Abstract

Corticospinal neurons (CSN) centrally degenerate in amyotrophic lateral sclerosis (ALS), along with spinal motor neurons, and loss of voluntary motor function in spinal cord injury (SCI) results from damage to CSN axons. For functional regeneration of specifically affected neuronal circuitry in vivo, or for optimally informative disease modeling and/or therapeutic screening in vitro, it is important to reproduce the type or subtype of neurons involved. No such appropriate in vitro models exist with which to investigate CSN selective vulnerability and degeneration in ALS, or to investigate routes to regeneration of CSN circuitry for ALS or SCI, critically limiting the relevance of much research. Here, we identify that the HMG-domain transcription factor Sox6 is expressed by a subset of NG2+ endogenous cortical progenitors in postnatal and adult cortex, and that Sox6 suppresses a latent neurogenic program by repressing proneural Neurog2 expression by progenitors. We FACS-purify these progenitors from postnatal mouse cortex and establish a culture system to investigate their potential for directed differentiation into CSN. We then employ a multi-component construct with complementary and differentiation-sharpening transcriptional controls (activating Neurog2, Fezf2, while antagonizing Olig2 with VP16:Olig2). We generate corticospinal-like neurons from SOX6+/NG2+ cortical progenitors and find that these neurons differentiate with remarkable fidelity compared with corticospinal neurons in vivo. They possess appropriate morphological, molecular, transcriptomic, and electrophysiological characteristics, without characteristics of the alternate intracortical or other neuronal subtypes. We identify that these critical specifics of differentiation are not reproduced by commonly employed Neurog2-driven differentiation. Neurons induced by Neurog2 instead exhibit aberrant multi-axon morphology and express molecular hallmarks of alternate cortical projection subtypes, often in mixed form. Together, this developmentally-based directed differentiation from cortical progenitors sets a precedent and foundation for in vitro mechanistic and therapeutic disease modeling, and toward regenerative neuronal repopulation and circuit repair.

Introduction

Whether toward functional regeneration of specifically affected neuronal circuitry in disorders of the central nervous system in vivo, or for appropriate disease modeling and/or therapeutic screening in vitro, reliable approaches to accurately differentiate specific types of affected and relevant neurons are required. Overly broad classes of generic or only regionally similar neurons do not adequately reflect the selective vulnerability of neuronal subtypes in most human neurodegenerative or acquired disorders. Molecular and therapeutic findings using broad or only regionally linked classes of neurons not affected in the disorder of interest are frequently not applicable for the neurons centrally involved.

Extraordinarily diverse neurons across the nervous system, in particular within the cerebral cortex, display many distinctive features, including cellular morphology, laminar and anatomical position, patterns of input and output connectivity, cardinal molecular identifiers, electrophysiology, neurochemical properties, and ultimately their functional roles (Fishell and Rudy, 2011; Greig et al., 2013; Harris and Shepherd, 2015; Ramón y Cajal, 1995; Sugino et al., 2006; Tasic et al., 2018; Veeraraghavan et al., 2024). Diversity exists not only between broad cell types (e.g. excitatory projection neurons vs. inhibitory interneurons; intratelencephalic vs. cortical output (‘corticofugal;’ projecting away from cortex) neurons; ipsilateral associative vs. commissural), but even within seemingly homogenous populations of neurons. For example, striking and sharp molecular, connectivity, and functional distinctions exist between both spatially separated subsets and interspersed subsets of CSN, with each molecularly distinct neuronal subpopulation programmed to project to distinct segments of the spinal cord, innervate topographically distinct gray matter areas, and synapse onto distinct subsets of interneurons (Sahni et al., 2021b; Sahni et al., 2021a; Itoh et al., 2023). Importantly, these diverse segmentally specific subsets have selective vulnerability and/or involvement in distinct human disorders (Sahni et al., 2020).

Such selective involvement reflects differences between specific neuronal subtypes in their molecular regulation during development and/or maturity. Specific subtypes of neurons are thus affected in distinct developmental, neurodegenerative, and acquired disorders of the central nervous system (CNS), typically resulting in irreversible functional deficits (Saxena and Caroni, 2011; Durak et al., 2022). Particularly relevant to the work presented here, corticospinal neurons (CSN; sometimes termed ‘upper motor neurons,’ UMN) centrally degenerate in amyotrophic lateral sclerosis (ALS) and other motor neuron diseases, along with spinal cord ‘lower motor neurons,’ of entirely different developmental origin and function. Furthermore, loss of voluntary and skilled motor function in spinal cord injury results from damage to CSN axons in the corticospinal tract (Rösler et al., 2000; Hains et al., 2003).

Notably, no appropriate in vitro models exist with which to investigate CSN/UMN selective vulnerability and degeneration in ALS, critically limiting the relevance of much research. In contrast, the availability of useful in vitro models of at least immature spinal motor neurons has enabled substantial success in the spinal muscular atrophy (SMA) field, with both modeling and therapeutics (for more detailed discussion, see Sances et al., 2016).

Importantly, and in parallel to in vitro modeling, one potential regenerative approach for neurodegenerative or acquired disorders is to restore elements of the affected circuitry with new neurons that are engineered to re-establish circuit-appropriate input and output connectivity (Qian et al., 2020; Czupryn et al., 2011; Wuttke et al., 2018). Previous studies have demonstrated that active and quiescent progenitors exist in restricted regions of the adult brain (Rietze et al., 2001; Lois and Alvarez-Buylla, 1993; Kuhn et al., 1997; Reynolds and Weiss, 1992), and that new neurons can integrate into preexisting neural circuitry, supporting the feasibility of cellular repair in the CNS (Czupryn et al., 2011; Kempermann et al., 2015; Feliciano et al., 2015; Magavi et al., 2000; Brill et al., 2009; Ohira et al., 2010; Chen et al., 2004). Although transplantation of in vitro generated neurons, either from pluripotent stem cells (PSC) or from other developmentally distant cell types, is one potential approach (Michelsen et al., 2015), either ex vivo directed differentiation or in situ generation of type- or subtype-specific neurons from optimally appropriate, regionally specified resident progenitors offers several advantages. First, either approach is potentially more likely to recapitulate appropriate neuronal identity than pluripotent stem cell approaches, since presumptive partially fate-restricted resident progenitors and the desired neurons share common developmental lineage, originate from the same neural progenitor domains, and were exposed to the same diffusible and local signaling during embryonic development, thus are likely to share significant epigenomic and transcriptomic commonality (Roessler et al., 2014; Treutlein et al., 2016; Cahoy et al., 2008). Avoiding transplantation via in situ neurogenesis would offer the additional advantage of circumventing the requirement for new neurons to migrate long distances to their sites of ultimate incorporation from an injection site with favorable local growth conditions, potentially enabling desired integration of newly recruited neurons at the single-cell level (Wuttke et al., 2018; Michelsen et al., 2015; Espuny-Camacho et al., 2013), emulating endogenous adult neurogenesis (Gage, 2019; Bond et al., 2015; Kempermann, 2016); and avoiding pathological heterotopias.

Substantial progress has been made in efforts to reprogram reactive glia in vitro and in vivo to acquire some form of neuronal identity (Gascón et al., 2016; Rivetti di Val Cervo et al., 2017; Wu et al., 2020; Heinrich et al., 2014; Torper et al., 2015; Grande et al., 2013; Niu et al., 2013; Heinrich et al., 2010; Felske et al., 2023; Herrero-Navarro et al., 2021). However, functional repair of specific circuitry requires highly directed differentiation of specific neuronal subtypes (beyond a generic neurotransmitter identity, e.g.), so new neurons can form circuit-appropriate input and output connectivity (Mattugini et al., 2019). Work from our lab and others have advanced this goal by identifying central molecular programs that first broadly, then increasingly precisely, control and regulate specification, diversity, and connectivity of specific cortical projection neuron subtypes during the period of their differentiation (Greig et al., 2013; Veeraraghavan et al., 2024; Sahni et al., 2021b; Sahni et al., 2021a; Arlotta et al., 2005; Lodato and Arlotta, 2015; Ozkan et al., 2020; Shibata et al., 2015; Nord et al., 2015; Taverna et al., 2014; Srinivasan et al., 2012; Lui et al., 2011; O’Leary et al., 2007; Greig et al., 2016; Woodworth et al., 2016; Galazo et al., 2016; Galazo et al., 2023). According to an emerging model, complementary and exclusionary sets of proneural and class-, type-, and subtype-specific transcriptional controls act in a subtype-, stage-, and dose-dependent manner to direct distinct projection neuron differentiation trajectories, while repressing alternative fates (Ozkan et al., 2020). This sharpens subtype identities and distinctions.

In the work reported here, we build on prior work from our lab (Azim et al., 2009a) identifying Sox6 as a unique stage-specific, combined temporal and spatial, control over all cortical projection neuron development that is both expressed by all cortical-pallial/excitatory projection neuron progenitors and excluded from subpallial/interneuron progenitors, and that effectively represses the transcriptional expression of the proneural gene neurogenin 2 (Neurog2). We identify that a subset of NG2+ (Nerve-Glial antigen 2 is a transmembrane chondroitin sulfate proteoglycan, with the protein component encoded by the gene Cspg4) endogenous cortical progenitors continue to express Sox6, which continues to repress Neurog2 expression and neuronal differentiation. We take advantage of genetic access to FACS-purify these endogenous cortical progenitors and establish a culture system to investigate the potential for their directed differentiation into cortical output neurons, the type of clinically relevant neurons that centrally includes CSN.

We then synthesized and applied a multi-component gene expression construct with complementary and differentiation-sharpening transcriptional controls (activating Neurog2 and Fezf2, while antagonizing Olig2 with VP16:Olig2) to these purified, partially fate-restricted progenitors from postnatal mouse cortex. We find that this approach directs highly specific acquisition of many cardinal morphological, molecular, and functional characteristics of endogenous corticospinal neurons, and not of the alternative intracortical or other CNS neuronal subtypes. We further investigate these results in several directions, finding, e.g., that Neurog2 alone is not sufficient to induce a specific neuronal identity; that neurons induced by Neurog2 instead exhibit aberrant multi-axon morphology and express molecular hallmarks of alternate cortical projection subtypes, often in mixed form.

As a proof of concept, we employ synthetically modified RNAs to control timing and dosage of the exogenous transcription factors, finding that a single pulse of Neurog2 combined with Fezf2 induces projection neuron differentiation from cultured SOX6+/NG2+ endogenous cortical progenitors, further highlighting the seemingly ‘poised’ and already partially cortical neuron fate-directed potential of these specialized progenitors. Our results demonstrate the feasibility of achieving molecularly directed, subtype-specific neuronal differentiation from a widely distributed endogenous progenitor population, with significant implications for both in vitro disease modeling and efforts toward therapeutic in situ repopulation of degenerated or injured cortical circuitry.

Results

Identification of SOX6+/NG2+ cortical progenitors in postnatal and adult neocortex

Progenitors and glia in postnatal and adult cortex share a common ancestry with cortical neurons (Elsherbiny and Dobreva, 2021). Therefore, we hypothesized that at least some of these progenitors and glia might have dormant neurogenic potential, and that a subset might have molecular characteristics that might enable their enhanced and potentially appropriate differentiation into cortical projection neurons (Elsherbiny and Dobreva, 2021; Zhang et al., 2014).

To identify this potential subset, we labeled proliferative cells in postnatal and adult cortex with an injection of BrdU (see Methods), and immunolabeled for PAX6, TBR2, SOX6, and FEZF2 –transcriptional controls that play key roles in embryonic pallial progenitors (Greig et al., 2013; Hevner, 2006; Woodworth et al., 2012). This experiment revealed that many BrdU+ proliferative cells continue to express SOX6 in postnatal and adult mouse cortex (Figure 1A, Figure 1—figure supplement 1). Sox6 controls molecular segregation of dorsal and ventral telencephalic progenitors during telencephalon parcellation in important part by blocking ectopic proneural gene expression by pallial progenitors and subpallial mantle zones (Azim et al., 2009a). To investigate whether Sox6 has parallel function in postnatal proliferative cells, we investigated proneural gene expression in Sox6 null brains. Strikingly, the proneural gene Neurog2 is ectopically expressed throughout Sox6-null cortex at postnatal day 6 (P6) (Figure 1B, Figure 1—figure supplement 2). This result indicates that a subset of postnatal cortical progenitors maintains latent neurogenic programs that are actively suppressed by Sox6, similar to its function in embryonic progenitors.

Identification and culture of SOX6+/NG2+ cortical progenitors with high purity and fidelity.

(A) Confocal micrograph of mouse neocortex at postnatal day 7 (P7) showing expression of SOX6 by a subset of BrdU+ proliferative cells. See also Figure 1—figure supplement 1. (B) In situ hybridization of Neurog2 in Sox6 wild-type (wt) (left) and knockout (KO) (right) cortex at P6. (B’) Insets showing the boxed areas in B. Loss of Sox6 results in widespread ectopic expression of Neurog2 (B’). See also Figure 1—figure supplement 2. (C) Immunofluorescence showing expression of SOX6 by DsRed+ NG2+ progenitors in cortex at P5. (D) Immunostaining of NG2 proteoglycan in NG2-DsRed cortex at P5 shows expression of DsRed by NG2+ progenitors. Inset: a cell with a strong DsRed signal in the cell body and NG2 proteoglycan around the main cell body and in cellular processes. See also Figure 1—figure supplement 3A–C. (E) Representative FACS plot of neocortical cells from NG2-DsRed transgenic cortex showing distinct DsRed-Bright, -Dim, and -negative populations. (F) qPCR analysis of Sox6, Olig2, and Cspg4 from acutely sorted DsRed-Negative, -Dim, and -Bright populations, as well as cultured DsRed-Bright cells (5DIV), demonstrates that SOX6+/NG2+ progenitors are enriched in DsRed-Bright population and maintain key gene expression in vitro. See also Figure 1—figure supplement 3D–G. Data are presented as mean ± SD, n=4, biological replicates, Actb normalized data relative to DsRed-negative population. ∗∗∗∗p<0.0001, p≥0.05, no statistically significant difference (n.s.); ANOVA Tukey’s post hoc test. (G) Volcano plot comparing fold difference in average expression of progenitor genes between acutely sorted DsRed-Bright and -Dim populations (RNA-seq, n=5, biological replicates). See also Figure 1—figure supplement 4A. (H) Representative brightfield image of cultured SOX6+/NG2+ (DsRed-Bright) progenitors at 5 DIV showing preserved progenitor multipolar morphology. See also Figure 1—figure supplement 3J–P. (I, J) Cultured progenitors continue expressing the key progenitor-specific molecules NG2, SOX10 (I), OLIG2, and SOX6 (J) at 7 DIV. (K) Quantification of TUJ1+ and GFAP+ cells at 3-, 5-, and 7 DIV shows essentially no contaminant cells in culture. Data are presented as mean ± SD, n=2, biological replicates. See also Figure 1—figure supplement 3Q. (L) Pearson correlation analysis of progenitor genes shows high similarity between acutely sorted and cultured SOX6+/NG2+ (DsRed-Bright) progenitors (R=0.84, p<2.2e-16). Data points represent log2 fold differences in gene expression relative to acutely sorted DsRed-Dim population. See also Figure 1—figure supplement 4A. (M) Heatmap of the top five marker genes for seven major cell types in brain shows that SOX6+/NG2+ progenitors are enriched in DsRed-Bright populations and that progenitor cultures are free of potential contaminants. Counts are variance-stabilizing transformed (vst) normalized data in log2 scale. (N) Volcano plot comparing fold differences in average expression of the top 500 genes for major cell types between cultured SOX6+/NG2+ (DsRed-Bright) progenitors and acutely sorted cells. n=5/6, biological replicates. See also Figure 1—figure supplement 4. Scale bars (A, C, H) 50 μm; (D, I, J) 100 μm. cc: corpus callosum, ctx: cortex.

We then focused the investigation on SOX6+ cells by immunocytochemistry (ICC) and by using genetically labeled progenitors (NG2-DsRed) (Zhu et al., 2008a). We identify that SOX6+ cells are a subset of NG2-proteoglycan-expressing proliferative cells resident across the CNS (Figure 1C). These data indicate that at least a subset of SOX6+/NG2+ progenitors resident in the neocortex possess some level of dormant neurogenic competence, which might be activated with relatively focused molecular manipulation. Therefore, we targeted SOX6+/NG2+ progenitors for directed differentiation into clinically relevant cortical output neurons, including CSN.

Purification and culture of SOX6+/NG2+ cortical progenitors

We established a culture system of purified SOX6+/NG2+ cortical progenitors to evaluate candidate transcriptional regulators for their ability to direct differentiation of SOX6+/NG2+ progenitors into cortical output neurons in vitro, thus enabling rigorous and iterative experimentation under controlled conditions. We used a transgenic NG2-DsRed mouse line (Figure 1D, Figure 1—figure supplement 3A–C); Zhu et al., 2008a to isolate DsRed-positive cells by FACS from micro-dissected dorso-lateral neocortex at P2-P6 (Figure 1, Figure 1—figure supplement 3D). Three distinct DsRed populations were identified based on fluorescence intensity: ‘DsRed-Bright’ (2–5%), ‘DsRed-Dim’ (~20%), ‘DsRed-negative’ (~75%) (Figure 1E). Quantitative PCR (qPCR) (n=4) and ICC (n=2) revealed that DsRed-Bright cells are progenitors with high expression of Sox6, Cspg4 (NG2), and Olig2 (Figure 1F), whereas DsRed-Dim cells are a heterogeneous population that includes GFAP+ astrocytes, NESTIN+ progenitors, and a subset of NG2+ progenitors (Figure 1—figure supplement 3E–G). To further investigate these DsRed populations, we performed RNA-seq on acutely sorted DsRed-Bright, DsRed-Dim, and DsRed-negative populations (n=5–6), and evaluated expression of a focused set of 500 genes most enriched in major cortical cell lineages (Supplementary file 1; Zhang et al., 2014). Cortical NG2+ progenitor-enriched genes are highly expressed by the DsRed-Bright population (Figure 1G, Figure 1—figure supplement 4A), whereas neuronal, astroglial, and microglial genes are depleted (Figure 1—figure supplement 4B–D). Together, these data indicate that DsRed-Bright cells are canonical SOX6+/NG2+ progenitors, potentially optimally suited for use in subsequent directed differentiation experiments.

We FACS-purified DsRed-Bright SOX6+/NG2+ progenitors with stringent gating and cultured them for 5 days (days-in-vitro, DIV) until they reached optimal confluency for transfection (Figure 1H). To promote the preservation of endogenous progenitor characteristics in culture, we performed a pilot experiment varying morphogen composition to broadly optimize serum-free medium formulation based on previously published protocols (Figure 1—figure supplement 3H; Lyssiotis et al., 2007). When cultured in this medium, progenitors proliferate robustly in response to the mitogens PDGF-A and FGF2 (Figure 1—figure supplement 3I–N). They maintain their cardinal molecular hallmarks, including expression of SOX6, NG2, OLIG2, and SOX10 (Figure 1F, I, J, Figure 1—figure supplement 3J-M), and conserve characteristic branched morphology with non-overlapping territorial processes (Figure 1—figure supplement 3N–P; Hughes et al., 2013).

We next investigated the extent of spontaneous oligodendrocyte differentiation from these progenitors in culture, since a substantial subset of broad NG2+ progenitors produces oligodendrocytes in vivo (Zhu et al., 2008a). Previous work demonstrated that Sox6 is expressed by at least some proliferating NG2+ progenitors, and is down-regulated upon differentiation (Baroti et al., 2016; Stolt et al., 2006). Under our culture conditions, FACS-purified cortical SOX6+/NG2+ progenitors continue to express Sox6 (Figure 1F, I, J), indicating maintenance of their progenitor state. ICC for O4 expression (a marker for pre-myelinating oligodendrocytes) revealed that only ~0.15% of these cells express O4 at 3 and 5 DIV (~51 and~49 O4+ cells/cm2, respectively). Similarly, qPCR for myelin basic protein (Mbp), a canonical oligodendrocyte marker, demonstrated that Mbp expression does not increase when cells are cultured for 3 or 5 days, compared to acutely sorted progenitors (n=4) (Figure 1—figure supplement 3E). Together, these data indicate that our culture conditions are not permissive for oligodendrocyte differentiation, and that the purified SOX6+/NG2+ progenitors maintain their progenitor state.

Next, we applied multiple analyses to identify whether there exist contaminant neurons or astrocytes in these cultures of SOX6+/NG2+ progenitors. To identify non-progenitor cells in culture, we immunolabeled for TUJ1 (antibody against TUBB3, a common immature neuronal marker) and GFAP (expressed by astrocytes and some other types of neural progenitors) at 3, 5, and 7 DIV (Figure 1K). At 3 DIV, among ~12,000 total cells/cm2, there were 7 TUJ1+ cells and 3 GFAP+ cells. At 5 DIV, among ~32,000 total cells/cm2, there were 11 TUJ1+ cells and 0 GFAP+ cells. At 7 DIV, among ~70,000 total cells/cm2, there were 6 TUJ1+ cells and 14 GFAP+ cells (Figure 1K). These data reveal the exceptional purity (>99.9% pure) of these primary cultures of SOX6+/NG2+ cortical progenitors. Reinforcing these immunocytochemical results, qPCR revealed that neither Tubb3 nor Gfap are detected in these cultures at 5 DIV, nor in acutely sorted DsRed-Bright cells (n=4) (Figure 1—figure supplement 3E). In striking contrast, and reinforcing that these culture conditions maintain progenitor competence of SOX6+/NG2+ progenitors, supplementing medium with serum resulted in downregulation of Sox6 and NG2 and increased expression of Gfap (n=4) (Figure 1—figure supplement 3Q). Together, these results identify that there is essentially no contamination under these culture conditions at any time point investigated, and that progenitors maintain their molecular and functional characteristics in vitro.

We further investigated the progenitor cultures for potential pericyte contamination, since pericytes express NG2 proteoglycan (Ozerdem et al., 2001), so they are DsRed-positive in NG2-DsRed cortex (Figure 1D, Figure 1—figure supplement 3A and B). qPCR for pericyte markers Pdgfrb and Mcam (CD146) revealed that pericytes are abundant in acutely sorted DsRed-Bright cultures, but are absent in culture at 5 DIV (n=4) (Figure 1—figure supplement 4F), indicating that pericytes do not survive in these culture conditions. Validating these results by ICC, there were no PDGFRB+ cells in culture at either 3 or 5 DIV (0 cells/cm2, n=2), unless DsRed-Bright cells were cultured in serum-supplemented media (Figure 1—figure supplement 4J–L). Together, these results reveal that these culture conditions do not support pericyte survival, and that progenitor cultures are pericyte-free.

To even further investigate by independent means whether progenitors maintain their in vivo molecular features in vitro, we performed RNA-seq on these cultures at 5 DIV (n=6), evaluating expression of 500 genes most enriched in the major alternative cell lineages (Supplementary file 1; Zhang et al., 2014). The purified SOX6+/NG2+ progenitor cultures express progenitor-enriched genes (Figure 1L), but, appropriately, do not express neuronal-, astroglial-, microglial-, pericyte-, or vascular-enriched genes (Figure 1M, N, Figure 1—figure supplement 4A–D and G–I), confirming the ICC and qPCR results. Similarly, oligodendrocyte-enriched genes are not upregulated in culture compared to acutely sorted cells (Figure 1M, N, Figure 1—figure supplement 4E). Importantly, gene expression profiles of cultured progenitors were highly consistent and reproducible across biological replicates (n=6) (Figure 1—figure supplement 4A–I). Together, these data further confirm that cortical SOX6+/NG2+ progenitors maintain their molecular characteristics in vitro, enabling establishment of a robust in vitro culture system in which to reproducibly manipulate progenitors under controlled conditions.

Multi-gene construct ‘NVOF’ induces neuronal differentiation and unipolar pyramidal morphology from SOX6+/NG2+ cortical progenitors

To direct differentiation of corticospinal neurons from cortical SOX6+/NG2+ progenitors, we designed a tandem construct containing three transcriptional controls (Neurog2, VP16:Olig2, and Fezf2 – collectively termed ‘NVOF’) based on their developmental functions (Figure 2A and B; Tang et al., 2009). The expression of the polycistronic construct is driven by the CMV-β-actin (CAG) promoter, with the open reading frames separated by 2A linker sequences (Supplementary file 3; Tang et al., 2009), also including a GFP reporter to identify transfected cells.

NVOF induces mature glutamatergic neurons from SOX6+/NG2+ cortical progenitors in vitro.

(A–C) Strategy for directed differentiation of SOX6+/NG2+ progenitors into cortical output neurons (A), the NVOF multigene construct (B), and the experimental outline (C). (D) Representative images of control-GFP and NVOF-transfected cells at 1-, 3-, 7-, and 16-days post-transfection (DPT). Unlike control-transfected cells, NVOF-transfected cells lose progenitor morphology at 1 DPT and progressively exhibit complex neuronal morphology, including a primary axon-like process and multiple dendrite–like processes. (E) Percentage of control-GFP and NVOF-transfected cells with neuronal morphology and TUJ1 expression (~42% at 3 DPT and ~74% at 7 DPT for NVOF, n=4, >200 cells/n). (F) Quantification of primary process length for NVOF-induced neurons at 3 and 7 DPT (n=3,>100 cells/n). (G) Representative morphology of NVOF-induced, TUJ1+ neurons at 7 DPT. Note the single axon, dendrite-like structures, and multiple axonal collaterals. (H) Representative images of NVOF-induced neurons at 16 DPT showing acquisition of elaborate dendritic morphology and highly intercalated axonal processes. (I) High-power representative images of individual NVOF-induced neurons at 16 DPT showing dendritic complexity and a single primary axon-like process for each neuron (red arrows). (J) Representative images of Neurog2-induced neurons with multiple atypical axon-like processes. (GFP is pseudocolored for enhanced clarity of cell morphology). (K) Representative images of Neurog2-induced neurons expressing the axonal marker ANKYRIN-G (ANK3) by multiple neurites (n=2). (L) Quantification of neurons with single versus multiple axons in Neurog2- and NVOF-induced neurons. At 7 DPT, 49 ± 16% of Neurog2-induced neurons have multiple, long axon-like processes, whereas a small number of such neurons exist after NVOF induction (9 ± 5%) (n=5, >100 cell). See methods for details. (M–N) Representative images of NVOF-induced neurons at 7 DPT showing compartmentalized expression of the somato-dendritic marker MAP2, the somato-axonal marker Neurofilament-M, and the mature neuronal marker, NeuN. (O) Quantification of NVOF-induced, TUJ1+ neurons expressing MAP2 at 3 DPT (~48%, n=3, >200 cells) and 7 DPT (~93%, n=4, >200 cells), as well as NeuN at 7 DPT (66 ± 16%, n=4, >100 cells). (P) Volcano plot showing upregulation of neuronal genes in NVOF-induced neurons compared to control-transfected cells at 7 DPT (RNA-seq, n=3, biological replicates). (Q) Bar graph of RNA-seq data displaying upregulation of neuronal genes and downregulation of progenitor genes in NVOF-induced neurons at 7 DPT. Neurons exclusively upregulate glutamatergic genes, but not genes specific to alternate neuronal identities. Scale bars (D, G, H, J, M, N) 100 μm; (I) 50 μm. Error bars show standard deviations. ∗∗∗∗p<0.0001, ***p<0.001, **p<0.01, t-test in (E, F, L). n.f. (no TUJ1+ cell found).

First, to drive glutamatergic neuronal identity, we selected the pallial proneural transcription factor neurogenin2 (Neurog2) (Schuurmans et al., 2004; Mattar et al., 2008). Previous data showed that forced expression of Neurog2 reprograms cultured postnatal glia and human ESC/iPSCs into synapse-forming glutamatergic neurons in vitro (Heinrich et al., 2010; Zhang et al., 2013; Hulme et al., 2022), and can induce neuron-like cells from postnatal glial cells (Felske et al., 2023; Herrero-Navarro et al., 2021) and injury-induced reactive glial cells in the adult mouse brain (Gascón et al., 2016). We tested Neurog2 alone in cultured progenitors and found that, in line with previous reports, Neurog2 is sufficient to induce neurons with long axons in vitro (Figure 2—figure supplement 1A).

Second, to overcome the predominant gliogenic potential in NG2+ progenitors, we complemented Neurog2 with VP16:Olig2 (VP16 transactivation domain from herpes simplex virus fused to an OLIG2 DNA binding domain) (Mizuguchi et al., 2001). This activator form of Olig2 functions as a dominant negative transcriptional regulator to counteract Olig2 gliogenic function (Mizuguchi et al., 2001; Novitch et al., 2001; Zhou et al., 2001). Olig2, a bHLH transcription factor, is necessary for the specification of a broad population of NG2+ progenitors and for their differentiation into oligodendrocytes (Li and Richardson, 2016). In addition, OLIG2 has been shown to antagonize NEUROG2 activity during neurogenesis to maintain progenitors for subsequent gliogenesis during spinal cord development (Lee et al., 2005). Misexpression of Olig2 in the cortex broadly represses proneural and neurogenic genes and increases oligodendrocyte precursor cell numbers (Liu et al., 2015). Intriguingly, antagonizing OLIG2 function in reactive glial cells after injury results in a substantial number of immature neurons in the cortical or striatal parenchyma (Buffo et al., 2005; Kronenberg et al., 2010). To confirm whether VP16:Olig2 is able to suppress glial differentiation capacity of cortical SOX6+/NG2+ progenitors in our experimental paradigm, we transfected progenitors with either VP16:Olig2 or control GFP constructs. At 1 DPT, the cultures were treated with thyroid hormone (T3) to induce differentiation of oligodendrocytes. At three days post-T3 treatment, as expected, control cells differentiated into oligodendrocyte-like cells, whereas VP16:Olig2 transfected progenitors had remarkably turned into neuroblast-like bipolar cells, indicating that VP16:OLIG2 successfully blocks endogenous OLIG2 function (Figure 2—figure supplement 1B and C).

Third, to induce cortical output neuronal fate, we selected Fezf2, an upstream transcriptional regulator that controls specification and development of cortical output neurons during cortical neurogenesis (Greig et al., 2013; Chen et al., 2004; Arlotta et al., 2005; Galazo et al., 2023; Figure 2—figure supplement 1D and E, see also Discussion). Fezf2 is capable via single gene over-expression of generating cortical output neuronal fate from alternate cortical progenitors (Molyneaux et al., 2005), from progenitors of striatal neurons in vivo (Rouaux and Arlotta, 2010), and from intracortical projection neurons post-mitotically in the early postnatal brain (De la Rossa et al., 2013).

We first verified expression of individual proteins from the polycistronic construct (Figure 2—figure supplement 1H–J), then assessed the construct’s functionality in mouse embryonic cortical progenitors in vivo (Figure 2—figure supplement 2). Previous work has shown that misexpression of Fezf2 in late-stage embryonic cortical progenitors modifies their fate to cortical output neurons, re-routing the intracortical axonal trajectories of layer II/III neurons to subcortical targets (Molyneaux et al., 2005). To investigate whether this FEZF2 function persists in the presence of NEUROG2 and VP16:OLIG2, we electroporated NVOF into embryonic ventricular zone progenitors in utero at E15.5, the peak production of upper layer intracortical neurons, and found that forced expression of NVOF induces cortical output identity in electroporated neurons (n=3) (Figure 2—figure supplement 2). Unlike control GFP-only neurons (Figure 2—figure supplement 2A and B), many NVOF+ axons descend through the internal capsule, to or past the thalamus (Figure 2—figure supplement 2C), with some extending into the cerebral peduncle (Figure 2—figure supplement 2D). These data demonstrate that the NVOF construct is expressed by electroporated neurons*,* and that Fezf2 continues to specify cortical output identity when co-expressed with Neurog2 and VP16:Olig2.

We transfected NVOF into cultured cortical SOX6+/NG2+ progenitors at 4–5 days after FACS purification and analyzed their morphology and expression of cardinal ICC markers of cell type identity over two weeks of differentiation (Figure 2C). Progenitors began to lose multipolar morphology within 24 hr (Figure 2D). By 3 days post-transfection (DPT), many extended a single axon-like neurite (Figure 2D and F) and expressed the broad neuronal marker TUJ1 (42%, n=3, >200 cells/experiment) (Figure 2E, Figure 2—figure supplement 3A and B). This morphological transformation was coupled with the loss of the progenitor markers NG2 and SOX10 (Figure 2—figure supplement 3A and B). By 7 DPT, ~73% of NVOF-transfected cells expressed TUJ1, acquired neuronal morphology with dendrite-like features, and extended a single prominent axon-like process (n=4, >200 cells/experiment) (Figure 2D-G, Figure 2—figure supplement 3C and D). Consistent with pyramidal neuron morphology, the primary axon-like processes of NVOF-directed neurons underwent significant extension between 3 DPT and 7 DPT, often extending further than 500 μm from the soma (>40%, n=3) (Figure 2F). By 16 DPT, the morphology of these putative neurons became more elaborate; the single long axon-like neurite was maintained, their dendrite-like structures became more tufted, and axon-neurite branches of neighboring cells became intercalated (Figure 2H and I).

In striking contrast, progenitors transfected with a control GFP-only construct displayed glial morphology throughout the culturing period, and no GFP+/TUJ1+ cells were present at all (n=4, 250–350 cells/experiment) (Figure 2D and E). Furthermore, even among non-transfected, GFP-negative cells, only 5 cells/cm2 out of ~30,000 progenitors/cm2 were TUJ1+, and these GFP-/TUJ1+ cells did not increase over time (n=4). These results further reinforce the absence of contaminating progenitors with spontaneous neurogenic characteristics in these cultures, and the lack of spontaneous differentiation by cultured cortical SOX6+/NG2+ progenitors.

Neurog2 is widely used to induce generic excitatory neurons from somatic and pluripotent stem cells (Heinrich et al., 2010; Zhang et al., 2013). We directly compared Neurog2-induced and NVOF-induced neurons to determine whether Neurog2 might be sufficient for induction of equivalent neuronal differentiation from cultured SOX6+/NG2+ cortical progenitors. We transfected cultured progenitors with either Neurog2-GFP or NVOF and analyzed cells at 7 DPT. Though superficially similar in some respects to NVOF-induced neurons (Figure 2J), Neurog2 induces multipolar neuronal morphology with many dendrite-like structures and multiple long axon-like processes. While almost all NVOF-induced neurons extend a single primary axon (90%), ~50% of Neurog2-induced neurons aberrantly extend multiple axon-like ANKYRIN-G+ processes originating from their cell bodies (Figure 2J–L) (n=5, >100 cells/n). This aberrant, over-exuberant neuritogenesis by Neurog2-induced neurons indicates defective polarization, potentially due to a lack of negative feedback signaling for inhibition of surplus axon formation (Funahashi et al., 2020).

NVOF-induced neurons exhibit cardinal features of mature functional neurons

We investigated further whether NVOF-induced, TUJ1+ cells acquire the cardinal molecular hallmarks of mature neurons. At 7 DPT, NVOF-induced neurons express the somato-dendritic marker MAP2 (>90%, n=4, 130–200 cells/n) and the somato-axonal marker NF-M (Figure 2M and O), indicating clear polarization and dendritic compartmentalization. Dendrite formation was confirmed by high-power imaging at 16 DPT, revealing that the NVOF-induced neurons have dendrite-like processes with filopodial protrusions and a single axon-like primary process lacking dendrite-like structures (Figure 2I, highlighted with red arrows). Further, at 7 DPT, NVOF-induced neurons express neuronal nuclear antigen (NeuN) (66 ± 16%, n=4, >100 cells/n) (Figure 2N and O), polysialylated neural cell adhesion molecule (PSA-NCAM or Ncam1) (Figure 2—figure supplement 3F and G), the presynaptic molecule synapsin (Figure 2—figure supplement 3H), with some displaying synaptophysin in axonal branches and tips of axonal protrusions (Figure 2—figure supplement 3I), and vGLUT1 (vesicular glutamate transporter 1) (Figure 2—figure supplement 3J), indicating glutamatergic identity. Together, these data indicate that NVOF robustly induces neuronal differentiation and maturation by cortical SOX6+/NG2+ progenitors in vitro.

To determine whether neuronal differentiation from these cortical SOX6+/NG2+ progenitors requires an intermediate proliferative step, we pulsed cultures with BrdU for 15 hr after transfection and labeled GFP+ cells (NVOF-transfected or GFP-only controls) for BrdU by ICC at 3 DPT (n=2). Previous work has reported that cell division is not required for neuronal differentiation from resident glia (Heinrich et al., 2010). While a majority of cells transfected with GFP-only were BrdU+, only rare NVOF-transfected cells were BrdU+. This result indicates that NVOF causes rapid cell cycle exit, and that chromatin reorganization during cell division is not required for NVOF-induced neuronal differentiation and maturation from SOX6+/NG2+ progenitors.

To more broadly investigate the molecular identity and specificity of neurons induced from cortical SOX6+/NG2+ progenitors transfected with NVOF, we performed RNA-seq on control GFP-transfected and NVOF-transfected cells at 7 DPT (n=3) (Figure 2—figure supplement 4A and B). NVOF-induced neurons have decreased expression of progenitor genes and increased expression of neuronal genes, relative to GFP-transfected cells (Figure 2M, Figure 2—figure supplement 4C). Upregulated neuronal genes include proneural transcription factors, neuron-specific cytoskeletal molecules, and molecules that function in synaptic transmission, dendritic specialization, glutamatergic signaling, axon guidance, and neuronal connectivity (Figure 2P, Q, Figure 2—figure supplement 4D). Neurog2 is required for the differentiation of multiple neuronal types across regions of the nervous system and overexpression of Neurog2 in somatic and stem cells generates neurons with mixed identities (Hulme et al., 2022; Lin et al., 2021; Kempf et al., 2021; Chouchane et al., 2017; Sheta et al., 2022; Ang et al., 2024). We, therefore, confirmed that NVOF-induced neurons express exclusively genes typical of glutamatergic neurons, but not genes specific for alternate neuronal types (e.g. GABAergic interneurons, striatal projection neurons, or serotonergic, dopaminergic, hindbrain, or spinal motor neurons) (Figure 2Q, Figure 2—figure supplement 4E and N).

We co-cultured NVOF-transfected cells at 1 DPT with primary forebrain cells from mouse cortex in astrocyte-conditioned media (Figure 2—figure supplement 3K) (see Methods) to investigate whether such a potentially permissive and/or instructive environment might even further enhance neuronal differentiation and maturation. It is known that neurons cultured below critical density, or in the absence of glial-derived trophic factors, often survive poorly and/or do not mature (Kaech and Banker, 2006; Pfrieger and Barres, 1997). Indeed, culture with primary neurons increased morphological maturation of NVOF-induced neurons, resulting in elaborate dendrites with abundant synapses (n=2) (Figure 3A–D, Figure 2—figure supplement 3L–N), demonstrating synaptic input from surrounding neurons and functional integration into neuronal networks. Quite notably, the morphology and density of dendritic synapse-like structures in NVOF-induced neurons were essentially indistinguishable from those of primary cortical neurons cultured under identical conditions (Figure 3).

NVOF-induced neurons are electrically active and have spontaneous synaptic currents.

(A–B) Representative high-magnification images of NVOF-induced neurons at 14 DPT (pseudo-colored GFP) with and without coculture of forebrain primary neurons and astrocyte-conditioned media. (A’-B’) Insets showing the boxed areas. Note differences in morphology of presumptive synaptic structures between the two conditions. See methods for details. (C) Representative high-magnification image of a primary cortical neuron at 14 DPT (pseudo-colored tdTomato) from in utero electroporated wild-type mice as a positive control. (C’) Note similarity in morphology of presumptive synaptic structures between primary neurons and NVOF-induced neurons in B’. See methods for details. (D) Representative high-magnification image of a SYNAPSIN-positive NVOF-induced neuron co-cultured with forebrain neurons, indicating abundant connections from surrounding neurons. Arrows show the presumptive single primary axon with no synapsin staining. (E) A representative NVOF-induced neuron at 10 DPT showing depolarizing steps evoking a train of action potentials (red highlighted trace: step 6, 50 pA). 10 min after break-in, or following a resting Vm stabilization greater than 1 min, cells were injected with 10 current steps ranging from –40 pA to 95 pA in 15 pA increments, for a duration of 500 ms each. (F) The first evoked action potential in response to po