Adult Hooded Long Evans Rats (n = 4) were deeply anesthetized, guillotined, and brains removed. Brains were cut into 2 mm slabs with a coronal brain matrix and the neocortex carefully dissected to avoid inclusion of white matter under a dissecting microscope in chilled HBSS buffer. Tissue was dissociated using a papain-based neural dissociation kit (Miltenyi) and cells were selected for the O4 antigen via Magnetic Activated Cell Sorting (MACS) and cultured as described previously (Dincman et al., 2012). OPCs were grown and passaged on PDL/laminin coated plastic dishes or glass coverslips in OPC growth medium (DMEM/F12, BSA (0.1%), N2 (1%), PDGFaa (10 ng/mL), FGF2 (20 ng/mL), and Insulin (5 μg/mL)) at 5% CO₂. Cells were passaged, cryopreserved, and maintained as a primary cell line with no later than passage 7 used in experiments.

Differentiation potential was assessed by replacing OPC growth media with Oligodendrocyte Differentiation Media (DMEM/F12, N2 (1%), B-27 Supplement (B27 1%), Penicillin/Streptomycin/Fungizone (PSF 1%), Insulin (50 ng/mL), Triiodothyronine (T3 40 ng/mL)) or Astrocyte Differentiation Media (DMEM/F12, Fetal Bovine Serum (10%), PSF (1%)). Experiments were performed in triplicate.

OPCs were transduced with OPC growth media containing retroviral constructs carrying the transgene for candidate neurogenic transcription factors (Supplementary Table S1). In vitro screening studies initially used a mixture of the constitutive CAG promoter or the OPC-specific NG2 promoter in different constructs. No advantage was conferred by NG2-promoter expression and to avoid potential silencing with neuronal differentiation, we chose to standardize on the CAG promoter for our studies. After 24 h, transduction media was replaced with Neuronal Reprogramming Media (Neuralbasal Media, B27 (2%), Glutamax (1%), PSF (1%)) at 10% CO2 with BDNF (20 ng/mL) added on Day 5 and every 4 days thereafter with no media changes (Heinrich et al., 2011). For co-culture experiments, PN1 rat neurons (derivation described in Sun et al., 2008) were plated onto reprogrammed neurons within 48 h of their transduction. Experiments were performed in triplicate. Time-lapse imaging studies were conducted using a Leica SP8 resonance confocal microscope equipped with an environmental stage (Okolab) and cultures were maintained at 37°C and 5% CO₂ for the duration of the imaging session.

Coverslips were transferred and rinsed in a well of PBS and fixed in 4% paraformaldehyde (0.1 M phosphate buffer). Three rinses in Tris-buffered saline (TBS) preceded a 1 h block in 5% donkey serum/0.25% TritonX-100/TBS solution (TBS++). For multiple immunostaining, sections were incubated overnight (at 4°C on a shaker table) with primary antibodies diluted in TBS with 0.25% TritonX-100 (TBS+). Primary antibodies used are listed in Supplementary Table S2. Following two 30 min block/rinses with TBS++, sections were incubated for 2 h with Alexa Fluor 488 (Molecular Probes), Cy3, or Cy5 (Jackson Immunoresearch) diluted in TBS+ (1:500) for 2 h at room temperature. Following two 15-min rinses with TBS, cells were counterstained with DAPI, rinsed with TBS, and coverslips were mounted onto slides, with a glycerol-polyvinyl alcohol plus DABCO (1,4 diazabicyclo [2.2.2] octane; Sigma D2522, Sigma-Aldrich, Inc., St. Louis, MO, United States) PVA-DABCO solution to self-seal and prevent fading, and stored at 4°C in the dark. Negative controls lacking primary antibodies were run along samples.

To assess the efficiency of both viral transduction and neuronal conversion, the number of total cells (DAPI), virally transduced cells (GFP), and converted neurons (beta-III-tubulin) was estimated at each time in the screening assay by systematic sampling of coverslips in triplicate. Coverslips were imaged on an Olympus Spinning Disk (DSU) confocal microscope under the control of stereological software (StereoInvestigator, MBF Bioscience, Inc.) to achieve systematic, fractionated sampling. The entire coverslip was defined as a contour and the sampling density adjusted to acquire images at 10 sites with a randomized start site. This approach assured that all regions of the coverslip had an equal probability of being sampled. Confocal stacks (20 focal planes at 1 μm intervals) were automatically collected by the software at each site for each channel of fluorescence. Cells of each phenotype (DAPI, GFP, and beta-III-tubulin) were counted using the software to obtain an estimate of total cells (DAPI), transduction efficiency (GFP cells over DAPI cells), and conversion efficiency (beta-III-tubulin cells over GFP cells) per coverslip. The extent of colocalization of beta-III-tubulin was also calculated. As the physical parameters of the coverslip did not change, the results are expressed as the sum of the 10 fields sampled. Efficiency calculations were defined as the mathematical ratio of the mean values.

Cells were rinsed with PBS and harvested with no more than 5 × 10⁵ cells resuspended in 1 mL CCM, then lysed and extracted with the RNeasy Plus Micro kit (Qiagen) using larger volumes for the purification of RNA from cells protocol. RNA concentrations were obtained using a Nanodrop 1000 (Thermo Scientific) using 1.5 μL RNA. Each sample was then assessed for quality using the Experion Capillary Electrophoresis System (Bio-Rad). Samples were diluted, denatured and prepared according to the manufacturer’s directions for the Experion RNA HighSens analysis kit (Bio-Rad). They were run on the same chip as an extraction negative control and a positive control and compared using the Bio-Rad Experion software (v3.2.243.0) to a supplied denatured, diluted RNA ladder. Only samples with an RQI (RNA Quality Indicator) number greater than 7.0 were used for arrays. 400 ng of RNA from each sample was reverse transcribed using the ImpromII Reverse Transcription system (Promega, manufacturer’s directions). 91 μL of RNase-free water was added to the cDNA for each sample. The PCR component mix was made up using iTaq Universal SYBR Green supermix (Bio-Rad) for 96-well format A for use in a Bio-Rad iCycler as instructed by the manufacturer (Qiagen). Each sample was pipetted into a Rat Neurogenesis array (Qiagen PARN-404Z) and performed on a Bio-Rad iCycler. The same threshold was set for all arrays and controls checked. The rat genomic DNA contamination control indicated the absence of genomic DNA. The positive PCR control wells were within 2 cycles of a Ct of 20, indicating a lack of inhibitors. Data was analyzed using the RT2 Profiler PCR Array Data Analysis software (v3.5) from Qiagen. Three reference genes that had a difference in Ct values of <2 across all samples were chosen to analyze the data. Delta-delta Ct was calculated for each well to find genes that were under and over expressed by more than 4 fold in the 7dpt and 15dpt day groups when compared to the OPC control cells.

Electrophysiological recordings were conducted at room temperature (22°C) using standard whole-cell patch-clamp techniques, extracellular artificial cerebrospinal fluid gassed with 95% O₂/5% CO₂ and containing (in mM) 130 NaCl, 2.5 KCl, 2.0 CaCl₂, 1.2 MgSO₄, 1.25 KH₂PO₄, 25 NaHCO₃ and 10 dextrose, and intracellular solution containing 135 K-gluconate, 2.0 MgCl₂, 4.0 Na₂-ATP, 0.4 Na-GTP, 10 Na-phosphocreatine and 10 HEPES adjusted to pH 7.3 with KOH. Cells were selected based upon neuron-like morphology and GFP marker expression (n = 15 from triplicate preparations). Recordings were conducted in both current-clamp and voltage-clamp modes with cell potential adjusted to −70 mV using Axon Instruments Multiclamp 700B amplifier, Digidata 1,440 analog-digital converter and pClamp 10.2 software. Input resistance was measured with −20 pA, 500 msec current steps and neuron-like excitability (action potential firing) was tested using 500 msec current steps ranging from −50 to +100 pA in current-clamp. Sodium currents and spontaneous excitatory postsynaptic currents (sEPSCs) were measured in voltage-clamp. sEPSCs were detected and measured offline using Minianalysis software with threshold amplitude set to 7.5 pA to avoid false positives from noise. Datasets of three 1-min recording epochs before, during and after CNQX per cell (n = 4) were analyzed by 1-way ANOVA followed by Tukey’s multiple comparisons test.

3rd generation (self-inactivating) VSV-G pseudotyped Murine Moloney Leukemia Virus vector were produced by transfection of packaging (Gag/Pol, VSV-G) and vector plasmids into 293 T cells with the retroviral vectors collected into the cell culture supernatant or into Opti-MEM medium (serum free) for concentration by ultracentrifugation (Tiscornia et al., 2006). All vector transgenes were driven by the CAG promoter and co-expressed eGFP or dsRed under an internal ribosomal entry site (IRES) sequence. See Supplementary Table S1 for the origins of each retroviral vector plasmid.

Seven to eight week-old young adult male Long Evans Hooded rats were purchased from Harlan Sprague Dawley (Indiana, United States) for sole use in this study. Rats were housed two to three per cage under standard laboratory conditions with a light/dark cycle of 12 h with food and water ad libitum. All experiments were performed according to and approved by national and institutional guidelines (NIH, IACUC). Intracranial injection procedures adopted to minimize mechanical trauma to the cortex included careful drilling of a skull opening so that no mechanical compression of the dura or parenchyma occurs, use of a small gage needle to superficially create an opening in the dura for insertion of the delivery needle without contact with the cortical surface, and use of a specially designed 33G needle (Hamilton Neuros Syringe) to minimize needle track volume combined with slow delivery of viral volume. With these procedures, neuronal loss is confined to the volume of the 33G needle and hypertrophy of astrocytes is confined to the lining of the needle track and the cortical surface at the site of insertion. Rats were randomly assigned to viral delivery groups for the different transgenes. Viral injections were done for each transgene in groups of three to minimize wastage of the virus. Rats within each group were randomly assigned for collection of brains at the different analysis times. Rats were anesthetized with Isoflurane inhalant and injected with 3ul of concentrated retroviral CAG-eGFP (3 × 10⁸) or CAG-Neurogenin2-IRES-eGFP (3 × 10⁸) using a motorized stereotaxic injector (Stoelting) with delivery at 0.2 μL/min into either the right motor cortex or the right entorhinal cortex. The following coordinates were used: for motor cortex [from Bregma (mm): nosepiece −0.3; A/P: +0.5; M/L: −2.5; D/V(from dura): −2.0 raised 0.1 every minute] and for entorhinal cortex [from Bregma (mm): nosepiece: −7.0; A/P: ±9.2; M/L: −5.0; D/V(from dura): −4.5 raised 0.1 every 1 m 30 s].

Brains were collected at 1 and 2–3 weeks post injection following transcardial perfusion with 0.9% saline wash, followed by 4% paraformaldehyde (PFA) in phosphate buffer, then post-fixed in 4% PFA overnight at 4°C, equilibrated in 30% sucrose, and sectioned at 40 μm on a freezing stage microtome (Leica). Using unstained wet-mounted sections, all brains were screened for the detection of native GFP expression; brains that did not contain fluorescent cells or brains where the injection reached the corpus callosum or otherwise were not entirely within the cortex were not included in the study. Final group sizes were GFP (27) and Ngn2 (27). For immunofluorescence staining, net inserts (Corning Costar) were used to transfer free-floating sections between solutions and minimize handling. To remove cryoprotectant, all sections were rinsed multiple times in TBS followed by 3 h block in TBS++. For multiple immunostaining, sections were incubated for 72 h (at 4°C on a shaker table) with primary antibody diluted in TBS+. Primary antibodies and dilutions used are listed in Supplementary Table S2. Following two 1 h blocks with TBS++, sections were incubated for 48 h with Alexa Fluor 488 (Molecular Probes), Cy3, or Cy5 (Jackson Immunoresearch) diluted in TBS+ (1:500) for 2 h at room temperature. Following two 15-min rinses with TBS, all sections were mounted onto slides, cover slipped with PVA-DABCO mounting media, and stored at 4°C in the dark. Negative controls lacking primary antibodies were run along samples. The use of different reporter genes in some cases prevented the blinding of investigators to group identity.

Confocal images were acquired using an Olympus DSU, Olympus Fluoview XV, Fluoview 500, or Leica SP8 resonance confocal microscope. Appropriate apertures and Nyquist sampling were used to acquire optimal resolution confocal stacks. Signal intensity was set objectively against a signal distribution histogram and with reference to positive and negative imaging controls to ensure an appropriate acquisition without oversaturation. Three-dimensional colocalization of signal was performed by generation of orthogonal views or through rendered three-dimensional visualization with rotation and orthogonal sectioning of the image volume using NeuroLucida software (MBF Bioscience, Inc.). Figures were composed in Adobe Photoshop with minimal adjustment to normalize signal distribution between panels and images were not otherwise manipulated.

The Navigator feature of the Leica LASX software was used to acquire a three-dimensional virtual tissue section of all sections containing GFP-positive cells. Imaging was performed on a Leica SP8 resonance confocal microscope using the 25 × 0.95 NA water immersion lens with a z-axis sampling interval of 1 μm to generate a focal series throughout the entire section thickness. All image stacks were stitched together to generate a virtual section. The series of virtual sections for each site of gene delivery were sampled using the StereoInvestigator software (MBF Bioscience, Inc.) to implement Optical Fractionator sampling to produce an estimate of total cell number (Peterson, 2004; Peterson, 2014). The sampling parameters used were: section series- every third section; area fraction 65 × 65 μm counting frame and a 125 × 125 μm sampling grid; optical disector height of 12 μm with a mean section thickness of 24.6 (±1.04) μm. This sampling density generated a mean coefficient of error (CE) value of 0.16 (±0.02). Cells were counted as they first came into focus using the optical disector counting rules and scored for single or multiple expression of fluorescent labeling. Data was summarized and analyzed using Prism 9.4 software (GraphPad, Inc.) with an initial ANOVA followed by a Bonferroni post-hoc test for significance with significance accepted at p ≤ 0.05.

To prepare for future studies of neurogenic engineering in rat models of injury and dysfunction, we isolated and cultured adult rat cortical OPCs for use as an in vitro assay to identify successful transcription factors for subsequent in vivo cortical delivery to induce neurons from resident non-reactive OPCs (Figures 1A–C). To screen developmentally relevant pro-neuronal transcription factors (Imayoshi and Kageyama, 2014) (Supplementary Table S1), adult OPCs were isolated from cortical gray matter to replicate the target cell population for in vivo reprogramming. OPCs were selected for expression of the O4 antigen (Dincman et al., 2012) from dissociated adult rat neocortical gray matter, and could be cultured with the ability to expand, passage, and cryopreserve cells (Figures 1B,D). All adult cortical-OPCs expressed pan-oligodendroglial markers, Olig2 and Sox10 (Nishiyama et al., 2009) (Figure 1E) and NG2 (Figure 1F), but displayed maturational heterogeneity within the OPC lineage with a concomitant loss of nestin and increased O4 expression as cells progressed from early bipolar to later multipolar morphologies (Figure 1F). Under OPC growth conditions, most cells retained an OPC phenotype and the premyelinating oligodendrocyte marker RIP was rarely observed (Figure 1G). However, under oligodendrocyte differentiation conditions, RIP became strongly expressed (Figure 1H) confirming that the isolated OPCs were indeed within the oligodendroglial lineage and could proceed to terminal differentiation as oligodendrocytes with appropriate instruction.

Delivery of neurogenic transcription factors (Supplementary Table S1) cloned into retroviral vectors resulted in robust expression of the neuronal lineage marker ß-III-tubulin by 7 days (Figures 2A–E) following delivery of the single factors Neurogenin2 (Ngn2; 50% efficiency) or NeuroD1 (50% efficiency), and to a lesser extent with the combined delivery of Ascl1 and Dlx2 (15% efficiency), but not in any of the other combinations, including GFP-reporter control (0%) and Pax6 (0%). Ngn2-induced neurons (Ngn2-iNs) expressed ß-III-tubulin as early as 3 days post transduction (dpt) and expressed Map2 (Figure 2E′). Non-infected cells continued to express Olig2 and O4, while these were downregulated in Ngn2-infected cells (Figure 2F). Initial transduction efficiency was approximately 60% and of those GFP-expressing cells, neuronal conversion efficiency was approximately 25% (i.e., 15% of total cells; Figures 2G,H).

To confirm lineage progression from OPCs to induced neurons, we performed time-lapse imaging over 70 h from four to 6 days post-infection and conducted a trajectory analysis on the lineage progression of individual cells (Figure 3). Single frame images from each day during the imaging period (Figures 3A–C) illustrate morphological progression from multipolar short processes of OPCs to the eventual long processes typical of neuronal morphology that contact neighboring cells. The dynamic nature of this interaction can be seen in the Supplementary Video S1 (https://doi.org/10.5281/zenodo.8189544). Immunostaining immediately following confirmed the expression in these cultures of the mature neuronal marker, NeuN (Figures 3D–F).

As an assessment of neuronal function following Ngn2 delivery, a neurogenesis-themed PCR array found that GFP-control vector largely matched the expression profile of OPCs while Ngn2-iNs upregulated neuronal genes (ie. DCX, Grin1) and the pro-neuronal transcription factor NeuroD1, but downregulated Olig2 by 7 dpt (Figures 4A,B). When co-cultured with P1 rat cortical neurons, Ngn2-iNs developed elaborate dendritic processes with punctate synaptophysin-positive contacts (Figures 4C–E), showed strong neuron-like voltage gated sodium current (Figure 4F), generated action potentials upon depolarization (Figure 4F), and displayed spontaneous excitatory post-synaptic currents that were inhibited by the glutamatergic AMPA receptor blocker, CNQX (Figures 4I,J). Thus, in addition to showing morphological and phenotypic features consistent with a neuronal identity, Ngn2-iNs evidence a similarity in transcriptional profile. Furthermore, Ngn2-iNs exhibited functional contacts and neuronal membrane properties consistent with a glutamatergic phenotype.

We next asked if in vivo retroviral delivery of Ngn2 could also reprogram non-reactive OPCs within the naive rat cerebral cortex into neurons. As predicted from the fact that OPCs are the primary proliferative cells in the naive adult cortex, we found that GFP-reporter labeled cells expressed the pan-oligodendroglial marker, Olig2, with characteristic OPC morphology (Figures 5A,B). Based upon viral delivery parameters, each injection infected a fraction of Olig2-positive cells in this cortical volume in the rat, reflecting the proportion of OPCs undergoing cell cycle at the time of retroviral availability. With GFP-reporter only delivery, infected cells rarely express markers for immature (DCX) neurons and then only weakly. However, GFP-positive cells did not co-express markers for mature (NeuN) neurons, mature astrocytes (S100ß), or microglia (Iba1) at 7 days post injection (dpi; Figures 5A–D). By using detection of soma-localized S100ß for astrocytes, we were able to unambiguously exclude co-localization of astrocytes with retroviral GFP-reporter (Figure 5D).

Olig2 is a transcription factor expressed throughout the oligodendrocyte lineage apart from mature oligodendrocytes, while Sox10 is a transcription factor expressed from early OPCs to terminal oligodendrocytes. OPCs begin to downregulate expression of NG2 as they progress to becoming mature oligodendrocytes. To further characterize the proliferating cell population targeted by retroviral gene delivery, which infects only dividing cells, we analyzed control eGFP-only in vivo retroviral delivery at the 7 day post-injection interval and found that 85.9% (±2.6%) were Sox10+/NG2+ OPCs and the remaining 13.8% (±2.7%) were Sox10+/NG2− oligodendrocytes. Thus, 99.7% (±0.03%) of GFP-expressing cells are OPCs or OPCs progressing toward terminal differentiation. As only proliferating OPCs were targeted by retroviral delivery, only 23.4% (±5.8%) of OPCs present at the point of retroviral delivery expressed GFP.

Retroviral delivery of Ngn2 induced DCX expression in nearly all Ngn2-GFP-positive cells by 7 days (Figures 6A–D). Ngn2-iNs began to express NeuN, although at weaker levels than pre-existing neurons by 7 dpi. Given the robust process extension and morphological remodeling observed of Ngn2-iNs in vitro at a similar time (Figure 3 and Supplementary Video S1), it is perhaps not surprising that in vivo Ngn2-iNs exhibited substantial process extension and complexity with cell polarity suggesting a transitional immature neuronal morphology (Figure 6E). These Ngn2-iNs also expressed the early neuronal lineage marker DCX, confirming their progression through a neuronal lineage (Figure 6E).

As Ngn2 reprogramming in vitro also drove very high expression of NeuroD1 (Figures 4A,B), we next asked if in vivo Ngn2 reprogramming alone was sufficient to generate pyramidal neuronal morphology with glutamatergic subtype specification. By 3 weeks, Ngn2-iNs were observed in superficial layers of the cortex (Figures 7A–C) showing a continuum of progression toward a mature neuronal morphology with distinct dendritic and axonal processes and, in all cases, distinctly co-expressing the mature neuronal marker NeuN. In all cases, signal co-expression was assessed in three-dimensions to verify that signals overlap in the same cellular compartments. All neurons, including the Ngn2-iNs, were closely associated with Iba-1-positive microglia (Figures 7A–C), but no microglia were observed to co-express GFP, indicating the absence of non-specific GFP uptake.

Ngn2-iNs were also observed by 3 weeks in deep cortical layers (Figures 7D–F) that distinctly co-expressed NeuN and exhibited cytoarchitecturally appropriate pyramidal cortical neuron morphology, including primary and basal dendrites, dendritic arbor elaboration, and dendritic spines. Evaluation of GFAP-positive cells confirmed the absence of GFP expression in astrocytes. Separate Ngn2-GFP gene delivery to the phylogenetically older entorhinal cortex (allocortex) also resulted in the generation of induced neurons by 3 weeks (Figures 7G–I) demonstrating that OPCs in regions other than neocortex may be amenable to neuronal reprogramming.

In vitro data had demonstrated that Ngn2-iNs established functional glutamatergic contacts (Figure 4). To determine if Ngn2-iNs generated following in vivo gene delivery adopt a glutamatergic neuronal subtype specification, GFP-positive Ngn2-iNs were examined for co-expression of the cortical glutamatergic neuronal markers Tbr1 and Cam-II-kinase-α. Evaluation of three-dimensional imaging (Figures 7J–L) revealed that both markers were detected, indicating that rat cortical OPCs had been reprogrammed by Ngn2 expression alone in the absence of prior injury into a glutamatergic neuronal subtype. We next evaluated dendrites and dendritic spines on Ngn2-iNs (Figures 7M–O) and identified synaptophysin-positive contacts on both GFP-positive dendrites and dendritic spines indicating the presence of morphological contacts with the pre-existing cortical circuitry.

We performed stereological quantitation (Figure 8) to assess the number of GFP-positive cells in total for neocortical retroviral gene delivery of both control GFP-only and Ngn2-GFP at three time points (7, 14, and 21 days post-delivery). Cells were counted based upon their expression of GFP alone, or coexpression with the early neuronal lineage marker DCX (7 and 14 day groups) or the mature neuronal marker NeuN (21 day group). The volume of cortex occupied by GFP-positive cells did not differ between control and Ngn2 induction at any time point but did reveal a decline over time (Figure 8A). The basis for this volumetric decline can be seen by the estimation of total GFP-positive cell number (Figure 8B). Although there was no difference at any time between control GFP-only and Ngn2-GFP conditions, overall the populations of cells declined. The extent of initial gene delivery distribution within the cortex varies from subject to subject. This variable, and likely other nuances of intracerebral delivery, contributed to substantial variance with the result that statistical evaluation by ANOVA did not produce adequate significance to proceed with subsequent between group tests.

Estimation of the number of GFP-positive cells that showed neuronal lineage commitment (Figure 8C) demonstrated that the Ngn2-induced groups significantly generated induced neurons compared to the control GFP-only group. In fact, the control GFP-only group had few GFP-positive cells that were weakly DCX-positive. Despite the weakness of staining, these cells were included in the counts for 7 and 14 day groups. In the 21 day group for the GFP-only control condition, no GFP-positive cells were detected that co-expressed NeuN apart from a single GFP-positive cell that showed weak positivity for NeuN. This cell was included in the count, which also enabled statistical testing to be performed for the 21 day group. When evaluated for neuronal lineage commitment as a percentage of GFP-positive cells (Figure 8D), the weakly DCX-positive cells in the control GFP-only condition represent less than 4% of GFP-positive cells in the 14 day group. The single weakly NeuN positive cell represents less than 1% of GFP-positive cells in the 21 day group. In contrast, neuronal lineage commitment is significantly higher in the Ngn2-induced condition, with more than 85% of GFP-positive cells identified as neurons by 21 days.

Retroviral transgene expression is restricted to cells undergoing cell division and thus also serves as a marker of cell cycle (Morshead and van der Kooy, 1992; Torper and Gotz, 2017). We therefore anticipated that retroviral delivery of reprograming factors to the naïve cortex would primarily target OPCs (Dawson et al., 2003), while also serving as a control for the inability to infect pre-existing post mitotic neurons. This straightforward delivery strategy also avoids potential false positive detection due to leaky expression of Cre under OPC-specific promoters that has been responsible for mistaken in vivo differentiation of OPCs to neurons (Richardson et al., 2011). Although a previous report (Guo et al., 2013) identified retroviral infection of astrocytes in mouse cortex, these cells only proliferate several days following an injury, and astrocyte proliferation may be misidentified and overstated due to the extent of GFAP upregulation (Burda and Sofroniew, 2014; Dimou and Gotz, 2014). This study design avoided any prior injury or activation of glial cells into a reactive state that would also stimulate proliferation in other cell types (Hampton et al., 2004), resulting in the absence of any retrovirally-delivered GFP detected in cells other than the constitutively proliferating OPC population. Thus although it is possible that occasional proliferating pericytes, endothelial cells, microglia, or astrocytes could potentially be included among infected cells, on a population basis, these other cell types were not detected to express GFP and the infected cell population was primarily OPCs, consistent with the non-reactive milieu at the time of viral infection. In control GFP-only conditions, we observed a portion of the GFP-expressing cells were Sox10-positive without NG2 expression. It is possible that some of these cells may have been oligodendrocytes proliferating at the time of retroviral delivery (Simon et al., 2011). The remaining population may have been OPCs at the time of retroviral delivery that subsequently downregulated NG2 as they progressed over the 7 days toward terminal differentiation as oligodendrocytes, a number and timeframe that is consistent with another report (Shimizu et al., 2020). Furthermore, we were able to exclude the participation of SVZ-derived neuroblasts as the source of our identified Ngn2-iNs based upon their absence in control GFP-only conditions and the identification of Ngn2-iNs in the entorhinal cortex, the most remote cortical region from the anterior SVZ.

Neurons have distinct phenotypic subtypes and the generation of new generic neurons without these properties and without appropriate morphology may be of limited value for repair. Thus, our goal was to achieve reprogramming of naive OPCs to distinct neuronal subtypes with cytoarchitecturally appropriate neuronal morphology. An important validation to the authenticity of neuronal reprogramming is the progression from earlier to more mature neuronal phenotypes (Calzolari and Berninger, 2021). Both our in vitro and in vivo data demonstrated that expression of Ngn2 in OPCs resulted initially in a transition in morphology and the expression of early neuronal lineage commitment markers. By following the morphological trajectory of cultured rat OPCs expressing GFP, the transition to a neuronal morphology could be confirmed in individual cells and the expression of mature neuronal markers subsequently confirmed.

Following in vivo delivery, Ngn2-iNs begin to adopt more mature neuronal morphologies and express more mature phenotypic markers and evidence of synaptic contact with the pre-existing neuronal circuitry. Newly Ngn2-engineered neurons were detected in both more superficial and deep layers of the motor cortex (neocortex) and also the phylogenetically distinct entorhinal cortex (allocortex). The appropriate pyramidal neuron phenotype of Ngn2-iNs was confirmed by their expression of the cortical glutamatergic neuronal markers that matched adjacent pre-existing mature neurons and indicates phenotypic subtype specification, consistent with the in vitro functional data indicating glutamatergic specification. The adoption of cytoarchitecturally appropriate, location specific morphology also suggests competence to respond to remaining local environmental factors in establishing final maturation. For example, inverted pyramidal neurons comprise a small projection neuron subset confined to the deep cortical layers (Mendizabal-Zubiaga et al., 2007), and this morphology is appropriately induced in a subset of Ngn2-iNs in the deep layers. Similarly, entorhinal Ngn2-iNs adopt a location appropriate morphology. Thus, our ability to target OPCs almost exclusively by using retroviral gene delivery and the demonstrated progression from earlier to mature neuronal lineage validate that these are newly generated neurons and argue against the concerns that have been identified with some previous studies (Calzolari and Berninger, 2021; Wang et al., 2021; Chen et al., 2022; Cooper and Berninger, 2022).

Apart from a single weakly NeuN-positive cell observed in one GFP-control subject at 21 days, no GFP-neurons were detected in control GFP-only injected animals. Nevertheless, the observation of morphologically appropriate neurons emerging by 21 days following retroviral delivery of Ngn2 led us to examine the possibility of a fusion event or other artifact, whereby GFP was simply transferred to a pre-existing neuron. Fusion events have been reported in neonatal brain, where it was mediated by infected microglia, but not in adult cortex (Ackman et al., 2006). We carefully examined Ngn2-iNs using high-resolution imaging and 3-dimensional rendering and found no cases of process fusion or double nuclei. Ngn2-iNs were contacted by Iba1-positive microglia, as were pre-existing neurons but in no cases were the Iba1 cells also GFP-positive. Furthermore, Ngn2-iNs passed through an immature, DCX-expressing neuronal state with advanced morphological complexity prior to the appropriate neuronal morphology seen later. In addition, Ngn2-iNs are not synchronous in their lineage progression and display a range of maturation, further arguing against transfer of GFP to pre-existing neurons.

The reduction in number of infected cells prior to maturation suggests that only fully reprogrammed Ngn2-iNs survive the reprogramming experience (Gascon et al., 2016). This low survival rate is not entirely surprising as 50–80% of adult newborn dentate granule cells die in an environment that is normally supportive of adult neurogenesis (Dayer et al., 2003; Sandoval et al., 2011). Similar outcomes in efficiency of neuronal reprogramming have been linked to regional heterogeneity of astrocytes (Hu et al., 2019) and this may also be true of OPCs. The efficiency of neuronal reprogramming is consistent with that seen in our in vitro data and in other reports (Guo et al., 2013), suggesting the possible death of cells that fail to achieve full re-specification. Furthermore, we estimate that only some 25% of available OPCs were targeted using retroviral delivery of the reprogramming factor. Increasing the number of cells targeted and the reprogramming efficiency to generate more induced neurons will be a goal for future studies.

There are two additional aspects that distinguish this report from most previous studies investigating cortical neuronal reprogramming. Previous studies in mouse cortex made the point that activation of astrocytes or glial progenitor cells by injury prior to delivery of reprograming factors was needed to achieve neuronal induction (Guo et al., 2013; Heinrich et al., 2014; Lentini et al., 2021). Using transcription factors alone for neuronal re-specification had not achieved this goal, suggesting that the prior injury or pathology was needed to generate a reactive or environmental state where the cells could be amenable to reprogramming (Grande et al., 2013; Guo et al., 2013; Heinrich et al., 2014; Torper and Gotz, 2017). In the present study, our retroviral targeting design avoided the establishment of an injury response state and provided insight as to the autonomous capacity of transcription factors to re-specify lineage (Guo et al., 2013; Heinrich et al., 2014). We reasoned these data could then also provide insight in regard to experimental or potential therapeutic modulation of local circuitry in disorders without a proximal injury response, such as psychiatric or addictive disorders (Southwell et al., 2014). This is not to say that an injury response after delivery of the reprogramming factor may not play a role in the process of lineage respecification. For example, some astrocytic hypertrophy was observed at the injection sites at 7 days post-injection and more severe cortical injury models note substantial proliferation of microglia and NG2-positive cells within 2–4 days (Hampton et al., 2004; Simon et al., 2011; von Streitberg et al., 2021). However, the data reported here demonstrate that neuronal reprogramming is possible without previously generating a reactive state in the target cell.

Another distinguishing feature of the present study is that the capacity for neuronal reprogramming is demonstrated in the naïve rat cortex, not the mouse cortex, which most studies to date have used. While mouse models are undeniably powerful research tools, many other important disease and behavior models have been developed using rats. Here, we demonstrate that cortical neuronal reprogramming is possible in naïve rat CNS, supporting the feasibility of extending studies of cell lineage specification into this important research tool.