MeCP2 is associated with several neurological disorders; of which, Rett syndrome undoubtedly represents the most frequent. Its molecular roles, however, are still unclear, and data from animal models often describe adult, symptomatic stages, while MeCP2 functions during embryonic development remain elusive. We describe the pattern and timing of Mecp2 expression in the embryonic neocortex highlighting its low but consistent expression in virtually all cells and show the unexpected occurrence of transcriptional defects in the Mecp2 null samples at a stage largely preceding the onset of overt symptoms. Through the deregulated expression of ionic channels and glutamatergic receptors, the lack of Mecp2 during early neuronal maturation leads to the reduction in the neuronal responsiveness to stimuli. We suggest that such features concur to morphological alterations that begin affecting Mecp2 null neurons around the perinatal age and become evident later in adulthood. We indicate MeCP2 as a key modulator of the transcriptional mechanisms regulating cerebral cortex development. Neurological phenotypes of MECP2 patients could thus be the cumulative result of different adverse events that are already present at stages when no obvious signs of the pathology are evident and are worsened by later impairments affecting the central nervous system during maturation and maintenance of its functionality.
Rett syndrome (RTT) is a devastating genetic disorder that worldwide represents the most common genetic cause of severe intellectual disability in females. Typical RTT patients appear to develop normally throughout the first 6–18 months of life when neurological development arrests and a regression phase leads to the loss of previously acquired skills. During and after the regression phase, patients develop a host of typical symptoms including intellectual disabilities, respiratory problems, hand stereotypies, absence of speech, and epilepsy. Genetic analyses show that most RTT cases are caused by mutations in the MECP2 gene. Moreover, the same gene is less frequently associated with neonatal encephalopathy and MECP2 duplication syndrome in males, and, in both genders, with autism, schizophrenia, Angelman-like syndrome, and mild to severe mental retardation (Amir et al. 1999; Bienvenu et al. 2000; Yusufzai and Wolffe 2000; Chen et al. 2001; Guy et al. 2001). All in all, such molecular data suggest that MeCP2 is a key protein in the brain, where its levels and functions cannot be altered without severe consequences.
The generation of several mouse models carrying different Mecp2 alterations has been instrumental to investigate MeCP2 functions and their involvement in pathogenic mechanisms. In particular, the so far mostly used Mecp2 null male appears normal until 3–4 weeks of age, when it develops gross abnormalities such as hind limb clasping, stiff and uncoordinated gait, hypotonia, reduced spontaneous movements, tremor, breathing irregularities, and often, seizures (Ricceri et al. 2008). Mecp2 null brain displays several structural abnormalities, including smaller size (Belichenko et al. 2008), while studies focusing in deeper details on brain cellular architecture showed that Mecp2 null neurons display reduced soma size, changes in density, size and morphology of dendritic spines, as well as in the orientation of axons (Smrt et al. 2007; Belichenko et al. 2009). Accordingly, many studies have shown that Mecp2 null hippocampal slices are characterized by significant deficits in synaptic transmission (Asaka et al. 2006; Zhang et al. 2008), as well as in synaptic plasticity (Della Sala and Pizzorusso 2014). Interestingly, ablation of Mecp2 in adulthood results in defects of neuronal functions resembling those displayed by animals constitutively missing Mecp2 (Gemelli et al. 2006; Fyffe et al. 2008; McGraw et al. 2011; Cheval et al. 2012; Nguyen et al. 2012). Altogether, these conditional animal models indicate that MeCP2 must play a role in the maintenance of brain functionality at post-natal ages. On the contrary, the investigation of the possible roles played by Mecp2 during embryonic development has been largely neglected, although it is now generally accepted that subtle but consistent impairments are present even at early post-natal ages, when typical symptoms are not yet overt (Nomura 2005; Picker et al. 2006; Chao et al. 2007; Santos et al. 2007; Fehr et al. 2011). Moreover, as already alluded, hemizygous MECP2 male patients display a severe pathological condition as early as at birth (Schüle et al. 2008; Fu et al. 2014), while MeCP2 null neurons derived from induced pluripotent stem cells (iPSCs) display defective structural and functional maturation (Marchetto et al. 2010; Okabe et al. 2010; Kim et al. 2011; Li et al. 2011; Farra et al. 2012).
The present study aims at defining MeCP2 expression pattern and roles in the embryonic cortex and at identifying the temporal windows through which lack of the protein leads to defects that could anticipate the impairments detected later in adulthood. The data presented here provide evidence that MeCP2 plays a pivotal role during cerebral cortex maturation. By analyzing global gene expression in the embryonic neocortex, we show the deregulation of several molecular pathways delaying maturation of Mecp2 null cortexes and impairing intrinsic neuronal responsiveness to external stimuli. Such early abnormalities, associated with the appearance of morphological defects, worsen during development, persist at later stages, and likely contribute to post-natal cortical functional defects. We hypothesize that, if properly unveiled and characterized, these impairments could, in principle, be exploited as possible early targets for either diagnosis or treatment.
Materials and Methods
Animals and Tissues
The Mecp2 null mouse strain, originally purchased from Jackson Laboratories (003890 B6.129P2(C)-Mecp2<tm1.1Bird>/J), was backcrossed and maintained on a clean CD1 background; these mice recapitulate the typical phenotype of the regular Black6 original mice, with the advantage of having a larger progeny. Mice genotypes were analyzed through PCR on genomic DNA purified from tails. The following primers, based on published sequences (Miralvès et al. 2007), were used for genotyping: 5′-ACCTAGCCTGCCTGTACTTT-3′ forward primer for null allele; 5′-GACTGAAGTTACAGATGGTTGTG-3′ forward primer for wild type allele; 5′-CCACCCTCCAGTTTGGTTTA-3′ as common reverse primer. Time-pregnant females were generated by overnight crossing wild type CD1 males with Mecp2+/− heterozygous CD1 females; the day of vaginal plug was considered E0.5. For RNA processing, embryos were collected from anesthetized (Avertin, Sigma-Aldrich), time-pregnant females; dissected cortexes were then rapidly frozen on dry ice and stored at −80 °C. Embryos dedicated to histology were transcardially perfused with 4% PFA under a dissection microscope and post-fixed in 4% PFA for additional 4 h; about 12 μm cryosections were then stored at −80 °C until further analysis. All procedures were performed in accordance with the European Community Council Directive 86/609/EEC for care and use of experimental animals; all the protocols were approved by the Italian Minister for Scientific Research and by the local Animal Care Committee.
About 12 μm sections were washed in PBS, treated to retrieve antigens (when necessary), and incubated for 1 h in blocking solution (10% horse serum, 0.1% Triton X-100, and PBS). Tissues were then incubated with the primary antibody diluted in blocking solution overnight at 4 °C. Primary antibodies were used at the following concentrations: anti-Mecp2 (1 : 1000; Sigma-Aldrich, M9317), anti-Mecp2 (1 : 200; Cell Signaling, D4F3), anti-Nestin (1 : 200; Millipore, MAB353), anti-Tuj1 (1 : 2000; Covance, MMS-435P), anti-Reelin (1 : 20; DSHB, R4B), anti-GFAP (1 : 300; Cell Signaling, GA5), and anti-GFP (1 : 1000; Molecular Probe, A10262). For bright field imaging, after incubation with biotinylated secondary antibodies (Vector Laboratory), sections were treated with the Vector ABC kit (Vectastain) followed by staining with the Vector VIP peroxidase substrate kit. For fluorescent imaging, Alexafluor secondary antibodies were applied for 2 h at room temperature; sections were then counterstained with DAPI (Invitrogen) and mounted in Prolong Gold (Invitrogen). Nikon microscopes, objectives, and software were used for imaging, and panels were composed into figures using Adobe Photoshop.
The procedure to generate neural stem cell (NSC) culture was first described by Gritti et al. (1996); our protocol is based on the one described by Magri et al. (2011). Briefly, E15.5 embryos were individually dissected in PBS and the neocortex was transferred in Dulbecco's Modified Eagle Medium/F12 (DMEM) and dissociated by extensive enzymatic digestion with Papaine (Sigma). Cells were then grown in medium containing DMEM/F12, 0.66% glucose, l-glutamine 1%, Pen/Strep 1%, 4 µg/mL of heparin, hormone mix, and also with the addition of either 20 ng/mL epidermal growth factor and 10 ng/mL basic fibroblast growth factor or bFGF alone. In such conditions, cells spontaneously formed neurospheres (Gritti et al. 1996). Neurospheres were then dissociated in single-cell cultures and plated on matrigel-coated coverslips (BD Bioscience). Differentiation was obtained by replacing growth factors with 2% serum. PFA fixation was used to produce samples to be analyzed by immunofluorescence experiments once the culture reached desired cell density, otherwise total RNA was purified as described below.
RNA Purification, cDNA Synthesis, Microarray, and qPCR Analysis
Total RNA was purified either from cultured neurospheres, neurons, or embryonic cortexes using commercial kits (RNeasy Mini and Micro Kit, QIAGEN) following vendor's protocols. Cortexes collected for the microarray analysis were obtained from 3 wild type and 3 Mecp2 null E15.5 animals. Isolated total RNA quality was checked for integrity with 2100 Bioanalyzer (Agilent Technologies); samples with RNA integrity number lower than 8 were discarded. Gene expression of whole transcriptome was performed with Illumina Direct Hybridization Assays according to the standard protocol (Illumina). For each sample, 500 ng of total RNA were reverse-transcribed according to the Illumina TotalPrep RNA Amplification kit (Ambion) and cDNA was generated. Washing, staining, and hybridization were performed according to the standard Illumina protocol; each sample was hybridized onto Illumina MouseRef-8 Expression BeadChip arrays. Hybridization and scanning were performed according to the manufacturer's indications on an Illumina iScan System, and data were processed with GenomeStudio; array data were processed with quantile normalization in the Genome Studio software before further analysis. Average signals were calculated on gene-level data according to the standard Illumina detection cutoff (detection P-value <0.001). RNA dedicated to qPCR was retro-transcribed using the RT2 First Strand Kit (QIAGEN). qPCR results depicted in Figures 1, 3, 5, 6, and 7 were performed using SYBR green (Life Technology) as a fluorescent dye; primers are listed in Supplementary Fig. 8. The results depicted in Figure 4 were obtained using either Neuronal Ion Channels RT2 Profiler PCR Array or Synaptic Plasticity RT2 Profiler PCR Array (QIAGEN) following vendor's recommendations.
Gene Set Enrichment Analysis
To analyze microarray results, the gene set enrichment analysis (GSEA) was used. The 2 lists of expression values (one for the Mecp2 null phenotype and one for the wild type phenotype) were inserted in the GSEA software, first ranking genes on the base of their expression values and then determining the distribution of a set of genes (gene set) throughout the ranked list (www.broadinstitute.org; Subramanian et al. 2005). To obtain a numerical measurement of the degree by which a gene set of interest is more represented in one experimental group or the other, the software calculates an enrichment score (ES). The ES increases when a gene that belongs to a gene set is found in the ranked list, or decreases when the gene is not in the gene set. The degree of the increment (or decrement) depends on the amplitude of differential expression in the analyzed dataset. The next step performed by the GSEA is to assess the statistical significance of the ES, determining the nominal P-value by an empirical gene set-based permutation test procedure. Gene sets are chosen by the users, they can derive from a custom-made list of genes, or can be downloaded from MSigDB (Molecular Signature Database; Subramanian et al. 2005).
We produced the gene sets described in Figure 2 using published data (Ayoub et al. 2011) where the authors micro-dissected the E14.5 cortex into 3 different groups including ventricular zone (VZ), subventricular zone (SVZ)–intermediate zone (IZ), and cortical plate (CP), and separately analyzed their transcriptional profile through RNA sequencing (RNAseq). To identify genes exclusively expressed in the proliferative compartment or in the late post-mitotic neurons populating the CP, expression data were downloaded and fold change values calculated between the VZ and the CP datasets. Setting a cutoff value at 20 folds, we obtained 221 genes exclusively expressed in the proliferative compartment and 162 genes specifically expressed in late post-mitotic neurons. To be noticed, the gene set defining the proliferative identity is comprehensive of genes expressed at different levels in both the VZ and SVZ, as laser dissection does not enable separating a mixed population such as the one composing the VZ/SVZ. This is demonstrated by the presence in the VZ of levels of expression of Eomes (Tbr2; Supplementary Fig. 1) a gene defining intermediate progenitor cells composing the SVZ (Englund et al. 2005). The results described in Table 1 and Supplementary Fig. 3 were obtained using canonical pathway gene sets (C2, curated gene sets, comprehensive of information from the database BIOCARTA, KEGG, and REACTOME; www.broadinstitute.org).
|Gene sets||SIZE||ES||Nominal P-value|
|Gene sets||SIZE||ES||Nominal P-value|
Note: Such gene sets were chosen for their relevance in neocortex maturation, arepresents glutamatergic receptors, bintracellular effectors and cionic channels.
E15.5 cortexes were isolated, washed in HBSS, digested with Trypsin 0.25%, and mechanically dissociated by pipetting. Cells were plated on poly-l-lysine 0.1 mg/mL at low density 30 000 cells/well in a 6-multiwell plate each containing a 24-mm diameter glass coverslip. Neurons were grown in culture medium containing Neurobasal, B-27 (1 : 50), Glutamax 1%, and Pen–Strep 1% (Invitrogen, Carlsbad, CA, USA).
Cortical neurons were loaded with calcium-sensitive dye by incubating in 10 mM Oregon Green 488 BAPTA-1 AM (Molecular Probes) in Neurobasal Medium for 1 h at 37 °C and then imaged for calcium response. Both glutamate and electrical field stimulation were performed in KRH (Krebs'–Ringer's–HEPES) containing (in μM): 125 NaCl; 5 KCl; 1,2 MgSO4; 1,2 KH2PO4; 25 HEPES; 6 glucose; 2 CaCl2; pH 7.4. Glutamate stimulation was performed in the presence of Nifedipine 20 μM, Tetrodotoxin (TTX) 1 μM, and Bicuculline 20 μM, by applying 25 μM of glutamate with a perfusion system (Warner Instruments, Hamden, CT, USA). Electrical field stimulation was performed in the presence of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) 20 μM, R-2-amino-5-phosphonopentanoate (APV) 50 μM, and Bicuculline 20 μM using a stimulation chamber (Warner Instruments). Electrical-evoked calcium transients were induced with a stimulus train of 2 s (duration 1 ms; amplitude 90 mA) at 30 Hz as previously reported (Pozzi et al. 2013), using a train generation unit (Digitimer Ltd, DG2A) connected to a stimulus isolation unit (SIU-102; Warner Instruments). Recording chambers were placed on the stage of an IX-71 inverted microscope (Olympus, Hamburg, Germany) equipped with an EMCCD (electron-multiplying CCD) camera (Quantem 512 × 512, Photometrics). Illumination was obtained using a light-emitting diode single LED (Cairn Research Optoled; 470 nm) and a related GFP filter; 16-bit images were captured using a ×20 objective (NA: 0.8). Regions of interest (ROIs) of about 15-pixel diameter (corresponding to ∼12 μm) were drawn on the cell cytoplasm of virtually all the cells in the recorded field. Time-lapse recording of calcium dynamics was performed with an acquisition rate of 5 Hz for 600 s and off-line analyzed with the MetaFluor software (Molecular Devices). Calcium responses were measured as ΔF (Fmax− F0) with respect to the baseline (F0). A total of 10–15 neurons were analyzed for each field. Cumulative data were then analyzed through Kolmogorov–Smirnov statistic to verify non-parametric distribution.
In Utero Electroporation
For in utero electroporation the plasmid pLentiLox3,7 GFP vectors expressing either a short hairpin RNA directed against Mecp2 (5'-GTCAGAAGACCAGGATCTC-3') or beta-galactosidase (control) were used. Timed-gestation CD1 females (E13.5) were deeply anesthetized with Avertin before exposing uterine horns by midline laparotomy. DNA (1 μg/μL) together with dast green for counterstaining was injected in the telencephalic vesicle using a pulled micropipette, and then platinum electrodes were placed outside the uterus over the telencephalon and a train of four 50 ms squared pulses of 40 V were applied at 1-s interval from each. The uterus was then placed back in the abdomen, and muscle and skin were sutured. Animals were sacrificed at E18.5. Brains were collected by first checking their fluorescence (under a fluorescent stereoscope); fluorescent tissues were then enzymatically disaggregated and sorted on the basis of their fluorescence content (deriving from the expression of GFP). Both GFP+ and GFP− cells were collected from animals treated with either LacZ controls or Mecp2-directed shRNAs, such that GFP− cells were used as negative controls for GFP+ samples deriving from both the experimental groups; total RNA was then purified as already described. Transcriptional levels depicted in Figure 7A are expressed as the percentage of expression versus the GFP− group (considered 100%, dotted line). Tissues dedicated to histology were handled as already described.
Morphological analyses were performed at DIV3 on cortical neurons (isolated from E15.5 embryos and prepared as described) plated at very low density (30 000 cells/well). Immunofluorescence was performed as previously described using antibodies raised against Tuj1 and DAPI. Neurites length was measured only in neurons not connected with others using the NeuronJ, a plugin of ImageJ. Neurites were considered as the Tuj1-positive structure starting from the soma of each neuron; they were classified as “long” when their length was above 100 μm, or “short” when shorter than 100 μm. 3 null and 4 wild type animals were used for these measurements. The size of neuronal nuclei was measured using ImageJ on images collected from both primary cultures and 12 μm cortical slices from E15 and E18.5 wild type and null animals, and such measurements were based on DAPI staining.
While the profile of Mecp2 distribution in the murine embryonic cerebral cortex has been previously analyzed, no conclusive data highlight whether it is expressed in all the cellular types composing the developing cortex. Several studies suggested that Mecp2 expression is restricted to late post-mitotic neurons residing in the CP and Cajal-Retzius (CR) cells composing the marginal zone, but it is absent from cycling progenitors in the VZ/SVZ (the proliferative compartment) and from maturing neurons in the IZ (Jung et al. 2003; Kishi and Macklis 2004; Schmid et al. 2008). However, the presence of Mecp2 in embryonic stem cells has been suggested in other studies (Caballero et al. 2009; Okabe et al. 2010; Kim et al. 2011). To draw a definitive picture of the pattern and timing of Mecp2 expression during cortical development, we first performed a detailed immunohistochemistry analysis (Fig. 1) using null tissues as negative controls to enable the specific detection of low levels of Mecp2. We revealed the presence of Mecp2 in a large set of cell types, spanning from cortical neuroblasts populating the neuroepithelium at E10.5 (compare Fig. 1A, wild type, with A′, Mecp2 null), progenitors, immature and differentiated neurons at E14.5 (wild type in Fig. 1B and null in Fig. 1B′), to virtually all the cell types of the cortex at E18.5 (Fig. 1C). The detection of Mecp2 transcript in proliferating and differentiating NSCs (Fig. 1D), together with the protein co-localization with Nestin (Fig. 1E,F), Tuj1 (Fig. 1G), and Reelin (Fig. 1H), demonstrates that Mecp2 is expressed in a very heterogeneous set of cells. In accordance with published studies (Jung et al. 2003), we found that Mecp2 levels vary according to the stage of cellular maturation (Fig. 1B,G), with Tuj1− cells in proliferative areas (Fig. 1g′) expressing lower levels of Mecp2 compared with Tuj1+ (Fig. 1g″,g‴).
Given the novel information gained through the analysis of Mecp2 expression in the developing cortex, we then investigated its transcriptional activity by purifying total RNA from both wild type and Mecp2 null embryonic cortexes collected at E15.5 and screening transcriptional levels through microarray (Fig. 2).
Using fold change as a selection criterion for differential expression, we identified 270 differentially expressed genes with a fold change cutoff of 1.3 and 11 genes with a fold change cutoff of 2 (Fig. 2A). Using a P-value of 0.05 as a cutoff criterion, we identified only 135 differentially expressed genes; nevertheless, these 135 differentially expressed genes were able to clearly segregate the 2 tested conditions (wild type and Mecp2 null mutants), as shown by the hierarchical cluster heatmap of expression data of the profiled 6 individual samples (Fig. 2B). However, using gene ontology (GO) analysis to functionally describe the 2 gene clusters separately (genes up- or downregulated in Mecp2 null mutant), we were not able to find any GO term enriched at a significant level (Supplementary Fig. 1).
Given that (1) differential expression analysis identified only a limited number of genes; (2) expression variations were only mild in amplitude, probably due to the fact that the neocortex is populated by an heterogeneous set of cell types each expressing specific levels of Mecp2, thus reducing the detection of possible differences in expression; and (3) simple GO analysis was not useful in functionally annotating the selected gene lists, we decided to use the GSEA to explore data in deeper details and extract biological insights. GSEA overcomes common analytical challenges inherent to microarray technology, such as detection noise, that could lead, when differential expression is minimal, to no or few genes passing the threshold for statistically significant results. GSEA takes into account gene sets that likely share common biological functions or regulation; thus, in contrast to single-gene analyses, this approach allows to emphasize the contribution of each expression difference (even mild) in a group of genes that roughly represent the whole molecular pathways (Subramanian et al. 2005).
First, we obtained a custom-made gene set extracted from data presented by Ayoub et al. (2011), in which the authors, through laser dissection and RNAseq, profiled transcription either in the VZ, SVZ/IZ, or CP of the E14.5 developing cortex. By setting a fold change cutoff of 20 on the transcriptional ratio of each gene analyzed in the dataset of either VZ or the CP provided by the authors, we identified 2 gene sets exclusively representative of either the proliferative compartment or late post-mitotic neurons (221 and 162 genes, respectively). As shown by the Venn diagram in Figure 2C, the 2 lists are not overlapping (the genes composing them are provided in Supplementary Fig. 2). These 2 groups of genes appeared fairly segregated between the 2 genotypes analyzed in our work (MA plot in Fig. 2D) and a significant enrichment (nominal P-value <0.001) was accordingly highlighted by analyzing the transcriptional profiling through the GSEA method: the proliferative compartment identity was enriched in the null sample (i.e., up-regulated in the null sample, Fig. 2E), whereas, on the contrary, the late post-mitotic identity was enriched in the wild type sample (i.e., downregulated in the null sample, as depicted in Fig. 2F). This was a first hint leading us to hypothesize that, in the E15.5 Mecp2 null cortex, the maturation of newly born cortical neurons might be somehow impaired. These 2 gene sets were chosen a priori to highlight whether one of the two transcriptional identities could be enriched in one of the two genotypes analyzed; in a further analytical approach, we used GSEA in a more canonical way, scanning an entire gene set collection to highlight the enrichment in specific molecular pathways. In order to do that, we superimposed on the wild type versus the null dataset the curated gene sets (C2) of the Molecular Signature Database (MSigDB, version 3.1, comprising 4850 different gene sets; Subramanian et al. 2005). Table 1 summarizes gene sets significantly enriched in the wild type samples (thus reduced in the null cortexes), with a nominal P-value cutoff of 0.001. These gene sets were selected because: (i) they all converge in 3 main groups: intracellular effectors, glutamatergic receptors, and ionic channels (a,b,c in Table 1); (ii) such mechanisms are all crucial for neuronal physiology, and (iii) they were all grouped at the top of the list ranked according to statistical significance (Supplementary Fig. 3 depicts all the gene sets showing a nominal P-value <0.001). Of interest, these pathways all share the ability to interact with and activate many other downstream effectors. To obtain a first degree of confirmation concerning the possible deregulation of such highlighted pathways, we measured, through qPCR, the levels of transcription of the most representative genes composing each gene set. As illustrated in Figure 3, the transcription levels of different genes common to each of the gene sets depicted as b in Table 1 were down-regulated at E15.5; notably, such down regulation persisted in most cases at later time points (E18.5 and P8).
Because our GSEA results suggested that Mecp2 might impact the transcription of different glutamate receptors and ion channels (Table 1a and c), we measured the cortical expression levels of AMPA, NMDA and metabotropic glutamatergic receptor subunits, whose expression is critical for proper activity-dependent signaling (Lujàn et al. 2005; Uchino et al. 2010; Guo et al. 2013).
We found that the levels of mRNAs encoding for ionotropic receptor subunits were significantly downregulated at both E15.5 and later time points in the null cortex when compared with wild type (Fig. 4). The transcriptional levels of metabotropic receptor subunits (not yet expressed at E15.5; Luján et al. 2005; Ayoub et al. 2011) were mostly normal at E18.5, whereas the expression of some of them (Grm3, 5, and 7) was downregulated at P8 in null mice. Similarly, mRNAs for different classes of ion channels (Ca2+, Cl−, K+, and Na+) were downregulated at E15.5 (Fig. 4), and again, such downregulation persisted at later stages of development (E18.5 and P8). Besides the deregulation of gene sets describing the glutamatergic phenotype, we found that gene sets describing GABAergic transmission were mildly enriched in wild type samples. Such effect was not confirmed by qPCR at E15.5 or E18.5, while some downregulation was detected at later time points (P8, Supplementary Fig. 4).
Taken together, our data indicate the occurrence of a fairly long-lasting impairment in the transcriptional regulation of genes involved in neuronal excitability. This may suggest that mechanisms modulating intracellular responses might be impaired in null mice as early as during embryonic neocortical development, and that such defects persist at later time points. To address this point, we produced cultures of cortical neurons established from E15.5 embryos. Neurons were cultured at low density (see Materials and Methods for details), in order to avoid (at least until DIV12) the establishment of a dense interconnected network and to focus on features ascribable, for the most part, to cell autonomous mechanisms. Moreover, by establishing neuronal cultures from E15.5 embryos, we drastically reduced the presence of glial cells, as clearly shown by the immunofluorescence analysis shown in Figure 5A,B, where no GFAP signal is detected and each cell, identified by DAPI staining, is a Tuj1-positive neuronal cell. These cultures were assayed at DIV3 through qPCR in order to evaluate the transcriptional levels of different subunits of AMPA and NMDA receptors as well as Ca2+, Cl−, K+, and Na+ channels. We found that the expression of these genes was significantly reduced in null neurons (Fig. 5C), thus in vitro confirming the in vivo data depicted in Figure 4. Interestingly, at such an early stage of differentiation, null cultured neurons display a significant reduction in nuclear size, as shown in Figure 5D. Such observation was only partially reproduced in vivo, where a mild, statistically not significant tendency toward the reduction in nuclear size was observed in null cortexes at E18.5 (Supplementary Fig. 5A; the effect is even milder at E15.5, Supplementary Fig. 5B). Morphological analyses using Tuj1 as a marker of neuronal structures revealed that in vitro neurites longer than 100 μm are on average shorter in null neurons compared with wild type (Fig. 5E), whereas no differences were detected in the length of short neurites (shorter than 100 μm; Fig. 5F).
Next, we used in vitro intracellular Ca2+ transients as readout of neuronal responsiveness to chemical and electrical stimulation of cortical primary neurons at early developmental stages. These experiments thus represent a second and functional validation of the GSEA predictions. At first, we characterized calcium responses of DIV3 neurons derived from wild type E15.5 embryos to either glutamate (25 μM; Fig. 6A) or electric stimulation (2 s at 30 Hz; Fig. 6B; Supplementary Fig. 6 depicts typical fields of view with ROI). Ca2+ responses evoked by glutamate stimulation (panel a) were strongly reduced by adding CNQX (AMPA receptors blocker, panel a′), and completely abolished by adding both CNQX and the NMDA receptors blocker APV (panels a″). The quantitative analysis of calcium changes, shown in panel a‴, clearly indicates that the glutamate-evoked calcium transients are due to ionotropic receptor activation, with AMPA receptors being responsible for most of the response. This observation implies that ionotropic receptors are predominant over metabotropic receptors in mediating glutamate-induced responses. On the other hand, electrically evoked calcium transients (Fig. 6B, panel b) are strongly reduced by the L-type Ca2+ blocker Nifedipine (b′; 20 μM) and completely abolished by 1 μM TTX (b″; Pozzi et al. 2013). The quantitative analysis (panel b‴) indicates that, at this developmental stage, L-type channels represent one of the major channel subtypes driving electrically evoked responses (Pravettoni et al. 2000), and that their opening is completely due to the activation of voltage-dependent sodium channels.
To assess whether the described transcriptional alterations (Figs 3–5) were associated with functional impairment, Ca2+ responses evoked by either glutamate or electrical stimulation were analyzed in wild type and null cultured neurons (Fig. 6C–J). In agreement with the transcriptional and GSEA results, indicating alterations of the levels of ionotropic glutamatergic receptors in Mecp2 null neurons, a significantly lower response to glutamate was recorded in null neurons compared with wild type (Fig. 6C–F); interestingly, this decrease was maintained at both DIV7 and DIV12 (Supplementary Fig. 7A,B). Accordingly, measurements of neuronal calcium responsiveness evoked by a short-field electrical stimulation revealed that electrically evoked calcium responses (recorded at DIV3) were significantly lower in null neurons compared with wild type (Fig. 6G–J), thus confirming that maturing neurons lacking Mecp2 exhibit a reduced neuronal responsiveness to electrical stimulation. These findings are clearly in line with the downregulation of genes encoding for both Na+ and L-type Ca2+ channels depicted in Figures 4 and 5C. However, this feature was not maintained over time, as, by DIV7, wild type and null neurons responded in an overlapping manner to the electrical stimulation (Fig. 6J and Supplementary Fig. 7C,D).
To further validate the hypothesis that early defects in Mecp2 expression affect the transcription of genes involved in activity-dependent signaling, we depleted Mecp2 expression in wild type cortical neurons by electroporating in utero short hairpin sequences targeting Mecp2 mRNA (Fig. 7). Given the low number of transduced neurons, this technique likely reveals cell autonomous effects of the gene knockdown.
E13.5 embryos from wild type timed-pregnant females were electroporated with either control (LacZ-directed short hairpin) or Mecp2-directed short hairpin (Zhou et al. 2006). Embryos were harvested at E18.5 and GFP-positive tissues were collected under a fluorescent stereoscope; GFP-positive neurons were then sorted from the GFP-negative internal controls. qPCR analyses on total RNA isolated from such cells confirmed the downregulation of Mecp2 and the reduced expression of different AMPA and NMDA receptor subunits, as well as L-type Ca2+ (Cacna1d and Cacng2) and Na+ channels (Scn8a) subunits, as depicted in Figure 7A.
It is known that the expression of proteins responsible for activity-dependent signaling, such as ion channels, neurotransmitter receptors, and Ca2+ channels, are critical for several aspects of neuronal maturation, including neuronal migration (Rakic and Komuro 1995; Komuro and Rakic 1998; Kumada and Komuro 2004). Such evidences prompted us to evaluate whether the transcriptional alteration observed in Mecp2 shRNA-electroporated neurons might affect cortical neuronal migration in vivo. As expected, we found that control neurons (electroporated with LacZ-directed sh) mostly populated the upper portions of the CP, the physiological target for neurons born at E13.5 (Angevine and Sidman 1961; Rakic et al. 1974; Fig. 7B). Conversely, GFP-positive neurons electroporated at E13.5 with Mecp2 shRNA mostly failed to properly migrate to the CP by E18.5, and rather, remained close to the VZ (Fig. 7C). The statistical evaluation of the migration of electroporated cells is depicted in Figure 7D. Given the transcriptional and migration defects depicted in Figure 7, we searched for a similar phenotype in Mecp2 null mice. Thus, newly born neurons were BrdU-labeled at E13.5 and their migration was analyzed at E18.5; however, in these experimental conditions, migration of BrdU-positive neurons was similar in Mecp2 null and wild type developing cortexes (Supplementary Fig. 5C).
Many mouse models carrying different Mecp2 alterations have been generated to improve our comprehension of MeCP2 functions and the mechanisms underlying RTT or other MECP2-related conditions (Bedogni et al. 2014). The use of such models demonstrated that: (1) the inactivation of Mecp2 at different post-natal ages (from late juvenile to adult) always causes the appearance of severe neurological phenotypes and premature death (McGraw et al. 2011; Cheval et al. 2012; Nguyen et al. 2012) and (2) the MeCP2-related conditions are reversible at least in mice (Giacometti et al. 2007; Guy et al. 2007; Jugloff et al. 2008; Garg et al. 2013). However, in spite of these recent enormous advances, the knowledge regarding the temporal steps through which the consequences of dysfunctional MeCP2 activity begin to manifest is still limited. In fact, most studies have been performed in pre-symptomatic (3–6 weeks of age) or symptomatic (adult) male animals. Nonetheless, the presence of subtle but consistent impairments at early post-natal ages, when typical symptoms are not yet overt (Nomura 2005; Picker et al. 2006; Chao et al. 2007; Santos et al. 2007; Fehr et al. 2011), strongly indicates that also the maturing brain depends on MeCP2 activities.
At the molecular level, most of the data obtained so far suggest that MECP2-related pathologies might be associated with its direct or indirect involvement in gene expression regulation. Unfortunately, however, the outcomes of defective MeCP2 activity are far from being understood and the identification of its specific transcriptional targets, especially at the early stages of the pathology, proved to be a difficult goal to achieve. Thus, we started off analyzing Mecp2 expression during embryonic maturation of the neocortex. The rationale was dual: first, the cortex has a clear involvement in MeCP2-related disorders (Armstrong 2001; Fukuda et al. 2005; Zhang et al. 2014); second, embryonic gene expression profile is a field only mildly approached in RTT that, however, should enable to avoid most if not all other indirect effects driven by either the health status of the animal or the environment. By superimposing microarray outputs to custom-made gene sets based on published data (Ayoub et al. 2011), we hypothesized that Mecp2 null cortexes do not properly acquire the transcriptional signature of mature neurons in the E15.5 CP (Fig. 2). The immaturity of the transcriptional phenotype well correlates with the defective morphology displayed by null primary cortical neurons (Fig. 5), a feature that is in line with previous data highlighting maturation impairments in neurons derived from human RTT iPS cells (Marchetto et al. 2010; Walsh and Hochedlinger 2010; Kim et al. 2011; Farra et al. 2012; Dajani et al. 2013). Noticeably, our GSEA output clearly suggests that such defects are due to many modest but consistent transcriptional alterations affecting genes involved in related or unrelated molecular pathways.
GSEA revealed the downregulation of a large set of genes involved in the control of activity-dependent signaling pathways, which then mediate transcriptional, translational, and plasticity mechanisms. Among these, genes controlling activity-dependent Ca2+ signaling represent a pivotal pathway present throughout the entire life of a neuron and indeed, intracellular Ca2+ dynamics have been described as early as E13.5 in proliferating progenitors (Desarmenien and Spitzer 1991; Gu and Spitzer 1995; LoTurco et al. 1995; Spitzer 2006), migrating neurons (Komuro and Rakic 1998; Behar et al. 1999; Spitzer 2006), and post-mitotic neurons (Platel et al. 2005). Moreover, it has also been reported that spontaneous neuronal activity contributing to maturation and refinement of neuronal circuits occurs in several brain regions at embryonic stages, including retina, thalamus, and cortex, thus even before experience-dependent activity driven by external sensory input (Meister et al. 1991; Yuste et al. 1992; Galli and Maffei 1998; Feller 1999; Weliky and Katz 1999). Given that Mecp2 null primary cortical neurons established at E15.5 exhibit lower calcium responsiveness to both glutamate and electrical stimuli at early stages of development, it is reasonable to hypothesize that early defects in neuronal excitability might play a critical role in the altered neuronal maturation observed during adulthood in different Mecp2 null models.
The ability of a neuronal cell to exhibit proper spontaneous activity and to respond to external stimuli reflects its stage of maturation; nonetheless, and more intriguingly, this ability is obviously part of the mechanisms through which a neuron differentiates during early development and functions later in adulthood. Accordingly, the defects described here are in line with the recently discovered morphological defects displayed by Mecp2 null neurons at DIV4 (Baj et al. 2014); importantly, they largely precede the onset of neurological outcomes in Mecp2 animal models. A tight link between early neuronal activity and late maturation of synaptic contacts has been shown both in vivo (Lin et al. 2010; Uchino et al. 2010) and in vitro (Perlini et al. 2011; Andreae and Burrone 2014). In fact, neuronal activity, through local and global changes of several activity-dependent signaling pathways, plays an important role not only during early phases of proliferation, migration, and differentiation but also at later stages of development, during synapse formation, and neuronal survival (Komuro and Rakic 1998; Crair 1999; Zhang and Poo 2001; Luján et al. 2005; Spitzer 2006; Rosenberg and Spitzer 2011). Therefore, our results likely foreshadow the reduction in the number of synapses in cultured Mecp2 null neurons compared with wild type, thereby linking transcriptional effects responsible for the reduced neuronal activity (this work) with altered synaptogenesis (Chao et al. 2007; Cohen and Greenberg 2008). Our observations are of particular interest in light of the recent hypothesis that several syndromes, including autism spectrum disorders, X-fragile, tuberous sclerosis, and Angelman syndrome, are all characterized by common alterations in the activity-dependent signaling networks controlling proper structural maturation of the brain (Ebert and Greenberg 2013). Intriguingly, the downstream effects of such alterations would likely lead to alterations in brain connectivity, a classical feature of RTT (Kishi and Macklis 2004; Degano et al. 2009).
Given the virtual lack of glial cells in both early in vivo and in vitro conditions, the low density neuronal seeding delaying the formation of contacts between neurons, and the early developmental stage at which calcium transients are recorded (DIV3), it is reasonable to infer that the link between the transcriptional defects and the intrinsic maturation impairments displayed by Mecp2 null neurons most likely arises from cell autonomous mechanisms. This is confirmed by the overlapping transcriptional alterations displayed by full null embryonic cortexes and Mecp2 silenced neurons sorted from otherwise normal cortical cells (Fig. 7A). Our observations are thus in agreement with the data presented by Kishi and Macklis (2010), in which the authors elegantly demonstrated that the transplantation of Mecp2 null projection neurons into wild type cortex does not prevent or rescue their impairments in morphology and complexity of dendritic structures. Of note, a reduced nuclear size was highlighted in young cultured null neurons, whereas E18.5 brain sections only displayed a non-significant tendency to nuclear size reduction, a typical feature of MeCP2-deficient adult neurons. Accordingly, we observed migration defects in the restricted cohort of cells depleted of Mecp2 through in utero electroporation, but not in null cortical tissues. We speculate that compensatory mechanisms might mask or even partly rescue defects that are evident only when analyzed in conditions enabling cell autonomous studies.
To conclude, our data suggest that newborn Mecp2 null cortical neurons display critical transcriptional defects that lead to reduced neuronal responsiveness, which, in turn, delays and disrupts proper neuronal maturation. Although we cannot rule out that other mechanisms could later determine at least a partial rescue of the phenotype here described, we speculate that the neurological phenotype displayed by Mecp2 null animals and RTT patients could thus be the cumulative result of different adverse events occurring ever since early embryonic maturation through either cell autonomous or non-autonomous mechanisms. Accordingly, it is well described that Mecp2 null glia and microglia, cell types continuously going through cycles of turnover and maturation throughout life, have detrimental effects on neuronal functionality due to their inability to exert trophic support (Ballas et al. 2009; Lioy et al. 2011; Derecki et al. 2012, 2013; Zeidán-Chuliá et al. 2014). Moreover, given the ubiquitous expression of Mecp2 in many different cell types, maturation defects could occur outside the brain as well, and contribute to other RTT non-neurological features.
Grant resources are as follows: Fondazione Cariplo (Grant 2010-0724 to N.L.); Fondazione Telethon (Grant GGP10032 to N.L.); Jerome Lejeune Foundation (to F.B.); ERC Consolidator Grant (617978 to M.P.); Health F2-2009-241498 Eurospin (to M.M.).
We are very thankful to ProRett Ricerca, an association of Italian parents of RTT girls whose efforts represent our major source of both strive and funding. F.B. and N.L. thank Vania Broccoli and his co-workers for kindly sharing reagents and knowledge. Conflict of Interest: None declared.