MicroRNA-375 Is Induced during Astrocyte-to-Neuron Reprogramming and Promotes Survival of Reprogrammed Neurons when Overexpressed

Abstract

While astrocyte-to-neuron (AtN) reprogramming holds great promise in regenerative medicine, the molecular mechanisms that govern this unique biological process remain elusive. To understand the function of miRNAs during the AtN reprogramming process, we performed RNA-seq of both mRNAs and miRNAs on human astrocyte (HA) cultures upon NeuroD1 overexpression. Bioinformatics analyses showed that NeuroD1 not only activated essential neuronal genes to initiate the reprogramming process but also induced miRNA changes in HA. Among the upregulated miRNAs, we identified miR-375 and its targets, neuronal ELAVL genes (nELAVLs), which encode a family of RNA-binding proteins and were also upregulated by NeuroD1. We further showed that manipulating the miR-375 level regulated nELAVLs’ expression during NeuroD1-mediated reprogramming. Interestingly, miR-375/nELAVLs were also induced by the reprogramming factors Neurog2 and ASCL1 in HA, suggesting a conserved function to neuronal reprogramming, and by NeuroD1 in the mouse astrocyte culture and spinal cord. Functionally, we showed that miR-375 overexpression improved NeuroD1-mediated reprogramming efficiency by promoting cell survival at early stages in HA and did not appear to compromise the maturation of the reprogrammed neurons. Lastly, overexpression of miR-375-refractory ELAVL4 induced apoptosis and reversed the cell survival-promoting effect of miR-375 during AtN reprogramming. Together, we demonstrated a neuroprotective role of miR-375 during NeuroD1-mediated AtN reprogramming.

1. Introduction

Despite controversies around the validation of reprogrammed neurons in vivo with lineage tracing techniques [1,2,3], glia-to-neuron reprogramming has been successfully demonstrated in several laboratories [4,5,6,7]. This has been achieved mainly through overexpression of neurogenic transcription factors, either alone or in combination, in various neurological disease or injury models [8,9,10]. One such factor, Neuronal Differentiation 1 (NeuroD1), has been shown to convert astrocytes, neuron-glia antigen 2-expressing glial cells (NG2 glia), and microglia into functional neurons both in vitro and in vivo [4,11,12,13]. Our previous study also demonstrated that forced expression of NeuroD1 reprograms reactive astrocytes to functional neurons in the injured spinal cord [11]. NeuroD1 is widely expressed in the developing central nervous system (CNS) and is critical to neuronal differentiation [14]. NeuroD1 is also a “pioneer” transcription factor [15], it and reprograms the chromatin landscape to elicit neuronal programming in embryonic stem (ES) cells [16] and microglia [12] when overexpressed. Re-expressed NeuroD1 can activate expression of downstream target genes such as Hes Family BHLH Transcription Factor 6 (Hes6) and NeuroD4 [16], which may serve as important effectors for neuronal conversion. While neuronal reprogramming has become a feasible approach for neuroregeneration, our understanding of the molecular mechanisms that govern this unique biological process is still incomplete.

MicroRNAs (miRNAs) are small non-coding RNAs that regulate gene expression post-transcriptionally [17], and they are crucial to the differentiation of neural cell types during CNS development [18]. Previous work from our lab and others demonstrated that miRNAs are important to cellular differentiation in the developing mouse forebrain [19,20,21,22] and cerebellum [23,24,25,26], and they are indispensable to reactive astrogliosis during spinal cord injury [27]. However, the function of miRNAs during neuronal reprogramming has not been systemically investigated.

In this study, we aimed to decipher miRNA function during NeuroD1-mediated neuronal conversion using human astrocyte (HA) culture as a model. Our RNA-seq analyses showed that NeuroD1 overexpression induced drastic upregulation of two miRNAs, miR-375 and miR-124. We experimentally demonstrated that miR-375 modulated the expression level of neuronal ELAV like RNA binding proteins genes (nELAVLs), which encode RNA-binding proteins and were also upregulated by NeuroD1. Interestingly, miR-375/nELAVLs were also induced by the reprogramming factors Neurogenin 2 (Neurog2) and Achaete-scute homolog 1 (ASCL1) in HA, and by NeuroD1 in the mouse astrocyte culture and spinal cord. Furthermore, overexpression of miR-375 by a retrovirus promoted cell survival during NeuroD1-mediated astrocyte-to-neuron (AtN) reprogramming at early stages without compromising the maturation of reprogrammed neurons in HA cultures. Thus, our data indicate that miR-375 facilitates NeuroD1-mediated reprogramming by modulating expression level of target genes including nELAVLs, and that combinatory overexpression of NeuroD1 and miR-375 may represent an improved reprogramming strategy with a greater translational value.

2. Materials and Methods

2.1. Animal Use

Twelve wild-type C57BL/6N mice (2–4 months old, eight females and four males) were used for Adeno-Associated Virus (AAV) injection experiments. Mice were housed in a 12 h light/dark cycle and supplied with sufficient food and water. All animal use and studies were approved by the Institutional Animal Care and Use Committee (IACUC) of Augusta University. All procedures were carried out in accordance with the approved protocols and guidelines of National Institute of Health (NIH).

2.2. Virus Production

Retroviruses expressing NeuroD1-GFP and GFP control have been previously described [4]. For NeuroD1-GFP, mouse NeuroD1-coding sequences amplified from a template of the pAd NeuroD-I-nGFP were inserted into a pCAG-GFP-IRES-GFP retroviral vector. For NeuroD1-RFP, mouse NeuroD1-coding sequences were subcloned into pCAG-Neurog2-IRES-DsRed vector [28] to replace Neurog2 by PCR-based strategy. For miR-375-GFP, a 450bp genomic sequence containing human miR-375 mature sequence was cloned into NeuroD1-GFP vector by replacing NeuroD1 at PmeI and NotI sites. For ELAVL2-GFP and ELAVL4-GFP, human gene coding sequences were cloned into NeuroD1-GFP vector by replacing NeuroD1 at PmeI and NotI sites. Lenti-miR-375-decoy was a gift from Dr. Brian Brown (Addgene plasmid #46617).

Retrovirus packaging was performed as described [4]. Briefly, HEK293T cells were co-transfected with the recombinant viral vectors containing transgenes and the plasmids encoding GAG-POL genes and VSV-G. The culture supernatants were collected at day 3 and day 4 after transfection. For retro-miR-375-GFP packaging, a plasmid containing human Dicer short hairpin RNA (shRNA) [pSicoR human Dicer1, a gift from Tyler Jacks (Addgene plasmid #14763)] was included to increase virus yield [29,30]. For lentivirus packaging, the procedure was similar to that of retrovirus except that the packaging plasmids psPAX2 and VSV-G were used. Virus-containing culture media were centrifuged and filtered to remove cell debris and then aliquoted and stored at −80 °C before use. Virus media were routinely assayed for titer by infecting HEK293T cells. Our retrovirus and lentivirus titers were usually at 10⁷ genomic copies (GC)/mL. AAV viral particles were produced using the established procedure [31]. HEK293T cells were transfected with three plasmids: a pAAV of interest (pAAV-Flex-GFP, pAAV-Flex-ND1-GFP, pAAV-GFAP-Cre), pAAV2/5 (Addgene Plasmid #104964), and pHelper (Cell Biolab) vector with 1:4:2 ratio, respectively. Medium containing viral particles was collected at 72 and 120 h after transfection. After centrifuging at 2000× g for 15 min, the supernatant was precipitated in 40% polyethylene glycol (PEG) in 2.5 M NaCl. Following centrifuging this PEG-media mixture at 4000× g for 30 min, we combined the PEG pellets with the cell pellets in lysis buffer (500 mM NaCl, 40 mM tris, 10 mM MgCl₂, and salt-active nuclease). The resulting lysates were extracted from an iodixanol step gradient following ultracentrifugation of 350,000× g for 2.5 h. Purified AAVs were concentrated with Amicon Ultra-15 centrifugal filter devices. The AAV titers were determined by quantitative PCR (qPCR) as described [32]. AAV mixtures with a titer of 1 × 10¹² GC/mL were used for spinal cord injection experiments.

2.3. Human Astrocyte Culture, Virus Infection, and Cisplatin Treatment

Human Astrocytes (HA) were purchased from ScienCell Research Laboratories (Catalog No. #1800) and cultured in HA culture medium as described previously [4]. HA cells were cultured in HA culture medium (HAM), which contains DMEM/F12 (GIBCO) supplemented with 10% fetal bovine serum, 0.4% penicillin/streptomycin (GIBCO), 3.5 mM glucose, B27, 10 ng/mL epidermal growth factor (EGF, Invitrogen), and 10 ng/mL fibroblast growth factor 2 (FGF2, Invitrogen). For infection, HA cells were seeded on the poly-D-lysine (MP Biomedicals, #150175, Irvine, CA, USA) coated coverslips (Bellco Glass, #1943-10012A, Vineland, NJ, USA) with the confluency of 70–80%. In co-infection experiments, spinfection was performed to increase infection efficiency [33]. Briefly, virus-containing media were mixed with polybrene (2 µg/mL) before being applied onto HA cultures. The resulting culture plates were centrifuged at 1000× g for 45 min at room temperature and then cultured in CO₂ incubator at 37 °C. After 5 h, the virus medium was replaced with HAM. At one day post virus infection, cultures were subject to medium change from HAM to neuron differentiation medium (NDM), which contains DMEM/F12 (GIBCO, Waltham, MA, USA), 0.5% FBS, 0.4% penicillin/streptomycin (GIBCO), 1 μM Y27632, 5 μg/mL vitamin C, B27 and N2 supplement (GIBCO) [34]. For long-term cultures, NDM was changed every four days. Brain-derived neurotrophic factor (BDNF, 20 ng/mL, Invitrogen, Waltham, MA, USA) was supplemented into NDM to support survival of reprogrammed neurons [4]. Treatment of Cisplatin (Sigma-Aldrich, Burlington, MA, USA) was performed at indicated concentrations for 24 h in HA cultures before fixation and immunostaining.

2.4. RNA-seq and Bioinformatics Analyses

For RNA-seq, HA cells were seeded on the poly-D-lysine (MP Biomedicals, #150175) coated 6-well plates with the confluency of 70–80%. Total RNAs were extracted from HA cells infected with retro-GFP or ND1-GFP for 2 and 6 days by using TRIzol^® reagent (Invitrogen, Waltham, MA, USA) (Figure 1a). The integrity of the total RNAs was assessed using the Bioanalyzer 2100 (Agilent). We then made a uniquely indexed library from each sample using the Illumina TruSeq^® Stranded mRNA library kit (Illumina). We also separately made a uniquely indexed miRNA library from each sample using the Illumina Small RNA library kit (Illumina). We made an equimolar pool of the mRNA libraries and a separate equimolar pool of the miRNA libraries, and then mixed the two library pools together at a ratio of 85 parts mRNA library pool:15 parts miRNA library pool. A 75-nt single-read sequencing run on this combined pool was performed on the Illumina MiSeq^® sequencer (Illumina) at Penn State Genomics Core Facility. The reads were divided up as ~25 million reads per sample for the mRNA libraries and ~5 million reads per sample for the miRNA libraries.

For both mRNA and miRNA sequencing, FastQC (v0.11.9) was used to check the quality and adaptor contents of the reads. For mRNA sequencing, adaptor contents were low, and no trimming was needed. The Fastq read files were quantified using kallisto (v0.46.1) with the transcriptome indices built on the cDNA FASTA files from human genome assembly GRCh38. Genes with lower than 10 total read counts were filtered out for subsequent differential expression analysis. Differential expression analysis was performed using Deseq2 (v1.32.0) by comparing different samples, and genes with fold changes >2 and p-value < 0.05 were called as differentially expressed. Gene Ontology (GO) Consortium and KEGG pathway databases were used to perform GO and pathway enrichment analysis in DAVID (https://david.ncifcrf.gov). Gene set enrichment analysis (GSEA) was employed to assess cell type signatures in ND1 group from cluster markers identified in single-cell sequencing studies of human tissue (C8: cell type signature gene sets). Venn diagram analyses were performed in InteractiVenn (http://www.interactivenn.net/). The protein–protein interaction networks were predicted using STRING database (http://string-db.org). For miRNA sequencing, the reads were trimmed using Trim Galore! 9 (v0.6.5) for adapter sequences and any trimmed reads greater than 30 bp were filtered. The remaining reads were aligned to the genomic DNA sequence from human genome assembly GRCh38 using Bowtie2 (v2.4.3), and the aligned reads for each miRNA were quantified with featureCounts in the Subread package (v2.0.3) according to the sequences of all miRNA hairpins from miRbase. MiRNAs with lower than five total read counts were filtered for subsequent differential expression analysis. Differential expression analysis was performed using Deseq2 (v1.32.0) by comparing different samples, and miRNAs with fold changes >1.5 and p-value < 0.05 were called as differentially expressed.

2.5. Laminectomy, Spinal Cord Injury, and Stereotaxic Viral Injection

Mice were anesthetized using a SomnoFlo™ Low-flow electronic vaporizer (Kent Scientific Corp., Torrington, CT, USA) connected with isoflurane. A laminectomy and stab injury were then performed as previously described [11]. Briefly, after exposing the dorsal surface of the spinal cord at the T11–T12 vertebrae, the longitudinal stab injury was conducted with a 26-gauge needle with 0.8 mm in depth and 2 mm in length. The lesion site is 0.3 mm lateral to the central artery. By using a 10-μL Hamilton syringe with a 34-gauge injection needle, 1 μL of AAVs was immediately injected at the proximal and distal lesion sites at a depth of 0.7 mm and at a rate of 0.1 μL/min and the needle was gradually moved up to a depth of 0.3 mm during the injection. The injection needle was kept in place for 5 min after injection to prevent drawing out the virus during withdrawal and then slowly withdrawn. The mice were kept on a heating pad and treated with carprofen (5 mg/kg) for pain relief via subcutaneous injection.

2.6. Mouse Astrocyte Isolation and Culturing

For mouse astrocyte culture, postnatal (P1–P3) mouse cortical tissues were carefully separated from the meninges, cut into small pieces using sharp blades, and subject to digestion using 2 mL of 0.25% trypsin at 37 °C for 30 min with frequent shaking. The digestion was stopped by adding 10 mL of astrocyte culture medium followed by dissociation into single cell suspension by pipetting. The resulting cell suspensions were plated onto 25 cm² flasks with the density of about 50,000 cells/cm² [4]. Cells were cultured for 7 days, and flasks were rigorously shaken to remove small non-astrocytic cells. After reaching confluence, astrocytes were passaged and plated on poly-D-lysine coated coverslips before being infected by retroviruses the following day. One day later, the medium was changed to NDM for reprogramming. Astrocyte culture medium contained DMEM/F12 (Gibco), 10% FBS (Gibco), penicillin/streptomycin (Gibco), 3.5 mM glucose (Sigma), and supplemented with B27 (Gibco), 10 ng/mL EGF and FGF2 (Invitrogen).

2.7. Fluorescence Activated Cell Sorting (FACS)

HA cells were cultured with HAM in 5% CO2 incubator at 37 °C. On the day before infection, HA cells were passaged and plated at 70–80% confluence in 6-well plates. Co-infection with ND1-RFP and miR-375-GFP retroviruses was performed with spinfection. After 5 h, the virus-containing medium was replaced with HAM. Three days after infection, cells were dissociated with 0.05% trypsin and then resuspended in medium containing 2% FBS. To exclude non-viable cells, 1 μL of 7-AAD staining solution (1 mg/mL, Enzo Life Sciences, Inc., Farmingdale, NY, USA) was added to cell suspension prior to sorting. GFP and RFP cell sorting was performed on the MoFloXDP 2-laser, 7-color cell sorter using Summit software (Beckman Coulter Life Sciences). The sorted RFP+ only and RFP+/GFP+ cell populations were collected and replated on coverslips with pre-seeded HA cells in NDM. The sorted cells were maintained in culture and fixed with 4% paraformaldehyde (PFA) for 15 min at the indicated time for immunocytochemistry.

2.8. Immunocytochemistry and Immunohistochemistry

Immunocytochemistry was carried out as previously described [4]. Briefly, fixed cell cultures with 4% PFA were incubated with monoclonal antibodies against ELAVL4 (E1) (mouse IgG, 1:200, #sc-28299, Santa Cruz, Dallas, TX, USA) and NeuroD1 (mouse IgG, 1:200, #ab60704, Abcam, Boston, MA, USA); polyclonal antibodies against GFP (chicken IgY, 1:400, Aves Labs, Davis, CA, USA), mCherry/RFP (rabbit IgG, 1:500, #ab167453, Abcam), NeuN (rabbit IgG, 1:400, #ab104225, Abcam), microtubule associated protein 2 (MAP2) (rabbit IgG, 1:400, #ab5392, Abcam), ELAVL2 (rabbit IgG, 1:100, #PA5-97702, Invitrogen, Carlsbad, CA, USA), Ki67 (rabbit IgG, 1:400, #ab15580, Abcam), cleaved caspase-3 (Asp175) (rabbit IgG, 1:1000, #9661, Cell Signaling, Danvers, MA, USA), Synaptophysin1 (SYP) (rabbit IgG, 1:100, #101002, Synaptic Systems, Goettingen, Germany), and Doublecortin (DCX) (rabbit IgG, #ab18723, 1:1000, Abcam), followed by appropriate species-specific secondary antibodies (Molecular Probes). DAPI (10 µg/mL, Sigma, Cibolo, TA, USA) was included in the secondary antibody incubations to label nuclei. The stained cells were then mounted in mounting medium and analyzed by conventional or confocal fluorescence microscopy.

For immunohistochemistry, the target region of the spinal cord (~0.5 cm in length) was surgically dissected after perfusion, fixed in 4% paraformaldehyde (PFA) in PBS for 1 day, cryo-protected in 30% sucrose solution for 1 day, and sectioned into 20 μm horizontal slices using a Leica CM1950 cryostat. The spinal cord sections were collected serially onto Superfrost™ Plus glass slides (Fisher Scientific, Waltham, MA, USA) and air-dried. Sections were washed in PBS three times for 5 min per wash, permeabilized with 0.5% Triton X-100 in PBS for 60 min and blocked using a 5% normal donkey serum (NDS) and 0.5% Triton-X in PBS for 90 min to reduce non-specific binding of the antibodies. The samples were then incubated with primary antibodies diluted in the same blocking buffer at 4 °C for overnight, washed in PBS three times for 10 min per wash, and incubated with secondary antibodies diluted in blocking buffer for 120 min. Finally, the samples were washed in PBS three more times for 10 min per wash and mounted with coverslips using anti-fading mounting solution (Invitrogen). DAPI (10 µg/mL, Sigma) was included in the secondary antibody incubations to label nuclei. The immunostained sections were examined and imaged using conventional or confocal fluorescence microscopes. The levels of cellular fluorescence from fluorescence microscopy images were determined in ImageJ software and corrected to mean fluorescence of background readings.

2.9. Western Blot Analysis

Cell cultures were harvested in RIPA buffer (Alfa Aesar, # J63306.AK, Ward Hill, MA, USA) following manufacturer’s instructions. Protein concentrations were determined by Coomassie Plus (Bradford) Assay Kit (#23236, Thermo Scientific, Rockford, IL, USA). Forty µg of protein were boiled in SDS sample buffer for 5 min and loaded on each lane of Any kD™ Mini-PROTEAN^® TGX™ precast polyacrylamide gels and transferred onto PVDF membranes. Western blot analysis was performed as previously described [19]. The primary antibodies were anti-ELAVL4 (rabbit IgG, 1:500, #AB5971, Millpore Sigma-Aldrich, Burlington, MA, USA), anti- ELAVL2 (rabbit IgG, 1:1000, #PA5-97702, Invitrogen, Carlsbad, CA, USA), anti-ELAVL2/4 (rabbit IgG, 1:500, #ab72603, Abcam, Boston, MA, USA), and anti-GFP (rabbit IgG, 1:500, Abcam, #290). The primary antibodies were detected by appropriate species-specific DyLight 700 or 800-conjugated secondary antibodies (1:10,000, Thermo Scientific). Anti-GAPDH (6C5) (mouse IgG, 1:10,000, #ab8245, Abcam) and anti-GAPDH (rabbit IgG, 1:1000, #bs-2188R, Bioss, Woburn, MA, USA) were used to normalize sample loadings. Quantification of relative protein expression levels was performed by measuring signal intensity of target bands on a LI-COR Odyssey^® Infrared Imaging System and normalizing to that of GAPDH. For quantifications on western blots, data were collected from three biological replicates.

2.10. Quantitative Reverse Transcriptase PCR (qRT-PCR)

The procedures for total RNA extraction and qRT-PCR analysis have been previously described [19]. Briefly, total RNAs were purified from cell cultures by TRIzol^® reagent following manufacturer’s instructions (Invitrogen). Quality and concentration of total RNA samples were examined on Nanodrop 8000 (Thermo Scientific). RT reactions were carried out with 0.5 µg of total RNAs using qScriptTM cDNA SuperMix (Quanta Biosciences). In the case of detecting miRNA expression, a Poly-A addition step was included and RT reactions were carried out with 0.5 µg of total RNAs using a miR-RT-primer (caggtccagtttttttttttttttvn) (where v is a, c and g, and n is a, c, g and t) following a protocol previously described [35]. PCR was performed on a BIO-RAD CFX96™ Real-time System (BIO-RAD) using SYBR Select Master Mix (Applied Biosystems, Foster City, CA, USA). Two-step PCR reaction was performed with initial denaturation for 3 min at 95 °C, denaturation for 10 s at 95 °C, and the combination of annealing and extension for 30 s at 60 °C for 40 cycles. To test each reaction product for specificity, post-PCR melt curve was performed by increasing temperatures from 65 °C to 95 °C with the speed of 0.5 °C/5 s. Primer sequences for qRT-PCR are shown in

Table 1. Primers used for qRT-PCR.

GenBank Accession	Gene Name	Forward	Reverse	Size (bp)
MIMAT0000728	miR-375-3p	gtttgttcgttcggctc	cagtttttttttttttttcacgc	40 *
MIMAT0000422	miR-124-3p	taaggcacgcggtga	ccagtttttttttttttttggcat	40 *
NR_138085	U6	gcttcggcagcacatatactaaaat	cgcttcacgaatttgcgtgtcat	89
NM_001419	ELAVL1	gatcctctggcagatgtttgg	cgcggatcactttcacattg	59
NM_001171195	ELAVL2	acaagcgaattgaggcagaag	tggcaccgggaggtttc	63
NM_001420	ELAVL3	gctgagcccatcacagtcaa	gcctgccccgtcttctg	56
NM_021952	ELAVL4	catggagcctcaggtgtcaa	ctggagggtccattgcttgt	57
NM_010894	NeuroD1	gaggaacacgaggcagacaag	tgcattcatggcttcaagctc	51 #
NM_000555	DCX	tgcttgggcctcacactagc	catataccgcaatcaaggaaatactc	51
NM_001082575	NeuN	tcgtagagggacggaaaattga	gccgttggtgtaggggttc	85
NM_001256799	GAPDH	ggtgaaggtcggagtcaacgg	gaggtcaatgaaggggtcattg	112
NM_008085	GAPDH	aggtcggtgtgaacggatttg	tgtagaccatgtagttgaggtca	123 #

* Primer sequences for detecting miRNAs are the same for both human and mouse. ^# For detecting mouse mRNAs.

. The relative gene expression levels were normalized to that of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). MiRNA-specific primers were designed according to a strategy previously described [35] and are shown in Table 1. The relative expression levels of individual miRNAs were normalized to that of U6 small nuclear RNAs (snRNAs).

2.11. Measurements and Statistical Analysis

The data were presented as mean ± standard error of the mean (SEM) for n = 3 or 4 replicates. Statistical analysis was performed in GraphPad Prism 9 using Student’s t-test for data involving only two groups, and one-way analysis of variance (ANOVA) with Bonferroni t-test for data involving more than two groups. Pearson’s chi-squared test was performed in Figure 1h. p < 0.05 was considered a significant difference.

3. Results

3.1. NeuroD1 Activates Downstream Neuronal Genes to Initiate AtN Reprogramming

To reveal the molecular mechanisms underlying AtN reprogramming, we performed RNA-seq analysis on HA cultures during NeuroD1-mediated reprogramming. Samples were harvested at different days post infection (DPI) for characterizations and analyses including RNA-seq at 2 and 6 DPI (Figure 1a). Florescence live imaging indicated that morphological changes occurred as early as 2 DPI by NeuroD1 compared with the GFP control, and that neuronal morphology became apparent at 6 and 14 DPI (Figure 1b). The expression of NeuroD1 as well as the early neuronal marker doublecortin (DCX) and the mature neuronal marker neuronal nuclear protein (NeuN) was detected during reprogramming (Figure 1c and Figure S1). Furthermore, the high efficiency of NeuroD1-mediated neuronal reprogramming in HA cultures was confirmed by the high percentages of immunoreactive cells for DCX, NeuN, and microtubule-associated protein 2 (MAP2) (another mature neuronal marker) at 6 and 14 DPI (Figure 1d,e).

For RNA-seq, two separate libraries were prepared on each total RNA sample with one being for mRNAs to detect coding genes, and the other for small RNAs to detect miRNAs (see Experimental Procedures). The global gene expression profile underwent drastic changes during neuronal reprogramming. As demonstrated by principal component analyses (PCA), while replicates were clustered together, NeuroD1 and GFP control groups were already separated at 2 DPI and even more so at 6 DPI indicating a reprogramming progression over time at the global gene expression level (Figure 1f). Differentially expressed genes (DEGs) were obtained by comparing the NeuroD1 groups to the GFP control groups at different time points. Interestingly, the number of up-regulated genes (URGs) was much greater than that of down-regulated genes (DRGs) at both 2 and 6 DPI while a comparison between the two GFP control groups (i.e., GFP_D6 vs D2) showed less of a difference (Figure 1g,h). This is consistent with the fact that NeuroD1 is a transcription activator turning on neuronal genes to initiate the AtN conversion program [36]. Similar patterns for the number of URGs and DRGs were also reported in ES cells and microglial cultures during NeuroD1-mediated neuronal reprogramming [12,16]. In contrast, chemically induced neuronal reprogramming in the same HA cultures showed an equivalent number of URGs and DRGs, at least during first week of induction [37].

We performed Gene Ontology (GO) analysis to identify major functional categories of the DEGs. As expected, the GO terms of URGs were mostly neuron-related ones containing “neurogenesis”, “development”, and “synaptic” (Figure 1i). Interestingly, we observed more terms containing “development” in URGs at D2 and more terms containing “synaptic” at D6 suggesting a progression from immature to mature neuronal state during reprogramming (Figure 1i). In contrast, the GO terms of DRGs were all non-neuronal ones including interferon and virus-related ones, which could be due to differential cellular responses to retrovirus infection between different conditions (Figure 1i). We then performed Gene Set Enrichment Analysis (GSEA) of cell type signature gene sets with single-cell sequencing data of human tissue (C8: Cell Type Signature). The results showed that the top 10 gene sets enriched in the NeuroD1_D6 group were various types of human neurons, while the top 10 gene sets enriched in the GFP_D6 group were all non-neurons (Figure 1j). This further confirms that NeuroD1 can quickly convert HA cells to neuronal-like cells.

3.2. Identification of NeuroD1-Induced “Core” Genes during Glia-to-Neuron Reprogramming

NeuroD1 has been utilized to reprogram other glial cell types including microglia to neurons and drive ES cells into neurons when overexpressed [12,16]. To identify conserved transcriptional responses during NeuroD1-mediated neuronal reprogramming, we compared our RNA-seq data with publicly available datasets from others and found an overlap of 70 genes, all of which were URGs (Figure 2a). There were no overlapping genes among the DRGs. Functional annotation clustering analysis showed that these overlapping “core” genes possess unique GO terms contributing to the neuronal reprogramming process (Figure 2b). These included “synapse” and “glutamatergic synapse”, which are essential functional structures of mature neurons and are consistent with the fact that NeuroD1-reprorammed neurons are mostly glutamatergic subtypes [4,8]. Other GO terms such as “neurogenesis” and “neuron projection morphogenesis” indicate the developmental process of neuronal differentiation. “Axons” and “plasma membrane” contained important genes such as potassium-channels KCNC1, KCNQ2, and sodium channel SCN3A, that are critical to functional properties of neurons. There was also a group of genes encoding “RNA binding” proteins, suggesting the importance of post-transcriptional regulation during neuronal reprogramming process (Figure 2b). Gene network analysis of the core genes by STRING program showed that most of these genes were interconnected, with top representations of GO terms being “neurogenesis”, “synapse”, “excitatory synapse”, and “RNA recognition motif” (Figure 2c).

3.3. NeuroD1 Induces miRNA Expression Changes during AtN Reprogramming

To determine the function of miRNAs during AtN reprogramming, we first examined expression changes of miRNAs in the small RNAs portion of our RNA-seq data. In PCA, miRNA profiles of NeuroD1 and GFP control groups were clustered closely at D2 and diverged at D6 (Figure 3a). Like mRNAs, there were more upregulated miRNAs (URmiRs) (5 at D2 and 19 at D6) than downregulated miRNAs (DRmiRs) (1 at D2 and 9 at D6) by NeuroD1 at both time points as shown by the volcano plots (Figure 3b). The two most significant URmiRs during NeuroD1-mediated neuronal conversion in HA were hsa-miR-375-3p (miR-375) and hsa-miR-124-3p (miR-124) (labeled red in Figure 3c). Based on average read counts (RC), the highly expressed miRNAs in HA cultures included miR-21-5p that has been shown to function in astrocytes [38,39] (Figure 3d). However, these highly expressed miRNAs were not changing their expression during AtN reprogramming. In contrast, both miR-375 and miR-124 had a minimal expression level in HA and were induced drastically by NeuroD1 (Figure 3d) suggesting their important functions during the reprogramming process. We further confirmed the expression changes of these two miRNAs by qRT-PCR (Figure 3e). Of note, the detection of significant upregulation of miR-124 at D2 by qRT-PCR, but not by RNA-seq, indicates a higher sensitivity of qRT-PCR method.

To estimate the impact of DEmiRs on the global gene expression changes, we extracted predicted target genes of these DEmiRs by TargetScan program and overlaid them with the DEGs. About half of the DEGs (381/782) were predicted DEmiR targets at D2, and even more (679/877) at D6 (Figure 3f). Therefore, DEmiRs likely made a significant contribution to the changing genetic program during AtN conversion. MiR-375 and miR-124 also showed overlapping target genes with the DEGs at both time points (Figure 3g); these target genes have highly enriched neuronal GO terms (Figure 3h). Given the inhibitory mechanism of action for miRNAs on gene expression, one would expect an inverse correlation of expression levels between DEmiRs and their predicted DEG targets. Surprisingly, most of the target DEGs of URmiRs were URGs during AtN reprogramming, and the ratios of target URGs/DRGs from ND1 vs GFP_D6 were significantly higher than those from GFP_D6 vs D2 (Figure 3i). These results suggest that miRNAs could be fine tuners of DEG expression to ensure an optimal level for successful reprogramming.

3.4. MiR-375 Regulates nELAVLs’ Expression Level during NeuroD1-Mediated AtN Reprogramming

While there is an abundance of literature on the role of miR-124 in neuronal differentiation and reprograming [33,40,41,42,43], relatively little is known about miR-375′s contribution to the same processes. Therefore, we decided to investigate the function of miR-375 during NeuroD1-mediated neuronal reprogramming in our HA culture system. There were only 10 miR-375 targets overlapping with the DEGs between two time points (Figure 3g); these targets were listed with their mean expression levels from different groups by RNA-seq (Figure 4a). Among the very few miR-375 targeted DEGs, we identified three members of an RNA-binding protein family (ELAVL), i.e., ELAVL2, 3, and 4 (Figure 4a). While ELAVL1 is regarded as universally expressed by many cell types, ELAVL2-4 are preferentially expressed in neurons [44]. Interestingly, the 3′-UTRs of ELAVL2-4, but not ELAVL1, contain one or more predicted miR-375 binding sites potentially allowing gene expression regulation by miR-375 (Figure 4b). Indeed, ELAVL4 has been experimentally confirmed as a miR-375 target in other systems [45,46]. We first confirmed ELAVL2-4 gene expression changes by qRT-PCR (Figure 4c) and changes of ELAVL2 and 4 at protein level by western blot (Figure 4d) during AtN reprogramming. ELAVLs as RNA-binding proteins stabilize target mRNAs for longer half-life and thereby increase protein translation [47]. ELAVLs recognize an AU-rich element (ARE) sequence allowing prediction of their target genes [44]. Since three ELAVLs were highly upregulated during NeuroD1-mediated reprogramming, we wanted to see how many of the ARE genes were also upregulated that were potentially the result of elevated ELAVLs. Indeed, more than 26% of the DEGs are predicted ARE genes at both time points, and their GO terms were enriched in neuronal ones (Figure 4e).

To determine if miR-375 can regulate nELAVLs’ gene expression level during NeuroD1-mediated reprogramming, we adopted overexpression and knockdown strategies. For overexpression, we constructed a retrovirus expressing miR-375 under the CAG promoter. For knockdown, we introduced a lentivirus expressing a miR-375 inhibitor (miR-375-decoy) that has been validated to decrease the miR-375 level by more than 50% [48]. Our design was to coinfect HA with each of these miR-375 constructs and NeuroD1 and compare with NeuroD1 alone to see if changing the miR-375 level in the presence of NeuroD1 could modulate the expression level of nELAVLs. For this, we decided to detect changes at protein level by immunostaining. We found that overexpression of miR-375 decreased ELAVL2 and 4 protein levels in coinfected cells compared with NeuroD1 alone, while a reduction in miR-375 by decoy increased them (Figure 4f,g, and Supplementary Figure S2). Therefore, our data indicate that a NeuroD1-induced increase in nELAVLs can be further regulated by the miR-375 level.

3.5. MiR-375/nELAVLs Are also Induced by Neuronal Reprogramming Factors ASCL1 and Neurog2

To determine if miR-375/nELAVLs’ upregulation is unique to NeuroD1-mediated neuronal reprogramming, we analyzed their expression in AtN induced by two other reprogramming factors, ASCL1 and Neurog2. At 6 DPI, the neuronal reprogramming by three factors was confirmed by a morphological change of infected cells in HA, albeit ASCL1 was less potent in reprogramming than NeuroD1 and Neurog2 (Figure 5a,b). We further confirmed overexpression of these reprogramming factors in infected HA cultures by qRT-PCR (Figure 5c). Consistent with the morphological change upon reprogramming, qRT-PCR analysis also showed induction of neuronal markers DCX and NeuN as well as nELAVLs by all three factors at 6 DPI except that NeuN was not significantly upregulated by ASCL1 (Figure 5d). Furthermore, both miR-375 and miR-124 were markedly upregulated by all three reprogramming factors, interestingly with a much higher level of miR-375 induced by ASCL1 than the other two factors (Figure 5e). These data suggest that miR-375/nELAVLs’ interaction could be a common mechanism during neuronal reprogramming induced by different reprogramming factors.

3.6. NeuroD1 Induces ELAVL4 Expression in the Injured Mouse Spinal Cord and miR-375/ELAVL4 Expression in Mouse Astrocyte Cultures during AtN Reprogramming

We next wanted to determine if the interaction of miR-375/nELAVLs is relevant during AtN reprogramming in vivo. To target astrocytes in the mouse spinal cord, we applied AAV-Cre-Flex expression system as previously described [11]. A mixture of AAV5-GFAP-Cre and AAV5-Flex-NeuroD1-GFP was injected into the mouse spinal cord after a stab injury. Based on our previous characterization of AtN reprogramming in the injured spinal cord, 2-week post injection (WPI) represents an intermediate state of reprogramming when we detected mostly “transitional cells” that were positive for both GFAP and NeuN [11]. At 2 WPI, we first observed many AAV-infected astrocytes that were GFAP+ around the injury sites (asterisks in Figure 6a); interestingly, AAV-ND1-GFP-infected astrocytes showed smaller cell bodies than AAV-GFP-infected ones as revealed by GFP fluorescence (Figure 6a), indicating ongoing AtN reprogramming. We also confirmed NeuroD1 expression in AAV-ND1-GFP-infected astrocytes by immunostaining (Figure 6b) and the presence of “transitional cells” in the AAV-ND1-GFP-infected spinal cord as expected (Figure 6c). Importantly, among the AAV-ND1-GFP-infected astrocytes, we were able to identify some cells also expressing ELAVL4 protein (arrowheads, Figure 6d). The expression level of ELAVL4 in these cells was much weaker than the nearby neurons (Figure 6d), suggesting an early onset of neuronal gene expression at this stage. To detect miR-375 expression increased by NeuroD1, we turned to mouse neonatal astrocyte cultures for the convenience of analysis. NeuroD1-expressing retrovirus could readily reprogram mouse-cultured astrocytes into DCX+/NeuN+ neurons within 2 weeks (data not shown). We confirmed the morphological change and NeuroD1 expression of ND1-GFP-infected cells compared with the GFP control at 9 DPI and the induction of ELAVL4 by NeuroD1 by immunostaining (Figure 6e). Lastly, qRT-PCR data showed that both miR-375-3p and miR-124-3p were significantly upregulated by NeuroD1 in these mouse astrocyte cultures (Figure 6f). Therefore, NeuroD1 can induce miR-375/ELAVL4 expression in mouse astrocyte cultures as in HA and induce at least ELAVL4 protein in astrocytes of the injured spinal cord during AtN reprogramming.

3.7. Overexpression of miR-375 Elicits a Protective Effect during NeuroD1-Mediated AtN Reprogramming by Reducing Apoptosis

Given that miR-375 can potentially regulate many DEGs during NeuroD1-mediated reprogramming either directly or indirectly through nELAVLs (Figure 3g and Figure 4e), we set out to determine if manipulating the miR-375 level would affect the AtN reprogramming outcome in HA. We set up a coinfection experiment combining NeuroD1-RFP with GFP control, miR-375-GFP, or miR-375-decoy (Figure 7a). The typical coinfection efficiency was around 50% in retrovirus combinations and much higher with lentivirus (i.e., miR-375-decoy) (Figure 7b). We realized that many miR-375-GFP retrovirus-infected cells exhibit a much weaker GFP signal than the control GFP virus-infected ones (Figure 7a). The GFP signal of miR-375-GFP became so weak in long-term cultures and it could not be reliably detected even by immunostaining. Therefore, we performed our quantitative analyses among NeuroD1-infected cells in all the coinfection experiments without distinguishing singly and doubly infected cells. We examined AtN reprogramming by immunostaining (Figure 7c) and showed that reprogramming efficiency as measured by the neuronal markers DCX and NeuN did not differ among the coinfected cultures (Figure 7d), and that 2-fold more marker-positive cells per field were observed in the miR-375 coinfection than the GFP control (Figure 7e). The knockdown of miR-375 by the decoy construct in the presence of NeuroD1 did not show a significant difference from the control coinfection by the analyses mentioned above (Figure 7e). Live imaging also indicated that there were significantly more NeuroD1-infected cells in the miR-375 coinfected HA culture than in the other coinfections at 28 DPI, even though all the cultures started with equivalent numbers of infected cells at 2 DPI (Figure 7f,g). qRT-PCR analysis showed that the miR-375-GFP retrovirus induced a much higher level of miR-375 than NeuroD1+GFP at both early and late stages (Figure 7h). Therefore, simultaneous miR-375 overexpression resulted in more NeuroD1-converted neurons in HA cultures.

We reasoned that the increased number of converted neurons in the miR-375 coinfected HA cultures could be due to decreased apoptosis and/or increased cell proliferation during the reprogramming process. We first assessed cell proliferation during reprogramming in HA cultures with anti-Ki67 antibody. As expected, NeuroD1 as a neuronal differentiation factor inhibited cell proliferation and drastically decreased Ki67+ cell number at 3 and 7 DPI when compared with the uninfected cells, and yet no difference was observed between the two coinfection cultures (Figure 8b). To obtain an overview of the apoptosis status during NeuroD1-mediated neuronal reprogramming, we stained the NeuroD1-infected HA cultures with an antibody against cleaved caspase 3 (CCasp3). We found that ~8% of NeuroD1-infected cells were CCasp3+ at 3 DPI, and this percentage dropped down to a minimal level at 7 and 14 DPI (Figure 8c,d). Surprisingly, the miR-375 coinfected HA cultures showed a drastic decrease in the number of CCasp3+ cells when compared with the GFP coinfected ones at 3 DPI (Figure 8d), suggesting a cell protective effect of miR-375 at early stages of reprogramming. To mimic the stressed condition in disease/injury, we exposed the reprogramming HA culture to the chemotherapeutic agent Cisplatin [49]. First, we treated normal HA with various doses of Cisplatin and showed a dose–response of HA in expressing CCasp3 (data not shown). We then treated NeuroD1-infected HA cultures at 2 DPI with a low and high dose of Cisplatin (5 and 250 μM, respectively) for 24 h. Strikingly, at both doses of Cisplatin treatment, overexpression of miR-375 reduced not only the percentage of CCasp3+ cells (Figure 8e,f) but also the cellular CCasp3 expression level as measured by florescence intensity when compared with the control groups (Figure 8g).

To assess the potential long-term effect of miR-375 overexpression in reprogrammed neurons, we cultured these neurons for 30 days. Since the miR-375 overexpressing retrovirus loses reporter GFP expression over time, we enriched double-infected HA by FACS at 3 DPI when the miR-375-GFP level is still high (Figure 9a). The sorting efficiency is quite high, and we detected almost 100% coinfection efficiency in the ND1-RFP+/miR-375-GFP+ group one day after plating while no GFP signal was observed in the ND1-RFP+ group (Figure 9b). After 30 days in culture, we examined the expression of ELAVL4 and the mature neuronal markers MAP2, synaptophysin (SYP), and NeuN in these reprogrammed neurons by immunostaining (Figure 9c and Figure S3). We found that the levels of these markers were not reduced but even increased in the ND1+miR-375-reprogrammed neurons compared with the ND1-reprogrammed ones (Figure 9c,d), suggesting that miR-375 overexpression does not interfere with neuronal maturation. The slight but significant increase in ELAVL4, MAP2, and SYP levels is likely due to the healthiness of neurons in the double-infected group (Figure 9c,d). Taken together, miR-375 overexpression improves NeuroD1-mediated AtN reprogramming efficiency in HA cultures by inhibiting apoptosis at early stages and does not compromise the maturation of the reprogrammed neurons in long-term cultures.

3.8. Overexpression of ELAVL4 Triggers Apoptosis in HA and Reverses miR-375-Induced Survival-Promoting Effect during NeuroD1-Mediated AtN Reprogramming

The survival-promoting effect of overexpressed miR-375 at the early stage of AtN reprogramming could be due to the expression inhibition of its target genes. As we have shown that miR-375 can regulate nELAVL levels during neuronal reprogramming (Figure 4), we decided to examine the roles of nELAVLs in this cell survival effect. We generated retroviral constructs constitutively expressing ELAVL2 and 4 genes without 3′-UTRs (miR-375-refractory). The expression of these constructs was confirmed by western blot (Figure 10a). First, when we overexpressed nELAVLs in HA cultures, we found that both nELAVLs constructs induced neurite-like process extension from the infected cells at 6 DPI (Figure 10b,c). However, these nELAVL-infected HA did not express the neuronal markers DCX and NeuN (data not shown) and disappeared soon after the morphological change. Indeed, we detected a significantly higher percentage of CCasp3+ cells among the ELAVL4-infected HA compared with control ones at 3 DPI (Figure 10d and Figure S4), suggesting that overexpression of nELAVLs induced apoptosis in HA. Next, we overexpressed these miR-375-refractory nELAVLs constructs along with NeuroD1-RFP and miR-375. We could not distinguish between nELAVLs and miR-375 infected cells since both expression constructs have a GFP reporter. However, relatively high coinfection efficiency by two retroviruses (Figure 7b) implied that a significant population was coinfected by three constructs. We then examined CCasp3+ cells among the NeuroD1-infected (RFP+) population and found that additional overexpression of ELAVL4 but not ELAVL2 induced significantly more CCasp3+ cells among RFP+ cells when compared with NeuroD1 and miR-375 coinfection (Figure 10e and Figure S5). Thus, these data indicate that overexpression of miR-375-refractory ELAVL4 reverses the miR-375-induced survival-promoting effect during AtN reprogramming.

4. Discussion

In this study, we examined both mRNA and miRNA expression changes during NeuroD1-mediated AtN reprogramming by RNA-seq analysis. In addition to significantly upregulated mRNAs related to neuron differentiation, two major upregulated miRNAs, miR-375 and miR-124, were identified. We further investigated the function of miR-375 in reprogramming by overexpression and knockdown approaches and found that miR-375 regulates the expression level of nELAVLs, which were also upregulated by NeuroD1, in the reprogramming astrocytes. Surprisingly, we observed a cell survival-promoting effect by overexpression of miR-375 during reprogramming. Finally, we propose a mechanistic model, in which NeuroD1 turns on miR-375 to modulate its downstream effector genes such as nELAVLs to achieve successful reprogramming accompanied by apoptosis, and this miRNA-mediated gene expression modulation can be further tweaked by miR-375 overexpression for a better reprogramming outcome (Figure 10f). To our knowledge, this is the first study to systematically examine miRNA expression and decipher their functions during glia-to-neuron reprogramming.

4.1. MiR-375 Is a Transcriptional Target of NeuroD1

Our RNA-seq data indicate that miR-375 is one of the most significantly upregulated miRNAs during AtN reprogramming. This result is not completely unexpected since miR-375 has been identified as a direct transcriptional target of NeuroD1 in the pancreas [50] where it plays a critical role in pancreas development and insulin secretion [51,52,53]. However, miR-375 is not well studied in the CNS. It has been reported that miR-375 is transiently expressed in the developing telencephalon at early embryonic stage (E13) before it is downregulated [45], and yet the regulation of this transient expression of miR-375 is not well understood. The expression level of miR-375 is minimal in the adult CNS except for the pituitary gland [54]. In a human ES cell culture system, miR-375 has been identified as one of the repressor element-1 silencing transcription factor (REST) downstream miRNAs, and its expression is upregulated in REST-null ES cells and ES-derived neural stem cells (NSCs) [55]. In the same study, the authors have also shown that miR-375 expression increases during motor neuron differentiation from ES cells [55]. To our knowledge, our study is the first to show that miR-375 is upregulated by overexpression of NeuroD1 in neural cells, and this upregulation of miR-375 is likely through direct transcriptional activation by NeuroD1 as is the case in pancreas. Interestingly, miR-124 (another major significantly upregulated miRNA in our RNA-seq analysis) and nELAVLs are also expressed in the pancreas, suggesting that common mechanisms of gene regulation and function may exist between the CNS and the pancreas [56,57].

4.2. MiR-375 Regulates Levels of nELAVLs during AtN Reprogramming

The function of nELAVLs in neuronal differentiation during development has been documented. Gene knockout studies have shown that deficits in axonal transport and abnormalities in neuronal polarity are observed in ELAVL3-null Purkinje cells in the cerebellum [58], and that loss of ELAVL4/HuD results in a defective dendritic overgrowth in both the neocortex and the CA3 region of the hippocampus [59]. On the other hand, overexpression of ELAVL4/HuD accelerates dendritic outgrowth in cortical neuron culture via GAP43 [47]. ELAVL4/HuD has also been shown to promote neuronal differentiation of neural stem/progenitor cells (NSPCs) in the adult subventricular zone (SVZ) by stabilizing the mRNA of SATB1, a critical transcriptional regulator during neurodevelopment [60]. Furthermore, post-transcriptional regulation of neuronal mRNAs by ELAVL4/HuD has been shown to mediate synaptic plasticity in mature neurons [61]. With all the above findings, it is not surprising that we observe a drastic upregulation of nELAVLs during NeuroD1-mediated AtN reprogramming. We do not have experimental evidence that nELAVLs are direct transcriptional targets of NeuroD1, although it is possible since upregulation of these genes occurs as early as 2 DPI (Figure 4). Alternatively, this upregulation by NeuroD1 is indirect and occurs via other intermediate transcription factors. Consistently, Neurog2 has been shown to bind two E-boxes in the ELAVL4/HuD promoter region and regulate its transcription [62], and Neurog2 is also upregulated during NeuroD1-mediated AtN reprogramming (data not shown).

ELAVL4/HuD is a validated target gene of miR-375, overexpression of which reduces ELAVL4/HuD gene expression and in turn results in decreased dendritic branching in the mouse hippocampus [45]. MiR-375 is transiently expressed in the neocortex during embryonic development, and its temporal expression pattern is complementary to that of ELAVL4/HuD, suggesting a negative regulatory control between miR-375 and ELAVL4/HuD [45]. Our data in this study further confirm that miR-375 can regulate nELAVLs during neuronal reprogramming process, especially at early stages (Figure 4). We postulate that the significance of this miR-375-mediated regulation is to tune NeuroD1-induced nELAVLs’ expression to an optimal level for successful reprogramming. Consistently, it has been reported that, although ELAVL4/HuD expression is increased after learning and memory, constitutive ELAVL4/HuD overexpression in the adult brain does not improve, but rather impairs animal behavior in a battery of learning and memory tests [63], indicating that tightly controlled ELAVL4/HuD expression level is critical for its function.

Neuronal reprogramming is not a “natural” biological process [8], and constitutive overexpression of the reprogramming factors may trigger distinct cellular and molecular responses in the reprogrammed cells. In our case, forced expression of NeuroD1 in astrocytes initiates neuronal reprogramming by turning on downstream genes including nELAVLs. However, high level of nELAVLs’ expression not only facilitates the reprogramming process but also brings detrimental effects such as apoptosis. The NeuroD1-induced expression of miR-375 is probably an adaptive response of the reprogrammed cells to trim down nELAVLs expression level for a better outcome. We propose that this “incoherent” [64] regulation between miRNAs and their target genes is not a rare event during NeuroD1-mediated reprogramming since we observed many miRNA/target gene pairs that are both upregulated by NeuroD1 (Figure 3). Furthermore, this expression “trimming” function of miRNAs may be common in many transcription factor-induced reprogramming scenarios. Another point worth noting is that nELAVLs are also upregulated in other glial cell types including microglial and glioblastoma cell lines during NeuroD1-mediated reprogramming [12,34]. It would be interesting to examine if miR-375 is also upregulated in these glial cells and modulates nELAVLs’ expression level during these neuronal reprogramming processes.

4.3. MiR-375 Is Neuroprotective during AtN Reprogramming

AtN reprogramming is an event involving drastic changes in gene expression and signaling pathways. Many cells die during the reprogramming process if they cannot pass the metabolic “checkpoint” [65]. Cells that manage to survive will proceed to convert into neurons. Our previous research has indicated that NeuroD1 can convert reactive astrocytes into functional neurons in the spinal cord with high efficiency [11]. This high efficiency is partially due to the fact that NeuroD1 is not only a critical neuronal differentiation transcription factor, but also a survival factor during development, especially in developing granule neurons of the cerebellum and hippocampus where NeuroD1 is highly expressed throughout adulthood in mice [14]. Despite this fact, we did observe cell apoptosis at the early stage (3 DPI) of reprogramming in cultured HA after NeuroD1 overexpression and surprisingly miR-375 help reduce apoptosis when co-expressed with NeuroD1 (Figure 8). In fact, the survival-promoting effect of miR-375 has been documented previously. For instance, the miR-375 level is reduced in the degenerating motor neurons of the ALS mouse and human patients [55,66], and knockdown of miR-375 in culture induces motor neuron death via p53 pathway [55]. Surprisingly, we did not detect any functional effects of inhibitor-mediated miR-375 knockdown during reprogramming process in HA cultures except for an increase in ELAVL2/4 protein levels (Figure 4f). This could be due to only partial knockdown of miR-375, and thus miR-375 gene knockout may help reveal potential loss-of-function effects if any. Nevertheless, our data show that high levels of nELAVLs (i.e., by overexpression) induces apoptosis in HA culture (Figure 10) and miR-375 can regulate nELAVLs’ expression level during NeuroD1-mediated reprogramming (Figure 4). Therefore, it is likely that miR-375 helps suppress NeuroD1-induced nELAVLs’ expression to elicit cell survival-promoting effect (Figure 10f). However, we do not exclude other pathways (such as p53) that may also be affected by miR-375 overexpression [55] and are independent of nELAVLs. Regardless of the potential mechanisms, our observation that miR-375 is neuroprotective during neuronal reprogramming has translational values. Overexpressing miR-375 along with NeuroD1 could further elevate the cellular miR-375 level (Figure 7h) and increase neuronal reprogramming efficiency by promoting survival of reprogrammed neurons. We are currently constructing the combinatory vector to co-express NeuroD1 and miR-375, and the reprogramming efficiency of this co-expression vector will be characterized in our future investigation.

5. Conclusions

In conclusion, we examined both mRNA and miRNA expression changes during NeuroD1-mediated astrocyte-to-neuron reprogramming by RNA-seq analysis. In addition to significantly upregulated mRNAs related to neuron differentiation, two major upregulated miRNAs, miR-375 and miR-124, were identified. Surprisingly, we observed a cell survival-promoting effect by overexpression of miR-375 during reprogramming even under cisplatin-induced stress conditions. We further showed that miR-375 regulates the expression level of nELAVLs, which were also upregulated by NeuroD1, in the reprogramming astrocytes. Finally, we propose a mechanistic model, in which NeuroD1 turns on miR-375 to modulate the level of its downstream effector genes such as nELAVLs to achieve successful reprogramming, and this miRNA-mediated gene ex-pression modulation could be conserved in neuronal reprogramming mediated by other factors. Further investigation is warranted to verify if this miR-375/nELAVLs’ mechanism also works during neuronal reprogramming in vivo.