MECP2 duplication and mutations impair NSCs differentiation via miR-197 regulated ADAM10

How MECP2 (Methyl-CpG-binding protein 2) duplication affects cortex development remains elusive. We found that elevated MeCP2 expression promotes neurogenesis during cortex development in Tg(MECP2) mouse brain. Ectopic expression of MeCP2 in NPCs inhibits ADAM10 and hence compromises the NOTCH pathway during NPC differentiation. MeCP2 up-regulates miR-197 to down-regulate ADAM10. The enhanced NPC differentiation/migration in Tg(MECP2) embryonic brain can be repressed by overexpression of ADAM10 or a miR-197 inhibitor. Consistently, the reduced neurogenesis induced by three rare MeCP2 missense mutations (H371R, E394K, G428S) identified in a Han Chinese autism spectrum disorders (ASD) cohort, can be reversed by miR-197 both in vitro and in vivo. Our results revealed that a regulatory axis involving MeCP2, miR-197, ADAM10, and NOTCH signaling is critical for neurogenesis, which is affected by both MeCP2 duplication and mutation.


Introduction
MECP2 (methyl-CpG binding protein 2), an X-linked gene encoding the methyl-cytosine binding protein MeCP2, is associated with two severe neurological disorders, Rett syndrome (RTT) and MECP2 duplication syndrome (MDS), which result from loss and gain of function of MECP2, respectively. Mutations in MECP2 were first identified as the monogenic cause for RTT (1). Although clinically distinguished from Autism Spectrum Disorder (ASD) by the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) (2), autistic features are often observed (>60%) in RTT patients (3,4). The Mecp2 knockout mice (Mecp2-/y) also showed excellent replication/representation of RTT and ASD phenotypes (5,6). Recently, MDS, a severe male intellectual disability syndrome, caused by MECP2 duplication was identified (7,8), which has ~100% penetrance of autistic-like behaviors (9,10). Transgenic mouse models for MDS (11,12) and lentivirus-based transgenic monkey (13) both showed severe autistic-like and compulsive repetitive behaviors. Therefore, it has been suggested that the functional and dosage variations of MeCP2 are tightly associated with autistic features (4,14). Exploring the functional conversions of MECP2 duplications and mutations may provide useful insight into the mechanisms underlying the development of ASD phenotypes (4,14).
MeCP2 was originally identified as a transcriptional repressor (15,16), but recent studies have shown it has diverse functions, including transcription activation (17), mRNA splicing (18,19), and microRNA (miRNA) processing (20,21), to name but a few. As both RTT and ASD patients show symptoms shortly after birth, most functional studies of MeCP2 have been focused on postnatal events such as dendritic arborization (22)(23)(24), synapse formation and plasticity (24-26), and adult neurogenesis (27-29). However, the function of MeCP2 during embryonic CNS development is still elusive. 4 MeCP2 is widely and highly expressed in the developing central nervous system (CNS), including both the early neural tube and embryonic forebrain, in zebrafish, chicken, and mouse (30-33). However, embryonic development seems to be unaffected in Mecp2 knockout mice (6,34).
But study in Xenopus showed that MeCP2 promoted neurogenesis of Xenopus embryos, while the RTT mutant R168X failed to do so (35). A recent study in the monkey showed that Talent-edited mutation of MECP2 caused embryonic lethality in male mutant monkeys (36). Furthermore, human iPSCs generated from RTT patients with dysfunctional MeCP2 have abnormal neurogenesis and gliogenesis (37, 38). Therefore, it is likely that MeCP2 has an important role during early CNS development.
In the present study, we demonstrated that overexpressed MeCP2 promoted neurogenesis in the embryonic brain in the MDS model Tg(MECP2) mouse. Our results revealed a novel mechanism involving miR-197, ADAM10 (A disintergrin and metalloprotease 10), and NOTCH signaling as a critical regulatory axis for the enhanced neurogenesis induced by MeCP2. Furthermore, we identified three rare missense MECP2 mutations (H371R, E394K, G428S) in ASD patients in our Han Chinese ASD cohort, which are novel in the East Asian population according to ExAC (39). These

Neurogenesis is enhanced in Tg(MECP2) mouse fetal brain
To determine the effect of MECP2 duplication on cell fate in vivo in fetal brain, we investigated a Tg(MECP2) mouse line which had previously been used as a model for MDS (11), as it contains an extra copy of human MECP2 and exhibits approximately doubled MeCP2 levels in the brain (Fig. 2D).
During cortical neurogenesis in the mouse, NSCs/NPCs in the ventricular (VZ) and subventricular (SVZ) zones are differentiated into immature neurons and migrate radially to the cortical plate (CP) to become mature neurons (40,41). For quantification of cell fate in wild-type (WT) and Tg(MECP2) cortex, immunofluorescent staining on E18.5 and P7 brain sections with different markers were performed. The results revealed that there are significantly more Satb2 + (a cortical neuron marker) cells in CP layer and significantly less Sox2 + and Tbr2 + (progenitor markers) cells in VZ/SVZ layer of E18.5 Tg(MECP2) mice compared to WT littermates (Fig. 1A). By P7, there are still significantly more Satb2 + and Tbr1 + neurons present in P7 Tg(MECP2) mice cortex than WT cortex (Fig. 1B).
Primary NPCs were isolated from E12.5 mouse cortex to further examine the in vitro effect of MeCP2 overexpression on NPCs differentiation. WT NPCs from C57BL/6 mice were infected with MeCP2 lentivirus after seeding on the culture dish. The exogenous expression doubled the expression levels of MeCP2 in those NPCs ( Fig. 2A). Both transiently overexpressed MeCP2 in C57BL/6 NPCs and elevated MeCP2 in transgenic Tg(MECP2) mice significantly up-regulated the level of the neuronal marker MAP2, and down-regulated the level of the glia marker GFAP in NPCs ( Fig. 2C&D). Immunofluorescent staining on cultured NPCs infected by lentivirus also showed significantly more MAP2 + and less GFAP + cells with MeCP2 overexpression (Fig. 2B), which was consistent with previous reports (42). Taken together, the results demonstrate that elevated MeCP2 expression affects NPCs cell fate and promotes neurogenesis.

ADAM10 is downstream of MeCP2 in the control of neurogenesis
Since NOTCH is a key molecule for NSCs fate (43, 44), we next examined the levels of several key molecules in the NOTCH pathway in primary cultured NPCs after overexpressing MeCP2. The levels of two NOTCH ligands JAG1 and DLL1, the full-length receptor NOTCH1, the activated NOTCH intracellular domain (NICD), and the two rate-limiting S2 enzymes for NOTCH cleavage, ADAM10 and ADAM17, were examined by Western blot analysis in primary NPCs transfected with either MeCP2 or an empty control vector. The results showed that the levels of NICD and ADAM10 were significantly down-regulated by MeCP2 expression in NPCs (Fig. 3A), while ADAM17, full-length NOTCH1, JAG1, and DLL1 were not significantly affected ( Fig. S1A-E). These results suggest that MeCP2 might inhibit NOTCH signaling through the down-regulation of ADAM10, in order to regulate NPC development.
To study the relationship between ADAM10 and MeCP2, we expressed ADAM10 together with MeCP2 either in primary cultured NPCs or in embryonic brain. Expression of ADAM10 together with MeCP2 in NPCs reversed the effect of MeCP2 as indicated by the restoration of MAP2 and GFAP levels close to those of controls (Fig. 3B). The in vivo effect of ADAM10 on MeCP2 promoted neurogenesis was also tested in the developing cortex by in utero electroporation (IUE) experiments.
There were significantly more EYFP + cells in CP layer in the Tg(MECP2) cortex (Fig. 3D) as well as in C57BL/6 WT cortex ectopically overexpressing MeCP2 (Fig. 3C). Further, immunofluorescent staining showed that those EYFP + cells migrated into the CP layer were positive for the neuronal markers NeuN and Tuj1; whereas those cells remaining in the VZ/SVZ layer were positive for the progenitor markers Sox2 and Tbr2 (Fig. S2). When ADAM10 was co-expressed with MeCP2, it could 7 reverse the effect of MeCP2 on neurogenesis, resulting in significantly fewer cells reaching the upper CP layer, compared to MeCP2 overexpression cortex (Fig. 3C&D). Such effects were not observed when ADAM17 was co-expressed with MeCP2 (Fig. S1F&G). These results suggest that ADAM10 is functionally downstream of MeCP2 during neurogenesis.

MeCP2 down-regulates ADAM10 expression through up-regulating miR-197
We next investigated how MeCP2 regulates ADAM10. Initially, we confirmed the regulation of ADAM10 by MeCP2 in human cells. Similar to cultured mouse NPCs, the ADAM10 protein level was also significantly down-regulated (~53%) by MeCP2 overexpression in a human glioblastoma cell line U251 (Fig. 4A). The mRNA level of ADAM10 was down-regulated ~19% by MeCP2 (Fig. 4B).
Since ADAM10 was not identified as a transcriptional target of MeCP2 (17), we screened the potential microRNAs (miRNAs) that might be involved in ADAM10 regulation by MeCP2. Seven miRNAs, identified as the most significantly regulated miRNAs by MeCP2 in the adult NSCs of the Mecp2-/y mouse (45), were used for our studies. When inhibitors for each miRNA (e.g. i-197 as an inhibitor of miR-197) were transfected into U251 cells, ADAM10 protein levels were significantly up-regulated by i-222, i-193 and i-197; while the other four inhibitors had no statistically significant effect (Fig. S3A). However, when MeCP2 was co-transfected with miRNA inhibitors, only i-197 and i-193 were able to reverse the MeCP2-induced down-regulation of ADAM10 protein (Fig. 4D & Fig.   S3B). The expression profile from miRNAMap shows that hsa-miR-193 is primarily expressed in the muscle, and hsa-miR-197 is highly expressed in the brain (http://mirnamap.mbc.nctu.edu.tw) (46); therefore, miR-197 is the most likely target of MeCP2 in the regulation of ADAM10 expression in the brain.
We went on to investigate the potential target site on ADAM10 3'-UTR for miR-197. A 7mer-m8 8 miR-197 binding site at position 1568-1574 of human ADAM10 3'-UTR was predicted by TargetscanHuman (http://www.targetscan.org/vert_71/) (47), but it is poorly conserved between human and mouse (position 1563-1568 of mouse Adam10 3'-UTR) (Fig. 4E). A luciferase reporter A10-I-WT was constructed with a 305bp fragment of human ADAM10 3'-UTR covering this predicted position 1568-1574. The G to A point mutation in the seed sequence was also constructed as A10-I-Mut (Fig. 4E). Both constructs were transfected into U251 cells with either miR-197 mimics or mimic control. Surprisingly, the results showed that miR-197 had no effect on either A10-I-WT nor  We subsequently sought to determine how MeCP2 regulates miR-197. RT-PCR experiments showed that the levels of pre-and matured miR-197 were significantly up-regulated by approximately 5.4 and 2.7 fold by MeCP2, respectively ( Fig. 4J-K), while the level of pri-miR-197 was not affected by MeCP2 (Fig. 3I). This up-regulation of pre-miR-197 by MeCP2 does not affect the down-regulation of pre-miR-134 by MeCP2 (Fig. 4L). Taken together, MeCP2 down-regulated ADAM10 by up-regulating pre-miR-197 (Fig. 4M).

MiR-197 plays a critical role in MeCP2 mediated neurogenesis
Since MeCP2 expression is capable of up-regulating the expression of miR-197, we overexpressed miR-197 in primary cultured NPCs to better understand its effect on NPCs differentiation. Immunofluorescent staining showed that, like MeCP2, miR-197 overexpression promoted neurogenesis and inhibited gliogenesis, as evidenced by an elevated ratio of MAP2 + and decreased ratio of GFAP + cells in the differentiated NPCs (Fig. 5A). The inhibitor of miR-197 (i-197) had opposite effect with decreased ratio of MAP2 + and elevated ratio of GFAP + cells (Fig. 5A).
Furthermore, the neurogenesis promoting effect of MeCP2 could also be blocked by i-197 when they were co-expressed in NPCs (Fig. 5B). MeCP2 induced down-regulation of ADAM10 was also blocked by i-197 (Fig. S5A).
Similar results were observed in the mouse fetal brain. The effect of overexpressing miR-197 alone was almost identical to MeCP2 overexpression in WT C57BL/6 mouse cortex, with more EYFP + cells reached the upper CP layer and less EYFP + cells stayed in the VZ/SVZ layer than in the controls (Fig. 5C). More importantly, i-197 blocked the effect caused by MeCP2 overexpression in WT brain. That is, the number of EYFP + cells reaching the upper CP layer was reduced by i-197 (Fig.   5C). Expression of i-197 in Tg(MECP2) mice could also reverse the effect of elevated MeCP2 expression, as significantly more EYFP + cells remained in the VZ/SVZ and IZ layers and fewer EYFP + cells reached the upper CP layer (Fig. S5B). These data strongly suggest that miR-197 plays a key role in MeCP2 mediated neurogenesis during cortical development.

MECP2 mutations identified in a Chinese ASD cohort failed to promote neurogenesis
A targeted-sequencing of MECP2 gene exons was performed in a Han Chinese cohort consisting of 288 ASD patients and 369 controls. Five rare missense MECP2 mutations were identified in six male ASD patients that were not observed among any of our controls (Table 1).
These mutations were c.590C>T, c.695G>C, c.1112A>G, c.1180G>A, and c.1282G>A, which correlated to T197M, G232A, H371R, E394K and G428S in the human MeCP2-e1 protein (mouse MeCP2-e2) (Fig. S6), represents the isoform highly expressed in the brain (32, 49). The patients carrying these mutations showed a spectrum of symptoms, and the patient carrying the G428S mutation presented the most severe symptoms including severe intellectual disability with no functional language skills, and other medical conditions such as sleep disturbances (Table 1).
The recurrent mutation H371R in our cohort represents a novel mutation in all ExAC populations, a database which currently contains sixty thousand individuals (39). Therefore, we focused on those C-terminal mutations in the follow up studies.
We investigated the effect of three C-terminal mutants on NPCs differentiation. Although these mutations did not affect the expression levels of MeCP2 when overexpressed in NPCs (Fig. S7), they failed to promote neurogenesis in both NPCs and mouse embryonic cortex (Fig. 6B&C). Both MeCP2 H371R and MeCP2 E394K showed loss-of-function effects similar to that of MeCP2 380X , a 11 previously reported loss-of-function truncation mutant (20) in IUE experiments, as they resulted in fewer EYFP + cells reaching the upper CP layer, and more EYFP + cells remaining in the VZ/SVZ layer compared to the WT MeCP2 electroporated brain (Fig. 6B). To be noticed, MeCP2 G428S produced a more severe phenotype when overexpressed in mouse embryonic cortex with most of the MeCP2 G428S overexpressing cells remaining within the VZ/SVZ layers in the mouse cerebral cortex, and very few cells reached the CP at E18.5 (Fig. 6B).
Since the patient with the MeCP2 G428S mutation presented the most severe symptoms compared to other two mutations, and the MeCP2 G428S mutant had the strongest effect in IUE experiment, we went on to inspect the effects of MeCP2 G428S mutation in the in vitro NPC differentiation assay. Overexpression of the MeCP2 G428S mutant led to a significantly reduced percentage of MAP2 + cells (~21%) and a higher percentage of Nestin + cells (~42%), compared to either WT MeCP2 (~87% MAP2 + and ~15% Nestin + cells, respectively) or the control empty vectors (~49% MAP2 + and ~22% Nestin + cells, respectively) (Fig. 6C). The MeCP2 G428S mutant demonstrated a gliogenesis promoting effect during NPC differentiation as indicated by a higher percentage of GFAP + cells (~51%) than WT MeCP2 (~11%), or the control (~33%) (Fig. 6C).

MiR-197 reversed the neurogenesis defects caused by ASD-related MeCP2 mutants
We also investigated the effect of those MeCP2 mutants on the levels of miR-197 and ADAM10.
The results showed that three MeCP2 C-terminal mutants lost the ability to up-regulate pre-miR-197 and matured miR-197 (Fig. 7A&B), or to down-regulate ADAM10 and NICD expression (Fig. S8).
Again, MeCP2 G428S showed some gain-of-function effect evidenced as a significant up-regulation in NICD level in both mouse NPCs and human U251 cells (Fig. S8). However, when miR-197 was co-transfected with those three MeCP2 mutants, it was able to down-regulate the expression of ADAM10 even in presence of the MeCP2 mutants (Fig. 7C). Finally, when miR-197 was co-electroporated with MeCP2 G428S into E14.5 WT mouse brains, the abnormal distribution of cells detected in MeCP2 G428S expressing cortex at E18.5 could be almost completely reversed by exogenous miR-197 (Fig. 7D).

Discussion
Duplication or mutations of the MECP2 gene has been shown to cause MDS and RTT, respectively. 100% of MDS patients and >60% of RTT patients express autistic behaviors, yet the underlying mechanism remains elusive. It has been proposed that altered NSC/NPC differentiation may contribute to the etiology of ASD (50). Again, the mechanism by which MeCP2 affects NSC/NPC differentiation is unknown. In this study, we demonstrated that the dosage and function of MeCP2 is tightly related to embryonic NSC/NPC differentiation. We showed in a MDS animal model that elevated MeCP2 expression promotes neurogenesis by down-regulating ADAM10 and suppressing NOTCH signaling. Moreover, we identified three rare missense mutations in MECP2 in an ASD cohort and found that they are loss-of-function mutants involved in regulating NSC/NPC differentiation. Our results revealed that the defects in a regulatory axis involving MeCP2, miR-197, ADAM10, and NOTCH signaling is associated with the etiology of MDS and/or ASD. Not only can ADAM10 and/or the miR-197 inhibitor inhibit the neurogenesis promoting effect of elevated MeCP2 levels in the Tg(MECP2) mouse brain, but additionally miR-197 can reverse the differentiation defects caused by the ASD-related MeCP2 mutants both in vitro and in vivo. These results indicate that miR-197 must be a critical molecule for the etiology of MDS and ASD.
The role of MeCP2 in NSC/NPC cell fate determination is complex and varies by its dosage, spatial, and temporal pattern. For example, there are studies reporting that Mecp2 deficiency does 13 not affect NPCs fate in the Mecp2-/y mouse embryonic cortex (51, 52), rather it affects the neuron dendritic arborization and complexity in the mouse brain (51-53). However, overexpression of MeCP2 not only promotes neurogenesis of NPCs isolated from embryonic cortex (42), but also causes premature neurogenesis in the developing chick neural tube (33). Herein, we also observed enhanced neurogenesis in Tg(MECP2) mouse and in WT embryonic cortex with ectopic expression of MeCP2. Taken together, MECP2 duplication probably has more severe effects during the early embryonic stages. These observations are consistent with the clinical features of MDS patients, who have severe phenotypes starting from very early infantile stages (9,10).
A surprising finding is that miR-197, which we identified as a crucial molecule downstream of MeCP2, does not regulate human ADAM10-3'UTR through the conventional miRNA seed sequence.
Rather, miR-197 binds to an unconventional site on ADAM10-3'UTR. The fact that this unconventional binding site but not the predicted miR-197 seed sequence is conserved between human and mouse suggests that the interaction between miR-197 to ADAM10-3'UTR is more similar to a siRNA. Although the mouse miR-197 was removed from miRBase in 2014 based on the fact that "the sequence does not map in a stem-loop region of the genomic sequence or any known mouse transcript sequence"; however, we noticed that small RNA libraries were used to identify miR-197 in both human and mouse in the original paper by Landgraf et al. (54). Indeed, a small RNA could be PCR amplified from E18.5 mouse brain by using has-miR-197 specific primers (Fig. S4C&D). It is possible that there might be a small RNA with a similar sequence and function that exists in the mouse brain, even though it does not fit with the typical stem-loop structure of miRNAs.
Another interesting observation is that pre-miR-197 is up-regulated by MeCP2. Previous studies have shown that MeCP2 can both down-regulate and up-regulate miRNAs (20,21,55).
Down-regulation on miRNA, such as pre-miR-134, is due to MeCP2's binding to DGCR8 to interfere with the assembly of Drosha/DGCR8 miRNA processing complex (20). Indeed, pre-miR-134 is still down-regulated by MeCP2 in our system (Fig. 3L). Therefore, the MeCP2 regulatory mechanisms on miR-197 and miR-134 probably are unique to each miRNA. Recently, Tsujimura et al. found that miR-199a is up-regulated by MeCP2 through binding to Drosha and DDX5 (21). It is of interest that both pri-miR-199a (21) and pri-miR-197, but not pri-miR-134, can be experimentally pulled down by MeCP2 (Fig. S9). It is possible that when MeCP2 binds to a pri-miRNA, it facilitate the miRNA processing, but when MeCP2 cannot bind to a pri-miRNA, it interferes miRNA processing. Future studies to investigate whether other miRNAs regulated by MeCP2 also fit such rules are critically important.
It has been proposed that rare inherited mutations are important in the etiology of ASD (56-60).
Indeed, most of the ASD cases with the MECP2 mutations in our cohort inherited the mutation from their asymptomatic carrier mother ( Table 1). As all those ASD patients are boys, mutated MECP2 on the X chromosome would affect them more severely than their mothers. Although we cannot conclude that those mutations are causative without further genetic analysis of their whole genome, it is noticeable that the H371R mutation has two recurrent cases in our ASD patients (2/288), but does not exists in 61,075 controls from both ExAC and our cohort. Taken together with the fact that H371R has similar in vivo and in vitro effects as does the truncation mutation 380X, it is highly possible that H371R is a loss-of-function mutation with a high likelihood of being causative for ASD. The G428S mutation showed more severe effects on NPCs differentiation when compared to the 380X mutation.
How this mutation changes MeCP2's function will need further investigation. One possibility is the

Analyses of mouse cortical neurogenesis in vivo
For quantification of cell fate in WT and FVB-Tg(MECP2)1Hzo/J mice at E18.5 and P7, the regions of the primary somatosensory cortex were identified and the numbers of Sox+, Tbr2+, 16 Satb2+ or Tbr1+ cells were counted in each vertical column with 100 m width. All quantifications were performed with 4 brain sections from at least 3 animals. Data are presented as the mean ± SEM and statistical significance was assessed using unpaired Student's t test.

Primary mouse NPCs isolation and differentiation assay
NPCs were isolated from E12.5 C57BL/6 or FVB mouse embryonic cortex. NPCs isolation and culture methods are based on previous reports (42, 64). NPCs were either transfected with X-tremeGENE HP DNA transfection reagent (Roche, 06366236001) or infected with different lentivirus (Obio Technology (Shanghai) Corp., Ltd.). 72 hours post-transfection or infection, NPCs cells were either lysed for western blot or subjected to immunofluorescent staining and imaging with a Zeiss LSM700 microscope. All films from western blot were scanned and analyzed with Quantity ONE based on intensity or were directly measured with Tanon gel image software. The results were normalized to its corresponding loading control GAPDH.

In utero electroporation and cell counting
In utero electroporation was performed as previously described (65, 66). About 1.5 l of DNA mix was injected into each embryo. The ratio of either expressing plasmid or empty vector to pEYFP was 6:1. The ratio of (miR-197 inhibitor or ADAM10): (WT MeCP2): pEYFP was 3:3:1. The ratio of miR-197, MeCP2 G428S , and pEYFP was 3:3:1. After electroporation and recovery, E18.5 embryos were collected and sectioned at 20m and processed for further immunofluorescent analyses.
Nuclear cell staining with DAPI was used to define different sub-regions of the cerebral cortex based on cell density, as previously described (65). The percentage of EYFP + cells in each layer was calculated based on total number of EYFP + cells in the same brain section. At least three sections from one brain were collected and at least four different embryos obtained from 3-4 different pregnant dams were collected for each group for statistical analyses. .

Plasmids and Lentivirus
Human MECP2-e1 cDNA was purchased from Origene. Point mutations were generated by

qRT-PCR
Total RNA from NPCs or U251 was extracted with RNAeasy kit (QIAGEN, Cat# 74104). The levels of ADAM10 mRNA were determined using the qRT-PCR kit (Takara RR036A and RR820A) with GAPDH as internal control. The levels of mature miR-197 were determined with All-in-One™ miRNA qRT-PCR Detection Kit (GeneCopoeia, QP015). The levels of pri-miR-197 were detected with TaqMan pri-microRNA assay kit (Applied Biosystems) following reverse transcript with Toyobo ReverTra Ace- kit (Toyobo). The levels of pre-miR-197 were detected with miScript Precusor Assay (QIAGEN) following reverse transcript with miScript II RT Kit (QIAGEN).

RNA Immunoprecipitation
U251 cells were transfected with empty vector, WT MeCP2, or ASD-related MeCP2 mutant 18 expressing plasmids. 24 hours later, cells were collected for cross-linking, lysing, and sonication and were subjected to the RNA immunoprecipitation assay using previously described protocols (21,67).
Anti-MeCP2 antibody (ab2828) and rabbit immunoglobulin G (IgG) antibodies were used. The immunoprecipitated RNA was analyzed by qRT-PCR described before.

Statistical analyses
All experiments were repeated at least three times and the statistical significance was evaluated.
Data are expressed as mean ± SEM. Statistical differences were calculated by two-tailed unpaired t test for two datasets and ANOVA followed by Bonferroni post hoc test for multiple datasets using Prism (GraphPad Inc., La Jolla, CA). p<0.05 was considered statistically significant.

Human subjects
Blood samples from 288 ASD patients (mean age 6.1±3.1 years, 85% male) were collected between 2007 and 2010 at the Shanghai Mental Health Center of Shanghai Jiaotong University.
Protocols were reviewed and approved by the Ethics Committee of Fudan University and Shanghai 19 Jiaotong University prior to the commencement of the study. Written informed consent from the parents or guardians of the children was obtained prior to inclusion in the study. The 369 controls (mean age 19 years, 81% male) were unrelated healthy volunteers from the freshman student class at Fudan University (2010), which were ethnically and gender-matched from the same geographical area. DNA was extracted and all samples were pooled together for targeted sequencing on the 2kb upstream, 5'-UTR, 3'-UTR, and coding regions of MECP2 gene was carried out at GBP Biotechnology (Jiangsu, China). Sanger sequencing was performed to confirm those 5 missense mutations in the 6 ASD cases (Fig. S8A).         SEM. * or # p<0.05, ** or ## p<0.01, *** or ### p<0.001. Scale bar is 50m.     FVB mice were dissected out. RNA was extracted and subjected to RT-PCR with either hsa-miR-197 specific primer or control Rnu6 (U6 small nuclear RNA) primers. 8l products and the DNA ladder DL500 (Takara) were run on 12% PAGE gel without urea and stained with Gel-Red for 40min. (D) The PCR product was cloned into T-vector and sent for sequencing. The small RNA has 16 identical nucleotides to the 3' side of has-miR-197. was used for nuclear staining. Representative staining images were presented. Scale bar is 50m.