform to uncover the role of DNA methylation in gene expression.
In vitro differentiation of neural precursor cells/neural stem cells (NSC) into different brain cell types is utilized as an acceptable model system to mimic the in vivo neural development [31][32][33][34][35][36]. Previously, we used a similar in vitro NSC differentiation system to report the first preclinical MECP2 isoform-specific gene therapy vectors, for future gene therapy applications in Rett syndrome [35]. Further, we introduced differentiating NSC as a suitable in vitro model to study the expression and function of developmentally important genes such as Meis1 in neural development [37]. In the current study, we used this previously characterized system to study the expression and regulation of Mecp2 isoforms during NSC differentiation.
Investigation of MECP2/MeCP2 expression and function in neurodevelopmental disorders has been the focus of intensive research. However, despite the critical importance of precisely controlled levels of MeCP2 expression in the brain, the underlying regulatory mechanisms have been understudied. Here, we report the correlation between the expression of Mecp2 isoforms and DNA methylation patterns at the Mecp2 REs at different stages of NSC differentiation. Further, we demonstrate the effect of dynamic changes in DNA methylation induced by exposure and withdrawal of decitabine on the expression of Mecp2/MeCP2 isoforms.

Ethics statement
All experiments were performed in accordance with the standards of the Canadian Council on Animal Care with the approval of the Office of Research Ethics of University of Manitoba. All experimental procedures were reviewed and approved (protocol number 12-031) by the University of Manitoba Bannatyne Campus Protocol Management and Review Committee.

Neural stem cell isolation, culture and differentiation
Embryonic mouse forebrain-derived NSC were isolated from the forebrains of CD-1 mice at embryonic day (E) 14.5 and were cultured according to previously described methods [35,37]. Briefly, dissected forebrain tissues were mechanically homogenized in NSC media DMEM/F12 1:1 (Wisent, Quebec, Canada) containing HEPES, glutamine, antibiotic/antimycotic, glucose, recombinant human epidermal growth factor (rhEGF) (Sigma Aldrich, Oakville, Ontario, Canada, 20 ng/ml), basic fibroblast growth factor (bFGF) (Upstate (Millipore), Billerica, MA, USA, 20 ng/ml), heparin (Sigma Aldrich, Oakville, Ontario, Canada, 2 μg/ml) and hormone mix. Dissociated single cells were plated at a density of 10 5 cells/cm 2 in NSC media. The media were refreshed every 48 h and cells were cultured under these conditions for 7 days to generate neurospheres. Primary neurospheres were gently dissociated to single cells by accutase treatment. Dissociated cells were plated on plates coated with growth factor-reduced matrigel (BD Biosciences, Mississauga, Ontario, Canada) at a density of 10 5 cells/cm 2 in DMEM (GIBCO, Life Technologies Inc, Burlington, Ontario, Canada) and 10% Fetal Bovine Serum (Invitrogen, Life Technologies Inc, Burlington, Ontario, Canada) in the absence of rhEGF and bFGF. Cells were differentiated for 8 days, reported to be sufficient for differentiation of neuronal and glial cells [35,37], and media were changed every other day.

Decitabine treatment
At the onset of differentiation on day zero (D0), dissociated NSC were treated with 2.5 μM decitabine for 48 h. After two days (D2), the media were replaced with fresh media that was refreshed every other day for an extra 6 days (until D8). Control cells were cultured under similar experimental conditions, in the absence of decitabine.
Quantitative measurement of male/female contribution Genomic DNA from neural stem cells at D0, D2, D8 and decitabine-treated cells were extracted using the DNeasy Blood and Tissue kit (Qiagen, Ontario, Toronto, Canada), as per manufacturer's instructions. The contribution of male and female sexes were determined by semiquantitative PCR-based amplification of Sry (sex-determining region protein gene in the Y chromosome) and Il3 (autosomal gene as an internal control) genes, as described previously [38], using the primers listed in Table 1. The PCR program consisted of an initial denaturation at 95°C for 4.5 minutes, followed by 33 cycles of 95°C for 35 s, 50°C for 1 minute, 72°C for 1 minute, and a final extension step at 72°C for 5 minutes. The amplified products were run on 1.5% agarose gel and the bands were visualized by ethidium bromide staining. The Sry and Il3 PCR products were identified based on the corresponding sizes (Sry: 402 bp, and Il3: 544 bp). Intensity of the corresponding bands was quantified using Adobe Photoshop CS5 software. The contribution of either sex was further determined by quantitative reverse transcription PCR (qRT-PCR) for Xist (X-inactive specific transcript) gene, as previously described, with minor modifications [39]. The PCR program for Xist included an initial denaturation at 95°C for 5 minutes; followed by 35 cycles of 95°C for 30 s, 53°C for 30 s, 72°C for 30 s, and 78°C for 30 s.

Immunofluorescence experiments
Immunofluorescence (IF) experiments were performed according to previously described protocols [7,35,37]. Primary and secondary antibodies used for IF are listed

Nuclear extractions and western blotting
Nuclear extraction from D2 and D8 NSC were carried out using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Scientific, Ontario, Toronto, Canada), as per the manufacturer's instructions. Western blot (WB) experiments were conducted according to previously described protocols [49][50][51], and quantification of the signals was performed as reported [7]. ACTIN or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. Student t-test was used to determine statistical significance between control and treated cells. Primary and secondary antibodies used for WB are listed in the Tables 3 and 4, respectively.

DNA dot blot assay for 5mC and 5hmC
Genomic DNA was isolated using the DNeasy Blood and Tissue kit (Qiagen, Ontario, Toronto, Canada). DNA dot blot was performed using a previously described protocol [52], with minor modifications. The DNA blotted membranes were probed with either 5mC or 5hmC antibody (Tables 3 and 4). Total DNA levels were detected by staining the same membrane with 0.02% methylene blue (MB) in 0.3 μM sodium acetate (pH 5.2). Adobe Photoshop CS5 software was used to quantify dot blot signals.

Bisulfite pyrosequencing
Genomic DNA was isolated as described in the previous section. Primer design and bisulfite pyrosequencing experiments were conducted as a service by The Hospital for Sick Children (SickKids), Toronto, Canada, as reported elsewhere [53], using the primers listed in Table 5. The regions analyzed for the methylation patterns are shown in Figure 1B. Specific CpG dinucleotides that are analyzed within each region are shown in Additional file 1.

Correlation analysis
The correlation between DNA methylation at the Mecp2 REs and expression of Mecp2 isoforms was determined using the Pearson's correlation analysis and linear regression. The Pearson's correlation coefficient (r) was calculated for average methylation against each Mecp2 isoform, over entire regions and for individual CpG sites within each region. The strength of correlation was considered as follows: weak, 0 < r <0.3; moderate, 0.3 < r <0.4; strong,

Results
Dynamic expression of Mecp2 isoforms during NSC differentiation and DNA methylation patterns at the Mecp2 regulatory elements Primary neural stem cells were isolated from the embryonic forebrain at E14.5 and were cultured in the presence of growth factors to generate primary neurospheres. After 7 days in culture, primary neurospheres were dissociated and cultured under differentiation conditions for 8 days, reported to be sufficient for differentiation of both neuronal and glial cells [35,37] (Figure 2A). In agreement with our previous report [37] the proliferating primary neurospheres expressed NESTIN and KI67 ( Figure 2B). Similar to our previous reports [35,37] differentiated NSC at D8 consisted of a mixed population of neurons, astrocytes, and oligodendrocytes ( Figure 2C). The composition of the D8 population was determined by detection of cell typespecific markers (TUB III (4.7% ± 0.8 mean ± standard error of the mean (SEM)), Glial fibrillary acidic protein (GFAP) (54.4% ± 1.1), S100B (15.9% ± 2.7), 2',3'-Cyclic-nucleotide 3'-phosphodiesterase (CNPase) (1.6% ± 0.2), Myelin basic protein (MBP) (2.6% ± 0.42), Oligodendrocyte lineage transcription factor 2 (OLIG2) (36.2% ± 1.8) and KI67 (92.4% ± 1.3)) by immunofluorescence ( Figure 2C). KI67 is not a cell type-specific marker, but rather reflects the fraction of cycling cells within differentiating NSC. Indicating that the cells at an early stage of differentiation   are actively dividing, expression of KI67 (98.8% ± 0.8) was also detected in D2 cells (Additional file 2). Our detection of KI67 in the majority of cells at D8 indicates that although most differentiating NSC are actively dividing, fewer than 10% of cells are post-mitotic and may include neurons (TUB III + ) or non-proliferating cells committed towards neuronal cell fate. Taken together, this in vitro NSC differentiation system provided a suitable model system consisting of the three main neural cell types (neurons, astrocytes and oligodendrocytes) in the brain to study Mecp2/MeCP2 isoforms during neural differentiation.
First, we confirmed MeCP2 expression in neurospheres at D0 and differentiated NSC at D8 by IF studies. We used an antibody that was raised against the MeCP2 Cterminus and recognizes both isoforms. Characteristic punctate nuclear expression of MeCP2 was detected in 41% of neurosphere cells ( Figure 2D, a), which were positive for the NSC marker NESTIN ( Figure 2D, b). At D8 of NSC differentiation, MeCP2 protein was detected in all cell types in the D8 differentiated progenies, including neurons, astrocytes, and oligodendrocytes ( Figure 2E). Indicating the expression of MeCP2 in proliferating cells, we detected MeCP2 in KI67 + cells in the D8 population ( Figure 2E, g). The detected nuclear MeCP2 signals were enriched at the heterochromatin-rich regions of all three cell types. These observations are consistent with our previous reports on MeCP2 nuclear expression in in vivo differentiated primary neurons and astrocytes [7].
Pyrosequencing analysis of R1-R6 indicated that downregulation of Mecp2e1 and upregulation of Mecp2e2 from D0 to D2 were associated with slight, but significant demethylation of Mecp2 promoter R1 (2.3%, P <0.05). Similarly, upregulation of Mecp2e1 and downregulation of Mecp2e2 from D2 to D8 were associated with hypermethylation of Mecp2 intron 1 R5 (2.4%, P <0.01) ( Figure 3B). Detected expression changes in Mecp2 isoforms from D0 to D8 were associated with demethylation of Mecp2 promoter R1 (2.6%, P <0.05), and hypermethylation of Mecp2 intron 1 R4 (4.6%, P <0.05). In all cases, the differences in average percentage methylation between D0, D2 and D8 were relatively small, but statistically significant and ranging between 2 to 5%. Previous reports have shown that an increase in the overall MECP2 promoter methylation by approximately 2.0 to 2.5% in male autistic patients correlates with significantly reduced MECP2 expression levels [2]. In mouse brain exposed to maternal separation and stress, DNA methylation changes that are as little as 2 to 5% at individual CpG sites of the Mecp2 promoter are associated with significantly reduced MeCP2 expression [20]. As even slightly altered MECP2/Mecp2 promoter methylation (2 to 5%) affects MECP/Mecp2 gene expression in the human and mouse brain, it is likely that the statistically significant changes detected in the present study might be biologically important for Mecp2e1 and/or Mecp2e2 expression.
As mentioned, the ratio of Mecp2 splice variants was changed at different stages of NSC differentiation. Therefore, we performed Pearson's correlation analysis between Mecp2e1/Mecp2e2 expression ratio and DNA methylation at R1, R4 and R5 (the three regions that showed significant changes) during NSC differentiation. Pearson's correlation coefficient (r) represents the strength of correlation, with negative r indicating inverse correlation, and positive r indicating direct correlation between DNA methylation and the Mecp2e1/Mecp2e2 splice ratio. We detected a statistically significant positive correlation (r >0.9, P <0.01) between Mecp2e1/Mecp2e2 splice ratio at D2 and DNA methylation at intron 1 R4 (Additional file 3). Although it is possible that intron 1 (R4) may play a role in alternative splicing of Mecp2, further investigations are required to establish the involvement of DNA methylation in Mecp2 transcriptional splicing.
As MECP2/Mecp2 is an X-linked gene, it is possible that the observed changes in Mecp2 expression are due to a shift in the number of cells derived from male and female embryos. To exclude such a possibility, we determined the contribution from the male/female embryonic cells during NSC differentiation using a semiquantitative PCR-based method. Genomic DNA was extracted from each differentiation stage and subjected to PCR analysis for the presence of the Sry gene found in the Y chromosome. The autosomal gene Il3 was used an internal control. The adult male brain cortex was used as a positive control for the presence of the male genomic DNA (Additional file 4: Figure S4A). We did not observe statistically significant changes in the ratio of Sry/Il3 in the cells collected at different stages of differentiation, indicating that the ratio of male/female differentiating NSC were unchanged (Additional file 4: Figure S4B). To further confirm the contribution from the male/female gender, we tested the transcript levels of Xist gene (the gene is involved in X-chromosome inactivation) by qRT-PCR. We did not detect any significant change in Xist gene expression at different stages of NSC differentiation (Additional file 4: Figure S4C). These results indicate that our observed changes in Mecp2 expression are not due to altered contribution of male and female cells. Taken together, our results suggest a possible link between the Mecp2 isoform-specific expression and DNA methylation at the Mecp2 REs within the Mecp2 promoter and intron 1 during NSC differentiation.
Decitabine exposure leads to Mecp2e1 upregulation but its withdrawal downregulates both Mecp2 isoforms to different extents To further study the impact of DNA demethylation/ remethylation in Mecp2 isoform-specific expression, we treated dissociated neurosphere cells with 2.5 μM decitabine for 48 h, at the onset of NSC differentiation (D0) (Figure 4A). At D2, decitabine was withdrawn from the media and cells were kept in culture for another 6 days until D8, to study the effect of DNA remethylation ( Figure 4A). First, as a proof of principle, we verified whether decitabine acted as a DNA demethylating agent in our system. Global change in DNA methylation was determined by IF for 5mC and DNA dot blot assay for both 5mC and 5hmC. As expected, IF experiments showed that 5mC nuclear signals were noticeably lower in decitabine-treated NSC compared to D2 control untreated cells ( Figure 4B). DNA dot blot assays indicated that decitabine treatment resulted in reduced 5mC levels (3.79fold, P <0.05), with slight but statistically insignificant increase in 5hmC levels ( Figure 4C-D). In contrast, decitabine withdrawal led to re-establishment of global DNA methylation (5mC) at D8 as detected by IF ( Figure 4F). Furthermore, DNA dot blot assay showed DNA methylation reprogramming upon decitabine withdrawal, with elevated 5mC levels (1.5-fold, P <0.05), and relatively unchanged 5hmC levels in decitabine-treated NSC compared to controls ( Figure 4G-H). Although globally altered 5mC levels were expected following decitabine treatment in agreement with previous studies [54,55], the observed effect of decitabine to slightly increase 5hmC levels was novel and might be biologically important.
Next, we aimed to study whether the detected changes in Mecp2/MeCP2 expression were due to changes in cell population in response to decitabine treatment. Therefore, we studied the effect of decitabine on cell fate commitment of differentiating NSC at D2 and D8. After decitabine exposure at D2, we examined the expression of cell type-specific markers (neurons: Tub III, NeuN; astrocytes: Gfap, S100b; oligodendrocytes: Cnpase, Mbp) at the transcript levels by qRT-PCR. Comparing the control and decitabine-treated cells, we did not detect any statistically significant change in these cell type-specific genes, except for significant downregulation of Cnpase (9-fold, P <0.01) ( Figure 5I). In order to determine whether any of these detected changes in transcript levels are represented in the number of cells expressing each corresponding cell type-specific marker, we performed IF experiments with specific antibodies against these markers ( Figure 5J). IF experiments showed that there was no significant change in the number of TUB III + , GFAP + , CNPase + , or MBP + cells. However, we did not find any NEUN + , or S100B + cells in the control or decitabine-treated populations at D2, probably because these cells are still in the early stages of differentiation ( Figure 5J). In the D8 population, decitabine treatment led to insignificant changes in the transcript levels for all neuronal and oligodendrocyte markers compared to control untreated cells. Additionally, Gfap expression in decitabine-treated cells was downregulated 5.5-fold, whereas S100b expression was upregulated to a similar extent (6-fold) ( Figure 5K). Quantification of differentiated neurons, astrocytes and oligodendrocytes at D8 by IF did not show any significant change in the cell-fate commitment of these cells ( Figure 5L). However, reduced Gfap expression in decitabine-treated cells without any changes in the number of GFAP + cells might be explained by the reduced intensity of GFAP staining relative to the control astrocytes, since the images were taken at the same exposure level ( Figure 5M, a). Similarly, the significant upregulation of S100b transcript levels by decitabine with no change in the number of S100B + cells could be explained by the increased intensity of S100B in decitabine-treated cells, when the images were taken at the same exposure time ( Figure 5M, b). Taken together, these results suggest that decitabine has minimal effect on the differentiated number of neurons, astrocytes and oligodendrocytes under the described conditions. They further suggest that the detected changes in Mecp2 expression are not likely due to altered population of differentiating cell types.
Next, we investigated whether the observed altered Mecp2/MeCP2 expression was due to changes in the number of cells deriving from male and female embryos. Detection of Sry and Il3 by PCR indicated that the ratio of Sry/Il3 was relatively similar in D2 control and decitabine-treated populations (Additional file 4: Figure  S4D). Similar PCR analysis at D8 also showed no significant differences in the ratio of Sry/Il3 between D8 control and decitabine-treated cells (Additional file 4: Figure S4E). Furthermore, qRT-PCR analysis of the Xist gene expression in both D2 and D8 populations with and without decitabine treatment showed no significant change in Xist transcript expression levels between the control and decitabine-treated cells (Additional file 4: Figure S4F-G). Therefore, these results indicate that the observed changes in Mecp2/MeCP2 expression in response to decitabine exposure and withdrawal are not due to a shift in the number of cells deriving from male/female embryos.
Decitabine mediates altered DNA methylation patterns at the Mecp2 regulatory elements As mentioned earlier, DNA methylation changes at the overall MECP2 promoter and individual CpG sites within the MECP2/Mecp2 promoter are associated with altered MECP2/Mecp2 expression [2,19,20]. Therefore, we investigated whether altered expression of Mecp2 isoforms in our NSC system are associated with change in DNA methylation at the Mecp2 REs found within the Mecp2 promoter and intron 1. Bisulfite pyrosequencing analysis showed that decitabine treatment (D2) caused no significant change in the percentage DNA methylation at the Mecp2 promoter R1 and R3 ( Figure 6A, a, c). However, decitabine caused demethylation of all individual CpG dinucleotides at the R2 (CpG1, 3.5%; CpG2, 4.4%; CpG3, 3.1%; CpG4, 4.28%) ( Figure 6A, b), as well as the average R2 percentage DNA methylation (3.83%, P <0.05) ( Figure 6B, a). Similarly, decitabine caused demethylation of individual CpG sites at the intron 1 R4 (15.8%, P <0.05), R5 (13.08%, P <0.05), and R6 (CpG1, 8.01%, P <0.01; CpG2, 3.8%, P = 0.4) ( Figure 6A, d-f), with significant demethylation at the entire intron 1 (10.37%, P <0.05) ( Figure 6B, b). These results indicated that decitabine induced significant DNA demethylation at both the Mecp2 promoter and intron 1, at the individual CpG sites and the overall DNA methylation. As mentioned earlier, this detected DNA demethylation was associated with significant upregulation of Mecp2e1 isoform, but not Mecp2e2. Therefore, it is possible that the observed changes in DNA methylation at the studied REs contribute to the upregulation of Mecp2e1.
Similar bisulfite pyrosequencing analysis at D8 indicated that the three Mecp2 promoter regions (R1 to R3) and intron 1 regions (R4 to R6) were remethylated and DNA methylation was almost re-established following decitabine withdrawal ( Figure 6C). Analyzing the average DNA methylation over the Mecp2 promoter R2 and the entire intron 1 (which were demethylated at D2), we observed no significant differences in DNA methylation between D8 control and decitabine-treated cells ( Figure 6D). Despite the fact that DNA remethylation is expected to restore the gene expression levels, expression of both Mecp2 isoforms were significantly downregulated. This observation implies that at D8, other regulatory mechanisms apart from promoter/intron 1 DNA methylation might be involved in downregulating Mecp2 expression.
Taken together, these results show that the induced Mecp2e1 (but not Mecp2e2) expression is associated with reduced DNA methylation at the Mecp2 REs and decreased global 5mC DNA methylation. Hence, our findings implicate the possible role of Mecp2 genespecific DNA demethylation at the specific REs on the expression of Mecp2e1/MeCP2E1, and MeCP2 (total) at D2. Moreover, altered expression of Mecp2 isoforms without any change in DNA methylation at the Mecp2 REs at D8 imply that mechanisms other than DNA methylation could be involved in downregulating Mecp2 isoforms.
Mecp2 isoform-specific expression correlates with DNA methylation at the Mecp2 regulatory elements In order to establish a link between Mecp2 isoformspecific expression and DNA methylation, we performed Pearson's correlation analysis by comparing normalized (log2) expression of Mecp2 in each dataset to the respective average percentage methylation levels over entire regions, as well as methylation at individual CpG sites (from both control and decitabine-treated cells).
First, we tested whether average DNA methylation of the entire Mecp2 promoter (R1 to R3), and intron 1 (R4 to R6) regions correlate with Mecp2e1 and Mecp2e2 expression at D2 in control and decitabine-treated cells. We observed a significant negative correlation between Mecp2e1 expression and the average methylation at R1, R3 and R5 (r > −0.9, P <0.05). Correlation between Mecp2e1 upregulation and significant demethylation at R5, induced by decitabine at D2, suggests a possible contribution of R5 in upregulating Mecp2e1. On the other hand, Mecp2e2 showed significant negative correlation only with R3 methylation (r > −0.9, P <0.05) ( Figure 7A, a), that remained unchanged at D2, and this could explain the unaffected Mecp2e2 expression at D2.
Similar correlation analysis at D8 (in control D8 and decitabine-treated cells), indicated that Mecp2e1 shows a significant negative correlation with average DNA methylation at the promoter R1 (r > −0.7, P <0.05), R2 (r > −0.8, P <0.05) and R3 (r > −0.7, P = 0.06, close to significant) and a significant positive correlation with the average DNA methylation at the intron 1 R6 (r >0.9, P <0.001) ( Figure 7A, b). In contrast, Mecp2e2 did not show any significant correlation with any of the promoter regions (R1 to R3), but showed a significant positive correlation with the average methylation at intron 1 R4 (r >0.9, P <0.05) and R6 (r >0.9, P <0.01). This divergence in the correlation patterns (negative and positive depending on the stage of differentiation), might imply a potential dynamic role of DNA methylation in regulating Mecp2 isoforms at different stages of NSC differentiation.
Taken together, these results show a strong (r > 0.8, P <0.05) and dynamic (positive or negative) relationship between DNA methylation at the Mecp2 REs and expression of Mecp2 isoforms depending on the different stages of NSC differentiation. Therefore, these results implicate a possible dynamic role of DNA methylation at the Mecp2 REs in regulating Mecp2 isoform-specific expression.

Discussion
In the brain, precisely controlled MECP2/MeCP2 transcript and protein expression levels are critical, as even slightly altered expression is associated with severe neurological symptoms [2,16,[57][58][59][60]. However, so far little is known about how MeCP2 expression is regulated in the developing brain. MeCP2 is a major epigenetic regulator in brain, and its reduced expression in the autistic brain is associated with MECP2 promoter hypermethylation [2]. Surprisingly, the role of DNA methylation in MeCP2 expression during brain development is unclear. Currently, most diseases that are associated with aberrant MeCP2 function or expression deficits, including autism and Rett syndrome, have no cure or effective treatment. This underscores an urgent need for investigating how MeCP2 expression is regulated in the brain. Such knowledge for addressing this gap is essential for designing possible future therapeutic strategies. DNA methylation is a reversible epigenetic modification [22], which can be targeted by existing Food and Drug Administration (FDA)-approved drugs, including decitabine, which is suggested for use in autism [61,62]. Therefore, investigating the effect of such epigenetic drugs on MeCP2 expression is important. Therapeutic approaches such as gene therapy or restoring MeCP2 expression by genetic engineering have been suggested as possible therapeutic strategies for MeCP2-associated disorders [14,15,35]. However, even mild MeCP2 overexpression can lead to severe neurological complications, highlighting the importance of understanding MeCP2 regulatory mechanisms. Since both MeCP2 isoforms have been implicated in severe neurological disorders, investigating MeCP2 regulation is equally important for individual isoforms. This present study is the first report on the potential role of DNA methylation at the Mecp2 REs and the impact on the expression of Mecp2 isoforms.
We observed globally altered DNA methylation upon decitabine exposure and withdrawal. Since DNA methylation is a major epigenetic mechanism that is involved in modulating gene expression and chromatin architecture [22], these observed changes in 5mC levels may possibly lead to altered chromatin structure and genome-wide changes in gene expression. Furthermore, the presented findings highlight that exposure to drugs that disturb the epigenetic marks during differentiation of brain cells may lead to aberrant DNA methylation profiles. Our observations at D8 indicate that, even after the disturbance factor is withdrawn from the system, an epigenetic memory for this disturbance may be associated throughout cellular differentiation of brain cells. Thus, our findings highlight the biological importance of maintaining proper regulation of epigenetic factors/modifications during brain development with a clear focus on DNA methylation and MeCP2.
Our results show that decitabine alters Mecp2/MeCP2 expression at both the transcript and protein levels. Importantly, even minor changes in Mecp2 transcript expression led to nearly 2-to 3-fold altered protein expression, highlighting the biological significance of proper regulation of Mecp2 expression at the transcript levels. The observed correlation between the Mecp2/ MeCP2 (total) and Mecp2e1/MeCP2E1 transcript/protein expression at D2 reinforces the concept that potential changes in Mecp2 transcript levels may reflect possible changes at the protein levels. However, the noncorrelated Mecp2/MeCP2 (total) and Mecp2e1/MeCP2E1 transcript/protein expression at D8 indicates that decitabine withdrawal causes not only transcriptional but also, post-transcriptional regulation of MeCP2 expression, leading to reduced expression of MeCP2 (total)/MeCP2E1. One such post-transcriptional regulatory mechanism (See figure on previous page.) Figure 7 Correlation analysis between DNA methylation at the Methyl CpG binding protein 2 gene (Mecp2) regulatory elements and Mecp2 expression after decitabine treatment (day 2) and decitabine withdrawal (day 8). All graphs represent the Pearson's correlation coefficient (r) for Mecp2e1 (black), and Mecp2e2 (pink): statistical significance: *** P <0.001; ** P <0.01; * P <0.05; n = 3. could be the action of micro-RNAs such as miR132, expression of which has been shown to be increased by 5aza-2′-deoxycytidine/decitabine [63], and has the ability to repress MeCP2 expression [64].
Increased promoter methylation of autistic candidate genes such as RORA, BCL2 and MECP2 are shown to be associated with reduced expression of these genes in autistic patients [2,19,62]. Treatment with decitabine was shown to demethylate promoters and restore/induce the expression of the silenced RORA and BCL2 in autistic and patients with fragile X syndrome and hence, the use of DNA demethylating agents in drug therapy for autism and fragile X syndrome has been suggested [61,62]. A similar strategy to restore/induce MeCP2 expression might be extended to treat such diseases associated with reduced MeCP2 expression, including autism and RTT. Schematic representation of the correlation between Mecp2 isoform-specific expression and DNA methylation at the Mecp2 promoter regions (R1 to R3), and intron 1 regions (R4 to R6). The size of the signs, plus (+), and minus (−) represents the relative degree of correlation with either Mecp2 isoform. The statistically significant correlations are represented in red (red-circled plus, red-circled minus). After decitabine exposure Mecp2e1 expression negatively correlated with promoter R1 and R3, and intron 1 R5. Mecp2e2 isoform negatively correlates with promoter R3. In contrast, after decitabine withdrawal, Mecp2e1 expression negatively correlated with promoter R1, R2 and R3, and positively correlated with intron 1 R6. Hence, correlation between Mecp2e1 and DNA methylation at REs changed depending on the stage of NSC differentiation. Mecp2e2 isoform positively correlated with intron 1 R4 and R6.
Providing insights on such therapeutic strategies, the application of epigenetic drug therapy to induce nonmutated copy of MECP2 expression in Rett syndrome cell lines has been suggested and attempted previously [65]. Therefore, our findings on the ability of decitabine to induce MeCP2 expression in differentiating NSC provide further insights on designing possible drug therapies for autism. Even though the exposure of RTT cell lines (fibroblasts) to lower doses of decitabine for a longer period did not activate MECP2 expression [65], our results indicate that moderate dose of decitabine can induce Mecp2/MeCP2 expression within a shorter period. However, inhibition of MeCP2 by withdrawal of decitabine as well as other observed changes in DNA methyl marks implies that such drug therapy should be administrated with great caution.
Our findings on the changes in DNA methylation at the Mecp2 REs are in agreement with the previous reports on MECP2 promoter methylation, which demonstrate that an approximate difference of 2.0 to 2.5% overall methylation over a region −233 to −531 upstream of the MECP2 promoter is correlated with reduced MECP2 expression in autistic male brains. The authors report that within the 15 CpG sites found in this MECP2 promoter region, two CpG sites are specifically altered in the autistic males [2]. Furthermore, our results are in agreement with a previous report on significantly reduced MeCP2 expression in the postnatal mouse brain (under stress), which is associated with 2 to 5% increased methylation at the individual CpG sites within a 164-bp region of the Mecp2 promoter [20]. Supporting these observations, studies have also shown minor differences, such as 2 to 5% DNA methylation causing significant changes in the expression of other genes, such as RASSF1, in the human brain [66], AMOTL2 in the human heart [67], and PGC1α in human muscles [68]. Therefore, although the detected DNA methylation changes in this current study are not considerably high (they varied between 2 to 15%), they were statistically significant for average DNA methylation (within R1, R3 and R5) during NSC differentiation, and for several specific CpG dinucleotides subsequent to decitabine treatment (within R2, R4, R5, and R6), and are likely to be biologically important.
The Mecp2 promoter CpG island studied by Franklin et al., [20] overlaps with the R1 and R2 of the Mecp2 promoter that we studied here. The significantly methylated CpGs reported in their study coincides with the R2 CpGs, where we observed changes at individual CpG sites as well as average methylation upon decitabine treatment. However, in our study we did not see any significant change in the R1 CpG sites (both D2 and D8), where Franklin et al., reported DNA methylation changes. Importantly, the results we obtained for one of the promoter regions studied (R2) are in agreement with this previous report, which showed a biological and functional importance of the methylation changes in regulating MeCP2 expression in response to stress in vivo. Therefore, it is likely that the detected changes we observed in the Mecp2 REs in our study also have biological importance. The hypermethylation of this R2 region in mouse brain was associated with MeCP2 downregulation [20], and hence it is possible that the hypomethylation/demethylation of the same R2 region causes Mecp2/MeCP2 upregulation.
Our results on the ability of 2.5 μM decitabine to upregulate Mecp2e1 (but not Mecp2e2) suggest that the two isoforms may have different sensitivities to drugs/ chemicals. This observation is in agreement with the previous report on the higher sensitivity of Mecp2e1 than Mecp2e2 to Bisphenol A [69]. These observations further suggest that the differential sensitivity to drugs might be used to specifically induce only one Mecp2 isoform. This is also important because overexpression of Mecp2e2, but not Mecp2e1 causes neuronal cell death [10]. Hence, our study provides a functional relevance of DNA demethylation at the Mecp2 REs by decitabine causing upregulation of Mecp2e1, but not Mecp2e2.
The observed negative correlation between the expression of both Mecp2 isoforms and Mecp2 promoter elements are novel and are in accordance with previous correlation studies on the human MECP2 expression and promoter DNA methylation [2,19]. Furthermore, our study is novel in demonstrating a dynamic (positive/ negative) correlation between the intronic DNA methylation and expression of Mecp2 isoforms in differentiating brain cells. It is possible that the promoter regions analyzed in our study (which also overlap with the core Mecp2 promoter [13]) might be shared by both Mecp2 isoforms, whereas depending on the stage of neural differentiation, intron 1 regions may add another layer of regulation for Mecp2 isoform-specific expression. Supporting our findings, the role of intronic DNA methylation in regulating gene expression of other genes has been previously reported [70,71]. Several other reports also show evidence that gene expression negatively correlates with promoter methylation and positively correlates with gene-body methylation [67,72].
Intronic DNA methylation is reported to be involved in regulating alternative splicing [27,28]. Although, it is known that Mecp2 isoforms are generated by alternative splicing [4,5], the underlying molecular mechanisms are still unclear. We observed that the expression ratio of Mecp2e1/Mecp2e2 changed during NSC differentiation. The observed correlation between the splice ratio and intron 1 R4 DNA methylation in differentiating NSC at D2 would provide insights on the potential importance of this region in Mecp2 alternative splicing.
The intron 1 regions analyzed in this study were designated as part of a silencer element, which has been previously proposed to regulate MECP2 alternative splicing and tissue-specific expression [12]. Our findings are in agreement with possible involvement of these regions in Mecp2 isoform-specific expression. Although the link between DNA methylation and Mecp2 expression is supported by our results in the NSC system, the contribution of other epigenetic modifications such as histone acetylation and histone methylation should not be excluded [73,74].

Conclusion
The summary of the findings presented in our study is illustrated in Figure 8. First, expression of Mecp2 isoforms was significantly and reciprocally changed at different stages of NSC differentiation, in association with minor but significant changes in DNA methylation at selected Mecp2 REs, suggesting possible involvement of these regions in Mecp2 regulation. Second, treatment of differentiating NSC with decitabine for 48 h led to demethylation of specific Mecp2 REs (promoter R2 and all intron 1 regions) and subsequent upregulation of Mecp2e1/MeCP2E1 (but not Mecp2e2), implying the differential sensitivity of the two Mecp2 isoforms to decitabine. Such differential sensitivity of Mecp2 isoforms to decitabine might be useful in future drug therapies to specifically activate one isoform but not the other. Furthermore, the ability of decitabine to induce Mecp2e1/ MeCP2E1 at both transcript and protein levels provide insights for future therapeutic strategies for MeCP2 deficiency-related neurodevelopmental disorders such as autism and Rett syndrome. Finally, the significant and dynamic (positive or negative) correlation between the expression of Mecp2 isoforms and DNA methylation implies the potential contribution of these REs in regulating Mecp2 isoforms at different stages of neural differentiation. Collectively, our study contributes to the understanding of expression and regulation of Mecp2 isoforms during neural development and provides important insights for future therapeutic applications of decitabine for MeCP2-related neurological disorders.