Differentiation of multipotent neural stem cells derived from Rett syndrome patients is biased toward the astrocytic lineage

Mosaic expression patterns in fibroblasts procured from RTT-MZ twins

We recently established fibroblast cell lines from the RTT-MZ twins (patients RS1
and RS2). These RTT twins have a de novo frame-shift mutation in exon 4 (c.806delG) that truncates the MeCP2 protein within
the transcriptional repression domain (Fig. 1A). We also reported that fibroblasts generated from both patients exhibited random
XCI patterns 27], which were detected by the methylation-specific polymerase chain reaction (PCR)-based
HUMARA (human androgen receptor) XCI assay 28]. To examine the expression patterns of MeCP2 in RS1 and RS2 fibroblasts, immunostaining
was performed with a specific primary antibody against MeCP2. Consequently, the fibroblast
lines derived from the RTT-MZ twins included both MeCP2-positive and MeCP2-negative
cells (Fig. 1B). Such mosaic expression patterns for the MeCP2 suggests that the fibroblasts comprise
MeCP2-positive cells with the X chromosome harboring wild-type MECP2 as the active MeCP2 species, and MeCP2-negative cells with the X chromosome harboring
mutant MECP2 as the active MeCP2 species. The fractions of MeCP2-positive cells among the RS1
and RS2 fibroblasts were 0.64 and 0.60, respectively (Fig. 1C).

Fig. 1. MECP2 mutation in MZ twins with RTT and MeCP2 expression pattern in RTT fibroblasts. (A) Schematic representation of MECP2 gene structure and location of the MECP2 mutation. Direct sequencing of the four coding exons in the MECP2 gene detected a guanine deletion at position 806 (806delG) 27]. (B) Immunostaining for MeCP2 (red) and phalloidin (green) along with Hoechst staining
(blue) of wild-type MECP2- and mutant MECP2-expressing fibroblasts. Scale bar, 100 ?m. (C) Fraction of MeCP2-positive cells among wild-type MECP2- and mutant MECP2-expressing fibroblasts (n?=?5 experiments; 500 Hoechst-positive cells per experiment;
*p??0.05). WT, wild-type

Generation and characterization of RTT-MZ hiPSC lines

We utilized standard methods and transduction of OCT4-, SOX2-, KLF4-, and c-MYC-containing retroviruses to reprogram fibroblasts derived from the RTT-MZ twins into
RTT-hiPSCs. Each RTT-hiPSC clone was clonally isolated and selected by morphological
criteria and transgene silencing. We also verified pluripotency of the stem cells
by immunostaining of the hiPSCs with primary antibodies against pluripotency markers
(NANOG, OCT4, and TRA-1-81) (Fig. 2A and Additional file 1A).

Fig. 2. hiPSC lines and XCI patterns derived from the RTT-MZ twins. (A) MeCP2 (red) and OCT4 (green) expression in hiPSC lines derived from the RTT-MZ twins.
Scale bar, 150 ?m. (B) XCI patterns in the four hiPSC lines, as assessed by the methylation-specific PCR-based
HUMARA assay. Xi, X inactivation pattern based on the inactive X chromosome; Xa, X
inactivation pattern based on the active X chromosome; Xm, X chromosome inherited
from the mother; Xp, X chromosome inherited from the father

To evaluate in vivo pluripotency, we injected the RTT-hiPSCs into the testes of immunodeficient mice,
and confirmed the formation of teratomas containing derivatives of all three embryonic
germ layers (Additional file 1B). No abnormalities were found in the karyotypes of any of the hiPSC lines (Additional
file 1C). Notably, most of the selected hiPSC clones were either all MeCP2-positive or MeCP2-negative,
and putatively originated from a single MeCP2-positive or MeCP2-negative fibroblast
(Fig. 2A). Therefore, we isolated both wild-type MECP2-expressing (RS1-52 M and RS2-65 M) and mutant MECP2-expressing (RS1-61P and RS2-62P) clones from each patient (RS1 and RS2) to generate
four hiPSC lines.

We next analyzed the XCI patterns in the hiPSC lines, and found that the maternal
X chromosome was active in the MeCP2-positive (RS1-52 M and RS2-65 M) clones, while
the paternal X chromosome was active in the MeCP2-negative (RS1-61P and RS2-62P) clones
(Fig. 2B). Although we isolated several clones that partially include iPS cells with two active
X chromosomes, we only used clones with an inactive X chromosome (i.e., the “standard”
XCI status of undifferentiated hiPSCs 29]) in the present study.

Our previous study 27] revealed that the maternally-derived X chromosome carries the wild-type MECP2 allele, whereas the paternally-derived X chromosome carries the mutant MECP2 allele. These results were shown by sequencing the MECP2 gene in somatic hybrid cell clones carrying either the maternal or the paternal X
chromosome of the RTT-MZ twins. Accordingly, the RS1-52 M and RS2-65 M hiPSC lines,
in which maternal wild-type MECP2 was preferentially active, exhibited MeCP2 expression in the nuclei, whereas the
RS1-61P and RS2-62P hiPSC lines, in which paternal mutant MECP2 was preferentially active, did not (Fig. 2A).

MeCP2 expression in neural cells differentiated from RTT-hiPSCs

We next differentiated the four RTT-hiPSC lines into neural cells and examined MeCP2
expression in the progeny. Immunostaining revealed that all of the cells derived from
the RS1-52 M and RS2-65 M hiPSC lines (containing preferentially active maternal wild-type
MECP2) expressed MeCP2 in the nucleus, like the parental hiPSCs. In particular, MAP2-positive
neurons expressed MeCP2 more strongly than glial fibrillary acidic protein (GFAP)-positive
astrocytes (Fig. 3A) as previously shown in mice 30],31]. However, no MeCP2 expression was found in the progeny of the RS1-61P and RS2-62P
hiPSC lines (containing preferentially active paternal mutant MECP2), or in the nucleus of RS1-61P/RS2-62P hiPSC-derived MAP2-positive neurons or GFAP-positive
astrocytes (Fig. 3A). Therefore, neural cells differentiated from the RTT-hiPSC lines apparently retain
the XCI status of the undifferentiated cells.

Fig. 3. Immunostaining of neural cells derived from RTT-hiPSCs and gene expression analysis
of hiPSC-derived neural cells. (A) Immunostaining was performed to evaluate expression levels of MeCP2 (red), the neuronal
marker, MAP2 (green), and the astrocytic marker, GFAP (magenta) in RTT-hiPSC-derived
neural cells. Ho, Hoechst (blue). Arrowhead indicated MAP2, MeCP2 and Hoechst positive
cells. Scale bar, 50??m. (B) Results of PCA (performed by using MeV software (TIGR) Software) of microarray gene
expression in RTT-hiPSC lines and neural cells differentiated from RTT-hiPSC lines.
N, neural cells differentiated from hiPSC lines. (C) Scatter plots of microarray gene expression in hiPSC-RTT lines and neural cells differentiated
from hiPSC-RTT lines. Neural cells were co-cultured for ~30 days with mouse astrocytes

Global gene expression in RTT-hiPSCs and differentiated neural cells

We next investigated whether the loss of MeCP2 protein affects the transcriptional
network in undifferentiated hiPSCs and neural progeny differentiated from RTT-hiPSCs.
Global gene expression levels in wild-type MECP2– and mutant MECP2-expressing cells were examined by comparative microarray analyses of undifferentiated
hiPSCs and differentiated neural cells. The gene expression data were normalized and
subjected to principal component analysis (PCA) and hierarchical clustering, with
the exception of those genes with low expression in all samples.

The samples were clustered into hiPSC or neural cell (N) groups depending on the MECP2 expression pattern. In the PCA analysis, the original hiPSCs were clustered tightly
into one region, signifying few differences from one hiPSC clone to the next. On the
other hand, the neural cells were roughly clustered into two groups that reflected
the status of the activated X chromosome (Fig. 3B and Additional file 2). Next, we constructed scatter plots to compare wild-type MECP2– and mutant MECP2-expressing cells. In the undifferentiated hiPSCs, the correlation coefficient (R²) was??0.99 for any pair of hiPSC clones. However, in the differentiated neural cells,
higher correlations (R²??0.98) were only found in comparisons of hiPSCs with the same XCI pattern (i.e.,
paternal X vs. paternal X or maternal X vs. maternal X), while lower correlations
(R²??0.98) were only found in comparisons of hiPSCs with different XCI patterns (paternal
X vs. maternal X) (Fig. 3C). The clustering dendrogram (Additional file 3A) also showed that global gene expressions for all four RTT-hiPSC clones during the
undifferentiated stage. After this time, mutant MECP2-expressing hiPSCs became distinguishable from wild-type MECP2-expressing hiPSCs due to neural differentiation. Nevertheless, Tanaka et al. 25] recently characterized five patient-speific cell lines with different mutations to
show that hiPSCs with MECP2 gene mutations are distinguishable from normal hiPSCs and ESCs even during the undifferentiated
stage. By contrast, our isogenic cell lines might largely exclude the effect of donor
divergence, permitting the elucidation of MECP2 function during neural differentiation. While Tanaka and colleagues 25] demonstrated that the gene expression of a key mitochondrial transcription factor,
NR3C1, was upregulated in mutant MECP2-expressing cells after neuronal differentiation, we failed to detect such a difference
in the current study (Additional file 3B). These results suggest that MeCP2 plays a more important role as a transcriptional
regulator in differentiated neural cells than in undifferentiated hiPSCs.

Enhanced astrocytic differentiation of mutant MECP2-expressing neural stem cells derived from RTT-hiPSCs

Previous studies revealed that MeCP2 is involved in the regulation of astroglial gene
expression 32]-34]. Gfap and S100? are expressed at significantly higher levels in astrocytes derived from MeCP2-null
mouse embryonic stem cells than in those derived from wild type mouse embryonic stem
cells 35],36]. Moreover, a truncated form of RTT-associated MeCP2 (R168X) is reported to be unable
to promote neuronal differentiation in mice, and instead promotes an abnormally high
degree of astrocytic differentiation 37]. We characterized neurospheres derived from the four RTT-hiPSC lines (RS1-61P, RS2-62P,
RS1-52 M, and RS2-65 M), and found no significant differences in neurosphere number,
size, or expression of neural stem cell markers between the two groups derived from
mutant vs. wild-type MECP2-expressing hiPSCs (Additional file 4). Next, the neurospheres were differentiated into neural cells by using an adhesion
culture method without fibroblast growth factor-2 (FGF-2) at ~30 days in vitro (Fig. 4A). The differentiated cells were classified by their expression of ?III-tubulin and
GFAP. Immunocytochemical analysis revealed that the neural cells originating from
MeCP2-negative hiPSC lines (RS1-61P and RS2-62P) contained significantly higher proportions
of GFAP-positive cells than those originating from MeCP2-positive hiPSC lines (RS1-52 M
and RS2-65 M) (Fig. 4B). Most of the GFAP-positive cells failed to express neuronal markers; therefore,
we concluded that these cells were in fact astrocytes.

Fig. 4. Comparison of acquired neuronal and astrocytic properties between MeCP2-positive and
negative hiPSC-derived neural cells. (A) Immunostaining images of neural cells. Double labeling for GFAP (red) and ?III-tubulin
(green) is shown along with Hoechst staining (Ho, blue) in RS1-52 M, RS1-61P, RS2-65 M,
and RS2-62P RTT-hiPSC-derived neurons. Scale bar, 200??m. (B) Fraction of GFAP-positive cells relative to Hoechst-positive cells in neural cells
differentiated from RTT-hiPSC lines. (C) A qPCR-facilitated comparison of gene expression for astrocytic markers (GFAP, S100?) and neuronal markers (TUBB3, MAP2) in neural cells differentiated from RTT-hiPSC lines. Relative gene expression levels
were normalized to that of ACTB. Data in (B) and (C) were analyzed by Student’s t-test and Welch’s t-test (*p??0.05)

During the neural development of the mammalian central nervous system (CNS), neural
stem cells initially differentiate into neurons, followed by astrocytes and oligodendrocytes
at a later stage. We and others reported that MeCP2 binds to the promoter region of
astrocyte-specific genes, including S100? and Gfap, to inhibit the conversion of neurons into astrocytes in mammals 33],34],37]-39]. To determine how MeCP2-regulated astrocyte-specific genes are influenced by the
absence of MeCP2, we next examined the expression of S100? and GFAP in MeCP2-negative and McCP2-positive neural cells. As a result, MeCP2-negative RS1-61P
and RS2-62P neural cells showed significantly enhanced expression of the astrocyte-specific
genes, GFAP and S100? (Fig. 4C). However, decreased expression levels of the neuronal genes, TUBB3 and MAP2, were observed in several samples (i.e., TUBB3 in RS2-62P cells and MAP2 in RS1-61 cells) (Fig. 4C). These observations suggest that MeCP2 deficiency in the neural cell lineage increases
astrocytic differentiation from multipotent neural stem cells.

Lack of MeCP2 binding to the GFAP gene in RTT-hiPSC-derived neural stem cells

MeCP2 was previously reported to bind to the highly methylated exon 1 region of the
Gfap gene in mouse neurons 33],38]. To confirm these results in wild-type or mutant MECP2-expressing neurospheres, we performed chromatin immunoprecipitation (ChIP) assays
with a specific primary antibody against MeCP2. Binding data were quantitated via
real-time PCR by using primers spanning from the STAT3 binding site within the GFAP promoter to exon 1, relative to the transcription start site of GFAP (Fig. 5A, indicated by blue bidirectional arrows). For both the STAT3 binding site and GFAP exon 1, precipitated genomic fragments were scarcely detectable by quantitative PCR
(qPCR) in neurospheres derived from mutant MECP2-expressing hiPSC lines (RS1-61P, RS2-62P), while significant precipitation of genomic
fragments was detected in neurospheres derived from wild-type MECP2-expressing hiPSC lines (RS1-52 M and RS2-65 M) (Fig. 5B).

Fig. 5. MeCP2 binding analysis and DNA methylation status of the GFAP promoter region. (A) Schematic representation of the hGFAP genomic locus. ChIP/qPCR analysis was performed for the genomic regions (i.e., the
STAT3 binding site and GFAP exon 1) indicated by the blue bidirectional arrows. Bisulfite sequencing was performed
for the genomic region indicated by the black bidirectional arrows. (B) MeCP2 binding to the STAT3 binding site within the GFAP promoter region was quantified via ChIP/qPCR analysis by using an anti-MeCP2 antibody
in neural cells differentiated from RTT-hiPSC lines. (C) Methylation frequencies of (1) the CpG site within the STAT3 recognition sequence
and (2) seven other CpG sites around this sequence were analyzed in hiPSC-derived
neurosphres via bisulfite sequencing

Based on these results, we hypothesized that the increased expression of GFAP was
induced by lack of direct MeCP2 binding to the GFAP gene in the mutant MECP2-expressing cells. However, it is possible that MeCP2 dysfunction indirectly enhanced
the maturation of neural stem cells with truncated MeCP2, accompanied by decreased
methylation of gfap33], which might in turn modify MeCP2 protein recruitment. Thus, we performed bisulfite
sequencing at the STAT3 binding site within the GFAP locus (Fig. 5A, indicated by black bidirectional arrows) by using wild-type or mutant MECP2-expressing neurospheres. Nevertheless, the MeCP2 binding sites in the GFAP gene were similarly hypermethylated in both wild-type and mutant MECP2-expressing neural stem cells differentiated from the four hiPSC lines (Fig. 5C). These findings signify that aberrant GFAP gene expression in neural stem cells containing truncated MeCP2 is not caused by
changes in DNA methylation status (i.e., hypomethylation) within the genomic region
encompassing the STAT3 binding site. Instead, the aberrant gene expression may be
due to alterations in the amount of MeCP2 that binds to the STAT3 binding site within
the GFAP promoter and the hGFAP exon1.