Direct reprogramming of oligodendrocyte precursor cells into GABAergic inhibitory neurons by a single homeodomain transcription factor Dlx2

Oligodendrocyte precursor cells (NG2 glia) are uniformly distributed proliferative cells in the mammalian central nervous system and generate myelinating oligodendrocytes throughout life. A subpopulation of OPCs in the neocortex arises from progenitor cells in the embryonic ganglionic eminences that also produce inhibitory neurons. The neuronal fate of some progenitor cells is sealed before birth as they become committed to the oligodendrocyte lineage, marked by sustained expression of the oligodendrocyte transcription factor Olig2, which represses the interneuron transcription factor Dlx2. Here we show that misexpression of Dlx2 alone in postnatal mouse OPCs caused them to switch their fate to GABAergic neurons within 2 days by downregulating Olig2 and upregulating a network of inhibitory neuron transcripts. After two weeks, some OPC-derived neurons generated trains of action potentials and formed clusters of GABAergic synaptic proteins. Our study revealed that the developmental molecular logic can be applied to promote neuronal reprogramming from OPCs.

Primary cultures. Postnatal cortical OPCs. For culturing postnatal OPCs, we modified the previously published sequential immunopanning purification procedure 19 . We used postnatal day 2-3 (P2-3) CD1 mice from Charles River Laboratory strain #022. After removal of meninges, the cortices were minced and incubated at 35 °C for 60 min in a papain solution (20 U/ml; Worthington). Subsequently, the tissue was triturated in a solution containing ovomucoid and bovine serum albumin (BSA) (Sigma Aldrich) and filtered through a 70μm cell strainer (Falcon) to obtain a single cell suspension. The cell suspension was incubated in a petri dish coated with the O1 antibody (Gift from Dr. Steven Pfeiffer at University of Connecticut Health Center, 1:1 dilution) 20 to eliminate oligodendrocytes, with subsequent incubation of the supernatant in a petri dish coated with rat anti-mouse platelet-derived growth factor receptor alpha (PDGFRα) antibody (CD140a, 2 μg/ml, BD Biosciences, RRID:AB_397117) to capture OPCs. Non-adhered cells were washed off with phosphate-buffered saline (PBS), and adhered OPCs were collected by trypsinization. Cells were centrifuged for 5 min at 1100 rpm and re-suspended in DMEM (Thermo Fisher Scientific) supplemented with Sato's supplements (DMEM-Sato's) and 50 ng/ml PDGF-AA (Peprotech). Cells were plated on 12-mm glass coverslips (Fisher) coated with 100 μg/ml of poly-d-lysine (Sigma-Aldrich) and 15 μg/ml of laminin (Sigma-Aldrich) or on 35-mm tissue culture dishes (Fisher Scientific) coated with 30 µg/ml of poly-l-lysine (PLL, Sigma-Aldrich). At 3 days post transfection (dpt), cells were trypsinized and re-plated onto a layer of astrocytes at 125,000 cells per 12-mm glass coverslip. Media was switched to neuronal differentiation media consisting of 1% N2 and 1% B27 supplements, 1% l-glutamine, and 1% penicillin and streptomycin in DMEM/F12 (all from Thermo Fisher Scientific) and supplemented with 50 ng/ml human brain-derived neurotrophic factor (BDNF, Peprotech). At 7 dpt, media was completely switched to neuronal maturation media consisting of Neurobasal A (Thermo Fisher Scientific) and supplemented with 2% B27, 1% Glutamax (Thermo Fisher Scientific), and 1% penicillin and streptomycin supplemented with 50 ng/ ml human BDNF. Half of media was replaced with fresh media and BDNF every 3-4 days. On 10 dpt, 1% horse serum was added for astrocyte maintenance.
Postnatal cortical astrocytes. For culturing primary postnatal astrocytes, we isolated cortical astrocytes as described previously 21 . Briefly, we dissected and chopped the cortices of P1-P3 CD1 mice. Cortical pieces were transferred to a tube containing Hank's Balanced Salt Solution (Thermo Fisher Scientific) and 0.1% trypsin (Thermo Fisher Scientific) and incubated in 37 °C water bath for 15-30 min. Cells were collected by centrifuging for 5 min at 1100 rpm. Cells were resuspended in astrocyte media consisting of 10% horse serum (Hyclone) and 0.6% d-glucose (Thermo Fisher Scientific) in DMEM) and gently triturated into a single cell suspension. Cells were plated in T75 culture flasks coated with 30 μg/ml PLL. Medium was changed every 2-3 days. At 7-9 days post plating when astrocyte layer was confluent, flasks were shaken at 37 °C at 260 rpm overnight to detach microglia and OPCs. The next day, astrocytes were rinsed with PBS, trypsinized and frozen in astrocyte media with 10% DMSO and 50% horse serum for future use.
Immunofluorescence labeling. Coverslips were fixed with 4% paraformaldehyde (Electron Microscopy Sciences) in 0.1 M sodium phosphate buffer for 10 min, rinsed three times with PBS and then blocked and permeabilized in PBS containing 0.1% Triton X-100 (Sigma-Aldrich) and 5% normal goat serum (Thermo Fisher Scientific) or 5% normal donkey serum (Thermo Fisher Scientific) at room temperature (RT). The coverslips were then incubated with primary antibodies in blocking solution for 45 min. After three rinses, coverslips were incubated with the secondary antibodies in blocking solution for 30 min. Primary and secondary antibodies are listed in Table 1. Coverslips were mounted in Vectashield mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories).

RNA-sequencing (RNA-seq). RNA extraction, library construction, and sequencing. Immunopurified
OPCs were transfected with either control mCherry or Dlx2 plasmid vector. At 2 dpt, transfected cells were isolated by fluorescence-activated cell sorting (FACS) for mCherry with the BD FACSAriaII (Becton Dickinson), using the 488 nm diode laser in the Flow Cytometry Facility (University of Connecticut, Storrs). Three biological replicates consisted of mCherry-and Dlx2-transfected OPCs that had been isolated from different litters of mice. The sorted cells were collected and stored in Trizol reagent (Invitrogen), and total RNA was extracted using the RNeasy Plus Micro Kit (Qiagen). Genomic DNA was removed using gDNA Eliminator spin-columns provided by the kit. RNA concentration and integrity of each sample were measured using an Agilent 2100 Bioanalyzer (Agilent), housed in the Center for Genome Innovation (University of Connecticut, Storrs). Samples with www.nature.com/scientificreports/ an RNA Integrity Number ≥ 8 were processed. Ribosomal RNA was removed from samples using RiboGone™ (Clontech). cDNA libraries were prepared with 100 ng total input RNA using the SMARTer Stranded RNA-Seq Kit (Clontech). Libraries were multiplexed and subjected to a total of 18 cycles of PCR amplification, and the amplified cDNA library was sequenced at the Center for Genome Innovation (University of Connecticut, Storrs, CT) using Illumina NextSeq 500. Barcoded and multiplexed samples were run on four lanes to generate four technical replicates of 150-nucleotide paired end reads.
Gene ontology (GO) enrichment analysis. GO enrichment analysis of the differentially expressed genes was performed to identify biological processes enriched in differentially expressed genes using the enrichGO function of clusterProfiler v3.6 24 (RRID:SCR_016884). The GO over-representation test 25

Differentiation of Dlx2-transfected postnatal OPCs into GABAergic neuron-like cells. Dlx2
represses Olig2 during interneuron development and is not detected in OPCs 10,26 . Previously we showed that OPCs in the postnatal SVZ or the forebrain parenchyma do not express detectable levels of the key interneuron homeodomain transcription factor Dlx2 26 . To examine whether Dlx2 could alter the fate of OPCs that had been committed to the oligodendrocyte lineage, we transfected immunopurified OPCs from postnatal mouse cortex with either control mCherry (control) or Dlx2-mCherry (Dlx2) plasmid. One day after transfection (1 dpt), Dlx2-immunoreactivity was detected in Dlx2-but not in control-transfected OPCs (Supplemental Fig. 1A). As early as 2 dpt, 24.3 ± 4.9% of Dlx2-transfected cells had barely detectable Olig2 (Fig. 1A, arrowheads; Supplemental Fig. 2B) and had larger cell bodies and longer processes (see bottom left mCherry + cell in Fig. 1A and arrows in Fig. 1B), while the remaining Dlx2-transfected cells exhibited weak Olig2 immunoreactivity. This was not due to transfection of Dlx2 into Olig2-negative non-OL lineage cells, since at 1 dpt > 97% of mCherry + cells expressed NG2 and Olig2 in both control-and Dlx2-transfected cultures (Supplemental Fig. 1B). Using other oligodendrocyte lineage markers, we found that at 2 dpt, 45.4 ± 8.6% and 55.6 ± 11.6% of Dlx2-transfected cells showed no detectable immunoreactivity for PDGFRα and Sox10, respectively (Supplemental Fig. 2A-D). Furthermore, by 2 dpt, 76.0 ± 5.0% of mCherry + Dlx2-transfected cells expressed the neuronal cytoskeletal protein βIII-tubulin (Tuj1), while they continued to express the OPC antigen NG2 (Fig. 1B,C, arrows). By 7 dpt, 73.3 ± 6.6% and 76.9 ± 11.1% were negative for PDGFRα and Sox10, respectively, (Supplemental Fig. 2E-H), indicating that Dlx2-transfected cells had exited from the oligodendrocyte lineage. Thus, overexpression of Dlx2 in postnatal OPCs had caused them to switch their fate from oligodendrocytes to immature neurons. By 14 dpt, Dlx2-transfected OPCs expressed microtubule-associated protein 2 (MAP2), known to be present in neuronal soma and dendrites, and glutamic acid decarboxylase (GAD67), which converts glutamate into GABA in inhibitory GABAergic neurons. In Dlx2-transfected cultures, there were clusters of MAP2 + GAD67 + cells, which were 30-fold more abundant than in control-transfected cultures ( Fig. 1D- www.nature.com/scientificreports/ MAP2 + cells expressed GAD67. There were occasional isolated GAD67 + MAP2 + cells in control-transfected cultures. These were likely to have arisen from contaminating neuroblasts in the subventricular zone (SVZ), which were present in up to 2% of the immunopurified OPCs. Consistently, at 2 dpt, 0.7% ± 0.3% of the mCherry + cells in control-transfected cultures expressed Tuj1 (Fig. 1C). These observations suggest that forced expression of Dlx2 in postnatal oligodendrocyte lineage-committed cells led them to switch their fate toward GABAergic inhibitory neurons.

Dlx2 induces GABAergic neuronal genes in OPCs within 2 days.
To explore the molecular mechanism underlying the fate switch from OPC to GABAergic neurons by Dlx2, we compared transcriptomic profiles of control-and Dlx2-transfected OPCs after 2 dpt. Among the 740 significantly differentially expressed genes (p-adj < 0.01), 445 were upregulated and 295 were downregulated in Dlx2-transfected cells (Supplementary  Table 1). Figure 2A shows a heatmap of the 50 most significantly differentially expressed genes ranked by p-adj values. Among the most significantly downregulated genes were Cspg4, encoding NG2, and the oligodendrocyte transcription factors Olig1 and Olig2. Gene ontology (GO) term analysis revealed that gliogenesis was the largest represented biological process with the largest number of downregulated genes (Fig. 2B). Other significantly downregulated genes in this GO category were Pdgfra (ranked 209 by p-adj) and Sox10 (ranked 705). Other biological processes enriched in downregulated genes were related to chromatin assembly and organization, with the majority of the genes in these groups encoding histone proteins, suggesting changes in the chromatin structure during neuronal reprogramming of OPCs.  www.nature.com/scientificreports/ Among the significantly upregulated genes were those known to be expressed by immature neurons such as Dcx (Doublecortin) and Tubb3 (encoding βIII-tubulin) ( Fig. 2A, Supplementary Table1), as well as other genes important for neuronal function such as those encoding voltage-dependent K + channels (Kcnq3 and Kcng1) and synaptic proteins such as Grin2a (NMDA receptor 2A) and Sytl4 (synaptotagmin-like protein 4). Among the most significantly upregulated transcription factors was Sp9, which regulates the expression of Fgf8 (fibroblast growth factor 8) implicated in telencephalic patterning 27 . GO analysis of the upregulated genes revealed an enrichment of genes that are important for neurogenesis and neuronal differentiation (Fig. 2C). Nine out of ten GO terms that were enriched in upregulated genes in Dlx2-transfected cells were related to axonal growth.
We also examined genes that regulate the cell cycle and are known to be expressed in OPCs 28,29 . The majority of the positive regulators of the cell cycle were downregulated in Dlx2-transfected cells compared to controltransfected cells (Supplemental Fig. 3A), with Ccnd1, encoding the G1/S-specific cyclin-D1, being the most strongly downregulated. This suggests that Dlx2-transfected OPCs had become post-mitotic as they differentiated into inhibitory neurons, consistent with previous reports that neuronal differentiation from neural progenitor cells is initiated through downregulation of the activity of cyclins and CDK proteins 30,31 .
Upregulation of GABAergic but not glutamatergic neuronal genes. We next further examined neuronal genes that were upregulated in Dlx2-transfected OPCs ( Fig. 3A; Supplementary Table 1). There was a significant upregulation of Gad1 and its closely related gene Gad2, which encode GAD67 and GAD65, respectively, in GABAergic inhibitory neurons 32,33 , as well as upregulation of Slc32a1, which encodes vesicular GABA transporter (VGAT). However, we did not detect transcripts expressed by the major non-overlapping mature interneuron classes in the neocortex, Sst encoding somatostatin (SST), Pvalb encoding parvalbumin (PV), and Vip encoding vasoactive intestinal peptide (VIP), while we saw an upregulation of Calb2 encoding calretinin, also known as calbindin-2, expressed by a subset of VIP + and SST + interneurons 34 . Within the Dlx family of interneuron transcription factors, Dlx1, and Dlx5, to a lesser extent, were upregulated. By contrast, we did not detect the excitatory neuron bHLH transcription factor Neurod1 13 in Dlx2-or control-transfected OPCs. Nor did we see upregulation of other key genes expressed by cortical excitatory neurons such as Slc17A7 encod- www.nature.com/scientificreports/ ing vesicular glutamate transporter VGLUT1 35 or Tbr1 (T-box brain protein 1) encoding a transcription factor expressed in pyramidal neurons in the deep neocortical layers (Fig. 3A) 36 . These findings indicate that transient episomal Dlx2 misexpression in OPCs was sufficient to selectively upregulate inhibitory neuron transcription factor expression and initiate the program of differentiation towards GABAergic inhibitory neurons in 2 days.

Changes in chromatin modifiers during fate switch into GABAergic neurons.
To predict potential changes to the chromatin architecture during the early phase of neuronal reprogramming of OPCs, we next examined the RNA-seq dataset at 2 dpt to see whether Dlx2 overexpression altered the levels of transcripts encoding chromatin modifying factors. Among histone deacetylases (HDACs), there was a 3.7-fold upregulation of Hdac11 (Supplemental Fig. 3B), which encodes a class IV histone deacetylase and is implicated in spinal motor neuron survival by interacting with the mRNA splicing machinery 37 . Notably, there was no statistically significant change in Ezh2, which encodes the catalytic subunit of the polycomb repressive complex 2 (PRC2) responsible for depositing the repressive trimethylation mark on histone 3 lysine 27 (H3K27) at many inhibitory neuron genes in OPCs 15,38,39 .
Nor was there any change in expression of Kdm6b, which encodes H3K27 demethylase. Among other histone posttranslational modifiers, Kdm3a and 7a were significantly upregulated. Kdm3a demethylates mono-and dimethyl H3K9. It interacts with Neurogenin-2 and increases chromatin accessibility for NeuroD1 and Tubb2b during neuronal differentiation 40 . Kdm7a, which demethylates di-methyl H3K27 and H3K9, promotes neural differentiation by increasing FGF4 expression 41 . Interestingly, among the transcripts encoding ATP-dependent chromatin remodeling enzymes, chromodomain helicase DNA binding protein 3 (Chd3), implicated in cortical layer specification 42 , was significantly upregulated. DNA methylation also regulates chromatin structure and gene transcription through the deposition of methyl groups at CpG islands. Dnmt1, which encodes maintenance DNA methyltransferase essential for correct alternative splicing during OL differentiation 43 , was significantly downregulated in Dlx2-transfected OPCs (Supplemental Fig. 3C). By contrast, the level of Dnmt3a encoding de novo DNA methyltransferase was not changed in Dlx2-transfected OPCs. These changes triggered by Dlx2 overexpression could have caused nucleosome repositioning and facilitated activation or derepression of critical genes during differentiation of OPCs into GABAergic neurons. www.nature.com/scientificreports/

Dynamic changes in Dlx gene expression during neuronal reprogramming from OPCs.
To explore the mechanism of neuronal reprogramming from OPCs, we used qPCR to examine the temporal changes in transcription factor transcripts known to be critical for interneuron development in Dlx2-transfected cultures. We focused on other members of the Dlx family transcription factors, since they were upregulated in our initial RNA-seq analysis at 2 dpt and are known to be influenced by Dlx2. During development, Dlx2 is expressed early and transiently and directly activates Dlx5 and 6, which are critical for inhibitory neuron differentiation and maturation 44 . Dlx1 expression is co-regulated with Dlx2 45 . Its expression is transient in PV + interneurons but sustained in SST + interneurons. In Dlx2-transfected OPCs, Dlx1 expression peaked at 7 dpt, although it did not reach significance, likely due to large variability among samples. By 14 dpt, Dlx1 returned to a level only slightly above control (Fig. 3B). As predicted from developmental expression, the upregulation of Dlx5 expression in Dlx2-transfected OPCs was most prominent at14 dpt, although a small increase was observed by RNA-seq at www.nature.com/scientificreports/ 2 dpt (Fig. 3C). Dlx6 expression increased earlier than Dlx5, with peak expression at 7 dpt, and decreased but remained slightly elevated at 14 dpt (Fig. 3D). Interestingly, two long non-coding RNAs (lncRNAs) in the Dlx family were highly upregulated (Fig. 3E-F  and Supplementary Table 1). Dlx1as, which is an antisense lncRNA to Dlx1, has been shown to modulate Dlx1 expression 46,47 . Dlx6as, an antisense lncRNA to Dlx6, also known as Dlx6os1 or Evf2, has been shown to promote neurogenesis by inhibiting DNA methylation at the ultraconserved enhancer of Dlx5/6 48 . At 2 dpt, we detected a > 100-fold upregulation of Dlx1as. Dlx6as transcript was also elevated early at 2 dpt. Both transcripts declined by 7 dpt.

Functional and morphological differentiation of OPC-derived GABAergic neurons. OPC-de-
rived immature neurons generate action potentials. We noticed that episomal mCherry expression disappeared beyond the first week after transfection. To achieve persistent reporter expression in OPC-derived neurons, we transfected OPCs from NG2cre;Gad1-GFP mice [16][17][18] with a cre-dependent Tol2 transposon donor vector Tol2-Dlx2-ires2-mCherry (Tol2-Dlx2) or control Tol2-ires2-mCherry (Tol2-control) and the T2 transposase helper vector. This enabled identification of GABAergic neurons by GFP and Dlx2-transfected cells by mCherry expression. We detected Gad1-GFP in neuron-like cells as early as 3 dpt in Tol2-Dlx2-transfected cells, consistent with the early upregulation of Gad1 mRNA from our RNA-seq data. At 14 dpt, 93.6 ± 0.4% of mCherry + cells were also Gad1-GFP + in Tol2-Dlx2-transfected OPCs compared to 1.2 ± 1.1% in Tol2-control-transfected OPCs. Most of the latter remained in the oligodendrocyte lineage and expressed Olig2 (Supplemental Fig. 4A, arrowheads). The small number of Gad1-GFP + cells found in Tol2-control-transfected cultures were mCherry-negative (Supplemental Fig. 4B), also supporting the notion that they were likely to have arisen from untransfected neuroblasts in the original OPC culture.
Using the optimized method of sustaining reporter expression in Dlx2-transfected cells, we examined the excitability of mCherry + OPC-derived GABAergic neurons at 14 dpt using whole cell patch-clamp electrophysiological recordings. In current-clamp configuration, we examined whether cells could generate action potentials in response to a series of depolarizing current injections. When we recorded from Tol2-controltransfected mCherry + cells, we saw a graded voltage response but no action potentials in 8 out of 9 recorded cells (Fig. 4A,C,D). We categorized cells with this pattern of voltage change as type 1 cells. One cell in the control cultures exhibited a single action potential (categorized as a type 2 cell). We were unable to record from GFP + mCherry + cells in control OPCs as they were extremely rare and difficult to locate. When we examined mCherry + Gad1-GFP + cells in Tol2-Dlx2-transfected cultures, we observed a range of different types of responses (Fig. 4B-D). About half of the cells exhibited a single action potential (type 2 cells), and one-third of the cells displayed a burst of several action potentials that quickly adapted in the continuous presence of depolarizing current (type 3 cells), indicative of immature neurons. Notably, in 3 out of 15 cells we detected continuous non-adapting spiking trains of action potentials (type 4 cells). The heterogeneous action potential patterns likely reflect the transitional state of OPC-derived neurons from an electrically immature to a more mature state at 14 dpt after Dlx2 misexpression.

OPC-derived immature neurons develop GABAergic synaptic puncta.
To look for morphological evidence of synapse formation in OPC-derived neurons, we immunolabeled for synaptic proteins. At 14 dpt, we observed clustering of VGAT at the junction between a GAD67 + axon and a MAP2 + GAD67 + dendrite in areas of high cellular density (Fig. 4E arrowheads), indicative of an early stage of synapse formation. At 21 dpt, we saw clusters of VGAT and the inhibitory post-synaptic scaffolding protein gephyrin around the soma and proximal process of MAP2 + neurons derived from Dlx2-transfected OPCs, indicating the formation of morphological synapses (Fig. 4F, arrowheads). By contrast, none of the neurons in control-transfected cultures expressed VGAT. In Dlx2-transfected cultures we did not detect synaptic puncta that contained VGLUT1 or 2 and the excitatory post-synaptic scaffolding protein PSD-95, consistent with the lack of excitatory neuronal transcripts detected by the RNA-seq. Thus, OPC-derived GABAergic neurons formed exclusively inhibitory morphological synapses amongst themselves.

OPC-derived interneurons acquire a mixed neurochemical phenotype. We next examined
whether OPC-derived GABAergic neurons were differentiating into a specific class of inhibitory neurons by examining the presence of transcripts known to be expressed by PV + and SST + neurons 36,49 . We focused our analysis on signature genes of PV + and SST + interneuron classes, as they make up the majority of the cortical inhibitory neurons. We performed qPCR on Dlx2-transfected and control-transfected OPCs at 14 dpt to determine whether any of the transcripts were upregulated in Dlx2-transfected OPCs. For the genes known to be expressed by PV + neurons, we assessed, in addition to Pvalb, (1) Mef2c, a transcription factor critical for PV + interneuron development; (2) Erbb4, a receptor tyrosine kinase (RTK) that binds epidermal growth factor; and (3) Plcxd3, a phospholipase C protein 50 . We also included early markers of PV + interneurons: Ccnd2 encoding the G1/S cyclin D2 and St18, a transcription factor involved in interneuron migration 51 . Of these, only Erbb4 was significantly upregulated in Dlx2-transfected cells compared to control-transfected cells, although there was a trend for increased expression of the other genes in Dlx2-transfected (Fig. 4G).
We next examined genes that are known to be expressed by SST + interneurons 36,49 . At 14 dpt Dlx2-transfected cultures expressed a significantly higher level of mRNA encoding Sst and Satb1, a transcription factor that plays a critical role in the specification and maturation of SST + interneurons and persists in adult SST + interneurons 52,53 . We did not detect a significant change in Epha5, an early marker of SST + interneurons that disappears upon maturation 51 . Vip expression was not significantly upregulated. As expected, Gad1 expression, which is common to all interneuron classes, was significantly upregulated in Dlx2-transfected OPCs. To examine whether www.nature.com/scientificreports/ Dlx2-transfected OPCs expressed the signature interneuron proteins, we immunolabeled 14 dpt OPC-derived neurons for SST or PV. However, we did not observe any immunoreactive cells, despite strong immunolabeling with both antibodies on tissue sections (not shown). These data suggest that OPC-derived interneurons have not yet matured into an unambiguous class of interneurons. This could be due to class heterogeneity among OPC-derived neurons or that they were differentiating into a cell type with intermediate properties of SST + and PV + neurons.

Discussion
In this study, we show that misexpression of a single transcription factor Dlx2 caused postnatal neocortical OPCs to switch their fate from oligodendrocyte lineage to GABAergic inhibitory neurons. RNA-seq analysis of OPCs during the window of fate conversion into neurons revealed that the transcriptional program towards inhibitory neuron differentiation had already begun within two days of exposure to Dlx2. OPC-derived neurons exhibited clustering of GABAergic synaptic proteins, and the active membrane properties of some resembled those of PV + inhibitory neurons at 14 dpt. However, transcriptional analysis at 14 dpt did not reveal evidence for full maturation into a specific class of inhibitory neurons.

Intrinsic properties that facilitate reprogramming of OPCs into GABAergic neurons. Previous
reports that used Dlx2 to reprogram non-neuronal cells into inhibitory neurons invariably required additional transcription factors, particularly Ascl1, which functions as a "pioneer" transcription factor. Pioneer factors alter the chromatin landscape around closed chromatin regions by recruiting chromatin remodeling complexes and enzymes that catalyze histone posttranslational modifications, thereby making the key neuronal genes accessible to neurogenic transcription factors 54 . The presence of Ascl1 has been shown to increase the efficiency of Dlx2 to reprogram astrocytes into GABAergic interneurons from 35 to 93% 55 . By contrast, our results showed that Dlx2 alone could reprogram OPCs into inhibitory neurons at an efficiency of 94%. One possible explanation for the ability of Dlx2 alone to convert OPCs into inhibitory neurons without other transcription factors is that OPCs express a low level of Ascl1 (Supplementary Table 1) 56,57 , obviating the need for exogenous Ascl1. This is supported by our recent finding that in OPCs genes encoding the key critical transcription factors for inhibitory neurons exist in a more readily accessible chromatin state than those in astrocytes or fibroblasts 15 . This could also account for the selective activation of inhibitory but not excitatory neuronal genes. Whether the relatively de-repressed state of inhibitory neuron genes in OPCs is mediated by Ascl1 or other transcription factors is currently unknown. We previously showed that conditional knock-out of Olig2 in postnatal cortical OPCs converts their fate into protoplasmic astrocytes but not into neurons 58,59 . A similar fate conversion to astrocytes but not neurons occurs after deletion of HDAC3, which causes downregulation of Olig2 60 . By contrast, we did not observe astrocyte differentiation from Dlx2-transfected OPCs, despite an early loss of Olig2 (Fig. 1, Supplemental Fig. 1B). Although the exact mechanism underlying the fate choice of OPCs after loss of Olig2 remains unclear, it is likely that loss of Olig2 in the presence of Dlx2 triggers a transcription factor network that represses STAT3 and other astrocyte genes.
In OPCs transfected with Dlx2, there was a gradual loss of oligodendrocyte lineage antigens from 2 to 7 dpt, but roughly 25% still contained detectable Olig2, PDGFRα, and Sox10 at 7 dpt, suggesting that roughly 25% of OPCs are resistant to reprogramming by Dlx2 or take longer to be reprogrammed. The shutdown of oligodendrocyte lineage proteins could take longer in some OPC that are more "mature" and farther along the oligodendrocyte lineage than those that had newly differentiated from NPCs. This is supported by our recent finding that upon Olig2 deletion in OPCs, Sox10 expression disappears within a few days when Olig2 is deleted prenatally but remains for several weeks when deleted at P18 58,59 .

Partial recapitulation of developmental transcription network triggered by Dlx2 misexpression in OPCs.
During interneuron development in the medial ganglionic eminence, Dlx1/2 first specifies GABAergic lineages by repressing Olig2. Subsequently, in postmitotic cells, Dlx1/2 promotes further neuronal differentiation by inducing Gad1/2 and Dlx5/6 (reviewed in 61 ), as well as Sp9. These immediate targets of Dlx1/2,  www.nature.com/scientificreports/ as well as Dlx1, were all upregulated in OPCs after less than 48 h of exposure to Dlx2. We further used qPCR to examine the extent to which these secondary transcription factors were sustained in Dlx2-transfected OPCs. In our qPCR analysis, we could not attain the sensitivity obtained by RNA-seq at 2 dpt, since the OPC-derived neurons were too fragile for FACS purification at later time points, and many untransfected cells were mixed in the analyzed population. Nevertheless, the results from qPCR showed a trend for an early upregulation of Dlx1, though not statistically significant, and a significant upregulation of Dlx6 and 5, as well as significantly elevated and sustained expression of Gad1 through 14 dpt. Elevation of Dlx1 and Dlx6 was transient and declined by 14 dpt, whereas the most significant upregulation of Dlx5 expression occurred between 7 and 14 dpt. Interestingly, two lncRNAs in the Dlx family of genes, Dlx6as and Dlx1as, were significantly upregulated at 2 dpt in our RNA-seq data, and this was further validated by qPCR, suggesting a greater magnitude of elevation of these genes compared to the other Dlx genes. The upregulation of both Dlx lncRNA transcripts was transient, and by 7 dpt they returned to a level that was not significantly above baseline levels. The steep early upregulation of the two Dlx lncRNA species could play a significant role in modulating the temporal profiles of the levels of Dlx1/5/6 genes. The specific mechanism of action of Dlx1as and Dlx6as in this context remains unclear, and reports of studies in different cell types vary. Dlx6as has been shown to recruit and form a complex with Dlx2 to activate the Dlx5/6 intergenic enhancer 62 . Thus, it is possible that the elevated expression of Dlx6as is contributing to the elevated expression of Dlx 6 and 5 observed at 7 and 14 dpt, respectively. This is consistent with the report that knockdown of Dlx1as in adult SVZ cells decreased Dlx1 and Dlx2 expression 47 . On the contrary, Dlx1as knock-out mice have higher Dlx1 expression and exhibit the phenotype of gain-of-function of Dlx1 46 . If Dlx1as positively affects Dlx1 expression in OPC-derived neurons, the early induction of Dlx1as may have played a role in inducing Dlx1 expression, but the transient nature of Dlx1as expression may not have been sufficient to sustain Dlx1 expression.

Incomplete manifestation of interneuron class phenotypes in OPC-derived neurons. Dlx2-
transfected OPCs did not exhibit the fully differentiated phenotype of any of the three major interneuron classes known to exist in the neocortex. RNA-seq data at 2 dpt did not detect upregulation of Pvalb, Sst, or Vip. By 14 dpt, we detected a small but significant increase in Sst transcript by qPCR, suggesting an incomplete differentiation toward SST + interneurons. Interestingly, Sp9, which was found to be elevated at 2 dpt by RNA-seq, is known to promote the activation of Satb1 and St18 63 , both of which were elevated at 14 dpt in qPCR analyses. Satb1 is critical for the survival and maturation of SST + and some PV + interneurons 52,53 , and is characteristically expressed as one of the signature transcripts of SST + interneurons (database by the Linnarsson lab, http:// mouse brain .org/cellt ypes/).
During development, Dlx1 is expressed transiently in early PV + interneuron-fated progenitors and is downregulated upon PV specification, whereas Dlx1 expression is sustained as progenitor cells mature into SST + interneurons 64 . The decline of Dlx1 beyond 7 dpt in OPC-derived neurons could be related to the emergence of a transcriptional signature of SST + neurons. However, electrophysiological recordings of OPC-derived neurons revealed that 3/15 cells exhibited a train of action potentials with an average firing rate of ~ 85 Hz upon current injection with no sign of adaptation. The range of action potential firing frequency was higher than that of the fastest firing SST + neurons, but lower than that of mature PV + neurons and more similar to immature PV + neurons seen during the first postnatal week [65][66][67] . The most common action potential firing pattern seen among Dlx2-transfected cells was several spikes of decaying amplitude after current injection followed by inactivity. This is often seen in reprogrammed neurons 12,55,68 , suggesting a partially differentiated phenotype, possibly reflected by the density and type of Na + and K + channels. OPCs in different functional states have different Na + channel densities, though they are generally too low to generate regenerative action potentials 69,70 . Thus, types 2 and 3 firing patterns observed in Fig. 4 likely represent a gradient of neuronal differentiation from OPCs.
The health of the OPC-derived neurons began to deteriorate beyond 21 dpt under the culture conditions used, and we were unable to examine whether the cells developed more characteristic neurochemical and biophysical phenotypes of more mature interneuron classes over time. Since progenitor cells in the medial ganglionic eminence acquire competence to produce SST + neurons earlier than PV + class of interneurons 61 , it is possible that the detection of Sst, Satb1, and St18 is reflecting the relatively early stage of interneuron maturation. The lack of detectable PV or SST immunoreactivity in OPC-derived neurons at 14 dpt is also consistent with their immature state. We have also noted that Lhx6, a critical secondary interneuron transcription factor, was not upregulated in Dlx2-transfected OPCs. Additional transcription factors such as Nkx2-1 may be needed in parallel with the transcriptional network that is activated by Dlx2.
Of note, our electrophysiological recordings were conducted from OPC-derived neurons that had integrated Dlx2-IRES2-mCherry into the genome. While beneficial for lineage tracing of transfected cells, sustained Dlx2 expression may have impeded full neuronal maturation. Future studies may be conducted with temporal control of gene Dlx2 expression 71 to further promote maturation of OPC-derived interneurons.
Inputs from surrounding neurons, particularly those from excitatory neurons promote the survival of postmitotic immature interneurons [72][73][74] . Thus, it is possible that further maturation of OPC-derived neurons in a recent in vivo study in which striatal OPCs were reprogrammed into neurons with multiple transcription factors 14 may have been facilitated by the local in vivo environment. Our findings provide a logical framework for further testing strategies to promote interneuron reprogramming from OPCs under pathological conditions in which greater inhibitory neuron function is needed. www.nature.com/scientificreports/