MiR-124 and small molecules synergistically regulate the direct generation of neuronal cells from rat cortical reactive astrocytes


 Background：Irreversible neuron loss caused by central nervous system injuries usually lead to persistent neurological dysfunction. Reactive astrocytes, because of their high proliferative capacity, proximity to neuronal lineage, and significant involvement in glial scarring, are ideal starting cells for neuronal regeneration. Having previously identified several small molecules as important regulators of astrocyte-to-neuron reprogramming, our aim in this study was to explore whether other small molecules and miR-124, a key neural differentiation mediator, could co-regulate reactive astrocyte-to-neuron conversion.Methods: MiR-124, ruxolitinib, SB203580, and forskolin were used to induce postnatal rat cortex reactive astrocytes, and the neuronal phenotype of the induced cells was characterised. To understand the genetic changes, RNA-sequencing analyses were performed on reactive astrocytes, induced neurons, and rat neurons, and the mechanisms underlying the regulatory role of miR-124 during the neuronal conversion was explored.Results：MiR-124, ruxolitinib, SB203580, and forskolin could co-convert rat cortical reactive astrocytes into neurons. The induced cells had reduced astroglial properties, displayed typical neuronal morphologies, and expressed neuronal markers, reflecting 25.9% of cholinergic neurons. Gene analysis revealed that induced neuron gene expression patterns were more similar to that of primary neurons than of initial reactive astrocytes. On the molecular level, miR-124-driven neuronal differentiation of reactive astrocytes was via targeting of the SOX9-NFIA-HES1 axis to inhibit HES1 expression.Conclusions：Providing a novel approach for inducing endogenous rat cortical reactive astrocytes into neurons by co-regulation involving miR-124 and three small molecules, our research has potential implications for inhibiting glial scar formation and promoting neuronal regeneration after central nervous system injury or disease.

before. Considering that miR-124 drives neuronal differentiation by inhibiting NOTCH signalling via the SOX9-NFIA-HES1 axis, we veri ed that simultaneous modulation of NOTCH, JAK/STAT, and P38 MAPK signals is su cient to convert rat cortex RAs into mainly cholinergic neurons in vitro, opening new avenues for developing future CNS repair therapies.

Methods
Isolation and culture of primary cortical RAs Primary cortical astrocytes were isolated from neonatal Wistar rats (1-3 days old). Brie y, after removal of the blood vessels and pia mater, cerebral cortices were dissected and digested using papain (100 units, 40 penicillin-streptomycin [Invitrogen, 15140122]) in a 5% CO 2 humidi ed incubator at 37 °C. Culture medium was refreshed once every 2 days. About 1 week later, when cultured cells had reached 90% con uence, asks were moved to a shaker at 37 °C (280 rpm/min, 10 h/day) for 2 days. Thereafter, 20 μM cytarabine (Sigma, C1768) was added for 24 h to obtain high purity RAs. Subsequent experiments were performed when GFAP-positive cells exceeded 95% of the total cell number.

MiR-124 transfection of rat RAs
Puri ed RAs were seeded at a density of 5×10 5 cells/well in six-well plates for about 24 h before transfection. Thereafter, cells were transfected with different concentrations of ron-miR-124-3p mimic or corresponding negative control (RiboBio, China) using lipofectamine RNAiMAX (Invitrogen, 13778075).

Immuno uorescence staining
After harvesting, all culture samples were xed in 4% paraformaldehyde for 20 min, permeabilised with 0.2% Triton X-100 in phosphate-buffered saline (PBS) for 20 min, and then incubated with 10% goat serum in PBS at room temperature for 1 h. After blocking, samples were incubated with primary antibodies (Additional le 1: Table S1) overnight at 4 °C. After extensive washing, secondary antibodies (Additional le 1: Table S1) in PBS were added for 1 h in the dark, followed by nuclear labelling with Hoechst33342 for 15 min at room temperature. Five randomly sampled images were captured for each culture using a laser scanning microscope (Olympus) and analysed using ImageJ software.
Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis Total miR and mRNA were extracted using a miR extraction kit (QIAGEN, 217004) and Trizol (Invitrogen, 15596026). Reverse transcription of 1000 ng miRs from the total RNA was performed using the All-in-One miRNA rst-strand cDNA Synthesis kit (GeneCopoeia, QP014); mRNA reverse transcription was performed using the TransScript-Uni One-Step gDNA Removal and cDNA Synthesis SuperMix (Transgen Biotech, AU311). MiR and mRNA PCR ampli cations were performed on the 7300 Real Time PCR System using the All-in-One miRNA qPCR Detection kit (GeneCopoeia, QP011) and NovoStart SYBR qPCR SuperMix Plus kit (Novoprotein, E096-01B), respectively. Expression levels of rat U6 and GAPDH were used as internal controls for miRs and mRNAs; values were calculated as the reverse log of the -ΔΔCT from controls [28]. All assays were performed at least thrice. The primers used for the experiments (Additional le 1: Table S2) were synthesised by Shanghai Sangon Biotech (Shanghai, China).

Western blotting
Total protein (20 mg) was loaded onto 10% or 12% sodium dodecyl sulphate (SDS)-polyacrylamide gel electrophoresis (PAGE) gels and transferred to polyvinylidene uoride membranes (Pall, BSP0161), which were blocked with 5% fat-free milk in Tris-buffered saline containing 0.1% Tween-20 (TBST) for 1 h at room temperature, followed by incubation with primary antibodies (Additional le 1: Table S3) at 4 °C overnight. GAPDH was used as an internal control. After thorough washes with TBST, the membranes were incubated with peroxidase-conjugated secondary antibodies (Additional le 1: Table S3) at room temperature for 1 h. After washing, the signal was detected by chemiluminescence using the BeyoECL Plus (Beyotime, P0018) detection system and then analysed using ImageJ.
RNA-sequencing (RNA-seq) library preparation and data analysis Samples were sent to Sangon Biotech Co., Ltd. (Shanghai, China) for library construction, Illumina sequencing, data pre-processing, and gene mapping. Total RNA extraction was performed using the Total RNA Extractor kit (Sangon Biotech, Shanghai). After RNA purity, concentration, and integrity assessment, 2 μg high-quality RNA per sample were used as input material for RNA sample preparations. Sequencing libraries were generated using VAHTSTM mRNA-seq V2 Library Prep Kit for Illumina®, and index codes were added to attribute sequences to each sample. Subsequently, FastQC (version 0.11.2) was used for evaluating the quality of the sequenced data. Raw reads were ltered using Trimmomatic (version 0.36) and the obtained sequence reads were trimmed and mapped to the reference genome using HISAT2 (version 2.1.0), followed by RNA-seq assessment. StringTie (version 1.3.3b) was used to calculate the transcripts per million (TPM) of protein-coding genes in each sample. For samples with three biological replicates, DESeq (version 1.12.4) was used to obtain the signi cant differentially expressed genes (sDEGs). Functional enrichment analysis was performed using the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses using TopGO (version 2.24.0) and ClusterPro ler (version 3.0.5), respectively.

Co-immunoprecipitation
Appropriate amount of immunoprecipitation lysis buffer (Beyotime, P0013) supplemented with protease inhibitor was added, and the resultant cell lysates were subject to immunoprecipitation with a speci c antibody or IgG control (Beyotime, P7028) and protein A/G agarose beads (Beyotime, P2012) at 4 °C overnight, followed by thorough washes with lysis buffer. Subsequently, the bound proteins were eluted by heating the beads in 2% SDS gel-loading buffer at 95 ℃ for 5 min. Eluted proteins were resolved by SDS-PAGE and then detected by western blotting.

Statistical analysis
GraphPad Prism 7.0 was used for all statistical analyses, with the unpaired t-test for two-group comparisons and one-way or two-way analysis of variance for multiple-group comparisons. All experiments were repeated at least thrice. All data are represented as the mean ± standard deviation (SD). For analyses, p-values <0.05 were considered statistically signi cant.

Results
Identi cation and characterisation of RAs from the postnatal rat cortex To avoid neural progenitor cell (NPC) contamination, only cerebral cortex regions where NPCs were not abundant were used. Cultured cells were expanded in medium containing TGF-β1, which stimulated astrocytes to become RAs [29]. For about one week, until the cells reached 90% con uence, cytarabine was added to remove proliferating oligodendrocytes and microglia. After puri cation, most cells were typically star-shaped or polygonal, producing branching processes from the cell bodies ( Figure 1A). More than 95% of the cells expressed the RA markers GFAP and S100 without DCX or MAP2 positive neuronal contamination ( Figures 1B-1D). This suggested that the cultured cells were mainly RAs.

Characterisation of RAs overexpressing miR-124
To create RAs overexpressing miR-124, we transfected RAs with a miR-124 mimic (miR-124) or a negative control (NC-124). After a 48-h transfection, cell bodies in the miR-124 group became smaller and their branching processes slenderer compared with those in the NC-124 group (Figure 2A). qRT-PCR analysis demonstrated that miR-124 expression increased. The highest transfection e ciency was with 50 nM miR-124 for 48 h (about 23.4 times higher than in the control) ( Figures 2B and 2C). As expected, miR-124 overexpression reduced mRNA levels of the RA markers GFAP and S100 ( Figure 2D). Additionally, GFAP and S100 protein expression was also downregulated after miR-124 overexpression ( Figures 2E and 2F), which was consistent with the mRNA levels. These results suggested that we had successfully constructed miR-124 overexpressing RAs and that miR-124 overexpression suppressed the original astrocytic characteristics of RAs.
RAs synergistically treated with miR-124, ruxolitinib, SB203580, and forskolin acquired neuronal properties Although miR-124 inhibited RA characteristics, it alone could not convert RAs into neurons. We then treated RAs overexpressing miR-124 with ruxolitinib, SB203580, and forskolin ( Figure 3A). This synergistic combination dramatically converted RA morphology into neuron-like shapes after a 7-day treatment. The induced neuronal cell (iN) bodies became smaller and more compact and formed slender axon-like branching structures; no signi cant morphological changes were observed in the NC-124 control group ( Figure 3B). Moreover, miR-124 group cells had smaller territory sizes and longer branching processes ( Figures 3C and 3D). Immunostaining demonstrated that early neuronal markers DCX and TUJ1 and mature neuronal markers MAP2 and NEUN were detectable in miR-124 group on the 7 th day of induction ( Figures 3E-3G); the neurosynaptic junction marker SYN1 was detectable on the 21 st day ( Figure 3H). Based on neuronal marker expression and cell morphologies, the conversion e ciencies for DCX, TUJ1, MAP2, NEUN, and SYN1 were approximately 41.5%, 64.4%, 38.7%, 87.3%, and 0.7%, respectively ( Figure 3M). With regard to neuronal subtypes, our iNs were predominantly CHAT + (25.9%) cholinergic and VGLUT1 + (22.3%) glutamatergic, occasionally TH + (6.1%) dopaminergic, but rarely GAD67 + (1.4%) inhibitory neurons on the 7 th day of induction (Figures 3I-3L and 3N). These data suggested that rat cortical RAs acquired a neuronal fate after synergistic miR-124 and small molecules treatment, mainly forming cholinergic and glutamatergic neurons.
MiR-124, ruxolitinib, SB203580, and forskolin were indispensable for RA-to-neuron conversion To explore whether the same induction effects could be achieved using fewer components, different sets of miR-124 and small molecules were used for induction. The small molecules alone and miR-124 combined with at least one small molecule promoted the generation of DCX-positive neurons at day 7, but conversion e ciency was the highest when miR-124 and the three small molecules were added synergistically ( Figures 4A and 4B). MAP2 immunostaining results were also consistent (Figures 4C and 4D). As JAK/STAT3, P38 MAPK, and ERK1/2 signalling is involved in RA migration, astrogliosis, and proliferation [30][31][32], increased phosphorylation levels would enhance astrocyte activation. Therefore, we assessed the phosphorylation levels in the iNs. Western blot results at day 7 of induction revealed that phosphorylation levels were most downregulated when miR-124 and the three small molecules were added synergistically ( Figures 4E and 4F); this also reduced RA reactivity and bene ted neuronal transdifferentiation the most. These results suggested that the synergistic effects of miR-124 and the three small molecules best promoted RA-to-neuron conversion.
Gene expression pro le of iNs was closer to that of cortical neurons than of RAs To examine the differences between RAs, iNs, and rat neurons at the genetic level, we rst isolated rat cortical neurons and maintained them in a de ned, serum-free NM. At day 10 after puri cation, the cells exhibited a typical neuronal morphology ( Figure 5A). Most cells were immune-positive for TUJ1 and MAP2 ( Figure 5B), suggesting that the isolated cells were mainly neurons. As miR-124 expression gradually increases during neuronal differentiation and maturation [33], we assessed miR-124 expression in iNs. At day 7 of induction, total miR-124 expression was 23.9 times higher than that in control group ( Figure 5C). Considering that exogenous miR-124 decreases gradually ( Figure 2C), the total miR-124 overexpression was mainly caused by increased endogenous miR-124, further indicating RA-to-neuron conversion. To identify the gene expression pro les of RAs, iNs (day 7), and rat neurons (day 10), RNAseq analyses were performed. The number of sDEGs (TPM ≥ 5 in at least one sample, |FoldChange|> 2, qvalue < 0.05) between iNs and RAs (3065) was more than that between iNs and rat neurons (2695) ( Figure 5D), indicating fewer differences between the latter two. Consistently, hierarchical cluster analysis indicated that the gene expression pro le of iNs was more similar to that of rat neurons than of RAs ( Figure 5E), similar to the qRT-PCR results ( Figure 5F). These results suggested that RAs silenced their astrocytic transcriptional programs and adopted a neuronal identity during induction.
As the induction e ciency was incomplete, we were interested in whether prolonged induction time or environmental factors would affect the fate of the remaining RAs. Therefore, we cultured the 21-day iNs in AM and NM. On the third day of continuous culturing, no morphological changes were observed. On the 11 th day, cell morphology in NM was still unchanged, similar to that of rat neurons. However, cell numbers in AM increased, and the morphology was similar to that of RAs (Additional le 1: Figure S1A), which might be because serum exposure allowed the remaining RAs to proliferate. These results indicated that prolonged exposure and a neuronal environment could not change the fate of non-neuronal transformed RAs. Subsequently, we assessed the gene expression of these cells, RAs, and rat neurons. The genetic pattern of cells in AM resembled that of RAs and that of cells in NM resembled that of rat neurons (Additional le 1: Figure S1B).

Transcriptome analysis of iNs and RAs
To explore the molecular and cellular changes of iNs compared with their initial cells, we further analysed the transcriptome of iNs and RAs. At day 7 of induction, the transcriptome scatterplot analysis revealed that, of the 3065 sDEGs, 1820 were upregulated and 1245 were downregulated ( Figure 6A). The expression of RA-enriched genes (e.g., Gfap, Aldh1l1, Vim, Nes, Slit2, Axin2, and Mmp2) was decreased, while that of neuron-enriched genes (e.g., Dcx, Tuj1, Syt1, Kcnc1, Map2, Syn1, and Ascl1) was signi cantly increased. Additionally, expression of genes involved in brain development or cholinergic neuron differentiation (e.g., Foxg1, Bcl11b, and Myt1l) was also signi cantly upregulated. Further, qRT-PCR results con rmed expression patterns of the selected genes, which were consistent with the RNA-seq results ( Figure 6B). The GO analysis (Biological process, Cellular component, and Molecular function) showed that the upregulated genes were signi cantly enriched in neuron projection morphogenesis, neuron differentiation, axonogenesis, and synaptic vesicle ( Figure 6C), and the downregulated genes were signi cantly enriched in cell proliferation, cell migration, cell-substrate adhesion, and proteoglycan binding ( Figure 6D), suggesting that the astrogliosis and scar formation properties of RAs had weakened, while neuronal characteristics had increased. Additionally, KEGG analysis revealed that upregulated cyclic guanosine monophosphate-dependent protein kinase G and cyclic adenosine monophosphate (cAMP) signalling, soluble NSF attachment protein receptor interactions in vesicular transport, and other upregulated signals were mainly involved in neuron development, axon formation, and neurotransmitter release ( Figure 6E), while downregulated MAPK signalling, cell adhesion molecules, PI3K-AKT signalling, and other downregulated signals were related to astrogliosis and glial scarring of RAs ( Figure 6F), further pointing to RA-to-neuron conversion. Together, the transcriptional regulatory network of RAs was reprogrammed toward that of neuronal lineage under our induction.
MiR-124 regulated neuronal differentiation of RAs via the SOX9-NFIA-HES1 axis To clarify the molecular mechanism involved in the regulatory role of miR-124, we focused on the NOTCH downstream effectors SOX9 and NFIA, characteristic RA TFs [34], as well as HES1, whose oscillations are related to neurogenesis [35]. MiR-124 overexpression notably repressed SOX9, NFIA, and HES1 expression ( Figures 7A and 7B), indicating the regulatory effect of miR-124 on them. Considering that Sox9 is a target of miR-124 [36], we explored the relationship between SOX9, NFIA, and HES1 in RAs. After transfection with si-Sox9 (Additional le 1: Figure S2A), Sox9 mRNA and protein levels were signi cantly reduced ( Figures 7C-7E). Moreover, NFIA and HES1 expression was also downregulated with reductions in SOX9 levels ( Figures 7D and 7E), suggesting that SOX9 regulated NFIA and HES1. Additionally, a physical association between SOX9 and NFIA con rmed by co-immunoprecipitation further indicated that SOX9 directly downregulated NFIA ( Figure 7F). Subsequently, we transfected RAs with si-N a (Additional le 1: Figure S2B), and N a mRNA level and protein expression were remarkably decreased ( Figures 7G-7I). Moreover, HES1 also decreased upon reduced levels of NFIA ( Figures 7H and 7I), which reduces HES1 expression by binding to the Hes1 promoter region [37], suggesting the direct repression of NFIA on HES1. Altogether, miR-124 could downregulate HES1 expression by targeting the SOX9-NFIA-HES1 axis in RAs. To explore whether this axis was related to neurodifferentiation, we transfected RAs with si-Hes1 (Additional le 1: Figure S2C), which signi cantly downregulated Hes1 mRNA expression (Figures 7J-7L). Furthermore, the mRNA expression of Ascl1, a gene related to neuronal differentiation, was signi cantly elevated by Hes1 silencing (Figure 7J), revealing the negative regulation of Hes1 in neuronal differentiation. Simultaneously, we detected the effect of Sox9 or N a inhibition on Ascl1. The results showed that Sox9 or N a inhibition could also increase Ascl1 mRNA expression (Additional le 1: Figure  S2D), but not as much as Hes1 inhibition, further indicating the indispensable role of Hes1 in this regulatory axis. These results suggested that in RAs, miR-124 inhibited NOTCH signalling by targeting the SOX9-NFIA-HES1 axis, thereby promoting RA-to-neuron conversion.
HES1 was involved in the RA-to-neuron conversion coregulated by miR-124, ruxolitinib, SB203580, and forskolin To decipher whether the SOX9-NFIA-HES1 axis operated during the induction after the small molecules were added, we rst induced the aforementioned si-Hes1 RAs with the three small molecules for 3 days, and compared their HES1 expression with that in RAs, RAs treated with the three small molecules only, and RAs treated with both miR-124 and the three small molecules. Western blotting showed that HES1 expression was signi cantly suppressed in the miR-124 and small molecules group (Figures 8A and 8B). Immunostaining for DCX demonstrated that conversion e ciency in the miR-124 and small molecules group was better than those in other groups (Figures 8E [a-d] and 8F). Given that the miR-124 and small molecules group had the lowest HES1 expression but the best conversion e ciency; we speculated whether HES1 expression determined differences in conversion e ciency. We then applied DAPT, a NOTCH pathway inhibitor, to further suppress HES1 expression. For better effects, we also prolonged induction time to 7 days. As expected, DAPT application further suppressed HES1 expression in the miR-124 and small molecules group (Figures 8C and 8D), but conversion e ciency was still lower than that in the miR-124 and small molecules group without added DAPT ( Figures 8E [e-h] and 8G), demonstrating that apart from suppressing HES1 expression, the combination might also regulate neuronal conversion by other means. These results suggested that HES1 expression was involved in RA-to-neuron conversion co-regulated by the miR-124 and small molecule cocktail, but this was not the only mechanism.

Discussion
Astrocytes, a major cell population in the mammalian CNS, play multiple important roles in CNS pathophysiology [38,39]. When pathological conditions affect the CNS, astrocytes become reactive and are characterised by morphological changes (hypertrophy and process extension) and the upregulation of intermediate lament proteins such as GFAP and VIMENTIN [3,40]. Moreover, the production of chondroitin sulphate proteoglycans and release of pro-in ammatory cytokines further drives RAs to form glial scars, inhibiting neuronal regeneration [41].
Recently, the use of diverse small molecule cocktails to reprogram astrocytes into neuronal cells has achieved considerable success [13,14]. Yin et al. proved that using three to four small molecules (SB431542, CHIR99021, LDN193189, and DAPT) or their respective functional analogues to modulate the TGF-β/SMAD, BMP, NOTCH, and WNT pathways was su cient to reprogram human foetal astrocytes into neurons [14]. However, RA characteristics are also regulated by other signals, such as the JAK/STAT and P38 MAPK pathways and post-transcriptional levels of several non-coding RNAs. Here, we developed a new miR-124, ruxolitinib, SB203580, and forskolin cocktail, proving that this combination could signi cantly alter RA characteristics and promote RA-to-neuron conversion.
Studies have revealed the functional signi cance of miRs in regulating CNS development [42]. MiR-124, one of the most highly expressed miRs in the mammalian brain, is closely associated with the regulation of neurogenesis and neuronal differentiation [43]. Given that both astrocytes and neurons are derived from NSCs and that miR-124 is a neuronal fate determinant during neurodevelopment [44], we hypothesised that differential miR-124 expression might affect their distinct cell fates. Indeed, we found that miR-124 expression in astrocytes, especially RAs, was much lower than that in neurons (Additional le 1: Figure S3). Moreover, GFAP, S100, SOX9, and NFIA expression was signi cantly downregulated after miR-124 overexpression, indicating that miR-124 upregulation could e ciently suppress inherent RA characteristics and might promote RA-to-neuron conversion. Additionally, as each miR regulates hundreds of target genes [45], and Smad4 (involved in both the TGF-β and BMP pathways), Dact1, and Axin1 (negative regulators of the WNT/β-catenin cascade) are all targets of miR-124 [46][47][48][49], we hypothesised that miR-124 might replace TGF-β and BMP pathway inhibitors as well as WNT/β-catenin pathway activators in RAs and implement functions similar to those of the SB431542, LDN193189, and CHIR99021 combination during RA-to-neuron reprogramming. As expected, when we randomly selected SB431542 to add to the miR-124 and three small-molecule cocktail, no signi cant increase in DCXpositive conversion e ciency was observed (Additional le 1: Figures S4A and S4B). Considering its numerous targets, miR-124 might promote RA-to-neuron conversion through other unknown mechanisms, and eventually the effects of our cocktail were better than of the above three chemicals. However, miR-124 itself was not su cient to reprogram RAs into neurons in vitro, suggesting that signals other than TGF-β, BMP, and WNT, such as JAK/STAT [3] and P38 MAPK [31], that are crucial for maintaining RA characteristics, might be constantly activated and impair the transformation process. We thus hypothesised that additional inhibition of these signals might achieve more signi cant conversion e ciency.
JAK/STAT pathway activation is reported to be involved in CNS injuries, especially gliogenesis and glial scarring formation [30,50]. Moreover, Jak3 silencing or Stat3 ablation promoted NPC neuronal differentiation and NSC neurogenesis, respectively [51,52], revealing the neuronal promotion of JAK/STAT inhibition. Therefore, we speculated that JAK/STAT pathway suppression offers great possibilities for inhibiting inherent RA properties and promoting RA-to-neuron conversion. As expected, the conversion e ciency was signi cantly increased after treatment with ruxolitinib and miR-124. To the best of our knowledge, this is the rst time a JAK/STAT pathway inhibitor has been used for neuronal induction.
P38 MAPK and ERK1/2 activation is also associated with astrogliosis [53]. During reactive gliosis, ASK1 phosphorylation is enhanced, and P38 MAPK relays the signal to astrocytes for activation and glial scar formation both in vitro and vivo [54,55]. Thus, suppressing P38 MAPK signalling was more likely to attenuate RA characteristics and favour neuronal induction. Here we utilised SB203580, a common P38 inhibitor, to repress RA characteristics and promote neuronal conversion. Furthermore, forskolin is a wellknown cAMP and PRKACA activator and promotes neuron survival and neuronal conversion e ciency [56][57][58]. As expected, the three small molecules promoted RA-to-neuron conversion, albeit ine ciently.
However, when miR-124 was also added, e cient conversion was obtained, further indicating the importance of combinatory functions.
Although the functional mechanisms of the three small molecules are clear, the function of miR-124 is more complicated. Here, we mainly focus on miR-124 and NOTCH signalling, a highly conserved morphogenic pathway that regulates NSC proliferation, self-renewal, and neuronal differentiation [59].
Attenuated NOTCH1 signalling after stroke is necessary for striatal astrocyte neurogenesis in mice [60], indicating that RA-to-neuron conversion is related to NOTCH signalling. Here, we revealed that miR-124 promotes neuronal differentiation of RAs by targeting the SOX9-NFIA-HES1 axis and eventually elevates the expression of Ascl1 (Fig. 9), a proneural target gene of Hes1 that promotes neural differentiation [61]. This was consistent with previous studies showing that expression of the neural TF ASCL1 alone could reprogram astrocytes into neurons [62]. Nevertheless, persistent miR-124 overexpression alone did not produce DCX-or MAP2-positive RA-derived neurons. This might be because neurodifferentiation is promoted only when proneural activity is high enough [63]. Although Ascl1 was upregulated in RAs, the levels were not high enough for neuronal conversion. Moreover, HES1 expression disappears in differentiating neurons, leading to sustained ASCL1 expression, inducing neuronal differentiation [64]. In our study, miR-124 indirectly inhibited HES1 but failed to eliminate its expression. Thereafter, when the small molecules were added to regulate additional RA signals, we obtained DCX-and MAP2-positive iNs; however, when DAPT was added to further suppress HES1 expression in RAs, no more DCX-positive iNs were obtained. This might be because HES1 is a repressor-type basic helix-loop-helix protein, while ASCL1 is of the activator-type. When NOTCH signalling is inhibited, decreased HES1 expression upregulates ASCL1 to simultaneously promote neuronal fate determination and NOTCH ligands expression, activating NOTCH signalling and upregulating HES1 expression in neighbouring cells to maintain them in an undifferentiated state [65]. Therefore, HES1 and ASCL1 regulate each other, only allowing some cells to undergo neuronal differentiation. This may also explain why the remaining undifferentiated RAs in our study could not acquire a neuronal identity even with prolonged induction. Of course, the synergy between miR-124 and the other three small molecules is extremely complex, and this analysis is from the limited perspective of HES1 expression alone. Whether this synergism regarding RA-to-neuron conversion could function in other ways needs to be further explored.
The neuronal subtypes after conversion from astrocytes depend on the combinations of inducing chemicals. Previous nine-and four-molecule cocktails mainly converted astrocytes into glutamatergic neurons through upregulation of NEUROD1 and NGN2 [14,66]. Here, our RNA-seq results showed that Foxg1 and Myt1l mRNA expression was signi cantly upregulated. FOXG1 is one of the earliest TFs speci cally induced in forebrain NPCs [67]. MYT1L, the only known TF speci cally expressed in all neurons throughout life, is critical for maintaining neuronal identity [68,69]. Thus, our induction modulated RAs to obtain a neuronal identity via upregulating Foxg1 and Myt1l. Interestingly, unlike with other protocols, many cholinergic neurons appeared in our experiments. The qRT-PCR results also revealed that expression of genes involved in cholinergic neuron differentiation (e.g., Chat, Lhx8, Lhx6, and Tbr1) was signi cantly upregulated in these iNs (Additional le 1: Figure S5). The cholinergic hypothesis of Alzheimer's disease (AD) proposes that brain acetylcholine depletion is a primary cause of cognitive decline in advanced age and AD [70,71]. Moreover, cholinergic nuclei are differently affected by the pathogenetic mechanisms underlying major neurodegenerative diseases such as AD, PD, and Huntington disease [70]. Given that astrocyte reactivity is a hallmark of the pathological CNS, RA-derived cholinergic neurons might be potential targets for the treatment of neurodegenerative diseases. Recently, Liang et al. reported that MYT1L could directly reprogram human brain vascular pericytes into cholinergic neurons [72], indicating the Myt1l expression in our experiments might be essential for converting and maintaining the cholinergic neuron phenotype. Although these mechanisms remained to be further explored, we believe that the different effects depended on our unique induction cocktail.
Glial scars formed by RAs after CNS injuries ultimately impair neuronal regeneration; therefore, ameliorating the cellular properties of glial scars is essential for neural repair. Although 41.5% RAs in this study were successfully converted into DCX-positive neurons, with continuous miR-124 and small molecules treatment, the remaining undifferentiated RAs would lose their reactivity to form glial scars, and further contact the iNs to support their maturation. Collectively, utilising our cocktail simultaneously achieved the two goals of reprogramming RAs into neurons and repressing RA cellular characteristics, which is signi cant for the treatment of CNS injuries.
Although the use of small molecules to reprogram cells has achieved remarkable success, they should be non-toxic and approved after clinical safety trials. Ruxolitinib, the rst clinically effective, selective JAK1/2 inhibitor approved by the Food and Drug Administration, can penetrate the blood-brain-barrier and reduce astrogliosis in mice with HIV encephalitis when systemically administered [73]. Additionally, losmapimod, a P38 MAPK α and β isoform inhibitor functionally analogous to SB203580, has entered stage I clinical trials for treating facioscapulohumeral muscular dystrophy. Forskolin is a natural compound used for centuries in traditional medicine. These chemicals selected for our formula or their functional analogues are closer to clinical application and more feasible for use.

Conclusions
In summary, this study shows that the combination of miR-124, ruxolitinib, SB203580, and forskolin can Availability of data and materials: The datasets used and/or analysed during the current study are available from the corresponding author (Guangfan Chi, guangfan130@jlu.edu.cn) on reasonable request. GC, CQ, and YL conceived the study, designed the experiments, and together with YZ revised the manuscript. YZ performed the cellular and molecular experiments in vitro, with help from ZH and JX. KH, YY, SL, and LC provided reagents and conducted the data analysis and interpretation. YZ wrote the manuscript. All authors approved the nal version of the manuscript. shows that the majority of cultured cells expressed the RA markers, GFAP and S100, but not the neuronal markers DCX and MAP2. (D) Statistical analysis shows that >95% of all cells were GFAP-and S100positive RAs, but no DCX or MAP2-positive cells were detected. All data are reported as mean ± SD; n = 3 independent experiments. Scale bars, 50 μm.       The RA-to-neuron conversion co-regulated by miR-124 and small molecules was associated with HES1 expression (A-B) Western blot (A) and densitometric quantitation (B) of HES1 expression in RAs, RAs treated with small molecules for 3 days, RAs treated with si-Hesl and small molecules for 3 days, and RAs treated with miR-124 and small molecules for 3 days. (C-D) Western blot (C) and quantitation (D) by densitometry for HES1 expression in RAs treated with miR-124 and small molecules for 7 days, RAs treated with miR-124, small molecules, and DAPT for 7 days, RAs treated with small molecules and DAPT for 7 days, and RAs treated with miR-124 and DAPT for 7 days. (E-G) Immunostaining (E) and the percentage (F-G) of DCX-positive cells in RAs, RAs after induction with miR-124/si-Hes1/DAPT and/or small molecules for 3/7 days at different combinations. All protein expression levels were normalised to GAPDH. All experiments were repeated thrice, and all data are presented as mean ± SD (*P< 0.05; **P< 0.01). Scale bars, 50 μm. n.d., not detectable.