Generation of neural organoids for spinal-cord regeneration via the direct reprogramming of human astrocytes

Abstract

Organoids with region-specific architecture could facilitate the repair of injuries of the central nervous system. Here, we show that human astrocytes can be directly reprogrammed into early neuroectodermal cells via the overexpression of OCT4, the suppression of p53 and the small molecules CHIR99021, SB431542, RepSox and Y27632, and that the activation of signalling mediated by fibroblast growth factor, sonic hedgehog and bone morphogenetic protein 4 in the reprogrammed cells induces them to form spinal-cord organoids with functional neurons specific of the dorsal and ventral domains. In mice with complete spinal-cord injury, organoids transplanted into the lesion differentiated into spinal-cord neurons, which migrated and formed synapses with host neurons. The direct reprogramming of human astrocytes into neurons may pave the way for in vivo neural organogenesis from endogenous astrocytes, for the repair of injuries of the central nervous system.

One-sentence editorial summary:

Human astrocytes directly reprogrammed into early neuroectodermal cells, can be induced into spinal-cord organoids that survive, and integrate with host tissue after transplantation into mouse spinal-cord lesions.

Three-dimensional (3D) human neural tissue loss after complete spinal cord injury (SCI) is irreversible and often leads to functional paralysis of lower limbs. Currently, the way to regrow 3D spinal cord tissue remains difficult to achieve ^1–3. Previous studies have reported motor functional recovery of the spinal cord injured mouse after transplantation of multiple type neural cells, including astrocytes⁴, oligodendrocyte precursor cells⁵, neural stem/progenitor cells (NSCs/NPCs)^6–10 and neurospheres derived from human pluripotent stem cells (hPSCs)¹¹, indicating that cell transplantation is a promising therapeutic strategy for SCI. But hPSC derivatives remain to further improve for potential clinical applications¹². Recently, engraftment of 3D construct combined functionalized hydrogels with human NSC derived from human fetal tissues, have also improved behavioral recovery after spinal cord injury ¹³. Nevertheless, the use of fetal tissue faces ethical issues and difficulties to obtain. It is well known that the spinal cord is finely assembled and spatial specialized neural tissue with different types of neuronal and glial cells. It is a promising strategy to develop the in vivo spinal cord-like tissue for regeneration of SCI ^14–16. However, both hPSC-neurospheres and 3D engineered hNSCs have not been induced into specialized spinal cord architecture.

More recently, 3D brain organoids from human pluripotent stem cells (hPSCs) are attracting considerable attention to investigate neural development and diseases ^17–24. Previous studies have reported that 3D dorsal, intermediate and ventral developmental spinal cord-like organoids were induced from hPSCs using, sonic hedgehog (SHH), bone morphogenetic protein 4 (BMP4) and fibroblast growth factor (FGF) morphogens ^5,15. Although patient derived hiPSCs could be used to develop 3D organoids to avoid immune rejection, they need a well-defined neural differentiation protocol to eliminate non-neural cells, and also require cell transplantation. Now, in vivo direct reprogramming of endogenous glial cells is a promising approach for central nervous system (CNS) regeneration that can bypass cell transplantation and avoid immunosuppression ^25–28. Previous studies have reported that genetic manipulation or small molecules efficiently converted mouse and human astroglial cells into different kinds of functional postmitotic neurons in vitro and in vivo ^26,28–37. However, these induced neurons are not capable of proliferation and neural organogenesis. Thus, the preeminent strategy for CNS repair is to induce brain organogenesis from endogenous astrocytes. So far, efficient reprogramming of human astrocytes into 3D brain region-specific organoids for CNS repair has not been achieved.

Here, we present a unconventional method to generate 3D brain organoids by direct reprogramming of human astrocytes using OCT4 (O), p53 genes and small molecules cocktail CHIR99021 (C), SB431542 (SB), RepSox (R) and Y27632 (Y), termed as Op53-CSBRY. Human astrocytes treated with Op53-CSBRY were directly reprogramed into neural ectodermal cells that further generated human brain organoids (hAD-Organs). Importantly, these hAD-Organs can be patterned into spinal-cord organoids by activating FGF, SHH and BMP signaling. The grafts of human astrocyte-derived spinal-cord organoids (hADSC-Organs) survived, differentiated into spinal cord neurons, sprouted long-distance axons, formed synaptic connectivity with host neurons, and protected the mouse complete injured spinal cord in vivo, which may open up a promising avenue for organ therapy of SCI.

Results

Op53-CSBRY efficiently converted human astrocytes into neurons.

Currently, direct reprogramming of primary glial or fibroblast cells yields postmitotic neurons that cannot proliferate and form adequate cytoarchitectural CNS tissue in vitro and in vivo, which is a major disadvantage for CNS neural organ regeneration after brain damage ^30,34,36,38. The number of converted neurons largely depends on the expandability and non-senescent quality of the donor somatic cells as well as reprogramming efficiency. To obtain enormous proliferative neural cells and trigger organogenesis, we hypothesized that master reprogramming factor OCT4 may enhance human astrocyte reprogramming process ^39,40, and reduction of cell-cycle regulator p53 may promote the proliferation of intermediate reprogrammed cells ^41–43. We first over-expressed OCT4 and knocked down p53 synchronously in human primary astrocytes, hereinafter referred to as “Op53”, whereas only a few astrocytes can be reprogrammed into MAP2⁺ cells (Supplementary Fig. 1a), although astrocytes had demonstrated OCT4 overexpression and p53 knock-down efficiently (Supplementary Fig. 1b, c). To improve the astrocyte conversion efficiency, we selected 15 small molecules involved in neural differentiation or somatic cell reprogramming. The 15 small molecules for primary screening were as follows: CHIR99021, Valproic Acid (VPA), Y27632, SB431542, XAV939, Forskolin, JQ1, PD0325901, RepSox, ISX-9, LDN193189, TTNPB, Smoothened Agonist (SAG), SU5402 and DAPT. Through single small molecule screening (initial screening protocol, Fig. 1a, b), we found the efficiencies of human astrocyte reprogramming into MAP2⁺ neurons increased (>8%) significantly when treated with CHIR99021, SB431542, RepSox and Y27632 respectively (Fig. 1c, d). Additionally, MAP2⁺ neuronal clusters were also observed (Supplementary Fig. 1d) in these four groups, which indicated that astrocytes may transiently be re-entered into the cell cycle and proliferated during the reprogramming process. To further enhance the reprogramming efficiency, we then tested the different combinations of these four small molecules. We found that CHIR99021 (C), SB431542 (SB), RepSox (R) and Y27632 (Y), taken together, (henceforth known as CSBRY), robustly increased the astrocyte conversion efficiency (Fig. 2a, b). To further confirm this human astrocyte reprogramming method, we generated the developmental astrocytes from human embryonic stem cells (hESCs) (Supplementary Fig. 2a). About 96% of hESC-derived astrocytes expressed GFAP in this differentiation method (Supplementary Fig. 2b). hESC-derived astrocytes were significantly reprogrammed into MAP2⁺ neurons with similar efficiency to primary astrocytes treated with Op53-CSBRY (Fig. 2c, d). To further confirm the human fetal astrocyte results, we collected normal mature astrocytes containing >80% GFAP positive astrocytes from three 13–65 years-old patients with gliomas to further confirm the human fetal astrocyte results (Supplementary Fig. 2c). Astrocytes from patient astrocyte samples (2/3) were effectively reprogrammed into neurons, but with much lower efficiency than human fetal astrocytes (Fig. 2e, f), indicating different developmental stages of astrocytes may affect the reprogramming efficiency. Taken together, these results consistently demonstrated that human astrocytes can be robustly reprogrammed into neurons by Op53-CSBRY in vitro.

Fig. 1 |. Small molecule screening to improve astrocyte to neuron conversion efficiency.

a, Schematic for the direct reprogramming of human astrocytes into neurons. b, Schematic diagram of primary screening protocol to convert human astrocytes into neurons using each of 15 small molecules. AM, Astrocyte Medium; NB, Neurobasal medium. c, Representative images of MAP2⁺ cells treated with each of 15 small molecules and DMSO as control at day 21. Scale bar, 100 μm. d, Quantification of the percentage of induced MAP2⁺ neurons treated with 15 small molecules (CHIR99021, Y27632, SB431542, XAV939 and SAG, n = 4 independent experiments; the rest, n = 3 independent experiments) and DMSO (n = 4 independent experiments). Data are presented as mean ± SEM.

Fig. 2 |. Conversion of human astrocytes into neurons using Op53-CSBRY.

a, Representative images of MAP2⁺ cells treated with different combinations of small molecules and DMSO. C, CHIR99021; SB, SB431542; R, RepSox; Y, Y27632. Scale bar, 100 μm. b, Quantitative analysis of MAP2⁺ cells treated with small molecular combinations (CSBY, CSBRY group, n = 4 independent experiments; the rest, n = 3 independent experiments) and DMSO (n = 4 independent experiments). c, Representative images of MAP2⁺ neurons converted from hESCs-derived astrocytes treated by CSBRY cocktail. Scale bar, 100 μm. d, Quantitative analysis of MAP2⁺ neurons converted from hESCs-derived astrocytes treated by CSBRY (n = 3 independent experiments). e, Representative images of MAP2⁺ neurons converted from human mature astrocytes isolated from the patients with glioma treated by CSBRY combination. Scale bar, 100 μm. f, Quantitative analysis of MAP2⁺ neurons converted from human mature astrocytes isolated from the patients with glioma treated by CSBRY (n = 3 independent experiments). NP: gliomas patient. Data are presented as mean ± SEM. Unpaired Student’s t-test was used for comparing two groups.

Human astrocytes effectively reprogrammed into cerebral organoids by Op53-CSBRY.

We treated commercially acquired primary astrocytes isolated from the healthy human brain cerebral cortex with Op53-CSBRY. We observed that Op53-CSBRY induced lots of rosette-like clusters at day 14 (Extended Data Fig. 1a). This may suggest primary astrocytes re-enter into neurogenesis during reprogramming. To confirm this hypothesis, we transferred these reprogrammed cells into a suspension culture system (Fig. 3a, b). After suspension at day 14, induced cells were aggregated as spherical structures (Fig. 3c). After being embedded in Matrigel the spheroids showed neuroepithelial buds.The size of spheroids was about 3 mm at day 90 (Fig. 3c), and was up to 4–5 mm when the spheroids were cultured over 15 weeks (Extended Data Fig. 1b), which is similar to the size of cerebral organoids derived from hPSCs ^21,44. After 7 weeks culture, we observed neural progenitor cell marker SOX2⁺ and PAX6⁺ ventricular zone (VZ) -like structures, and that included a TUJ1⁺/MAP2⁺ neuronal layer ²¹ (Fig. 3d and Extended Data Fig. 1c, d). SOX2⁺ and PAX6⁺ VZ regions were continually enlarged at 10 weeks (Fig. 3e). To investigate whether these human astrocyte-derived organoids (referred to hAD-Organs) include multiple cortical layers, we performed immunostaining to examine protein expression of different cortical layer markers (Fig. 3f). We found that deep cortical layer marker CTIP2⁺ and TBR1⁺ cells were mainly distributed on the outside of the PAX6⁺ VZ region. Superficial layer neuronal markers SATB2⁺ and BRN2⁺ were located at the intermediate layers and were separated from Reelin⁺ cells which resembled the layer I marginal zone of the cerebral cortex (Fig. 3g and Extended Data Fig. 1e, f). These results indicated that hAD-Organs may have formed the structure of six cortical layers at week 10. In addition, Op53 treatment alone was not sufficient to support astrocytes reprogramming to form hAD-Organs (Fig. 3h, i and Supplementary Fig. 3a). Importantly, the small molecule cocktail CSBRY treatment induced rapid cell proliferation over 240 fold compared with the control at day 42 (Fig. 3h), and showed typical cerebral organoids with neuroepithelial buds that was larger in size compared to the control (Fig. 3i and Supplementary Fig. 3a). Immunostaining and FACS assay revealed that the majority of cells in hAD-Organs were TUJ1⁺/MAP2⁺ cells showing typical neuronal morphology in CSBRY treated group (Fig. 3j, k and Supplementary Fig. 3b). These findings indicated that CSBRY cocktail robustly induced and accelerated the growth and proliferation of organoids during human astrocyte conversion.

Fig. 3 |. 3D cortical organoids generated from Op53-CSBRY induced human astrocytes.

a, Schematic illustrating the generation of human cortical organoid from human astrocytes. b, Schematic diagram of 3D cortical organoids induction protocol. c, Representative images of cortical organoids induced from human astrocytes by Op53-CSBRY at day 8, 14, 42 and 90. Scale bars, 100 μm (Day 8), 400 μm (Day 14), 1000 μm (Day 42 and Day 90). d, Immunostaining of sections from cortical organoids at week 7 using the following markers: left: neuronal markers TUJ1 and MAP2, middle: neuronal marker TUJ1 and neural progenitor marker PAX6, and right: neural progenitor marker SOX2 and neuronal marker MAP2. Scale bars, 25 μm (left), 50 μm (center), 25 μm (right). e, Immunostaining of sections from cortical organoids at week 10 using the following markers: left: neuronal markers TUJ1 and MAP2, middle: neuronal marker TUJ1 and neural progenitor marker PAX6, and right: neural progenitor marker SOX2 and neuronal marker MAP2. Scale bars, 50 μm. f, Schematic representation of the neuronal markers for genes expressed in the six cortical layers (I-VI). g, Representative immunostains of sections from cortical organoids with neural progenitor marker PAX6 and cortical layer markers (from left to right: CTIP2, SATB2, TBR1, BRN2 and from second left to right: Reelin) at day 70. The dotted line delineates the VZ-like structures. Scale bars, 25 μm. h, Quantitative analysis showed that small molecules CSBRY treatment induced rapid proliferation compared with only Op53/DMSO control (n = 3 independent experiments). “Fold” is fold change in cell number compared to the control. Data are presented as mean ± SEM. i, Representative images of hAD-Organs induced by Op53/DMSO control and Op53-CSBRY (n = 3 independent experiments). Scale bar, 1000 μm. j, Immunostaining showed that 7-week-old hAD-Organs dissociated cells were MAP2⁺ and TUJ1⁺ neurons. Scale bar, 25 μm. k, Quantitative analysis the percentage of MAP2⁺ and TUJ1⁺ neurons dissociated from 7-week-old hAD-Organs treated by Op53-CSBRY (n = 3 independent experiments). Data are presented as mean ± SEM. Representative images from three sections of organoids with similar results (d, e, g).

To identify the origin of hAD-Organs in our reprogramming system, primary cultures of human astrocytes were analyzed by immunostaining and neurosphere assay. More than 80% of human primary astrocytes co-expressed GFAP and S100β, with almost no MAP2⁺ neuron (Extended Data Fig. 2a, b). However, some neural progenitor cells (NPCs) with weak expression SOX2 and PAX6 were found in the cultured human fetal astrocytes (Extended Data Fig. 2a, b). To test the ability of neurogenesis in these cells, we seeded 50 primary astrocytes per well into an ultralow attachment 24-well plate with neurosphere culture medium containing bFGF and EGF. After 14 days, we found that neurospheres were generated from the Op53-CSBRY reprogrammed astrocytes, but none in the primary cultured astrocytes (Extended Data Fig. 2c), indicating the contaminated SOX2⁺ and PAX6⁺ cells in primary human astrocytes were not the bonafide neural stem cells and had lost the ability of neurogenesis after long-term culture in serum containing medium. In addition, we could not find MAP2⁺ cells in primary cultured human astrocytes by treatment with the small molecules ISX-9 and RA, which were used to induce neural differentiation of NSCs (Extended Data Fig. 2d). Furthermore, Op53 only treated astrocytes could not generate brain organoids (Fig. 3h, i and Supplementary Fig. 3a). These results indicated small molecular cocktail CSBRY was necessary for the generation of hAD-Organs in this astrocyte reprogramming system. To trace the origin of the induced neurons, we infected human astrocytes with hGFAP::EGFP virus. We found that most of cells showed EGFP expression manifesting their astrocytic identities. Some of EGFP⁺ cells have demonstrated the neuronal morphology treated with CSBRY at day 19, and co-expressed MAP2, representing their neuronal identities (Extended Data Fig. 2e). Time-lapse imaging showed that EGFP⁺ astrocytes with flat morphology divided and converted into neuron-like cells (Extended Data Fig. 2g, Supplementary Video 1). After Time-lapse imaging, we fixed the cells and found the neuron-like cells were MAP2⁺ with weak expression of EGFP, which finally suggested that the converted neurons originated from human astrocytes (Extended Data Fig. 2e, f). In hGFAP::EGFP labeled cortical organoids at week 3, we found that some of GFP positive cells are co-labeled with MAP2⁺ neurons, and some GFP⁺ cells are FOXG1⁺ or SOX2⁺ neural progenitor cells, but most of them still maintained their astrocytic identities (S100ꞵ⁺) (Extended Data Fig. 1g-k). To further characterize the origin and cell population of the initial cultured human astrocytes which were isolated from the human fetal cerebral cortex, we conducted single-cell RNA-seqencing and identified these astrocytes based on cortical astrocyte layer specific markers, which were recently reported via a single-cell in situ transcriptomic map in the cerebral cortex⁴⁵. We found that OCT4⁺, NANOG⁺ and LIN28A⁺ hESCs were not observed, and also no NeuN⁺ and DCX⁺ neurons (Extended Data Fig. 3a-c). The majority of the primary astrocytes expressed cortical upper gray matter astrocyte markers ( such as 84.2% ITM2B, 88.8% BSG and 81.6% IGFBP2) and over 50% cells expressed deeper layer astrocyte markers ( 57.3% ID1, 55.1% ID3, 58.7% DKK3 and 57.6% EFHD2), indicating their developing cortical gray matter astrocyte identities (Extended Data Fig. 3a, d-f). Taken together, these data suggested hAD-Organs were generated from human developing cortical astrocytes reprogrammed by Op53-CSBRY, but not from contamination of NPCs. Finally, our results verified that human primary astrocytes can be directly reprogrammed into human brain organoids by Op53-CSBRY in vitro.

Generation of spinal-cord organoids derived from human Astrocytes.

Previous studies reported that FGF, SHH and BMP4 morphogens were involved in patterning spinal cord tissue during development ^9,15. It is unknown whether these Op53-CSBRY induced astrocytes could respond to spinal cord patterning signals. To confirm this, we designed the protocol by treating Op53-CSBRY induced astrocytes with bFGF, SAG and BMP4 to develop human astrocyte-derived spinal-cord organoids, referred to hADSC-Organs (Fig. 4a, b). After 4 weeks induction, hADSC-Organs grew up to about 1 mm, and the size could robustly increase to 2–3 mm after 7 weeks (Fig. 4c). We then performed immunostaining with different dorsal and ventral spinal cord progenitor markers (Fig. 4d). We observed NPC marker PAX6, ventral progenitor domain markers Nkx6.1 and Olig2, and dorsal progenitor domain marker Olig3 were highly expressed in these hADSC-Organs (Fig. 4e-h and Extended Data Fig. 4a, b). Specifically, Nkx6.1⁺ cells formed the ventricular zone-like structure (Fig. 4e, f and Extended Data Fig. 4a), and dorsal Olig3⁺ cells were well separated with ventral Nkx6.1⁺ cells (Fig. 4h). Furthermore, Olig2⁺ ventral motor neuron progenitor cells mainly clustered and co-localized with Nkx6.1⁺ cells (Fig. 4f). Those results demonstrated that this protocol induced both ventralized and dorsalized progenitor identities of developmental spinal cord.

Fig. 4 |. Generation of human astrocyte-derived spinal cord organoids.

a, Schematic illustrating the generation of human spinal cord organoid from human astrocytes. b, Schematic diagram for the induction of hADSC-Organs. c, Phase images of hADSC-Organs at different stages. Scale bars, 1000 μm. d, Schematic diagram showing the expression pattern of dorsal (pd1-pd6) and ventral progenitor domain markers (p0-p3) in the developing spinal cord. Left: BMP4 and shh concentration gradient, right: transcription factors expressed by the domains. e - h, Representative immunohistochemistry images of sections from hADSC-Organs at week 7 showing high expression of ventral progenitor domain marker Nkx6.1 and Olig2 in low power (e) and high power image (f), neural epithelium cell marker PAX6 colocalization with Nkx6.1 (g), and dorsal progenitor domain marker Olig3 were well separated with ventral marker Nkx6.1⁺ cells, the dotted line delineates the VZ area (h). Scale bars, e: 500 μm, f - h: 25 μm. i - l, Immunostaining of sections from hADSC-Organs at week 10 characterizing specific cholinergic spinal cord motor neuron markers ChAT (i) and HB9 (j), GAD67 positive GABAergic interneurons (k), and VGLUT1 positive glutamatergic neurons (l). Scale bars, 25 μm. Representative images from three sections of hADSC-Organs with similar results (e-h, i-l).

Next, we examined whether hADSC-Organs could generate different subtypes of spinal cord neurons. Notably, TUJ1⁺ / MAP2⁺ neurons and ChAT⁺ or HB9⁺ spinal cord specific motor neurons were detected in 10-week-old hADSC-Organs (Fig. 4i, j and Extended Data Fig. 4c, d). Major neuronal subtype GAD67⁺ inhibitory interneurons and VGLUT1⁺ excitatory neurons were observed in this stage (Fig. 4k, l). The populations of neurons showed 42.9±15.9% MAP2⁺ and 40.1±19.0% TUJ1⁺ neurons. Some of the cells especially expressed spinal cord motor neuron marker ChAT (38.7±11.4%) and HB9 (24.4±7.5%) (Extended Data Fig. 4e). VGLUT1⁺ and GAD67⁺ neurons showed 26.1±11.6% and 34.2±6.7% respectively in 10-week-old hADSC-Organs (Extended Data Fig. 4e). To better characterize hADSC-Organs, we dissociated and plated the 10-week-old hADSC-Organs. After plating for 5 days, most of the cells displayed typical neuronal morphology and only a few GFAP⁺ cells (Extended Data Fig. 4f-h), and ChAT⁺ spinal cord motor neurons showed high co-expression with MAP2 (Extended Data Fig. 4i). To further confirm human developmental spinal cord identities, we measured the global transcriptome of hADSC-Organs compared with hAD-Organs by RNA-seq analyses. At week 7, hAD-Organs showed high expression of cortical progenitor marker genes. In contrast, human developmental spinal cord specific transcriptional genes, such as HOX family genes, MNX1, ISL1, ChAT, LHX3, NKX2.2 and NKX6.1 were robustly expressed in hADSC-Organs (Extended Data Fig. 4j, k). Gene ontology analysis revealed the enrichment of many spinal cord developmental biological processes in 7-week-old hADSC-Organs, such as anterior/posterior pattern specification, and sensory perception of pain (Extended Data Fig. 4l).

To further investigate the spinal cord identities of hADSC-Organs, we performed single-cell RNA-seq for hADSC-Organs. We found that cells in hADSC-Organs were grouped into 17 main clusters, then characterized the dorsal and ventral domain-specific cell populations and gene expression profiles according to the reported markers Matrix of developmental spinal cord^5,46. We found that 7-week-old hADSC-Organs contained two main types of cells: spinal cord domain-specific neural progenitors and subtype neurons (Fig. 5a-c). The neural progenitors represent about 41.15% of the total number of cells and could be subdivided into RP, dp1, dp2–6, p0/1, p2, pMN and p3/FP progenitor regional clusters, while the neuronal cells (58.83%) contain dI1, dI2–6, V0/1, V2, MN and V3 spinal cord regional cell clusters (Fig. 5b). Cluster p0/1 is the largest proportion of progenitor clusters (18.18%). Cluster V0/1 (19.43%) and MN (19.29%) are the top two population of neuronal cells in 7-week-old hADSC-Organs. Bubble charts depicted the well-expression pattern of spinal cord dorsal and ventral domain-specific markers in each assigned cell cluster, indicating their developmental spinal cord identities (Fig. 5c). In addition, neural progenitor marker SOX2 and neural markers MAP2 and TUBB3 were highly expressed in 7-week-old hADSC-Organs (Fig. 5f). Spinal cord motor neuron markers MNX1 and ISL1 showed in similar neuronal clusters, but different with GATA3⁺ V2 interneurons (Fig. 5f). Furthermore, we also found GABAergic neurons (GAD1⁺, SLC32A1⁺ and GAD2⁺) were broadly expressed in multiple cell clusters, and glutamatergic (SLC17A6⁺), glycinergic (SLC6A5⁺), and cholinergic (SLC5A7⁺ and SLC18A3⁺) subtype neurons showed in the different specific-domains (Fig. 5d). To investigate rostro-caudal identities of hADSC-Organs, we found broad expression of HOXA2-HOXA5, HOXB2-HOXB8, HOXC4-HOXC5, HOXD3 and HOXD4 in hADSC-Organs, representing caudal cervical/thoracic spinal cord identities (Fig. 5e). So, our results demonstrated that hADSC-Organs possess the developing spinal cord identities with dorso-ventral and rostro-caudal spatial cytoarchitectures.

Fig. 5 |. Transcriptome profiling of hADSC-Organs.

a, tSNE visualization of single cell gene expression of hADSC-Organs at day 49 (n=12,341, cells from 4 hADSC-Organs). Roof plate (RP), dorsal progenitor cell types (dp1–6), ventral progenitor cell types (p0–3), Motor neuron progenitors (pMN), floor plate (FP), dorsal interneurons (dI1–6), ventral neurons (V0–3), Motor neurons (MN). b, Graph showing the percentage of cells in each cluster in hADSC-Organs. c - e, Bubble charts showing the expression of selected domain-specific markers used to identify DV domains progenitors and neuronal subtypes (c), and neurotransmitter identity-related genes in neuronal clusters (d), and expression of HOX family genes in each clusters in hADSC-Organs (e). Circle size represents the percent of cells expressing each gene per cluster. f, tSNE plots showing gene expression of selected hADSC-Organs neural progenitors, neuron markers, motor neuron progenitors. Colored scale shows normalized gene expression levels.

In summary, through in vitro defined conditions, we technically induced human astrocytes into spinal-cord organoids, which included both human ventral and dorsal spinal cord progenitors, major neuronal subtypes, and that resembled human developmental spinal cord-like tissue.

Transcriptional Analysis of converted neurons from human astrocytes.

To understand the mechanisms of direct reprogramming human commercial astrocytes into 3D organoids, we performed RNA-seq for the analysis of transcriptional profile after 5 days CSBRY treatment. Sample distance matrix showed the three independent biological replicates treated by CSBRY were well clustered (Extended Data Fig. 5a). Compared with CON treatments, 2250 genes were significant different expression (DE) (pVal<0.05, log₂FC>1, Fig. 6a, b). DE genes (71.2%) were upregulated by CSBRY. In contrast, only 28.8% of DE genes were downregulated (Fig. 6b and Extended Data Fig. 5b). Over 30% DE genes including both up and down-regulated genes were involved in signal transduction by KEGG pathway enrichment assay (Extended Data Fig. 5c). These results indicated that the CSBRY chemical cocktail mainly activated the external signaling pathways during the reprogramming process. Interestingly, we found that CSBRY treatment significantly up-regulated embryonic stem cell marker genes SOX2, LIN28A and SALL4 expression (Fig. 6c). However, overexpressed OCT4 were downregulated, another key pluripotent gene NANOG was not activated (Fig. 6c). We also found that the other embryonic regulation genes such as UTF1, ZFP42, TBX3, RIF1 and FZD1, showed no activation, even down-regulation. Human pluripotent stem cell markers NANOG⁺, SEEA4⁺ and TRA-1–60⁺ cells were not observed in both primary cultured astrocytes and Op53-CSBRY induced cells (Extended Data Fig. 5g). These results suggested that human astrocytes were not reprogrammed into full pluripotent stem cells by Op53-CSBRY. Importantly, we found specific neural ectodermal genes, such as OTX2, NOG, SOX1, FGF8, NES and PAX6 were significantly up-regulated (Fig. 6d). However, surface ectodermal genes MSX1 and MSX2 were significantly down-regulated (Fig. 6e). In addition, mesodermal and endodermal gene expression were not activated (Fig. 6f, g). These data suggested that induced cells possessed neural ectodermal characteristics. GO pathway analysis results showed that DEs significantly enriched in biological processes including cell proliferation in the forebrain, cell fate commitment and nervous system development (Fig. 6h). To further confirm their neural ectodermal cellular identities, we found normal cultured human astrocytes have almost no OTX2 expression, but slightly express FOXG1, SOX2 and NES (Fig. 6c-d, j-k and Extended Data Fig. 5e-f). However, CSBRY treatment robustly increased SOX2, SOX1, LIN28A, SALL4, OTX2, GBX2, FGF8, NES and PAX6 expression, most of them show high expression in neural ectodermal cells, and some of the induced cells highly co-express FOXG1 and OTX2 in one rosette-like cluster (Fig. 6c, d, l-n and Extended Data Fig. 5e-f), indicating that human astrocytes were actually reprogrammed by Op53-CSBRY into early neural ectodermal cells with heterogeneous characteristics, which may show different regional identities of neural tube (Fig. 6m, n). In summary, these results suggest that human astrocytes treated with Op53-CSBRY were directly reprogrammed into early neuroectodermal cells, which was amenable for regional specification and expansion (Fig. 6i). The cancer gene p53 was not down-regulated continually, and cancer genes KLF4 and MYC were not activated by CSBRY treatment (Extended Data Fig. 5d), which revealed a low ability to induce cancer generation. Taken together, Op53-CSBRY directly reprogrammed human astrocytes into early neuroectodermal cells, which enabled patterning of region-specific brain organoids by neural morphogens.

Fig. 6 |. Visualization of the RNA-seq dataset for transcriptional and subtype specification analysis of the conversion of human astrocytes to neurons.

a, Cluster analysis of differentially expressed genes comparing CSBRY with CON treatment (p < 0.05, log2FC > 1). b, Pie chart showing downregulated and upregulated genes in CSBRY group comparing to CON (Op53/DMSO) treatment. c, Human embryonic stem cell marker gene expression profiles with CSBRY treatment (n = 3 independent experiments) of Op53/DMSO control. d-g, Graphs showing neural ectoderm, surface ectoderm, mesoderm and endoderm marker gene expression profiles with CSBRY treatment (n = 3 independent experiments) of Op53/DMSO control. h, Graph showing KEGG enrichment pathways of CSBRY treatment. i, Working model of Op53-CSBRY induced astrocyte reprogramming. j and n, Representative images of neural ectodermal marker OTX2 (j) and forebrain marker FOXG1 (k) expression in human primary astrocytes, and OTX2 (l), FOXG1 (m) as well as neural epithelium cell marker PAX6, SOX2 (n) expression in reprogrammed cells after CSBRY treatment at day 14. Scale bars, 50 μm. j and k, Representative images of expression in human primary astrocytes at several randomaly chosen fields per coverslip of j neural ectodermal marker OTX2, k, forebrain marker FOXG1. l-n, Representative images of reprogrammed cells 14 days after CSBRY treatment expressing l, OTX2 m, FOXG1 n, and neural epithelium cell marker PAX6 and SOX2 (three coverslips of Op53-CSBRY induced cells with similar results). Scale bars, 50 μm. Data are presented as mean ± SEM. Multiple unpaired Student’s t-test was used for comparing each groups.

Functional characterization of hAD-Organs and hADSC-Organs.

We next investigated whether the induced neurons are functional. We performed whole-cell electrophysiological recordings from both dissociated hAD-Organs and hADSC-Organs (Fig. 7 and Extended Data Fig. 6). In hAD-Organs, recordings were performed at different culture stages (Fig. 7a, b). Among 167 recorded cells, 102 cells (60.1%) could generate action potentials. However, the ability to generate action potentials was dependent on reprogramming stages. We found that cells cultured for 100–140 days showed a higher probability of generating action potentials (65.2%, n = 135) than those cultured for a shorter period of time (80–100 days, 43.8%, n = 32). Among those cells with action potentials, 26.1% and 7.1% exhibited repetitive action potential firing (Fig. 7b). Spontaneous firing could also be detected occasionally (Extended Data Fig. 6a). For group 80–100 days, the resting membrane potential (RMP) of cells that could generate action potentials was significantly more hyperpolarized than those without action potential generation (Without action potential: −32.4 ± 3.8 mV, With action potential: −41.8 ± 1.7 mV; Extended Data Fig. 6b). For 100–140 days cultured cells, no significant difference in RMP could be detected between groups with and without action potentials (−34.7 ± 2.2 mV, −36.2 ± 1.1 mV; Extended Data Fig. 6b). However, among cells with action potentials, those with repetitive action potentials possessed a more hyperpolarized RMP (−41.3 ± 1.9 mV) than cells with single action potential (−34.4 ± 1.3 mV; Extended Data Fig. 6c).

Fig. 7 |. Electrophysiological characterization of hAD-Organs and hADSC-Organs.

a, IR-DIC (left) and fluorescence (right) images of an example recorded cell with APs (n = 24 cells). Scale bar, 10 μm. b, Left: membrane potential responses (top) to intracellular injections of step currents (bottom). The blue traces indicate the responses to +50/−30 pA current pulses. Notice the rebound AP occurred immediately after the negative pulse. Right: the percentage of AP generation at different ranges of culture days. The number of recorded cells in each group was provided. c, Example whole-cell currents before (left) and 5 minutes after TTX application (right). The blue traces show the evoked currents when the membrane potential was stepped to 0 mV. d, Group data showing the peak amplitude of Na⁺ currents with different culture days. Group 80–100 days: n = 14, Group 100–140 days: n = 69. Mann-Whitney test was used for comparing the two groups. e, IR-DIC (top) and fluorescence (color coded, bottom) images of a recorded cell with APs (n = 23 cells) from dissociated spinal cord organoids. Scale bar, top: 20 μm, bottom: 50 μm. f, Spontaneous rhythmic bursting was detected in the cell shown in e. g, Spontaneous regular firing from another cell. The trace also shows responses to +10/−10 pA current pulses. h, Top, the percentage of cells without AP (white), with single AP (blue) and repetitive AP (red) (top). Bottom, spontaneous firing rate. i, Top, example whole-cell currents before (left) and 5 minutes after TTX application (right). Bottom, group data showing the blockade of Na⁺ currents by TTX (left) and the current amplitudes in cells with single and repetitive Aps (right). Left: n = 4. Unpaired Student’s t-test was used for comparing with the control group. Right: Single APs: n = 26, Repetitive Aps: n = 22. Mann-Whitney test was used for comparing between single and repetitive APs. j, An example current trace showing spontaneous putative postsynaptic currents (PSC). A PSC event (arrow) was expanded for clarity (bottom left). k, Bright field image of 8-week-old hADSC-Organs covered on a 64-electrode MEA plate (top) and activity heat maps showing spike rates (spike/s) (bottom). l, Graphs showing neurons spikes were detected by single electrode in a 30 ms time frame. m, Representative traces of hADSC-Organs network activity in 65 s time frame. Network bursts were labeled by pink box. Data are presented as mean ± SEM.

In addition, we examined the voltage-gated Na⁺ currents at different culture stages (Fig. 7c, d). For cells cultured for less than 100 days, the peak amplitude of Na⁺ currents was −247.5 ± 68.6 pA (Fig. 7d), significantly smaller than that of cells cultured for 100–140 days was −355.5 ± 25.8 pA (Fig. 7d). The application of 1 μM TTX could block the Na⁺ currents completely (n = 5; Fig. 7c). In cells cultured for 100–140 days, we also compared the peak amplitude of Na⁺ currents between “single action potential” group and “repetitive action potentials” group, no significant difference was detected (Single action potential: −334.8 ± 28.4 pA, Repetitive action potentials: −397.1 ± 52.6 pA; Extended Data Fig. 6d). We also occasionally observed spontaneous putative postsynaptic events (Extended Data Fig. 6e) at 100–140 days. Indeed, we observed characteristic ultrastructure of synapses in 10-week-old hAD-Organs (Extended Data Fig. 6f), indicating that neurons in these organoids formed synaptic connections.

We further examined spiking activities in cells from dissociated hADSC-Organs (7 weeks culture, Fig. 7e-m). Cells grew multiple neurites (Fig. 7e) and 89.5% (51/57) of cells generated action potentials in response to step current injections (Fig. 7f-h). Among cells with action potentials, 45.1% (23/51) of cells could fire repetitively, 27.5% (14/51) exhibited spontaneously firing and some of them showed rhythmic bursting (Fig. 7f). The averaged spontaneous firing rate was 1.9 ± 1.5 Hz (Fig. 7h). The RMP of cells that could generate action potentials was −33.4 ± 7.1 mV. Similarly, 1 μM TTX could block Na⁺ currents completely (n = 4; Fig. 7i). In comparison with cells with single action potentials (−896 ± 571 pA, n = 26), those with repetitive action potentials possessed much larger Na⁺ currents (−1,818 ± 888 pA, n = 22; Fig. 7i). In some recorded cells, we again observed spontaneous putative postsynaptic events (Fig. 7j). In addition, to specially record action potential responses of motor neurons in hADSC-Organs, we labeled the neurons dissociated from hADSC-Organs by HB9::RFP lenti-virus. We found that 90% (18/20) of RFP⁺ motor neurons (Extended Data Fig. 6g) could generate action potentials, and 38.9% (7/18) of them could fire repetitively (Extended Data Fig. 5h). Moreover, consistent with the above whole-cells recordings, Microelectrode Array (MEA) recordings also revealed spiking activities in the whole hADSC-Organs (Fig. 7k-i), and further show neural network activities (Fig. 7m). Taken together, these results clearly demonstrated that the neurons from both hAD-Organs and hADSC-Organs induced by Op53-CSBRY from human astrocytes were fully functional.

Characterization of hADSC-Organs grafted into complete spinal cord injury of mice in vivo.

Because hADSC-Organs had spinal cord cellular characteristics, which included both human ventral and dorsal spinal cord progenitors, neuronal subtypes and cytoarchitectures, we explored whether hADSC-Organs support spinal cord regeneration in vivo after complete spinal cord injury. Acute transplantation may avoid growth inhibition and the formation of a glial scar to impede neural regeneration which was found after spinal cord injury ^2,3. Thus, we immediately grafted 7-week-old hADSC-Organs labeled with GFP into the spinal cord of NOD-SCID mice that were completely resected 2–3 mm at T10. To enhance the regeneration capacity of hADSC-Organs, we combined hADSC-Organs with Matrigel for promoting spatial integration with spared spinal cord tissue, and growth factors BDNF and GDNF to support cell survival and growth (Fig. 8a). We found that hADSC-Organs highly expressed GFP after transduction by LV-Ubi-GFP virus (Extended Data Fig. 7a), and could grow up to about 2 mm which is similar in size to a 1-month-old mouse spinal cord (Fig. 4c and Extended Data Fig. 7a, b). Importantly, 7-week-old hADSC-Organs have established their spinal cord cellular identities, including both ventral and dorsal progenitors and major subtypes of neurons (Fig. 4i-l, 5a-f), which indicated comparable cytoarchitecture with spinal cord tissue. After complete SCI, the mice showed hind limb paralysis suggesting a successful complete SCI model (Extended Data Fig. 7c and Supplementary Video 2). After 6 weeks transplantation, our results showed that hADSC-Organs had survived and integrated with the host mice spinal cords (Fig. 8b and Extended Data Fig. 7d, e).

Fig. 8 |. In vivo survival and neural connectivity of hADSC-Organs transplanted into complete spinal cord injury in mice.

a, Schematic illustrating experimental paradigm. hADSC-Organs at week 7 labeled with GFP were grafted immediately into T9 - T10 spinal cord after complete section in immunodeficient NOD-SCID mice. Matrigel was combined with hADSC-Organs for promoting hADSC-Organs spatial integration with spared spinal cord tissue, and growth factors BDNF and GDNF were added to support cell survival and growth. b, Images of vehicle control (left) and grafted hADSC-Organs integrated at the lesion epicentre (right). Scale bar, 250 μm. c and d, Immunostaining showing neuronal morphology of hADSC-Organs at week 6 after transplanted into mouse spinal cord using neuronal markers MAP2 (c) and TUJ1 (d). Scale bars, 25 μm. e - h, Immunostaining showing transplanted hADSC-Organs after 6 weeks could differentiate into GFAP⁺ astrocytes, VGLUT1⁺, GAD67⁺ and ChAT⁺ subtypes of mature neurons. Scale bars, e, g-h: 25 μm, f: 10 μm. i, The host specific presynaptic marker Bassoon (Bsn) was co-localized with GFP positive grafted neuron at the section of transplanted hADSC-Organs after 6 weeks. Scale bar, 10 μm. j, The human specific presynaptic marker hSyn showing co-localization with host ChAT⁺ neurons. Scale bar, 10 μm. k, Showing hNuMA⁺/GFP⁺ transplanted cells long-distance migration into host mouse spinal cord. The dotted lines outlined region where the hADSC-Organs were implanted. Scale bar, 250 μm. The expanded green box and the expanded red box, showing zoom-in images. Scale bars, 20 μm. l and m, Representative images of spinal cord implanted with vehicle control (left), and hADSC-Organs+MBG (right). Scale bars, 500 μm. MBG: Matrigel+BDNF+GDNF. n, Quantitative analysis of the size of the dystrophy of descending spared spinal cord of vehicle control and hADSC-Organs+MBG in each distance (per mm) from lesion (n = 3 mice). Data are presented as mean ± SEM. Represent images are from the section of spinal cord (n = 3 mice). Multiple unpaired Student’s t-test was used for comparing each groups.

Next, we further characterized subtypes of neuron and glia cells in the graft region. Immunostaining results showed that the grafted hADSC-Organs expressed neuronal markers TUJ1 and MAP2, but also included some GFAP⁺ astrocytes (Fig. 8c-e). We observed spinal cord mature neuronal subtype inhibitory GAD67⁺ and excitatory VGLUT1⁺ mature neurons in the area of transplanted spinal cord (Fig. 8f, g). Specifically, the hADSC-Organs contained motor neurons still maintained the characteristics of ChAT⁺ spinal cord motor neurons at 6 weeks after transplantation (Fig. 8h). These results indicated that hADSC-Organs maintained their spinal cord cellular identities after in vivo transplantation. Then, we further examined the integration and connectivity of grafted hADSC-Organs with host mice. We found that host spinal cord neurons formed the GFP⁺/mouse anti-Bsn⁺ regenerated input synapses in grafted hADSC-Organs region (Fig. 8i), and human synaptophysin⁺ synapses co-localized with ChAT⁺ host spinal cord motor neurons (Fig. 8j), demonstrating that synapses were formed between host neurons and the grafted hADSC-Organs. Notably, the grafted cells could migrate long distance into the host ascending spared spinal cord up to 2–3 mm from the lesion epicenter, and sprouted neurites at host mouse spinal cord after 6 weeks transplantation (Fig. 8k). We also observed that SMI312⁺ axons formed in the grafted hADSC-Organs, but no axons newly generated in matirgel only-treated mice (Extended Data Fig. 7g, h). Furthermore, we found that the grafted hADSC-Organs may reduce the size of the descending spared spinal cord (Fig. 8l-n), which may reflect the protective effects of the injured spinal cord. Together, these findings demonstrated that grafted hADSC-Organs could survive, differentiate into spinal cord neurons, re-established synaptic connections with host neurons, and bridge and protect the mouse complete injured spinal cord tissue.

Behavioral analysis is a very important indicator for evaluating the functional effects of organoid transplantation. Thus, we evaluated hindlimb locomotor movement after hADSC-Organs transplantation. We performed rigorous controlled experiments (four groups, in total 45 mice) and hADSC-Organs transplantation (25 mice) (Extended Data Fig. 7f). Unfortunately, we found no detectable improvement of spontaneous locomotor functions after hADSC-Organ transplantation at week 6 (Extended Data Fig. 7f and Supplementary Table 1). However, a few hADSC-Organs grafted mice showed hindlimb locomotor responses after strong stimulations (Supplementary Video 3), which may suggest that spinal-cord organoid transplantation has the potential to enhance locomotor functions after complete SCI. In summary, our studies show that grafted hADSC-Organs could maintain their spinal cord cellular identities, integrate with mouse host neurons and may help to bridge the complete SCI.

Discussion

Previous studies have reported that human astrocytes can be reprogrammed into different types of postmitotic neurons ^{29,30,33,34,36,37,47}, but whether 3D brain-region-specific organoids can be directly generated from human primary astrocytes remains unclear. Here, we first established a virus-free and integration-free chemically defined system to directly convert human astrocytes into neural ectodermal cells, which directly formed 3D cortical organoids and were guided into spinal-cord organoids that possessed the developing spinal cord identities with dorso-ventral and rostro-caudal spatial cytoarchitectures. The grafted hADSC-Organs can survive, differentiate into spinal cord neurons, migrate long-distance, form synaptic connectivity with host neurons, and may help to bridge the complete spinal cord injury. This study shows the potential for neural organogenesis of the central nervous system after injury via endogenous astrocyte reprograming in the future.

Currently, researchers have attempted to direct cellular reprogramming for tissue regeneration ^25,27,48 showing the potential to reduce immune rejection and thus overcoming this hurdle for transplantation. Direct somatic cell reprogramming has the potential for in situ organogenesis within damaged organs ^28,36. Tissue loss in the central neural system is irreversible after injury. So far, in situ neural organogenesis remains a challenge ⁴⁹. Previous studies have demonstrated that resident astrocytes can be directly reprogrammed into neurons by ectopic expression of transcription factor, microRNA or small molecules ^{26,28–35,47}. However, the methods of chemical reprogramming have shown human astrocytes bypass a proliferative state and directly convert into postmitotic neurons which have no ability for proliferation and neural organogenesis ^30,34. Moreover, the conversion efficiency is low compared with initial starting astrocytes. To improve on this, we found that small molecules CSBRY cocktail robustly enhanced human astrocyte reprogramming about 240 fold. Critically, these Op53-CSBRY reprogrammed cells specifically obtained a neural ectodermal state, which had the ability to rapidly proliferate, and spontaneously form cortical organoids. They could then be patterned into spinal-cord organoids by ventralizing and dorsalizing morphogens. Previous studies have showed that mature astrocytes could be directly converted into NSCs by OCT4 in rats, which was enhanced by Shh⁵⁰. Another study has reported that 6 small molecules including A83–01, CHIR99021, sodium butyrate, LPA, rolipram and SP600125, which combined with the over-expression of OCT4 could convert human neonatal fibroblast cells into hNSCs that also expressed PAX6 and OTX2⁵¹. Here, we first showed that a previously undescribed 4 compound cocktail CSBRY combined with OCT4 and knock-down p53 can directly reprogramm human astrocytes into neural ectodermal cells, which can be directly guided into brain region-specific organoids, displaying a high capability for organogenesis. Our findings may indicate that local astrocytes have the potential of in situ organogenesis after brain damage, such as stroke and spinal cord injury.

SCI, especially complete SCI leads to the loss of spinal cord tissue including extensive neural cell and fiber loss. Research into spinal cord neural regeneration has often shown conflicting or controversial results in recent decades ^3,12. The main strategies for restoring extensive lost spinal cord neural cells are reactivating neurogenesis using endogenous neural cells or cell replacement by transplantation of exogenous cells. Recent studies have demonstrated that a graft of hPSC-derived spinal cord NSCs helped axonal regeneration and promoted functional recovery after mice spinal cord injury ⁹, and also showed restorative effects after transplantation into the primate spinal cord ⁵². Previous studies have reported that transplanting hiPSC-derived neurospheres induced synapse formation with host neurons, promoted axonal growth and led to significant functional recovery in contusion spinal cord mice ¹¹. Here, we report that human endogenous astrocytes were directly reprogrammed into hADSC-Organs in vitro, which have developmental spinal cord cellular identities and show similar cytoarchitecture to the spinal cord. The grafted hADSC-Organs survived, differentiated into the major spinal cord neural cells, formed synapses with host neurons, and bridged and protected the host complete injury spinal cord tissue. However, we found no detectable improvement of spontaneous locomotor functions after 6 weeks transplantation, only some of hADSC-Organs treated mice showed locomotor responses after stimulations (Supplementary Video 3). One reason is that the grafted cells and host tissue are from different species, and thus may not form effective functional synapses. The other reason is hADSC-Organs had not shown in vivo architecture of the human spinal cord, such as without dorsal and ventral neuronal nuclei or white matter structure, which may better facilitate the connection with the host mouse corresponding spinal cord tissue.

Thus our induction method for hADSC-Organs may require optimization using biomaterials and engineering to mimic damaged spinal cord tissue for improved functional outcomes from neural organoid therapy of the SCI in the future.

In conclusion, we established a genetic integration-free and chemically defined method to directly generate cortical and spinal-cord organoids from human astrocytes through early neuroectodermal cells. Furthermore, we confirmed that these hADSC-Organs grafted into a complete spinal cord injury could survive, maintain their spinal cord cellular characteristics, form synaptic connectivity with host neurons and protected the injured spinal cord tissue. These studies indicate that neural organoid therapy may be a promising strategy for neural organogenesis. If further combined with a biomaterial delivery system or biomedical engineering, it will be open an advanced avenue of in situ organogenesis by in vivo human astrocyte direct reprogramming. Future studies will optimize the more delicate spinal-cord organoid-induced conditions and examine whether Op53-CSBRY protocol can in situ directly reprogram astrocytes into spinal cord organ in non-human primate SCI model in vivo, for further exploring the potential of clinical medical translation.

Methods

Human astrocyte culture.

Commercially acquired primary astrocytes isolated from human brain cerebral cortex (Catalog #1800, Sciencell) were cultured in Matrigel (Catalog #365230, BD)-coated plates using astrocyte medium (AM) (Catalog #1801 Sciencell). The medium was changed every other day. Cells were passaged with TrypLE^™ (Thermo Fisher Scientific), and neutralized with DMEM (Thermo Fisher Scientific) with 10% Fetal Bovine Serum (FBS) (Thermo Fisher Scientific), and were used for this study at passage 4–8. These human primary astrocytes were used in all the reprogramming experiments excluding in Fig. 2c-f.

Human embryonic stem cells (H9, WiCell) were maintained in Matrigel-coated plates with E8 medium (Gibco). For astrocytes differentiation, H9 cells were cultured in low attachment flasks in SRM medium (DMEM, 15% Knockout serum replacement, 2 mM L-glutamine and 10 μM β-mercaptoethanol) and treated with LDN193189 (100 nM, Stemolecule) and SB431542 (10 μM, MCE) from day 0 to day 7. Spheres were grown in N2aa medium (DMEM-F12 with 1% N2 and 200 μM L-ascorbic acid) and supplemented with 20 ng/ml bFGF and 20 ng/ml EGF from day 8 to day 14. From day 15, spheres were dissociated into single cells and reseeded in Matrigel-coated plate with AM medium and passaged every 3–5 days.

Human normal mature astrocytes were dissociated from the tissue of 13–65 years-old patients with gliomas sided to the focal lesion tissue. The samples were cut into 1 mm block and digested with papain for 1h at 37 °C. Then neutralized with DMEM+10% FBS and cultured in Matrigel-coated plate with AM. The protocol for collecting astrocytes was permitted by the scientific research sub-committee of the medical ethics committee of Zhongshan hospital affiliate to Xiamen University.

Human astrocyte reprogramming and small molecule screening.

Human Astrocytes were cultured in 6-cm or 10-cm plates until 90% confluence. 5–6 micrograms of pCXLE-hOCT3/4-shp53-F (addgene #27077) plasmid mixtures were electroporated into ~1 × 10⁶ cells using program T-020 (Lonza) with a 100 μl reagent kit（Lonza, VPD1001) according to the manufacturer`s instructions. On day 0, the cells were seeded onto 2×10⁴ cells/well in 24-well plate coated with Matrigel. The next day, the culture medium was replaced with fresh AM. After 3 days of transduction, the cells were switched to Neural Medium including Neurobasal medium (Thermo Fisher Scientific), Glutmax (1:100, Thermo Fisher Scientific), NEAA (1:100, Thermo Fisher Scientific), 2% B27 supplements (Thermo Fisher Scientific), 20 ng/ml BDNF (Peprotech), 20 ng/ml GDNF (Peprotech), 200 μM L-Ascorbic acid 2-phosphate (Sigma), and 100 U/mL Penicillin and Streptomycin (Thermo Fisher Scientific). From day 3 to day 8, small molecules including CHIR99021 (1.5 μM, Tocris), VPA (5 mM, Cayman), RepSox (1 μM, Tocris), Forskolin (10 μM, Cayman), JQ1 (50 nM, Sigma), ISX-9 (10 μM, Tocris), LDN193189 (100 nM, Stemolecule), TTNPB (0.5 μM, Tocris), Y27632 (10 μM, Tocris), DAPT (2.5 μM, Sigma), SAG (100 nM, EMD Millipore), SB431542 (10 μM, MCE), PD0325901 (10 μM, Tocris), SU5402 (10 μM, Biovision), and XAV939 (10 μM, Tocris) were added to induce neurons cultured with Neural Medium for primary screening. CHIR99021, SB431542, RepSox and Y-27632 were selected as the different combinations for second screening. Those small molecules were treated for 5 days. At day 8, the drugs were removed, changed into fresh Neural Medium. The medium was changed every 2–3 days. On day 21, cells were fixed for neuronal marker MAP2 immunostaining.

Generation of cortical organoids from human astrocytes.

For the generation of cortical organoids, on day 0, human astrocytes electroporated with pCXLE-hOCT3/4-shp53-F were seeded 2×10⁵ cells/well into 6-well plate coated with Matrigel. the next day, The culture medium was replaced with fresh AM. From day 3 to day 8, CSBRY small molecule cocktail treated in Neural Medium. From day 9 to day 13, the induced astrocytes were expanded in Neural Medium. On day 14, the induced astrocytes were transferred to ultralow-attachment 24-well plates (Corning) with 2–3×10⁵ cells/well for seupension culture, when large amounts of neural precursors appeared. ROCK inhibitor Y27632 (10 μM) was added to promote the survival of cells at the first day of reseeding. From day 15, the neural spheroids were generated in each well in 24-well plates on an orbital shaker and then continuely maintained in Neural Medium, which was changed every 2–3 days. At day 21, spheroids were embedded with Matrigel. Then the spheroids were transferred into ultralow-attachment 60 mm dish on an orbital shaker for long-term culture.

Generation of spinal-cord organoids from human astrocytes.

For induction of spinal-cord organoids, from day 0 to day14, the protocol was the same as the induced cortical organoids. From day 15 to day 18, the spheroids were induced with bFGF (20 ng/ml, Peprotech) and Retinoic acid (100 nM, Sigma) in the Neural Medium. On day 18, bFGF and RA were replaced with the sonic hedgehog signaling agonist SAG (500 nM) lasting for 10 days. Next, from day 28 to day 31, the spheroids were treated with BMP4 (15 ng/ml, Peprotech), with previous drug withdrawed. During this course, spheroids were embedded with Matrigel, then transferred into ultralow-attachment 60 mm dish on an orbital shaker on day 21 for long-term culture, with medium exchanged every 2–3 days. After 7 weeks, the spheroids were collected for further analysis.

For transplantation of spinal-cord organoids, the induced cells were infected with LV-Ubi-GFP virus at day 13 before reseeding to ultralow-attachment 24-well plate. The other induced procedures were the same as above.

Immunofluorescence staining.

For Immunocytochemistry, cells were fixed for 20 min at room temperature (RT) in 4% paraformaldehyde (PFA) in PBS, washed three times with PBS for 10 min each, and then blocked for 20 min in PBS containing 3% BSA and 0.3% Triton X-100, washed with three times with PBS for 10 min. Fixed cells were incubated overnight at 4 °C with the primary antibodies (Supplementary Table 2) in PBS with 3% BSA. Then the cells were washed three times with PBS for 10 min each and incubated with secondary antibodies in PBS with 1% BSA for 1 hour at RT. Cell nuclei were stained with DAPI.

Cortical organoids and spinal-cord organoids were fixed in 4% PFA overnight at 4 °C. Then transferred to 30% sucrose. After sinking into the bottom, organoids were embedded in optimum cutting temperature (OCT) embedding medium and sectioned in a cryostat with 20 μm thick slices. Serial sections of organoids were attached to slide glasses. Following air dry at 37 °C, the sample slices were permeabilized and blocked in PBS with 3% BSA and 0.3% Triton X-100 for 20 min at RT, and incubated with primary antibodies at 4 °C overnight. Next, the slices were washed with three times in PBS and incubated with secondary antibodies (Alexa Fluor 488-, 594- and 638-conjugated antibodies, 1:500) for 1 hour at RT. Cultured slices from dissociated hADSC-Organs were performed with similar protocol, instead incubated with streptavidin (1:1000, Thermo Fisher Scientific, S11227). The nuclei were stained by DAPI solution.

Western blotting.

Protein samples were collected on day 3 after the transduction of plasmids. Proteins (30 μg) were separated on 12% SDS-PAGE and transferred to a PVDF membrane. The membranes were incubated at 4 °C overnight with the primary antibodies: OCT3/4 (mouse, 1:1000, BD Bioscience, 561555), p53 (rabbit, 1:300, Epitomics, 1026–1) and GAPDH (rabbit, 1:5000, Abcam, ab181602). Horseradish peroxidase-conjugated anti-mouse or anti-rabbit IgG was used as secondary antibodies. Chemiluminescence (Yeasen, Super ECL Detection Reagent, 36208ES60) was used to develop the signals.

Electrophysiological recordings and analysis.

Culture slices prepared for whole-cell recording were transferred to an incubation chamber and perfused with ACSF. Whole-cell recordings were performed using patch pipettes with the impedance of 4–7 MΩ. The temperature was maintained at 34.5–35.5 °C. Infrared-differential interference contrast (IR-DIC) microscope (BX-51WI, Olympus) was used for the visualization of culture cells. A Multiclamp 700B amplifier (Molecular Devices) and a Power1401–3A analog-to-digital converter were used for data acquisition. Spike2 and Signal software (Cambridge Electronic Design) together with MATLAB (MATHWORKS) were used for data analysis. The ACSF contained (in mM): 126 NaCl, 2.5 KCl, 26 NaHCO₃, 1.25 NaH₂PO4, 2 MgSO₄, 2 CaCl₂, and 25 dextrose (315 mOsm, pH 7.4), and was bubbled with 95% O₂ and 5% CO₂. The pipette solution contained (in mM) 130 K-gluconate, 10 KCl, 4 MgATP, 0.3 Na₂GTP, 10 HEPES, 10 Na₂-phosphocrecreatine and 0.2% biocytin (288 mOsm, pH 7.29) or 140 K-gluconate, 3 KCl, 2 MgCl₂, 10 HEPES, 0.2 EGTA, 2 Na₂ATP and 0.2% biocytin (285–295 mOsm, pH 7.2). In some recordings, Alexa Fluor 488 dye (Invitrogen) was loaded for cell morphology.

To induce AP initiation in culture cells, positive current pulses (10 pA per step, 500 ms in duration) were injected to examine the electrophysiological properties when the cell was at the resting membrane potential (RMP). To obtain the Na⁺ currents, cells were held at −100 mV for 80 ms and then depolarized to different membrane potential levels (from −90 to 60 mV, 10 mV per step) for 50 ms. P/N subtractions were performed offline by using Signal software. Na⁺ current amplitudes were further analyzed by using MATLAB (MATHWORKS). The spontaneous firing rate was measured from a 30 s stable recording. The liquid junction potential (15 – 16 mV) was not corrected for the membrane potential shown in the text and figures. For group comparison, Shapiro-Wilk test was used for the data normality test. Two-sample Student’s t-test was used if they were normally distributed, otherwise Mann-Whitney U test would be employed. Data were presented as mean ± SEM in the text and figures.

Multielectrode Array (MEA) recording for hADSC-organs.

MEA were used to detect the neuronal spike. Each well of 6-well MEA plates (Axion Biosystems, Atlanta, GA, USA) contains 64 low-impedance platinum microelectrodes (0.04 MΩ/microelectrode) with 30 mm of diameter spaced by 200 mm. The plate was pre-coated with Matrigel 1:100 diluted in DMEM medium at incubator for 30 min. One 9-week-old spinal-cord organoids were placed into a coated MEA well to cover 64 microelectrodes. Matrigel was added to immobilized the spinal-cord organoid at incubator for 10 min and then added fresh Neural Medium. 10–15 min recordings were performed using a Maestro pro MEA system and AxIS Software Spontaneous Neural Configuration (Axion Biosystems).

For data analysis, Neural Metric Tool (Axion Biosystems) was used to classify active electrodes that are detected at least 5 spikes per minute. Based on an inter-spike interval (ISI) threshold requiring a minimum number of 5 spikes with a maximum ISI of 100 ms, spike bursts were then identified.

Virus package.

The 293T cells were plated to Collagen I (5 μg/cm²)-coated 10-cm dishes with 3 × 10⁶ cells and cultured for 24 hours in 10 ml of DMEM medium with 10% FBS, 4 mM L-Glutamine, 100 U/ml penicillin & streptomycin, 0.1 mM MEM NEAA. After 24 hours, the medium was replaced 1–3 hours prior to the transfection with 8 ml of pre-warmed fresh DMEM medium. For each 10-cm dish, 20 μg total plasmid DNA (10 μg LV-Ubi-GFP or LV-HB9::RFP and 7.2 μg PSPAX2 and 6.8 μg MD2G) was diluted in 0.5 ml of 0.25 M calcium chloride solution. The diluted plasmid DNA was mixed gently with an equal volume of 0.5 ml of 2 × HBS (PH 7.05) and immediately add into the dish drop by drop, then the cells were incubated at 37 °C in 5% CO₂ after swirling the plate. Next collected the supernatant to a 50 ml tube at 24 hours, 48 hours and 72 hours after transfection, followed by centrifugalizing the supernatant 3000 rpm at 4 °C for 10–15 min. Finally, aliquots of 100 μl supernatant were stored at −80 °C.

FACS analysis.

Cells were isolated using TrypLE and fixed by 4% PFA for 15 min, washed three times with PBS, and then resuspended in PBS buffer containing 2% normal serum (FACS buffer). Then, add the primary antibody MAP2 (1:500) to each sample and incubated on ice for 1 hour. Next, add secondary conjugated antibody to each sample and incubated on ice for 30–60 minutes, after that they were washed three times with PBS, resuspended in FACS buffer, and analyzed on a Beckman Gallios flow cytometer. Data Analysis was performed using FlowJo software.

Time-lapse imaging.

Human astrocytes were electroporated with pCXLE-hOCT3/4-shp53-F plasmid and then cultured in Matrigel coated 35 mm dish. After treated with CSBRY for 5 days, infected with 10 μl ssAAV-pGFAP-EGFP-WPREs virus suspension for 6 to 8 hours. For live cell tracing, GFP-positive cells were imaged under epifluorescent microscope (Zesi TE-2000-S) for 4 days with snaps shot every half an hour. After time-lapse imaging, the cells were fixed and did MAP2 immunostaining.

Electron microscopy (EM).

10-week-old organoids were immersed in fixative (2.5% glutaraldehyde in 0.1 M PBS buffer, pH 7.4) at 4 °C for at least 4 hours, wash three times with 0.1 M PBS buffer at 4 °C for 1 hour and postfixed in 1% osmium tetroxide in 0.15 M cacodylate buffer at 4 °C for 2 hours. And stained in 2% uranyl acetate for 1 hour. Organoids were dehydrated in an ethanol gradient (30%, 50%, 70%, 90% and 100% ethanol), embedded in Spur’s resin. Samples sectioned at 60 nm on a Leica Ultracut UCT (Leica, Bannockburn, IL), and transferred onto copper grids, stained with 2% uranyl acetate for 5 minutes and lead citrate stain for 1 minute. Grids were analyzed using an HT-7800 transmission electron microscope (Hitachi).

Thoracic complete section surgery and transplantation.

Adult male NOD/SCID mice (20 – 22 g) were anesthetized and maintained with 0.8% isoflurane throughout the whole surgical procedure. By performing laminectomy, we exposed the dorsal portion of spinal cord (T9 - T10 levels). The dorsal and ventral completed section with 2–3 mm length at T10 level was removed with spring scissors. GFP-labeled spinal cord organoids with Matrigel including BDNF (20 ng/ml) and GDNF (20 ng/ml) were transplanted into the lesion region by using a cut pipette after the complete section. While the equal volume of PBS or Matrigel was injected into mice as control. Cover muscles and skin at the site of injury were sutured with stylolite. Bladders of damaged animals were squeezed daily to help mice urinate during the experimental course. All experimental protocol was approved by the Institutional Animal Care and Use Committee at Xiamen University and compliance with all relevant animal-use guidelines and ethical regulations.

Behavioral and Histological Analyses.

Behavioral analyses were evaluated by two observers with a double-blind approach according to the criteria BBB score. For Immunohistochemistry, after transplantation 6 weeks, mice were deeply anesthetized, and through cardiac perfusion first with PBS and then 4% PFA (pH 7.4). The dissected spinal cords were fixed in 4% PFA at 4 °C overnight and dehydrated with 30% sucrose at 4 °C for one day. Then, using a cryostat we cut the dehydrated tissue into axial sections. And then following air dry at 37 °C, the sections were permeabilized and blocked in PBS with 3% BSA and 0.3% Triton X-100 for 20 min at RT, and incubated with primary and secondary antibodies according to the normal immunostaining protocol.

RNA-seq.

For the RNA-seq, three biological repeats were used including CSBRY treatments and control groups, and hAD-Organs comparing with hADSC-Organs. Total RNA was extracted using the RNAiso Plus (Takara, 9109) according to the manufacturer`s protocol. RNA purity and quantification were evaluated using a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA). RNA integrity was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Then the libraries were constructed using TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. The transcriptome sequencing and analysis were conducted by OE Biotech Co., Ltd (Shanghai, China). The libraries were sequenced on an Illumina HiSeq X Ten platform and 150 bp paired-end reads were generated. Raw data (raw reads) of fastq format were firstly processed using Trimmomatic, about 40–65M raw reads for each sample were generated. Then about 40–65M clean reads for each sample remained, which were obtained for downstream analyses by removing reads containing adapter, reads containing ploy-N and low-quality reads from raw data. The clean reads were mapped to the human genome using HISAT2. FPKM value of each gene was calculated using Cufflinks, and the read counts of each gene were obtained by HTSeq-count. Differential expression analysis was performed using the DESeq (2012) R package. P value < 0.05 and fold change > 2 or fold change < 0.5 was set as the threshold for significantly differential expression. Hierarchical cluster analysis of differentially expressed genes (DEGs) was performed to explore gene expression patterns. GO enrichment and KEGG pathway enrichment analysis of DEGs were performed respectively using R based on the hypergeometric distribution.

Single-cell RNA-sequencing.

To characterize the cell subtype population of primary human astrocytes (HA1800) and hADSC-organs, single-cell RNA-seq were performed using 10x Genomics. For identifying human astrocyte subtypes, the cultured astrocytes were dissociated into single cell suspension using TrypLE. For hADSC-organs population analysis, four hADSC-organs were transferred into fresh Neural Medium and cutted into small pieces. The hADSC-organs pieces were dissociated into single cells by incubated in Papain-based solution containing Papain (20 Unit/ml, LS03126, Worthington), DNase Ⅰ (10 μg/ml, 11284932001, sigma) and L-cysteine hydrochloride monochloride (180 μg/ml, C7880, Sigma) at 37 °C for 15 min. And then, the organoid pieces were triturated using cutted 1ml tips at incubation intervals of 5 – 10 min, amounting to about 45 min incubation. The dissociated cells were then centrifuged at 500 g for 5 min, and resuspended in Neurobasal Medium containing 10 % FBS, and then were filtered by a 30 μm cell strainer. After measuring the density and viability of astrocytes or hADSC-organs dissociated cells, the final cell density was modified into 300–600 cell/μl. Dissociated single cells mixed with enzymes and the beads containing barcode information were wrapped by oil droplets and loaded onto a Chromium Single cell 3’ chip to form GEMs (Gel Bead-In-EMulsions). cDNA products were obtained after cell lysis and reverse transcription reaction in GEMs, which was then connected with Barcode. Subsequently, GEMs was broken followed by PCR amplification of cDNA templates. Agilent 4200 were used for qualification of amplification products and single-cell RNA-seq libraries were prepared with the 10x Genomics Chromium Single Cell 3’ Library and Gel Bead Kit v3. Then, via the Illumina NovaSeq6000 the sequencing was performed to obtain paired-end 150 bp reads. Obtained reads were aligned to the human reference genome (GRCh38) using Cell Ranger. Then, filtering barcode and UMI counting were conducted. After counting reads for each feature in each cell, R package Seurat was applied for further analysis.

Genes that were not expressed in at least ten cells and cells expressed with more than 6000 and less than 1000 detected genes were excluded, as well as those with mitochondrial content more than 10%. And then, gene expression of cells was normalized using a global-scaling normalization methord (normalization.method = ‘‘LogNormalize,’’) and scaled using R package Seurat (scale.factor = 10000). To do the principle component analysis (PCA), the top 1,500 high variable genes were selected and scaled. The top 20 principal components were utilized to construct the nearest-neighbor graph using the FindNeighbors and clusters cells were identified by function FindClusters with a resolution of 0.8 in Seurat. Finally, we classfied cell clusters of human astrocyte and hADSC-organs based on the expression of known markers of astrocyte or spinal cord ^5,45,46.

Statistical analysis and Reproducibility.

Data are presented as mean ± SEM. Unpaired two-tail t-test, Mann-Whitney t-test, and Ordinary one-way ANOVA using GraphPad Prism software version 7.0 were used to determine the statistical significance. Significance level as *p < 0.05, **p < 0.01, ***p < 0.001. All the Statistical tests and biological replicates are shown in the figure legends. Direct reprogramming of human astrocytes by Op53-CSBRY is quite reproducible with in total over 25 independent experiments in this study, each of which can generate about 50 cortical organoids and 80 spinal-cord organoids.

Extended Data

ED_Fig.1 |. Characterization of hAD-Organs.

a, Phase images of Op53-CSBRY induced rosette-like clusters and the expanded red box (left) showing zoom-in images (right) of rosette-like clusters at 14 days. Scale bars, left: 400 μm, right: 100 μm. b, Phase image of hAD-Organs in suspension culture for over 15 weeks. c - d, Representative images of sections from hAD-Organs showing neuroepithelium-like structures at week 7. Showing neural progenitor marker SOX2 and PAX6, and neuronal marker MAP2. Scale bar, 250 μm. e and f, Immunostaining of PAX6+, CTIP2+ and TBR1+ and Reelin+ cells located in different cortical layers in 10-week-old hAD-Organs. Scale bar, 50 μm. g, Representive image of hGFAP::GFP and brightfield in 3-week-old cortical organoid (n = 3). Scale bar, 500 μm. h - k, Showing expression of MAP2, S100ꞵ, SOX2, and FOXG1 in 3-week-old cortical organoid. Scale bars, h and i: 500 μm, j and k: 50 μm. (Representative images from three sections of organoids with similar results (c-f, h-k).

ED_Fig. 2 |. Characterization of human astrocytes.

a, Characterizing human primary astrocytes (HA1800) by immunostaining for GFAP, S100ꞵ, SOX2, PAX6 and MAP2. Scale bar, 100 μm. b, Quantitative analysis of the population of human primary astrocytes at several randomaly chosen fields per coverslip (n = 3 independent experiments). Data are presented as mean ± SEM. c, Neurosphere assay of human primary astrocytes comparing with Op53-CSBRY induced astrocytes. Scale bar, 200 μm. d, Immunostaining of primary human astrocytes with GFAP and MAP2 treated with either ISX-9 (10μM) (left), or retinoic acid (RA, 100nM) for 7 days respectively. Scale bars, 50 μm. e, Immunostaining for MAP2 after astrocytes induced by Op53-CSBRY and infected with hGFAP::GFP virus at day 7 (n = 3 independent experiments). Scale bars, 50 μm. f, Immunostaining for MAP2 in Op53-CSBRY induced astrocytes infected with hGFAP::GFP virus at day 26 (n = 3 independent experiments). Scale bar, 25 μm. g, Live cell tracking of the conversion of human astrocytes during Op53-CSBRY reprogramming. Scale bars, 25 μm.

ED_Fig. 3 |. Characterization of human astrocytes by single-cell RNA-sequence.

a, Graph showing cell percentage of expression of selected pluripotecy (ESC), neuron layer (NL), white matter (WM), upper layer astrocyte, upper-layer-biased pan-astrocyte/gray matter astrocyte and deep layer astrocyte related genes in brain cortical astrocytes. b - f, tSNE plots showing gene expression map of representative pluripotecy marker gene LIN28A, NANOG and POU5F1 (b), neuron gene RBFOX3 and DCX (c), upper layer astrocyte gene ADIPOR2, EGOT and SPRY1 (d), upper-layer-biased pan-astrocyte/gray matter astrocyte gene ITM2B, BSG and IGFBP2 (e), and deep layer astrocyte gene EFHD2, DKK3 and ID3 (f).

ED_Fig. 4 |. Characterization of hADSC-Organs.

a and b, Immunostaining images of dorsal and ventral pMN progenitor markers Nkx6.1 and Olig3 labeled with neuron markers. Scale bar, 25 μm. c, The horizontal plane image of sections from hADSC-Organs (representative images from three sections of spinal-cord organoids with similar results) stained for neural markers TUJ1 and MAP2. Scale bar, 200 μm. d, Image of spinal cord motor neuron (ChAT⁺/MAP2⁺) in 10-week-old hADSC-Organs (representative images from three sections of spinal-cord organoids with similar results). Scale bar, 100 μm. e, Quantitative analysis of neuronal subtypes of hADSC-Organs at week 10 stained for neuronal markers ChAT, HB9, GAD67 and VGLUT (n = 3 sections of organoids). f and g, Immunostaining of GFAP⁺, TUJ1⁺ and MAP2⁺ cells in dissociated 10-week-old hADSC-Organs at several randomaly chosen fields per coverslip (n = 3 independent experiments). Scale bar, 50 μm. h, Quantitative the percentage of cells in dissociated 10-week-old hADSC-Organs for MAP2, TUJ1 and GFAP positive cells at several randomaly chosen fields per coverslip (n = 3 independent experiments). i, Representative images of ChAT⁺ spinal motor neurons co-localized with MAP2⁺ cells in dissociated 10-week-old hADSC-Organs at several randomaly chosen fields per coverslip (n = 3 independent experiments). Scale bar, 50 μm. j and k, Cluster analysis of differentially expressed genes show that the gene expression profile of hADSC-Organs was significantly different from that of hAD-Organs. The colour bar represents scaled gene expression value (j). Spinal cord specific markers MNX1, ChAT, ISL1, LHX3, Nkx6.1, NKX2.2 and HOX family were significantly high expression in hADSC-Organs. (n = 3 independent experiments). The colour bar represents FPKM value (k). l, KEGG Pathway enrichment assay demonstrating up-regulated genes comparing hADSC-Organs with hAD-Organs. Data are presented as mean ± SEM.

ED_Fig. 5 |. Gene expression profile of reprogrammed astrocytes by Op53-CSBRY.

a, Sample distance matrix showing three independent biological replicates treated by CSBRY. b,Volcano plots showing the results of RNA-seq after CSBRY/DMSO control treatment of Op53. Each dot represents a single gene, including upregulated, downregulated, and filtered genes. FDR ≤ 0.05 and log₂FC > 1. c, Showing significantly different expression (DE) genes were regulated by CSBRY/DMSO control treatment of Op53. FDR ≤ 0.05 and log₂FC > 1. d, CSBRY regulated the cancer genes TP53, KLF4 and MYC (n = 3 independent experiments). Multiple unpaired Student’s t-test was used for comparing each groups. e - f, Gray scale image and quantification of Nestin protein expression in human primary astrocytes, OP53 control group at day 1, 7 and 14, and OP53-CSBRY treat group at day 7 and 14 (n = 3 independent experiments). Data are presented as mean ± SEM. Ordinary one-way ANOVA with Dunnett’s multiple comparison test was used for multiple comparisons. g, Immunostaining of pluripotent markers of NANOG, SSEA4 and TRA-1–60 in human primary astrocyte (top row) and OP53-CSBRY treated cells (bottom row) at several randomaly chosen fields per coverslip (n = 3 independent experiments). Scale bar, 50 μm. Data are presented as mean ± SEM.

ED_Fig. 6 |. Functional characterization of induced neurons from human astrocytes.

a, Example trace showing spontaneous firing in the same cell shown in Fig. 7b. b, Group data comparing the RMP at different culture stages. For group 80–100 days: Without AP, n = 18 vs. With AP, n = 14; For 100–140 days: Without AP, n = 88 vs. With AP, n = 47. Data are presented as mean ± SEM. Two-sample Student’s t-test was used for comparing with AP and without AP in each group. c, Group data comparing the RMP in cells with single AP and repetitive AP. Single AP: n = 65 vs. Repetitive AP: n = 23. Data are presented as mean ± SEM. Two-sample Student’s t-test was used for comparing. d, Group data comparing the peak amplitudes of Na⁺ currents. Single AP: n = 46 vs. Repetitive APs: n = 23. Data are presented as mean ± SEM. Mann-Whitney test was used for comparing between single and repetitive APs. e, An example current trace showing two spontaneously occurring putative postsynaptic currents (PSC). The second PSC event (arrow) was expanded for clarity (bottom). f, An example synaptic ultrastructure (red arrow) in 10-week-old hAD-Organs (n = 3 independent experiments). Scale bar, 100 nm. g, Left, merged image of IR-DIC and HB9::RFP positive of an example recorded cells isolated from 9-week-old hADSC-Organs (n = 3 independent experiments). Scale bar, 10 μm. Right, membrane potential responses (top) to intracellular injections of step currents (bottom). The blue traces indicate the responses to +50/−10 pA current pulses. h, The percentage of cells without AP (white), with single AP (blue) and repetitive AP (red) (top). The number of recorded cells in each group was provided.

ED_Fig. 7 |. Transplantation of hADSC-Organs into complete spinal cord injury in NOD-SCID mice.

a, Representative images of 7-week-old hADSC-Organs expressed high GFP after transduction with LV-Ubi-GFP virus infection (n = 25 organoids). Scale bar, 500 μm. b, The procedure of hADSC-Organs transplantation showing the site of the spinal cord amputation. c, The posterior soles of mice with complete spinal cord injury showed hindlimb paralysis with the valgus phenomenon. d, Representative images survival of GFP⁺ cells in grafted hADSC-Organs in mice (n = 3). Scale bar, 50 μm. e, Representative images survival of GFP⁺ cells in the grafted field of hADSC-Organs were human nuclei positive of hNUMA (n = 3 independent experiments). Scale bar, 50 μm. f, Open field score of grafted hADSC-Organs mice with 5 different treatments. The number of experiments (n) are summarized in Supplementary Table 1. Each treatment for 10 – 25 mice and in total 5 groups for 70 mice. Data are presented as mean ± SEM. g and h, Representative images of the spinal cord from g, the Matrigel only group and h, hADSC-Organs +MBG group and an expanded image of the outlined region (white box) where the hADSC-Organs were implanted (n = 3 mice). Scale bars, g: 250 μm, h: left, 250 μm, right, 50 μm.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41551-01X-XXXX-X.

Data availability

The data supporting the results in this study are available within the paper and its Supplementary Information. The RNA-seq data were deposited at the Gene Expression Omnibus (GEO) under https://www.ncbi.nlm.nih.gov/geo/ and the accession number is GSE168635.The Single-cell RNA-sequencing raw data generated in this study were deposited at the Genome Sequence Archive (GSA) database under https://bigd.big.ac.cn/gsa-human/browse/HRA002541 and the accession number is HRA002541.