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 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^1–7. 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^8–10. Importantly, these diverse segmentally-specific subsets have selective vulnerability and/or involvement in distinct human disorders¹¹.

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^12,13. 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. Further, loss of voluntary and skilled motor function in spinal cord injury (SCI) results from damage to CSN axons in the corticospinal tract^14,15. For goals including regeneration of affected circuitry, and optimally informative cellular model systems in vitro, it is critical to reproduce the type or subtype of neurons involved, so insights and/or candidate therapeutic manipulations are valid for the actual neuron types involved pathologically. 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, see¹⁶).

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^17–19. Previous studies have demonstrated that active and quiescent progenitors exist in restricted regions of the adult brain^20–23, and that new neurons can integrate into preexisting neural circuitry, supporting the feasibility of cellular repair in the CNS^18,24–29. Although transplantation of in vitro generated neurons, either from pluripotent stem cells (PSC) or from other developmentally distant cell types, is one potential approach³⁰, either ex vivo directed differentiation or in situ generation of type- or subtype-specific neurons from optimally appropriate, regionally specified resident progenitors offer 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^31–33. 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^19,30,34, emulating endogenous adult neurogenesis^35–37; 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^38–45. 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 connectivity⁴⁶. 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^{2,7–9,47–59}. 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 fates⁴⁹. This sharpens subtype identities and distinctions.

In the work reported here, we build on prior work from our lab⁶⁰ 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+ endogenous cortical progenitors continue to express Sox6, which continues to repress inappropriate Neurog2 expression and neuronal differentiation. We take advantage of genetic access to FACS-purify these endogenous cortical progenitors and establish a pure 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 neurons⁶¹; they are exposed to the same morphogen gradients throughout embryonic development, and, thus, compared to distant cell types, have similar epigenomic and transcriptomic landscapes^62,63. 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.

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^2,64,65. This experiment revealed that many BrdU+ proliferative cells continue to express SOX6 in postnatal and adult mouse cortex (Figures 1A and S1). 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 zones⁶⁰. 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). 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.

Figure 1.

We then focused investigation on SOX6+ cells by immunocytochemistry (ICC), and identify that they 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 (Figures 1D and S2A-S2C)⁶⁶ to isolate DsRed-positive cells by FACS from micro-dissected dorso-lateral neocortex at P2-P6 (Figure 1E). Two distinct populations of DsRed-positive cells were identified based on fluorescence intensity: “DsRed-Bright” (2-5% of total) and “DsRed-Dim” (>20% of total) (Figure 1E). Quantitative PCR (qPCR) (n=4) and ICC (n=2) revealed that DsRed-Bright cells are progenitors with high expression of Sox6, NG2, and Olig2 (Figures 1F and S2D), whereas DsRed-Dim cells are a heterogeneous population that includes GFAP+ astrocytes, NESTIN+ progenitors, and a subset of NG2+ progenitors (Figures S2E-S2G). 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 Table S1)⁶³. Cortical NG2+ progenitor-enriched genes are highly expressed by the DsRed-Bright population (Figures 1G and S3A), whereas neuronal, astroglial, and microglial genes are depleted (Figures S3B-S3D). 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 preservation of endogenous progenitor characteristics in culture, we optimized serum-free medium formulation based on previously published protocols (Figure S2H)⁶⁷. When cultured in this medium, progenitors proliferate robustly in response to the mitogens PDGF-A and FGF2 (Figures S2I-S2N). They maintain their cardinal molecular hallmarks, including expression of SOX6, NG2, OLIG2, and SOX10 (Figures 1F, 1I, 1J, and S2J-S2M), and conserve characteristic branched morphology with non-overlapping territorial processes (Figures S2N-S2P)⁶⁸.

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 vivo⁶⁶. Previous work demonstrated that Sox6 is expressed by at least some proliferating NG2+ progenitors, and is down-regulated upon differentiation^69,70. Under our culture conditions, FACS-purified cortical SOX6+/NG2+ progenitors continue to express Sox6 (Figures 1F and 1J), 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 cultured for 3 or 5 days, compared to acutely sorted progenitors (n=4) (Figure S2E). 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 FACS-purified 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 S2E). 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 S2Q). 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 proteoglycan⁷¹, so they are DsRed-positive in NG2-DsRed cortex (Figures 1D, S2A, and S2B). 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 S3F), 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 (Figures S3J-SL). 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 Table S1)⁶³. 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 (Figures 1M, 1N, S3A-S3D, and S3G-S3I), confirming the ICC and qPCR results. Similarly, oligodendrocyte-enriched genes are not upregulated in culture compared to acutely sorted cells (Figures 1M, 1N, and S3E). Importantly, gene expression profiles of cultured progenitors were highly consistent and reproducible across biological replicates (n=6) (Figures S3A-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 2B)⁷². The expression of the polycistronic construct is driven by the CMV-β-actin (CAG) promoter, with the open reading frames separated by 2A linker sequences⁷², also including a GFP reporter to identify transfected cells.

Figure 2.

First, to drive glutamatergic neuronal identity, we selected the pallial proneural transcription factor neurogenin2 (Neurog2)^73,74. Previous data showed that forced expression of Neurog2 reprograms cultured postnatal glia and human ESC/iPSCs into synapse forming glutamatergic neurons in vitro^45,75,76, and can induce immature neuron-like cells from injury induced reactive glial cells in the adult mouse brain³⁸. 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 S4A).

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)⁷⁷. This activator form of Olig2 functions as a dominant negative to counteract Olig2 gliogenic function^77–79. Olig2, a bHLH transcription factor, is necessary for the specification of broad population of NG2+ progenitors and for their differentiation into oligodendrocytes⁸⁰. In addition, OLIG2 has been shown to antagonize NEUROG2 activity during neurogenesis to maintain progenitors for subsequent gliogenesis during spinal cord development⁸¹. Misexpression of Olig2 in the cortex broadly represses proneural and neurogenic genes, and increases oligodendrocyte precursor cell numbers⁸². Intriguingly, antagonizing OLIG2 function in reactive glial cells after injury results in a substantial number of immature neurons in the cortical or striatal parenchyma^83,84. 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 S4B and S4C).

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^2,29,47,59 (Figure S4D and S4E, see Discussion). Fezf2 is capable via single gene over-expression of generating cortical output neuronal fate from alternate cortical progenitors⁸⁵, from progenitors of striatal neurons in vivo⁸⁶, and from intracortical projection neurons post-mitotically in the early postnatal brain⁸⁷.

We first verified expression of individual proteins from the polycistronic construct (Figures S4H-S4J), then assessed the construct’s functionality in embryonic cortical progenitors in vivo (n=3) (Figures S4K-S4N). Previous work has shown that mis-expression 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 targets⁸⁵. 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 (Figures S4K-S4N). Unlike control GFP-only neurons (Figures S4K and S4L), many NVOF+ axons descend through the internal capsule, to or past the thalamus (Figure S4M), with some extending into the cerebral peduncle (Figure S4N). 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 hours (Figure 2D). By 3 days post-transfection (DPT), many extended a single axon-like neurite (Figures 2D and 2F), and expressed the broad neuronal marker TUJ1 (42%, n=3, >200 cells/experiment) (Figures 2E, S5A, and S5B). This morphological transformation was coupled with loss of the progenitor markers NG2 and SOX10 (Figures S5A and S5B). 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) (Figures 2D-2G, S5C, and S5D). 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 (Figures 2H and 2I).

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) (Figures 2D and 2E). Further, 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^45,75. 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 S2J), 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 (Figures 2J-2L) (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 formation⁸⁸.

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 (Figures 2M and 2O), 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) (Figures 2N and 2O), polysialylated neural cell adhesion molecule (PSA-NCAM or Ncam1) (Figures S5F and S5G), the presynaptic molecule synapsin (Figure S5H), with some displaying synaptophysin in axonal branches and tips of axonal protrusions (Figure S5I), and vGlut1 (vesicular glutamate transporter 1) (Figure S5J), 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 hours after transfection, and labeled GFP+ cells (NVOF-transfected or GFP-only controls) for BrdU by ICC at 3 DPT (n=2). Some previous work has reported that cell division is not required for neuronal differentiation from resident glia⁴⁵. 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) (Figures S6A and S6B). NVOF-induced neurons have decreased expression of progenitor genes, and increased expression of neuronal genes, relative to GFP-transfected cells (Figures 2M and S6C). 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 (Figures 2P, 2Q, and S6D). 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^76,89–93. 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) (Figures 2Q and S6E-S6N).

We co-cultured NVOF-transfected cells at 1 DPT with primary forebrain cells from mouse cortex in astrocyte-conditioned media (Figure S5K) (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^94,95. Indeed, culture with primary neurons increased morphological maturation of NVOF-induced neurons, resulting in elaborate dendrites with abundant synapses (n=2) (Figures 3A-3D and S5L-S5N), 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 (Figures 3A-3D).

Figure 3.

To investigate functional properties of NVOF-induced neurons, we performed whole-cell patch-clamp recordings at 10 DPT (without co-culture) and at 16 DPT (with primary neuron co-culture) (Figures 3E-3P). Consistent with their immunocytochemical and morphological characteristics, NVOF-induced neurons possess electrophysiological hallmarks of neurons, including trains of action potentials upon depolarizing steps (Figures 3E-3G), HCN-channel currents (Isag) upon hyperpolarization (Figure 3H), and spontaneous synaptic currents (Figures 3O-3P). NVOF-induced neurons also mature over time in culture, with overall increases in the action potential threshold (-35.9 mV at 10 DPT versus -30.9 mV at 16 DPT) (Figure 3K), decreases in action potential width (2.1 ms at 10 DPT versus 1.3 ms at 16 DPT) (Figure 3M), and increases in Isag (3.4 mV at 10 DPT versus 12.1 mV at 16 DPT) (Figure 3N). Cortical SOX6+/NG2+ progenitors transfected with the control vector possess membrane resistances and resting voltages that are inconsistent with neuronal identity (Figures 3I and 3J).

Vector-free induction of neuronal differentiation from cortical SOX6+/NG2+ progenitors with synthetic modified mRNAs

The results presented above reveal that NVOF-induced neurons express a quite comprehensive set of molecules that indicate faithful neuronal differentiation, and that they possess electrophysiological properties indistinguishable from those of primary neurons. However, previous work reports that sustained expression of Neurog2 can be deleterious to differentiating cortical neurons⁹⁶. To more closely reproduce the dynamics of developmental expression of Neurog2, we aimed to restrict Neurog2 expression to a short, early time period using synthetic, chemically-modified RNA in which one or more nucleotides are replaced by modified nucleotides. Previous work, in multiple systems, has revealed that synthetic modified mRNA mediates highly efficient, integration-free, transient protein expression in vitro and in vivo without eliciting an innate immune response^97,98.

In contrast to the transient expression of Neurog2 during neurogenesis in vivo, cortical output neurons express Fezf2 throughout development and adulthood⁸⁵. To emulate the distinct kinetics of endogenous developmental expression of Neurog2 and Fezf2, we devised a strategy by which Neurog2 is transiently expressed via synthetic modified mRNA, and Fezf2 is expressed on an ongoing basis as a plasmid DNA construct with a constitutively active CAG promoter. We first adapted our transfection protocol to transfect cortical SOX6+/NG2+ progenitors with mRNA at high efficiency (Figures S7A and S7B). To test feasibility of DNA-RNA co-transfection, we co-transfected tdTomato as a plasmid DNA, and GFP as a synthetic modified mRNA (Figure S7C). ∼50% of fluorescent cells were co-transfected with both reporters (n=3). We investigated dynamics of protein expression, finding that the GFP synthetic modified mRNA displays peak protein levels 12-24 hours post-transfection, then declines (Figures S7D-S7G).

Next, we directly compared the efficacies of a Neurog2 DNA construct and a Neurog2 synthetic modified mRNA. Strikingly, confirming the neurogenic competency of cortical SOX6+/NG2+progenitors, one dose of Neurog2 synthetic modified mRNA induces robust neurogenesis, albeit with lower efficiency than Neurog2 DNA or NVOF (n=3) (Figures S7H and S7I). We then co-expressed Neurog2 in synthetic modified mRNA form and Fezf2 as a plasmid DNA construct. Remarkably, this combination of synthetic modified mRNA plus plasmid DNA produced abundant neurons morphologically indistinguishable from NVOF-induced neurons (Figure S7J). These results reveal that synthetic modified mRNA transfection can be used to tailor more precise kinetics of developmental genes toward directed differentiation of neuronal subtypes.

NVOF-induced neurons acquire molecular hallmarks of cortical output neuron identity in vitro

We progressively focused our investigations to evaluate whether NVOF-induced neurons in vitro express cardinal molecular hallmarks of endogenous cortical output neurons, with a particular focus on the major output neuron subgroup of subcerebral projection neurons (SCPN, comprising neuronal subtypes that project to brainstem and spinal cord). Results reveal that ∼58% of NVOF-induced neurons at 7 DPT express BCL11b/CTIP2, a transcription factor that regulates outgrowth, guidance, and fasciculation of SCPN/CSN axons⁴⁷ (n=6, ave 177 cells/experiment) (Figures 4A, 4G, and 4H), whereas no control GFP-only cells express CTIP2 (n=2, ave 207 cells/experiment). NVOF-induced neurons also express PCP4 (Purkinje cell protein 4), a calmodulin-binding protein reproducibly expressed by SCPN/CSN⁴⁷ (∼83% at 7 DPT, n=4, ave 131 cells/experiment) (Figures 4B, 4G, and 4H). Importantly, the number of CTIP2+ NVOF-induced neurons continued to increase over time, indicating continued subtype differentiation after 7 DPT (Figure 4I).

Figure 4.

Next, we investigated NVOF-induced neurons for expression of corticothalamic projection (CThPN) neuron-enriched molecular controls. Intriguingly, most NVOF-induced neurons express FOG2 (ZFPM2) (∼79% at 7 DPT, n=4, ave 132 cells/experiment) (Figures 4C, 4G, and 4H), a critical regulator of CThPN axonal targeting and diversity⁵⁸. However, FOXP2, a transcriptional control required for CThPN specification⁹⁹, is expressed heterogeneously by NVOF-induced neurons, with minimal to no expression by many neurons (Figure 4D). These data indicate that NVOF-induced neurons acquire broad cortical output neuronal identity, but refinement of subtype identity (SCPN vs. CThPN) is incomplete, suggesting that additional controls are required for complete subtype refinement.

We also investigated the possibility of subtype “confusion” during directed differentiation by examining whether NVOF-induced neurons also or alternatively express cardinal molecular markers of callosal projection neurons (CPN) or other intra-cortical projection neurons. If identified, this would indicate either immature differentiation or mixed/hybrid identity that is commonly observed with ES/iPSC-derived neurons¹⁰⁰. Quite notably and appropriately, NVOF-induced neurons do not express SATB2 (0% at 7 DPT, n=4, ∼130 cells/n) (Figures 4E and 4G) or CUX1 (n=3) (Figure 4F), molecular controls that are expressed by CPN and other intracortical projection neurons.

Reinforcing and extending these ICC results, RNA-seq reveals that NVOF-induced neurons express many SCPN/CSN-enriched genes (Figure 4K) (see methods), including key molecules with central functions in subtype specification of SCPN/CSN (Figure 4L), along with some CThPN-enriched genes (e.g., Tbr1, Fog2, and Foxp2) (Figure 4L). In accordance with the ICC results, RNA-seq reveals that NVOF-induced neurons have no or minimal expression of genes specific to CPN or other intracortical projection neurons, including Satb2, Cux1, and Cux2 (Figure 4L). Together, these results indicate that NVOF-induced neurons acquire cortical output neuron identity, primarily of SCPN/CSN, but with some CThPN features, without fully refining molecular identity between these subtypes of cortical output neurons (see discussion).

We directly compared expression of key subtype-specific molecular controls between NVOF-induced and Neurog2-induced neurons (Figure S8). While Neurog2-induced neurons approximate elements of NVOF induction, with some expression of cortical output neuron markers CTIP2, PCP4, and FOG2 (Figure S8A), and not the CPN and other intracortical neuronal molecules such as SATB2 (Figure S8A) and CUX1 (data not shown), NVOF induction generated more neurons expressing CTIP2, PCP4, and FOG2, with higher average expression (n=>3 for e ach marker) (Figure S8A and S8B), indicating substantially enhanced subtype-specific differentiation by Fezf2. Reinforcing the interpretation from aberrant multipolar morphology of Neurog2-induced neurons that Neurog2 alone induces “confused” and unresolved differentiation (Figure 2J-2L), Neurog2-induced neurons simultaneously express CTIP1 (BCL11a), a CPN molecular control and antagonist of CTIP2^56,57 (Figures S8A-S8D). During cortical development, CTIP1 is initially expressed broadly by postmitotic neurons, but later, through its cross-repressive interaction with CTIP2, its expression resolves to CPN and CThPN, but not SCPN/CSN, at E17. Continued expression of CTIP1 by Neurog2-induced neurons further indicates incomplete and unresolved subtype differentiation.

To comprehensively and directly characterize subtype identities induced by NVOF compared with FACS-purified primary SCPN/CSN or the morphologically and molecularly “hybrid” Neurog2-induced neurons, we performed RNA-seq on FACS-purified GFP+ neurons generated by NVOF or Neurog2 at 7 DPT (n=3), and on FACS-purified SCPN/CSN or CPN from P2 mice (n=3) (Figure 5). Pearson correlation analysis for genes enriched in SCPN compared to CPN reveals that NVOF-induced neurons are substantially more similar to primary SCPN/CSN (R = 0.87) than are Neurog2-induced neurons (R = 0.77) (Figure 5A). Even more strikingly, NVOF induces higher expression of many SCPN/CSN genes relative to Neurog2 alone (Figure 5B), while Neurog2 simultaneously and aberrantly activates many typically CPN-specific genes that are expressed at E15 in mouse, the peak period of CPN birth and specification (Figure 5C)¹⁰¹. In particular, in line with the prior ICC results, NVOF-induced neurons express SCPN/CSN genes with known key functions in subtype-specific development of SCPN/CSN at higher levels (e.g. Ctip2 and Ephb1, both essential for SCPN/CSN axon guidance) (Figures 5D, 5E)^47,102. NVOF-induced neurons express Lumican and Crim1, recently identified to be expressed highly selectively by bulbo-cervical and thoraco-lumbar CSN, respectively, and to regulate their segmentally specific axon targeting^8–10. In striking contrast, Neurog2-induced neurons express many cardinal CPN genes at high levels (e.g. Epha3 and Satb2, which both regulate CPN connectivity)^103,104 (Figures 5D and 5E), further reinforcing that Neurog2 alone is insufficient for appropriate and resolved differentiation of SOX6+/NG2+ progenitors to cortical output identity.

Figure 5.

Together, these results highlight that optimized directed differentiation is achieved by emulating normal developmental steps of sequential subtype specification of neocortical neurons regulated by interactions between broad proneural programs and lineage-specific transcription factors with dynamic temporal expression and cross-regulatory activities.

Discussion

In the work presented here, we first FACS-purify and characterize a relatively rare subpopulation of postnatal cortical progenitors that are molecularly related to early developmental cortical projection neuron-specific progenitors, which express Sox6. These postnatal progenitors express both Sox6 and NG2, and have broad transcriptional profiles of quiescent progenitors. We next identify that developmental transcriptional controls can direct the differentiation of SOX6+/NG2+ cortical progenitors into CSN-like neurons in vitro. Fezf2, a molecular control over SCPN/CSN development, and transcriptional regulators Neurog2 and VP16-Olig2 (together, “NVOF”) are able to activate a dormant neurogenic program and overcome the default postnatal gliogenic differentiation program of these cortical progenitors. This directed differentiation generates neurons with a glutamatergic neuronal identity and specific morphologic, molecular, and electrophysiologic features of cortical output neurons resembling corticospinal/subcerebral projection neurons. Our results reveal that NVOF-directed neurons acquire the key molecular features of mature glutamatergic neurons (e.g. expression of NeuN, vGLUT1, CAMK2A, SYN1, SHANK1, ionotropic and metabotropic glutamate receptors), a cortical projection neuron-like morphology with a single long NF-M+ primary axon and a MAP2+ apical dendrite-like process, the expression of molecular controls specific for SCPN/CSN (e.g. BCL11B/CTIP2, CRYM, EPHB1, and PCP4), and, importantly, do not express molecular markers of alternate fates (e.g. SATB2, BCL11A/CTIP1, CUX1, GABA, DARPP32, TH, 5HT, ISL1). We identify that these critical specifics of differentiation are not reproduced by commonly employed Neurog2-driven differentiation. Together, our work indicates that directed differentiation via combinatorial and complementary action of central developmental transcriptional controls enables previously inaccessible specificity in generating defined neuronal subtypes for cellular regeneration or disease modeling of degenerated or damaged neuronal circuitry.

In contrast to Neurog2-only activation, NVOF-directed neurons acquire multimodal CSN identity

The neurons differentiated by NVOF closely resemble bona fide corticospinal neurons. Direct transcriptomic comparison with primary SCPN/CSN reveals that NVOF-directed neurons express a large number of SCPN/CSN-enriched genes (Figures 4J and 4K), with close similarity to SCPN/CSN (R=0.87) (Figure 5A), and their unipolar somatodendritic-axonal morphology also closely resembles that of purified CSN¹⁰⁵. In contrast to Neurog2- induced neurons, NVOF-directed neurons express multiple genes that typically identify CSN specifically. These include the general indicator Crymu, as well as Lumican and Crim1, expressed highly selectively by bulbo-cervical and thoraco-lumbar CSN, respectively, and regulate their segmentally specific axon targeting^8–10. Quite importantly, NVOF-directed neurons do not display substantial enrichment of key CPN-specific molecular controls (Figure 5C-5E), indicating that they do not acquire “mixed”, “confused” identity. This is all in stark contrast to Neurog2-only induced neurons, which display aberrant multipolar morphology, mixed transcriptomic signatures, and substantial co-expression of what are normally developmentally exclusionary differentiation regulators and CPN + SCPN molecular signatures.

While Neurog2 is expressed dynamically in cortical progenitors during generation of major neuronal subtypes¹⁰⁶, Neurog2 knockout does not show significant perturbations to the expression of molecular hallmarks of these neurons^107,108. Neurog2 misexpression by electroporation during the production of superficial layers does not induce characteristic molecular features of deep layer neurons, although a subset of axons of the transfected neurons are re-directed to the ventral telencephalon¹⁰⁸. Conversely, genetic deletion of Neurog2 or shRNA knockdown of Neurog2 from superficial layer intracortical neurons results in variable defects of midline crossing as well as misrouting of callosal axons toward aberrant cortical and subcortical targets¹⁰⁷. Together these data suggest that Neurog2 has only a limited lineage-instructive role over specification of cortical output neurons.

Neurog2 is also expressed by progenitors of spinal motor neurons, sensory neurons, and dopaminergic neurons in the mammalian brain, and regulates their specification and differentiation⁷⁶. Therefore, it is conceivable that Neurog2 expression will induce a subset of its genomic targets depending on the starting cell population, culture conditions, or in vivo context. Further, this subset might not be fully faithful to a particular Neurog2-dependent in vivo neuronal population. In agreement with this hypothesis, recent reports have identified mixed subtype features in neurons generated from ES/iPS cells by Neurog2 alone^89,93,109. Several approaches, including pre-patterning of progenitors, combinatorial expression of a cocktail of transcription factors, temporal control of Neurog2 expression, induction of signaling pathways with small molecules, and co-culture with astrocytes, have been successfully used to sharpen cell fate specification^{89,93,109,110}. Consistent with these results, our NVOF transcriptional regulator combination robustly generates cortical output neuron-like cells compared to Neurog2 alone.

Intriguingly, even though NVOF-directed neurons acquire both type-specific identity of cortical output neurons, and highly specific indicators of CSN identity, they do not fully resolve the subtype-specific identities of purely subcerebral vs. corticothalamic (CThPN) neurons. They express Fog2 and Tbr1, markers of corticothalamic neurons that are not normally expressed by most mature SCPN/CSN. SCPN and CThPN together comprise cortical output neurons. SCPN and CThPN are located in deep cortical layers V and VI, respectively, and both subtypes send their axons away from cortex via the internal capsule. Not only do these two subtypes share predominant portions of the molecular developmental programs regulating their specification, post-mitotic differentiation, and axon guidance, but approximately 5% are dual SCPN-CThPN that express both high-level Bcl11b/Ctip2 and Fog2, and that send dual projections to both thalamus and subcerebral targets^58,59. These dual-projecting neurons are thought to “share” cortical output information with multiple targets for sensorimotor integration. It is possible that the neurons generated here by NVOF directed differentiation are dual SCPN-CThPN. Recent results identify that the non-DNA-binding transcriptional co-repressor TLE4 forms a complex with transcription factor FEZF2 to epigenetically regulate Fezf2 expression level, thus the balance between SCPN and CThPN molecular and projection identity at least through the first postnatal week in mouse². This delineation between SCPN and CThPN follows multiple earlier regulatory steps, e.g. the control by the transcription factor SOX5 over sequential generation of CThPN and SCPN by progressively de-repressing Fezf2 expression. Thus, resolution between SCPN and CThPN subtypes normally occurs progressively through late differentiation in vivo.

More broadly, differential expression of key controls in terms of both their levels and timing of expression, in addition to combinatorial co-expression with other key regulators, delineates differentiation of cortical projection neurons into progressively distinct subtypes with distinct targets and functional circuitry^{2,49,56–58,111–113}. For example, Fezf2 and Ctip2 are expressed more highly by SCPN/CSN relative to CThPN, but both subtypes are severely affected by loss of Fezf2 function^85,114. In this normal developmental context, the partially unresolved state of NVOF-directed neurons might represent a mid-developmental stage of subtype identity acquisition, since early during normal development many molecular controls are expressed broadly, and their expression progressively resolves over time to produce more highly subtype-restricted expression in postnatal cortex^115,116. Consistent with this interpretation, the observed increase of Ctip2 expression over time by NVOF-directed neurons (Figure 4I) suggests ongoing subtype identity refinement.

An additional factor in the incomplete delineation of NVOF-directed neurons into SCPN/CSN might be the constitutive expression of Neurog2. Neurog2 expression is normally dynamically regulated in neural progenitors¹¹⁷. In addition to its well established role in activation of proneural genes, Neurog2 might activate some neuronal subtype specific genes, such as Fog2 and Ctip2^74,118. In this context, co-expression of Fog2 and Ctip2 by NVOF-directed neurons might be due, at least in part, to constitutive Neurog2 expression. To begin to overcome this issue, we applied synthetic modified RNA to enable fine-tuning of both level and temporal dynamics of expression of Neurog2, and observed robust neuronal induction. The regulation of both level and temporal dynamics of expression during normal development suggests that level- and temporal-controlled expression of Neurog2 coupled with sustained expression of Fezf2 (Fezf2 is expressed constitutively by SCPN/CSN in vivo) might enable more refined differentiation of SCPN/CSN from SOX6+/NG2+ progenitors.

Yet another contributing factor to the lack of full SCPN-CThPN delineation of NVOF-directed neurons might be that the basic neuronal induction medium lacks critical extrinsic factors (e.g. diffusible morphogens and growth factors) required for full neuronal maturation and identity refinement. We and others have reported similar but more severe “stalling” of developmental maturation of ES cell-derived cortical-like neurons under standard culture conditions¹⁰⁰. Supporting this hypothesis, co-culture of NVOF-directed neurons with primary cortical cells (including glia), and in the presence of astrocyte-conditioned medium, improves their survival, and both morphological and electrophysiological maturation (Figure 3).

Taken together, independent regulation over both level and temporal dynamics of individual transcription factor expression, along with culture in optimized induction medium, might likely generate neurons with even further refined identities and distinction between closely related subtypes.

SOX6+/NG2+ progenitors are a subset of cortical “NG2 progenitors” with distinct molecular and functional features

The broad group of cells often collectively characterized by shared expression of NG2 proteoglycan constitute ∼2-3% of neural cells in adult rodent cortex, and are the primary proliferative cell group from early postnatal stages through adulthood and in the aged CNS^123,124. Recent work reveals that this broad group of “NG2 progenitors” is not a homogeneous population; rather, it consists of at least several subpopulations with distinct molecular, cellular, and functional properties^125–135. While some NG2-expressing progenitors generate oligodendrocytes throughout life, most of them do not differentiate, and remain proliferative in the cortex^68,130. A subset of these cells generate protoplasmic astrocytes in the ventral forebrain and spinal cord^66,136, and a smaller subset has been reported to generate neurons in the piriform cortex^137,138 and dorsolateral cortex¹³¹.

During development, diverse sets of NG2-expressing progenitors arise from anatomically and molecularly distinct dorsal and ventral proliferative zones in sequential waves^139,140. A substantial proportion of the NG2-expressing progenitors in the cortex (∼80% in postnatal rodents) share a common lineage with cortical projection neurons in mice¹⁴¹ and are thus exposed during early- and mid-development to the same morphogen gradients and epigenetic landscaping. This shared origin and molecular history provides a strong developmental basis for understanding mechanistically why these SOX6+/NG2+ cortical progenitors that originate from the dorsal (pallial) cortical proliferative zone are especially competent for directed differentiation into cortical projection neurons, and cortical output neurons in particular.

Of particular note with regard to potential regenerative applications, repopulation of degenerated or injured neurons in particular, SOX6+/NG2+ progenitors, like NG2-expressing progenitors more broadly, are widely distributed in cortex in a tiled manner. Further, progenitors lost due to differentiation or cell death are replenished by cell division and migration of neighboring progenitors^68,142. Thus, SOX6+/NG2+ progenitors are already positioned local to sites of existing neuron degeneration or other pathology, thus theoretically avoiding the need for long-distance migration and appropriate positioning that would be necessary for transplanted exogeneous progenitors, induced neurons, or spatially restricted adult neuronal progenitors from adult neurogenic regions such as the anterior subventricular zone or hippocampal dentate gyrus. This broad, tiled distribution adds substantially to their potential for cellular repopulation and regenerative approaches.

Cortical SOX6+/NG2+ progenitors are developmentally poised to generate projection neurons

Our finding that loss of Sox6 de-represses the proneural gene Neurog2 strongly indicates that Sox6 continues to function importantly in regulation of proneural genes in cortical progenitors postnatally, and that SOX6+/NG2+ progenitors actively suppress neurogenic potential. Our observation of Neurog2 de-repression in the absence of Sox6 function suggests that downregulation of Sox6 might be considered as an additional or an alternate molecular regulator for future directed differentiation experiments. Reinforcing this interpretation, even transient expression of Neurog2 alone via a single dose of synthetic modified mRNA is sufficient to induce TUJ1+ neurons (Figure S7H), and, upon NVOF expression, substantial numbers of progenitors lose progenitor features and acquire unipolar neuronal morphology by three days-post-transfection (Figure 2E). Further, and quite remarkably, over-expression of the SCPN/CSN-molecular control Ctip2 (which has no known proneural function) in SOX6+/NG2+ progenitors is sufficient to induce unipolar neuronal morphology, TUJ1 expression, and down-regulation of glial genes (Figure S4F-4G). Together, these results indicate that SOX6+/NG2+ progenitors have substantial competence to differentiate into neurons, cortical projection neurons in particular, and that they are at a relatively advanced stage of progenitor fate acquisition.

Directed differentiation of type- or subtype-specific neurons from a developmentally related population of local progenitors might encounter fewer epigenetic blocks than with stem cell or less closely related progenitor populations, thus resulting in improved functional differentiation of type- or subtype-specific neurons. Recent studies have documented that residual transcriptional, epigenetic and chromatin domain signatures specific to cells of origin persist during derivation of iPSCs, e.g., especially during early passages^119–121. Such bias and/or blockade is likely to be suboptimal for differentiation of functional type- or subtype-specific neurons, and thus for either functional regeneration or reliable modeling of pathology. Circumstantially supporting this view of persistent effects of cellular origin, reprogramming of fibroblasts to neuronal lineage occurs at a much lower efficiency and more slowly compared to reprogramming of cultured postnatal astrocytes¹²², or to our results reported here. Intriguingly, we find that cortical SOX6+/NG2+ progenitors transfected with the single factor Fezf2 acquire a hybrid morphology, preserving glia-like cell body morphology while developing a neuron-like, single, long primary neurite (Figure S4D and S4E). These results suggest incomplete and heterogenous neuronal induction. Since Fezf2 has no known proneuronal function, and since it functions centrally in specification and differentiation of cortical output neurons with long axons, it is possible that some of the Fezf2’s target genes and their regulatory domains remain epigenetically accessible in cortical SOX6+/NG2+ progenitors. This partial, seemingly hybrid, differentiation driven by Fezf2 alone further reinforces both the competency of SOX6+/NG2+ progenitors to differentiate relatively efficiently into cortical output projection neurons, and the need for multi-component regulation to guide cortical output projection neuron differentiation while suppressing alternative fates and enhancing cell type distinction.

Summary

The work reported here substantially and uniquely advances the goal of induction of neurogenesis and directed differentiation of subtype-specific neurons from endogenous adult progenitors. We first identify the SOX6+/NG2+ cortical progenitor population and employ genetic access to pure cultures of these progenitors. We then develop NVOF, a first-generation multi-component transcriptional regulatory construct, that induces cortical output neuron directed differentiation while suppressing the otherwise default glial differentiation pathway. We next identify that NVOF-directed neurons derived from SOX6+/NG2+ cortical progenitors differentiate with remarkable fidelity to bona fide in vivo cortical output neurons with appropriate morphological, molecular, deep transcriptomic, and electrophysiological characteristics. Further, these neurons do not display characteristics of inappropriate, alternative neuron types, most notably not even of closely related by non-output-neuron cortical projection neurons. This sharp subtype delineation is in striking contrast to previously developed approaches (e.g. fibroblast or iPSC-derived iNs, or glial-derived neuron-like cells) that generate much more “generic” neuron-like cells with mixed molecular identity, multipolarity, and often continued expression of some genes residual from the cells of origin, further confusing the output cellular identity^143–146. Instead, SOX6+/NG2+ cortical progenitor-derived neurons closely resemble corticospinal/subcerebral projection neurons with some hybrid corticothalamic molecular markers (the two dominant and developmentally closely related subtypes of the specialized cortical output neurons), reminiscent of the ∼5% population of CSN/SCPN in vivo with hybrid corticothalamic molecular and projection features. Together, this developmentally based directed differentiation from genetically accessible, developmentally appropriate adult cortical progenitors sets a precedent and foundation for future optimizations of combinatorial levels, order, temporal dynamics, and subcellular localizations of an appropriate set of molecular controls over subtype-specific neuronal differentiation for in vitro mechanistic and therapeutic disease modeling, and toward regenerative neuronal repopulation and circuit repair.

Acknowledgements

This work was supported by NINDS grants NS045523, DP1 NS106665, and NS049553, and by grants from the Jane and Lee Seidman Fund for CNS Research, and the Emily and Robert Pearlstein Fund for Nervous System Repair (to J.D.M.). A.O. was partially supported by a fellowship from the Suna and Inan Kirac Foundation. H.P. was partially supported by an International Brain Research Organization Fellowship, and a McKnight Brain Research Institute/Regeneration Project Fellowship. We thank Jessica Kim, Jessica Wooten, Ioana Florea, and Ryan Humphries for technical assistance; David Dombkowski at MGH and Girijesh Buruzula, Joyce LaVecchio, and Silvia Ionescu at HSCRB for their help with FACS purification; Andrew Thompson for help with the cloning; Vibhu Sahni and Maria Galazo for assistance with retrograde labeling of SCPN/CSN and scientific discussions; Wataru Ebina for help with synthetic modified RNA experiments; Pratibha Tripathi for advice on astrocyte culture; and other members of the Macklis Laboratory for helpful suggestions and critical reading of the manuscript.

Author Contributions

J.D.M., H.P. and E.A. conceived of the project; J.D.M., A.O. and H.P. designed experiments with input from A.N.B.; A.O. and H.P. performed all culture experiments; P.K., C.S. and H.P. performed synthetic modified mRNA cloning; S.S. performed all electrophysiology recordings and analyzed these data; A.O., H.P. and J.D.M. analyzed and interpreted the data and wrote the manuscript. All authors read and approved the manuscript.

Materials and Methods

Mice

All mouse studies were approved by the Harvard University IACUC and were performed in accordance with institutional and federal guidelines. The date of vaginal plug detection was designated embryonic day (E) 0.5, and the day of birth as postnatal day (P) 0. Wild-type CD1 mice were purchased from Charles River Laboratories (Wilmington, MA). The NG2.DsRed.BAC mouse line was generated by Nishiyama and colleagues⁶⁶, and was procured from Jackson Laboratories (stock number: 008241). Sox6 knockout mouse was the generous gift of V. Lefebvre (Cleveland Clinic)¹⁴⁷, and was maintained on c57bl/6 background and separately crossed into an outbred CD1 background. Most Sox6 knockout embryos on a c57 background die perinatally⁶⁰, while outcrossing into the CD1 background resulted in live Sox6 knockout pups. These pups survived for several days, developed poor body condition, and died by about P14. Male and female pups were included in all retrolabeling, FACS purification, and culture experiments. All mice were maintained in standard housing conditions on a 12-hour light/dark cycle with food and water ad libitum. A maximum of four adult animals were housed per cage.

BrdU Labeling

To cumulatively label dividing cells in the cortex at P7 and P28, BrdU (Sigma, B5002) was injected intraperitoneally from P3 to P7 or from P23 to P28 (50 µg/mg/injection). To cumulatively label slowly dividing and/or quiescent populations in adult brain, BrdU was added to drinking water for 4-6 weeks (1.5 mg/mL). Brains were collected at corresponding ages and processed for BrdU immunocytochemistry.

Plasmids

CMV/β-actin promoter-driven plasmid pCBIG (derived from CBIG, a gift from C. Lois, Caltech) was used to drive expression of IRES-Gfp (control), single factors (Ctip2, Neurog2, VP16:Olig2, Fezf2-HA, and tdTomato) or NVOF construct. The NVOF construct was created by cloning GFP, Neurog2, VP16-Olig2, and Fezf2-HA coding sequences separated by 2A linker sequences into a pCBIG vector. In this system, genes linked to each other via viral 2A sites are transcribed as a single mRNA, but are translated into individual polypeptides^72,148,149. For synthetic modified mRNA synthesis, GFP, RFP, Fezf2-HA, and Neurog2 open reading frames were cloned into pORFin or pORFinB vectors (from D. Rossi Lab, HSCRB and Boston Children’s Hospital). pORFin vectors had the appropriate 5’ and 3’ UTR sequences flanking the cloning sites, and an upstream T7 promoter for in vitro transcription. RNA was synthesized in accordance with a published protocol¹⁵⁰.

Purification and Culture of Cortical SOX6+/NG2+ Progenitors

Heterozygous offspring pups (P2-P5) from the NG2-DsRed male and wild-type CD1 female crosses were used for FACS experiments. Pups were screened for red fluorescence under a dissecting microscope (Nikon, SMZ-1500), and anesthetized on ice. Brains were dissected, and meninges were removed in ice-cold Hank’s buffered salt solution (HBSS) (Gibco, 14025092). Neocortices were micro-dissected in ice-cold dissociation medium (pH 7.35), composed of 20 mM glucose (Sigma, G6152), 0.8 mM kynurenic acid (Sigma, K3375), 0.05 mM DL-2-amino-5-phosphonopentanoic acid (APV) (Sigma, A5282), 100 U/ml penicillin - 100 µg/ml streptomycin (Gibco, 15140122), 0.09 M Na₂SO₄, 0.03 M K₂SO₄, and 0.014 M MgCl₂ (pH 7.35±0.02).

Dissected cortices were enzymatically digested in dissociation medium containing 0.16 mg/ml DL-Cysteine hydrochloride (Sigma, C9768), 10 U/ml papain (Worthington, LS003126), and 30 U/ml DNAse I (Sigma, D5025) at 37 °C for 30 minutes, rinsed two times with ice-cold OptiMEM (Gibco, 51985034), and supplemented with 20 mM glucose, 0.4 mM kynurenic acid, and 0.025 mM APV to protect against glutamate-induced neurotoxicity¹⁵¹. Digested cortices were mechanically dissociated by gentle trituration using fire-polished glass Pasteur pipets to create a single-cell suspension. Dissociated cells were centrifuged at 100 g for 5 minutes at 4 °C, resuspended (5-10x10⁶ cell/ml) in OptiMEM with supplements, and filtered through a 35 μm cell strainer (Corning, 352235). All chemicals were purchased from Sigma-Aldrich unless stated otherwise.

Cells were purified based on DsRed fluorescence intensity using a BD FACSAria II cell sorter in four-way purity mode (85 μm nozzle). DsRed-positive cells from the NG2.DsRed BAC-transgenic mouse cortex consisted of two distinct populations: bright and dim. After qPCR and immunocytochemical characterization of both populations, only the bright population, which yielded 200-300K cells/brain, was purified for induced neurogenesis experiments. A previously published protocol was adapted to maintain cells in a proliferative progenitor state¹⁵². Purified cells were sorted into and cultured in growth medium, composed of DMEM/F12 with GlutaMAX (Gibco, 10565018), 15 mM HEPES (Gibco, 15630106), B27 without vitamin A (Gibco, 12587-010), N2-max (R&D Systems, AR009), 100 U/ml penicillin - 100 μg/ml streptomycin (Gibco, 15140122), 10 ng/ml PDGF-A (Peprotech, 315-17), and 20 ng/ml FGF2 (Peprotech, 450-33). Half of the medium in each well was replaced every other day. Cells were seeded (∼10K cells/cm²) on either 50-100 μg/ml poly-D-lysine (Sigma, P0899) plus laminin (Thermo, 23017015), or 0.01% poly-L-ornithine (Millipore, A-004-C) plus laminin-coated cover glasses (Fisher, 12-545-81) in 24-well plates for microscopy experiments (Corning, 353047), or without cover glass in 6-well plates for RNA experiments (Corning, 353047). Transfection was performed at ∼5 DIV after half-replacing the medium with fresh proliferation medium using Fugene 6 (Promega) with the following ratio: per 6-well plate, 600 μl DMEM/F12 medium (w/o supplements), 30 μl transfection reagent, and 8 μg of DNA was mixed, incubated for 15-30 minutes, and directly added into each well (∼100 μl/well), yielding ∼10% transfection rate at 24 hours. The same Fugene 6 transfection reagent was used for synthetic RNA transfections (20 μl media, 1.2 μl transfection reagent, and 0.2 μg RNA for each well of the 24-well plate). On the day following transfection, growth medium was replaced with neuronal induction medium, composed of a 1:1 mixture of DMEM/F12 and Neurobasal-A (Gibco, 10888022), GlutaMAX (Gibco, 35050061), 15 mM HEPES, B27 with vitamin A (Gibco, 17504044), N2 (Gibco, 17502048), and 100 U/ml penicillin - 100 μg/ml streptomycin (Gibco, 15140122). Medium was half-replaced every third day after transfection until fixation.

Retrograde Labeling and FACS Purification of SCPN and CPN

Retrograde labeling experiments were adapted from previously published procedures⁴⁷. Briefly, pups were anesthetized by hypothermia at P0/P1, and SCPN and CPN were retrolabeled from their corresponding axonal projections by pressure injection (Nanoject II, Drummond) of Alexa Fluor 555-conjugated cholera toxin, subunit B (CTB) (Invitrogen, C22843) (6-7 injections, 23 nl/injection, 2 μg/ul) using pulled and beveled glass micropipettes with a tip diameter of 30–50 μm. SCPN were labeled from the cerebral peduncle, and CPN were labeled from contralateral corpus callosum close to the midline (3-4 rostrocaudal levels). Injections were performed in deeply anesthetized pups using a Vevo 770 ultrasound backscatter microscopy system (VisualSonics). Brains were collected at P2 for FACS purification, and retrograde labeling success was verified under a fluorescence-equipped dissecting microscope (SMZ-1500; Nikon). Cells were purified with stringent fluorescence gating using a BD FACSAria II cell sorter (85 μm nozzle) in four-way purity mode.

In utero Electroporation

Timed pregnant CD1 dams were anesthetized with isoflurane, and an incision was made in the abdomen. The uterine horns were exposed and gently positioned on a sterile piece of gauze.

1.0 μg/μl of plasmid DNA was mixed with 0.005% Fast Green in sterile PBS and injected in utero into one lateral ventricle of each embryonic brain. The injections were performed with beveled glass micropipettes (tip diameter of 30–60 um) via mouth pipetting with an aspirator tube assembly (Sigma, A5177). Plasmid electroporations were performed by placing a positive electrode (tweezer electrodes, 5 mm diameter) above the cortex and a negative electrode behind the head, and applying five pulses of current at 40V for 50 milliseconds per pulse with 1 second intervals between pulses (CUY21Edit Electroporator, Bex Co. Ltd.). Brains were collected at P7 for NVOF misexpression analysis and at P0-P1 for primary neuron culture.

Astrocyte-conditioned Media

Production of astrocyte-conditioned media was based on the published protocol for primary culture of postnatal cortical astrocytes¹⁵³. Briefly, cerebral cortices were micro-dissected from wild-type P5-P7 CD1 pups, gently dissociated without enzymatic digestion using fire-polished glass Pasteur pipets, and centrifuged at 100 g for 5 minutes at 4 °C. Dissociated cells were seeded in T25 flasks, and cultured in astrocyte growth medium (DMEM/F12 with GlutaMAX (Gibco, 10565018), 10% fetal calf serum (Seradigm, 97068-091), 5% horse serum (Invitrogen, 26050070), B27 (with vitamin A), 100 U/ml penicillin – 100 μg/ml streptomycin (Gibco, 15140122), 10 ng/ml EGF (Peprotech, 315-09), and 10 ng/ml FGF2 (Peprotech, 450-33). Medium was fully changed 24 hours post-culturing, and half of the medium was replaced three days post-culturing. Culture fidelity was verified by morphology and GFAP expression of the differentiated cells. To obtain astrocyte-conditioned media, astrocytes were passaged at ∼5 DIV using trypsin (Gibco, 25200056), centrifuged at 100 g for 5 minutes at room temperature, diluted 1:4, re-seeded in T75 flasks containing astrocyte growth medium, and cultured for 24 hours. Growth medium was subsequently replaced with neuronal induction medium (described above). The conditioned medium was collected at days 10, and 20, and aliquots were stored at -80°C.

NVOF-induced and primary neuron co-culture

To co-culture of induced neurons with primary neurons, primary forebrain neurons were obtained from P0-P1 CD1 wild-type pups using the dissociation protocol described above, and directly added onto progenitor cell cultures at 24 hours after transfection (25K/cm²). One half of the medium was replaced with fresh astrocyte-conditioned media every third day. For dendritic morphology comparison, cortical projection neurons were labeled via in utero electroporation (at E14.5) of a tdTomato reporter plasmid driven by CMV-beta-actin promoter (see above). Neurons were dissociated at P0-P1, cultured in 24-well plates with cover glass (50K cell/cm²), and cultured in parallel with induced neurons using the same neuronal media that is described above.

Histology and Immunocytochemistry

Immunocytochemistry (ICC) for tissue sections was performed following standard protocols. Briefly, mice were transcardially perfused with PBS then 4% PFA, dissected, and post-fixed overnight at 4 °C in 4% paraformaldehyde. Brains were embedded in 4% low melting temperature agar (Sigma-Aldrich) and sectioned at 50 μm on a vibrating microtome (Leica). Fixed tissues were stored in PBS with 0.025% sodium azide. Floating sections were blocked with 0.3% BSA (wt/vol) (Sigma, A3059), 0.3% Triton X-100 (Sigma, T8787), and 0.025% sodium azide (Sigma, S2002) in PBS for 30 minutes. Primary antibodies were diluted in the same blocking solution and incubated with sections for 4 hours at room temperature, or overnight at 4 °C. Sections were rinsed three times with PBS for 10 minutes and incubated with appropriate secondary antibodies diluted in blocking solution for 2-3 hours at room temperature. Sections were rinsed three times with PBS, and mounted using Fluoromount with DAPI (SouthernBiotech, 0100-20) for image acquisition. ICC for BrdU was preceded by 2 hours of treatment with 2 N HCl at room temperature for antigen retrieval.

ICC for cultured cells was performed by first fixing cells in 4% paraformaldehyde at room temperature for 10 minutes, rinsing three times with PBS, and storing in PBS with 0.025% sodium azide at 4 °C. Cells were blocked in the blocking solution for 15 minutes, incubated with primary antibodies for 2 hours, rinsed with PBS three times for 5 minutes, incubated with secondary antibodies for 45 minutes, rinsed with PBS three times for 5 minutes (all reactions at room temperature), and mounted using Fluoromount with DAPI.

The following primary antibodies and dilutions were used: mouse anti-ANK3 (ANKYRIN-G), 1:250 (Santa Cruz, sc-12719); rat anti-BrdU, 1:500 (ACSC, OBT0030); rabbit anti-CSPG4 (NG2), 1:500 (Millipore, AB5320); rabbit anti-CTIP2, 1:500 (Abcam, ab28448); rat anti-CTIP2, 1:250 (Abcam, ab18465); rabbit anti-CUX1, 1:200 (Santa Cruz Biotechnology, sc-13024); rabbit anti-DARPP32, 1:250 (Cell Signaling Technology, 2306S); rabbit anti-FOG2, 1:250 (Santa Cruz Biotechnology, sc-10755); rabbit anti-FOXP2, 1:2000 (Abcam, AB16064); mouse anti-GABA, 1:200 (Sigma, A0310); mouse anti-GFAP, 1:1000 (Sigma, G3893); rabbit anti-GFAP, 1:1000 (Sigma, G9269); chicken anti-GFP, 1:1000 (Invitrogen, A10262); rabbit anti-GFP, 1:1000 (Invitrogen, A11122); mouse anti-HA, 1:1000 (Covance, MMS-101R); mouse anti-ISL1, 1:250 (Novus, H00003670); mouse anti-MAP2, 1:500 (Sigma, M1406); chicken anti-NESTIN, 1:2000 (Novus, NB100-1604); mouse anti-NeuN, 1:500 (Chemicon, MAB377); rabbit anti-NF-M, 1:200 (Millipore, AB1987); mouse anti-NEUROG2, 1:100 (R&D Systems; MAB3314); goat anti-OLIG2, 1:200 (R&D Systems, AF2418); rat anti-RFP, 1:500 (antibodies-online, ABIN334653); rabbit anti-PCP4, 1:500 (Proteintech, 14705-1-AP); rabbit anti-PDGFRB, 1:100 (Cell Signaling, 3169); mouse anti-PSA-NCAM, 1:200 (Chemicon, MAB5324); mouse anti-SATB2, 1:200 (Abcam, ab51502); rabbit anti-SATB2, 1:500 (Abcam, ab34735); rabbit anti-SOX6, 1:500 (Abcam, AB30455); goat anti-SOX10; 1:200 (Santa Cruz, sc-17342); rabbit anti-SYNAPSIN, 1:500 (Synaptic Systems, 106002); mouse anti-SYNAPTOPHYSIN, 1:500 (Millipore, MAB5258); rabbit anti-TH, 1:250 (Millipore, AB152); rabbit anti-TUBB3 (Tuj1), 1:1000 (Sigma, T2200); mouse anti-TUBB3 (Tuj1), 1:1000 (Biolegend, MMS-435P), rabbit anti-vGLUT1, 1:500 (Synaptic Systems, 135302); rabbit anti-2A-peptide, 1:1000 (Millipore, ABS31), rabbit anti-5HT, 1:3000 (Immunostar, 20080). Alexa Fluor-conjugated secondary antibodies (Invitrogen) were used at a dilution of 1:1000. Positive controls were included in all ICC experiments with negative results. All ICC experiments utilized different batches of FACS-purified cells from independent litters to yield a minimum of three true biological replicates. Primary data were analyzed by one investigator (AO), then confirmed by a second independent investigator (HP).

Image Acquisition, Quantification, and Statistical Analysis

Wide-field image acquisition was performed with a Nikon 90i epifluorescence microscope equipped with a Clara DR-328G cooled CCD digital camera (Andor Technology) running NIS Elements software (Nikon). Brightfield images were acquired using a Nikon ECLIPSE Ts2R-FL inverted microscope. For optimal data visualization, images were adjusted for contrast, brightness, and size in Adobe Photoshop and Illustrator (2019). Identical procedures were applied across different experimental conditions. For cell quantifications, a cover glass area of ∼50 mm² (7x7 tile) was imaged using a 10x objective. The acquired image was binned as 1 mm² boxes, individual boxes were randomly selected, and all GFP+ cells in each selected box were quantified using NIS-elements software (Nikon). To quantify the immunofluorescence intensity of target molecules, nuclei were identified via DAPI, and the average intensity of the outlined nuclear area was measured on Nikon-NIS. The following criteria were used to mark neurons with multiple axons: If the second longest neurite originating from the cell soma was at least half the length of the longest neurite, that cell was marked as multipolar. A minimum of four independent biological replicates were used for each experimental condition across the study unless otherwise mentioned in the text. Microsoft Excel, RStudio (version 1.3.959), and GraphPad Prism 8 were used for data analysis, plotting graphs, and statistics. Statistical details of the experiments can be found in the figure legends. Significance is based on the p value indicated on the graphs as * p % 0.05, ** p % 0.01, ***p % 0.001, ****p % 0.0001.

Electrophysiology

Electrophysiological recordings were performed at 20-25 °C on an Olympus BX51WI microscope. Cells were bathed in artificial cerebral spinal fluid (ACSF) containing 119 mM NaCl, 2.5 mM KCl, 4 mM CaCl₂, 4 mM MgSO₄, 1 mM NaH₂PO₄, 26.2 mM NaHCO₃, and 11 mM glucose. ACSF was continuously saturated with 95% O₂/5% CO₂. Intracellular recordings were obtained using glass micropipettes filled with an internal solution containing 136 mM KMeSO₃, 17.8 mM HEPES, 0.6 mM MgCl₂, 1 mM EGTA, 4 mM Mg-ATP, and 0.3 mM Na-GTP. Traces were collected using a Multiclamp 700B amplifier (Molecular Devices), filtered with a 2 kHz Bessel filter, digitized at 50 kHz using a Digidata 1440A digitizer (Molecular Devices), stored using Clampex 10 (Molecular Devices), and analyzed off-line via customized procedures written in Igor Pro (WaveMetrics). Series resistance was monitored throughout the experiment. Cells at DPI/DIV 15-16 were identified visually by fluorescence. Action potentials were evoked by injection of current steps, ranging from –140 pA to 400 pA in 60-pA increments, with a duration of 600 ms. Action potential parameters were quantified for the first action potential evoked at the lowest current injection that resulted in an action potential. The threshold potential was defined as the voltage at which dV/dt of the action potential waveform reached 10% of its maximum value, relative to a dV/dt baseline taken 10 ms before the peak. Action potential amplitude was defined as the difference between the threshold value (in mV) and the maximum voltage of the action potential. Width was measured at half-maximum amplitude. Sag current was measured during a -140 pA step current for a duration of 600 ms.

RNA sequencing

A minimum of three independent biological replicates was used for each experimental group (i.e., mouse litters, cell culture batches, FACS purifications, etc. were different for each biological replicate). RNA isolation was performed using a Qiagen RNeasy Plus Mini Kit with the gDNA elimination step. FACS-purified cells were collected directly into RLT Plus buffer with β-mercaptoethanol. RNA concentration, purity, and integrity were measured by a Nanodrop (ThermoFisher), an Agilent TapeStation 2200, and an Agilent Bioanalyzer 2100. Only high-quality RNA samples were used for library preparation. For the 32 samples used in this study, the minimum RNA integrity number (RIN) was 8, the average was 9.7, and the median was 10.

Library preparation and sequencing were performed by the Bauer Core Facility at Harvard University. RNA was fragmented at 94 °C for 6 minutes with a final size range of 200-300 bp. The library was prepared from 50 ng total input RNA per sample using a Kapa mRNA HyperPrep kit (14 cycles) with PolyA enrichment (stranded via dUTP addition, and first-strand preserved). Unique dual 8 bp adapters (1.5 μM) (IDT for Illumina) were used for indexing. The library quality and concentration were confirmed by an Agilent TapeStation 2200 and a Kapa qPCR library quantification kit. The pooled samples were run on Illumina NextSeq High flow cells (75bp, paired-end reading). Sequencing quality was assessed by FASTQC (version 0.11.9). STAR-aligned counts were used for quality control metrics¹⁵⁴.

The quasi-aligned counts from Salmon with default options were used to perform downstream gene expression analyses¹⁵⁵. Transcript-level count matrices were produced via the Bioconductor package “tximport”¹⁵⁶. Ensembl gene IDs were generated using the GRCm38 reference genome (Ensembl v98). DESeq2 was used to perform differential expression analyses¹⁵⁷. Low count genes (total reads <10) were pre-filtered before DESeq2 functions. Gene names and other information were annotated using the Bioconductor package “AnnotationDbi”. Variance-stabilizing transformed (vst) normalized counts (log2 scale) were used for data visualization¹⁵⁸. The code used to perform subsequent analyses of the sequencing data was an adaptation of standard R packages. More detailed information is available upon request. Gene ontology (GO) enrichment analysis was performed using the PANTHER online database^159,160. Raw FASTQ files and processed counts are available upon request.

Quantitative PCR

cDNA was prepared using the Superscript IV first-strand synthesis system (Thermo, 18090050) and random hexamers (Thermo, SO142) following the manufacturer’s standard protocol. Random hexamers were used for amplification. qPCR was performed using the iTaq Universal Sybr Green Supermix (Bio-Rad) on a BioRad CFX96 thermal cycler following standard procedures. For all qPCR primers used in this study, reaction efficiency was calculated by standard curve analysis, and only primers with high efficiency (90-105%) were used. See Supplementary Table S2 for the primer list. Four independent biological replicates were used for each experimental group in all qPCR experiments.