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.

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 10⁵ cells/cm² 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 10⁵ cells/cm² 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.

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.

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.

RNeasy Mini Kit (Qiagen, Ontario, Toronto, Canada) was used for RNA extraction as per the manufacturer’s protocol. Preparation of cDNA and qRT-PCR were carried out as described previously 40 41 42 . Transcript levels of Mecp2 (total), Mecp2e1 [NCBI: NM_001081979.1], Mecp2e2 [NCBI: NM_010788.3], Dnmt genes (Dnmt1, Dnmt3a and Dnmt3b), neuronal genes (Tubulin III (Tub III), NeuN), astrocytic genes (Gfap, S100b), and oligodendrocyte-specific genes (Cnpase, Mbp) were examined by using gene-specific primers (Table 2), as described previously 37 43 . The relative expression and fold changes were calculated as described previously 37 . Two-way analysis of variance (ANOVA) and the Student t-test were used to calculate significant differences between untreated control and decitabine-treated samples.

Immunofluorescence (IF) experiments were performed according to previously described protocols 7 35 37 . Primary and secondary antibodies used for IF are listed in the Tables 3 and 4, respectively. Immunofluorescence signals were detected by an Axio Observer Z1 inverted microscope and LSM710 Confocal microscope from Carl Zeiss. Images were obtained with AxioVision 4.8 (Carl Zeiss Canada Ltd. Ontario, Toronto, Canada) and Zen 2009 software and assembled using Adobe Photoshop CS5 and Adobe Illustrator CS5. For quantification analysis in neurospheres, three neurospheres were randomly selected and all of the cells within each neurosphere were counted based on 4',6-diamidino-2-phenylindole (DAPI) staining. For cell quantification of differentiating cells at D2 and D8, 8 to 10 random fields were selected under the microscope. Approximately 250 cells from the D2 population and 750 cells from the D8 population were counted based on DAPI labeling. The cell counting was done using the ImageJ program.

MeCP2, Methyl CpG binding protein; GFAP, Glial fibrillary acidic protein; OLIG2, Oligodendrocyte lineage transcription factor 2; CNPase, 2',3'-Cyclic-nucleotide 3'-phosphodiesterase, MBP, Myelin basic protein; NEUN, NEUronal Nuclei; 5mC, 5-methylcytosine; 5hmc, 5-hydroxymethylcytosine; IF, Immunofluorescence; WB, Western blot.

IF, Immunofluorescence; WB, Western blot.

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.

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.

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.

F: Forward PCR primer, R: Reverse PCR primer (Biotinylated), S: Sequencing primer.

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, 0.4 < r <0.7; or very strong, 0.7 < r <1.0. A negative r-value indicates an inverse/negative correlation whereas a positive r-value indicates direct/positive correlation. Statistical significance was determined at P <0.05.

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 type-specific 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 C-terminus 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 .

Next, we investigated Mecp2 isoform-specific transcript expression at three stages of NSC differentiation: undifferentiated cells (D0), cells at an early stage of differentiation (D2), and cells at a later stage of differentiation (D8). Distinct and mirror-like (reciprocal) transcript expression patterns for Mecp2 isoforms were observed during NSC differentiation (D0, D2, and D8) (Figure 3A), suggesting possible differential regulation of these isoforms during NSC differentiation. Expression of Mecp2e1 was reduced from D0 to D2 (2.7-fold, P <0.001) and was slightly elevated from D2 to D8. In contrast, Mecp2e2 expression increased from D0 to D2 (3.1-fold, P <0.05), but declined from D2 to D8 (4.2-fold, P <0.01). At each of these studied differentiation stages, the expression ratio between the two isoforms (Mecp2e1/Mecp2e2) varied significantly (D0, 5.99; D2, 0.69; D8, 4.62) (Additional file 3). At D0 and D8, Mecp2e1 expression was significantly higher than Mecp2e2 (D0, P <0.01, and D8, P <0.05). In contrast at D2, Mecp2e2 expression was significantly higher than Mecp2e1 (D2, P <0.05). These observations imply differential regulation of Mecp2 isoforms and possible changes in alternative splicing of Mecp2 at different stages of NSC differentiation.

MECP2/Mecp2 expression is known to be regulated by REs found within its promoter and intron 1. The altered MeCP2 expression in autistic patients and in mouse brains subjected to stress is correlated with increased MECP2/Mecp2 promoter methylation 2 12 13 18 19 20 . Therefore, we hypothesized that DNA methylation at the Mecp2 promoter and intron 1 might impact Mecp2 isoform-specific expression. For DNA methylation analysis by bisulfite pyrosequencing, we selected three regions within the Mecp2 promoter (named R1 to R3, hereafter) and three regions within the intron 1 (named R4 to R6, hereafter) (Figure 1B). These regions harbored different numbers of CpG sites; promoter regions R1: CpG island contains 13 CpG sites; R2: 4 CpG sites; R3: 2 CpG sites, and intron 1 regions R4: 1 CpG site; R5: 1 CpG site; and R6: 2 CpG sites.

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.

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 (3.79-fold, 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.

DNA demethylating agents can function as cytosine analogues and/or as Dnmt/DNMT inhibitors 56 . Therefore, we investigated Dnmt expression levels in decitabine-treated differentiating NSC by qRT-PCR. In accordance with reduced DNA methylation levels at D2, decitabine treatment caused significant inhibition of transcript levels of all three DNA methyltransferases (Dnmt1, 1.7-fold, P <0.05; Dnmt3a, 1.5-fold, P <0.05 and Dnmt3b, 2.5-fold, P <0.01) (Figure 4E). Even though we anticipated that decitabine withdrawal would restore Dnmt levels, only Dnmt1 levels were elevated (2.2-fold, P <0.05), whereas both Dnmt3a (1.4-fold, P <0.05) and Dnmt3b (1.8-fold, P = 0.06) levels remained inhibited (Figure 4I). In summary, these results indicate that decitabine functions as a DNA demethylating agent in differentiating NSC and globally affects DNA methyl marks. Additionally, our data indicate that decitabine withdrawal would lead to DNA methylation reprogramming in differentiating NSC.

Next, we investigated possible changes in Mecp2/MeCP2 expression induced by decitabine. Quantitative RT-PCR experiments indicated that decitabine treatment at D2 caused slight but statistically significant upregulation of Mecp2e1 (1.41-fold, P <0.05), with minimal and insignificant increased levels of the total Mecp2 (1.2-fold, P = 0.5), and unchanged levels of Mecp2e2 (Figure 5A). Analysis of protein levels by WB showed that decitabine upregulated total MeCP2 (2.5-fold, P <0.05), and MeCP2E1 (3.1-fold, P <0.05) protein expression (Figure 5B-C). However, the lack of an antibody specific for MeCP2E2 limited our investigation of MeCP2E2 protein levels. Correlation analysis of transcript and protein levels of Mecp2/MeCP2 at D2 indicated significant correlation between the detected transcript and protein expression (Mecp2/MeCP2 (r = 0.97, P <0.05) and Mecp2e1/MeCP2E1 (r = 0.98, P <0.01); Figure 5G).

In contrast to D2, withdrawal of decitabine at D8 significantly downregulated the transcript expression levels of Mecp2e1 (1.92-fold, P <0.001), Mecp2e2 (1.39-fold, P <0.05) and the total Mecp2 (1.52-fold, P <0.01) (Figure 5D). Similar to the transcript levels, decitabine withdrawal resulted in downregulation of total MeCP2 (3.2-fold, P <0.001), and MeCP2E1 (4.3-fold, P <0.05) protein expression levels (Figure 5E-F). A similar correlation analysis between transcript and protein levels of Mecp2/MeCP2 at D8 did not show any statistically significant correlation (Mecp2/MeCP2 (r = 0.6, P = 0.2), and Mecp2e1/MeCP2E1 (r = 0.6, P = 0.3); Figure 5H). These observations emphasize that even minor change in Mecp2 transcript levels are biologically important and can result in significantly altered MeCP2 protein expression levels.

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.

Taken together, our data so far indicate that a single administration of decitabine for 48 h induces Mecp2e1/MeCP2E1, MeCP2 (total) expression, whereas its withdrawal downregulates Mecp2 (total)/MeCP2 (total), Mecp2e1/MeCP2E1, and Mecp2e2 expression with minimal change in NSC differentiation.

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 gene-specific 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.

In order to establish a link between Mecp2 isoform-specific 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.

Last, we investigated whether individual CpG sites within the studied regions (R1 to R6) showed specific correlation with either Mecp2 isoform (Figure 7B-C). Implicating the possible role of promoter R1 and R2 in mainly regulating Mecp2e1 (major isoform) at both D2 and D8, we observed a negative correlation between CpG methylation and Mecp2e1 expression at several CpG sites (r > -0.8, P <0.05) (Figure 7B, a, and 7C, a). At D2, unlike other REs, the average methylation over R3 showed an equally strong negative correlation with both Mecp2e1 (r = -0.94, P <0.05) and Mecp2e2 (r = -0.98, P <0.01) (Figure 7A, a). Therefore, we studied the two individual CpG sites located within R3 far apart from each other, which were differentially methylated (CpG1, approximately 10% and CpG2, approximately 30%) (Figure 6A, c). Interestingly, CpG1 showed a significant negative correlation with Mecp2e1 (r = -0.9, P <0.05), while CpG2 showed a significant negative correlation with Mecp2e2 (r = -0.9, P <0.01) (Figure 7B, c). Further confirming the potential role of these two CpG sites within R3 in Mecp2 isoform-specific expression, a similar correlation (CpG1: Mecp2e1, r = -0.85, P <0.05; CpG2: Mecp2e2, r = -0.87, P <0.05) was observed at D8 (Figure 7C, c). The studied intron 1 regions seemed to have preferential correlation with individual isoforms. For instance, at D2, the only CpG site within R5 showed negative, significant correlation with Mecp2e1 (r > -0.8, P <0.05) (Figure 7B, e). Interestingly, at D8 intron 1 R4 and R6 showed positive, significant correlation with the Mecp2e2 isoform (r >0.8, P <0.05) (Figure 7C, d-f). The observed correlations for all REs are represented in Figure 7B-C and are summarized in Figure 8.

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.