Engineered thoracic spinal cord organoids for transplantation after spinal cord injury

Abstract

Stem-cell-based neural tissue engineering and spinal cord organoids show promises for spinal cord injury repair. However, the native spinal cord presents cell heterogeneity and a stereotypical spatial structure that makes difficult their recapitulation within an organoid architecture, which requires an assembly encompassing cellular composition, segmental organization and dorsoventral features. Here we engineer a thoracic vertebral segment-specific spinal cord organoid (enTsOrg) model that can precisely match the transplantation site, establish synaptic connections and enhance in vivo neuroelectric conduction. The organoids are generated from fibroblasts-derived induced pluripotent stem cells and a layered double-hydroxide matrix in a basement membrane hydrogel (Matrigel). Grafted in a spinal cord injury mouse model, enTsOrg presents advanced maturation, functionalization and organized distribution of critical neuronal subtypes with thoracic segmental heterogeneity, including various motor neuron and interneuron subtypes, that serve essentially to restore motor functions. Transplantation of enTsOrg can restructure neural circuits in paralysed animals and restore hind-limb motor function. The robust neurological function and therapeutic efficacy of enTsOrg highlight a potential avenue for organoid designing for specific anatomical regions in neurological injury treatments.

Main

Pluripotent and multipotent stem cells have an extraordinary ability to self-organize into three-dimensional (3D) aggregates that encapsulate the developmental stages and microenvironments of mammalian tissues and suborgans. Internal and external signals guide the differentiation and maturation of these clusters of stem cells, leading to the formation of specialized structures known as organoids that can form complex and functional analogues of corresponding tissues¹. Currently, organoids are driving the development of therapeutic strategies in medicine. It can be used not only as models for the study of disease and development but also as grafts to replace pluripotent or embryonic stem cells and can be chimerically inserted into the host for therapeutic functional remodelling¹. The principle of therapeutic intervention using human cells derived from stem cell technology is to match the cells to the anatomical site of injury as accurately as possible to favour microenvironmental integration². Recent research has also led to the progress of related organoid applications, as complex brain regions with more compartmentalized features have been modelled with organoids and implanted into a host, partially connecting with host nerves to achieve a certain degree of neural circuitry remodelling³.

Spinal cord injury (SCI) has been the focus of intense research efforts aimed at elucidating the complex cellular heterogeneities that arise after injury^4,5. The intricate microenvironment and spatially distinct cellular changes that occur following spinal cord damage have limited the therapeutic potential of organoids. A key aspect of spinal cord cell development and injury response is the formation and functionalization of diverse neuronal subtypes, which play a crucial role in limb function recovery. Recent studies have shed light on the functions of these neuronal subtypes^6,7,8. Given cellular similarity related to the spatially specific anatomical location, it is reasonable to suggest that refining organoids to enhance limb function may be best achieved by targeting regional heterogeneity and neuronal subtype specification. However, current approaches for constructing spinal cord organoids (sOrgs) lack a comprehensive framework for recapitulating thoracic segmental regional heterogeneity. Our work is dedicated to simulating the regional heterogeneity of neuron subtypes in the thoracic region with high precision, examining gene and protein expression profiles as well as the functional properties of key motor neurons (MNs).

The spinal cord contains more than 20 subclasses of neurons that form well-organized neural circuits capable of sensing the environment and generating motor behaviour⁸. In work related to the construction of cultured sOrgs, researchers have induced the differentiation of stem cells into several spinal cord-specific neuronal cells, including ISL LIM homeobox 1-positive (ISL1⁺)/choline acetyltransferase-positive (ChAT⁺) MNs, ventral spinal cord ascending commissural (V2a) interneurons and gamma-aminobutyric acid (GABA)-ergic and glycine neurons^9,10,11. In addition to facilitating the generation of diverse neuronal cells, reproducing the anatomical structures of the spinal cord is crucial in the construction of sOrgs to achieve the integration of sensory and motor functions. The grey matter of the spinal cord contains cell bodies organized into distinct neuronal populations, dorsal interneurons receive and transmit peripheral sensory signals through dorsal roots, while motor neuronal populations release motor information via ventral roots. Furthermore, distinct spinal cord segments exhibit varied structures and functions, and nerve injuries manifest divergently across different spinal cord segments⁸. Considerable efforts have been made to emulate anatomical features within sOrgs, and researchers have successfully cultured homeobox C5-positive (HOXC5⁺) and HOXC6⁺ neurons with cervical segment ventral features, as well as HOXC9⁺ neurons in thoracic segments and HOX10 high-expressing neurons in lumbar segments, in spinal cord-like organs¹².

To date, sOrgs can replicate the initiation of neural tube formation in early spinal cords. However, progress has been made in the dorsoventral (D–V), segmental and functional construction of sOrgs, and culturing regulatory factors has yielded organoids with specific characteristic cells and gene expression patterns¹¹. Nevertheless, existing sOrgs fall short of replicating the spatial distribution and functionalization of D–V neuron populations, thereby hindering the realization of the intricate and well-organized circuitry found in a native spinal cord. Furthermore, previous studies on sOrgs have focused predominantly on the de novo regeneration of MNs and the restoration of motor function in vitro. However, comprehensive structural remodelling of the spinal cord and integrated reconstruction of motor functions in transplanted organoids have not been performed. Thus, the realization of sensory and motor functions within organoids has yet to be achieved. It is also worth noting that in the context of organoid-based SCI repair, the distinctive structural characteristics of different injured segments have not been adequately addressed. The current focus on organoid construction needs to sufficiently account for the precise replication of structural features specific to different injured segments. The intricate nature of attaining the multifaceted development of a well-organized spatial distribution of functional neurons amplifies the complexity of organoid-based SCI repair¹³.

Studies have indicated that spinal cord tissues adjacent to injury sites undergo various pathological alterations; neurons proximal to the lesion undergo dynamic processes involving a range of genes related to neurotransmission and synaptic structure, and these cell populations activate molecular pathways with the potential to support recovery¹⁴. Thus, transplantation therapeutic strategies must consider the heterogeneity of cells and the environment in complex tissues, which requires a more precise match to the site of host injury⁴. For stem-cell-derived therapies for SCI, repair and restoration functions should be applied to post-developmental processes in vertebrates. Previously reported SCI therapies have involved developmentally mismatched pre-stem cells, which by default are programmed to target the brain or early-developmental neural tube, posing an additional challenge to these organ tissues; their gene expression must be readjusted to adapt to new neurophysiological signals of the spine in a complex cytokine-injury microenvironment¹⁵. The development of the in vivo regenerative potential of sOrgs following transplantation has been the subject of considerable interest recently, including the influence of cellular dynamics within the organoids on their capacity to repair SCI¹⁶. However, existing studies of sOrgs lack sequentially remodel neural heterogeneity of target the SCI for repair. To construct functional spinal organ tissues, it is necessary to not only induce the acquisition of spinal neurons expressing high levels of specific genes but also to anatomically and developmentally form spatial structures that correspond to the transplantation site with a high degree of anatomical and developmental accuracy to form precise and highly bionic circuit structures. The establishment of functional regenerative systems tailored to specific injury sites is of paramount importance. This approach represents an essential step towards developing spinal cord-like organ constructs capable of robustly repairing SCI.

In instances of thoracic (T3–T12) SCI, a constellation of symptoms encompassing radicular pain, upper extremity motor-neuron-related paralysis, and urinary and bowel dysfunctions is manifested. Furthermore, the loss of lower-extremity MNs results in paralysis, urinary incontinence and faecal incontinence¹⁷. Treating thoracic SCI is very difficult clinically. Thus far, the construction of a spinal cord-like organ tailored explicitly to the thoracic segments for treating thoracic SCI has yet to be achieved.

We introduce an engineered thoracic segment spinal cord organoid (enTsOrg) from induced pluripotent stem cells (iPSCs) using a bioactive material scaffold to generate organoid grafts that recapitulate the distribution of diverse neurons with thoracic segment heterogeneity, exhibit a well-defined neural circuit structure and display electrophysiological properties reminiscent of native spinal cord tissue. After transplanting enTsOrg into a rodent model of thoracic segment complete SCI, we observed a more pronounced enhancement in motor function, neuron subtype diversity and motor neuronal electrophysiological conductivity in the injured animals than in the non-segmentally specific sOrgs. The spatial transcriptome of the chimeric human–mouse injury and adjacent sites showed that, after transplantation, enTsOrg formed more refined functional neurons which tended to be more dorsally and ventrally characterized in terms of neuronal arrangement. In addition, upon analysing the progression of thoracic MN development, enTsOrg was found to exhibit more mature MNs after transplantation, including alpha and gamma neurons, which are valuable for muscle contraction and extension functions in paralysed animals¹⁸. Layered double hydroxide (LDH) promotes the formation of regionally specified thoracic sOrgs by activating patched 1 (PTCH1) and modulating retinoic acid (RA) signalling. Following transplantation into SCI models, LDH regulates phosphoinositide 3-kinase (PI3K)/glycogen synthase kinase 3 beta (GSK3β) and mitofusin 2 (MFN2) expression, synergistically enhancing neuronal survival and promoting the differentiation and maturation of both MNs and interneurons. This leads to improved functional integration and restoration of hind-limb motor function. Overall, we introduce a potent approach that enhances the precision and pertinence of sOrgs, offering a valuable tool for restoring SCI and revealing the intricacies of central nervous system development.

Results

Generation of enTsOrg with thoracic segmental characteristics

We present the engineering design, fabrication and application of a thoracic sOrg (Fig. 1a and Extended Data Fig. 1a). Building on our previous work, which demonstrated that the inorganic laminar material LDH can facilitate neural stem cell differentiation into neurons both in vitro and in vivo following SCI¹⁹, we investigated the potential of LDH to promote the differentiation of iPSCs towards thoracic segment-specific neurons. LDH treatment significantly increased the proportion of thoracic MNs (HOXC9⁺/ISL1⁺) in vitro (Fig. 1b). During the construction of enTsOrg, iPSC spheroids were encapsulated in LDH–Matrigel (a mixture of LDH with a basement membrane hydrogel) at neural differentiation phase. Neuroepithelial bud formation was observed after encapsulation (Fig. 1a and Supplementary Fig. 2a). The prepared LDH was a hexagonal nanomaterial with sheet-like morphology, measuring approximately 100 nm in diameter (Extended Data Fig. 1e). X-ray photoelectron spectroscopy analysis revealed elements Mg 1s (1,296.2 eV) and Fe 2p (700.2 eV) (Extended Data Fig. 1b). The characteristic peaks of LDH at 003, 006, 009 and 110 in X-ray diffraction patterns were also detected (Extended Data Fig. 1c). Modification of Matrigel with LDH resulted in a rougher surface characteristic compared with Matrigel alone (Extended Data Fig. 1e), while the protein structure resembled an α-helical conformation, akin to Matrigel (Extended Data Fig. 1d). The above results show that the LDH–Matrigel exhibited favourable physicochemical properties for neuronal growth (Supplementary Fig. 1). LDH also showed high biocompatibility for organoid formation (Supplementary Fig. 2c). Scanning electron microscopy (SEM) revealed dense neuronal cell bodies at the centre of enTsOrg, surrounded by elongated nerve fibres; sOrg exhibited a similar spatial distribution of cells but with relatively short nerve fibres (Fig. 1c and Supplementary Fig. 2b). After 21 days of incubation, the diameter of the organoids reached approximately 1.5 mm (Fig. 1a and Supplementary Fig. 2a). RNA sequencing (RNA-seq) and spatial transcriptomics sequencing (ST-seq) analyses revealed that compared with sOrg, enTsOrg exhibited an elevated degree of neuronal maturation (Fig. 1d and Supplementary Fig. 3a). On day 21, organoids showed sex-determining region Y-box transcription factor 2 (SOX2, a maker for neural progenitors) expression and a rosette structure with neurofilaments (Supplementary Fig. 3b). By day 42, mature neurons (NeuN⁺) and dense nerve fibres (NF200⁺) were abundant (Fig. 1e).

Fig. 1: Generation of enTsOrg recapitulating thoracic neural tube segment morphogenesis and firing properties.

a, Schematic diagram of sOrg generation from iPSCs with representative images of different stages of enTsOrg development. b, Representative immunohistochemistry images of thoracic segment marker (HOXC9) and MN marker (ISL1) in the control and LDH-treated 2D-iPSC induction medium on day 14 and day 21. Bar graph showing the percentage of HOXC9-positive and ISL1-positive cells in control and LDH-treated group. Two-sided Welch’s t-test was used (n = 3 biological replicates, 14 days LDH and 21 days Ctrl; n = 4 biological replicates, 14 days Ctrl and 21 days LDH). c, Representative SEM images of enTsOrg and sOrg at day 14. d, Distribution of neurons and progenitors in the enTsOrg and sOrg on day 21 determined by ST-seq with pie diagrams visualizing the proportions of these cell types. e, Representative immunohistochemistry images of neurons (NeuN) and neural fibres (NF200) in the sOrg and enTsOrg on day 21 and day 42. The magnified areas are shown in the white boxes; white arrowheads indicate NeuN⁺/NF200⁺ cells. f, t-SNE plot of the different partitions of the A–P axis in enTsOrg and sOrg on day 21 by transcriptome analysis referring to human embryonic spinal cord dataset (GSE171892), with pie diagrams visualizing the proportions, n = 2. g, Heat map of thoracic Hox gene expression in enTsOrg and sOrg according to ST-seq, n = 2. h, qPCR analysis of thoracic Hox genes in the enTsOrg and sOrg on day 21; expression levels were normalized to the sOrg group using the 2^−ΔΔCt method. Two-sided Welch’s t-test (n = 4 biological replicates). i, Representative immunohistochemistry images of HOXC9 and ISL1 in the enTsOrg and sOrg on day 21 and day 42. Bottom: bar graph showing the percentage of HOXC9⁺ and ISL1⁺ cells in the enTsOrg and sOrg. Two-sided Welch’s t-test (n = 6 biological replicates, 21 days sOrg; n = 7 biological replicates, 21 days enTsOrg; n = 11 biological replicates, 42 days sOrg; n = 16 biological replicates, 42 days enTsOrg). j, t-SNE plot of the neuron cluster expressing specific neurotransmitter marker genes in the enTsOrg group on day 21; the pie diagram shows the proportions of these cell types, n = 2. k, Heat map of the expression of genes associated with neurotransmitters in the enTsOrg and sOrg on day 21. These selected genes were used to annotate the neural subtype clusters in the t-SNE plot. l, Bright-field image of organoids on 64-electrode MEA plates (upper row) and activity heat map of the MEA recording (bottom row). m, Plots of average spike amplitude in the sOrg and enTsOrg on day 35. The shaded areas indicate the wave range. n, Representative traces of the network activities in enTsOrg and sOrg; network bursts are labelled by blue boxes. o, Quantitative analysis of spike size and spike number measured by MEA with a two-sided unpaired t-test. p,q, Characteristic curves (left) and quantitative analysis (right) of sAP (p) and iAP (q) from sOrg and enTsOrg on day 35 measured by whole-cell patch-clamp analysis. Unpaired t-test (n = 10 biological replicates, sOrg; n = 11 biological replicates, enTsOrg, in p; n = 17 biological replicates, sOrg; n = 16 biological replicates, enTsOrg, in q). r, Current-clamp characteristic recordings of organoids (left) with the current–voltage (I–V) relationship of the Na⁺ current, which are normalized for cell capacitance (right). The shaded areas indicate s.e.m., two-way ANOVA, n ≥ 2. s, Intracellular [Ca²⁺] transients of Fura-2 AM-loaded neurons in organoids. Calcium fluorescence signals are normalized by ΔF/F₀: ΔF is the transient amplitude, and F₀ is the resting fluorescence value. The bold curve indicates the average ΔF/F₀. Bar plots showing the [Ca²⁺]_i and area under the curve (AUC, 340/380 ratio) statistics. Two-sided paired t-test, n = 10. Scale bars, 50 μm in b, e and i. Data represent the mean ± s.e.m in b, h, i, o–q and s. D0, Day 0; D10, Day 10; D14, Day 14; D17, Day 17; D21, Day 21; D35, Day 35; D42, Day 42; PUR, purmorphamine; LDN, LDN-193189; CHIR, CHIR99021; BDNF, brain-derived neurotrophic factor; GDNF, glia-derived neurotrophic factor; Ctrl, control; a.u.·s, arbitrary units × second.

Source data

The LDH-engineered enhancement of thoracic MN expression was considerably more pronounced in organoid cultures, where neuronal maturation occurred over time. ST-seq analysis showed that 86.61% of spots from enTsOrg exhibited thoracic spinal cord characteristics, more than the 65.91% observed in sOrg, compared with human embryonic spinal cord data (Fig. 1f). Increased expression levels of the thoracic marker genes were shown in enTsOrg compared with sOrg (Fig. 1g), as corroborated by quantitative PCR (qPCR), immunofluorescence staining and RNA-seq analysis (Fig. 1h,i and Extended Data Fig. 6a). To restore motor function after injury, we investigated spinal MNs in organoids, focusing on choline acetyltransferase (ChAT)-positive cells, which showed substantially increased expression levels compared with organoids lacking regional heterogeneity (Fig. 1j,k). Furthermore, enTsOrg exhibited elevated expression of MN-associated genes, including synaptophysin (SYP) and ISL1, indicative of enhanced synaptic plasticity and MNs identity (Supplementary Fig. 3a). Cell state analysis during neural tube development showed that MNs were primarily present in the enTsOrg, followed by motor neuron progenitors (pMNs). A smaller population of dorsal interneuron 4 (dI4) cells, which typically emerge later and contribute to complex neural circuits, was also observed. (Supplementary Fig. 3f). Immunostaining revealed that ISL1⁺ MNs exhibited concentrated localization at the apical regions of rosette structures (Supplementary Fig. 3g). MNs and pMNs were mainly situated in ventral regions of the neural tubes, as supported by the RNA-seq results showing elevated expression of ventral neural tube-associated genes, which exhibited a distribution pattern closer to ventral characteristics along the D–V axis (Supplementary Fig. 3d,e).

In the adult spinal cord, a variety of neurotransmitters collectively contribute to motor control and sensory transmission and are necessary for establishing comprehensive neural circuits and efficient signal transduction²⁰. RNA-seq analysis revealed notable gene expression related to cholinergic neurons, while the presence of genes associated with GABAergic, glutamatergic, dopaminergic and serotonergic neurons was also evident. As indicated in Supplementary Fig. 3e, the expression of specific genes increased as the organoids developed. ST-seq of enTsOrg consistently revealed a predominant population of cholinergic neurons, followed by GABAergic, dopaminergic, glutamatergic and serotonergic neurons, and the expression of genes related to these neurons was greater than that of related genes in sOrg (Fig. 1j,k). These results indicate that enTsOrg provides a developmentally mature and comprehensive assemblage of spinal neurotransmitter-secreting neurons.

While spinal organoids mimic the characteristics of spinal cord tissue in terms of structure and cellular composition, the presence of neural electrophysiological activity is essential for their functional efficacy²¹. On day 35 of organoid growth, we performed functional assessments on neurons in situ. Multi-electrode array (MEA) technology was used to comprehensively evaluate the electrophysiological activity and network firing patterns¹⁰ of the organoids. The enTsOrg organoids exhibited a heightened spike firing frequency, indicative of enhanced neuronal excitability (Fig. 1l). The recorded spikes in enTsOrg also displayed significantly greater amplitudes than did those in their sOrg counterparts (Fig. 1m), implying more robust and potent neuronal responses. Network trace analysis further revealed substantial distinctions, with enTsOrg exhibiting a more intricately interconnected and dynamically active network than sOrg (Fig. 1n). Furthermore, the number of spikes per unit time was substantially greater in the enTsOrg than in the sOrg, reflecting an increase in the overall firing rate among the neuronal ensembles within the network (Fig. 1o). These findings provide profound insights into the augmented neuronal activity and network dynamics exhibited by enTsOrg.

According to whole-cell patch-clamp recordings, cells in the enTsOrg showed a significantly greater frequency of spontaneous action potentials (sAPs) (1.84 ± 0.44 Hz) than cells in the sOrg (0.56 ± 0.12 Hz), indicating a greater level of neuronal excitability within this group (Fig. 1p). When subjected to a 500 ms current injection, neurons in the enTsOrg exhibited a greater propensity to generate induced action potentials (iAPs) compared with the sOrg group, with 10.31 ± 1.20 times in enTsOrg and 5.12 ± 1.07 times in sOrg (Fig. 1q). In addition, we applied a ramp stimulus ranging from −80 mV to +60 mV in 1 s to the neurons in the organoids to test the voltage-sensitive currents. In addition to exhibiting both potassium currents (I_K) and sodium currents (I_Na), neurons in the enTsOrg displayed a greater peak current in I_Na (1.11 ± 0.22 nA) than did those in the sOrg (0.50 ± 0.14 nA) (Supplementary Fig. 8a). In addition, we examined the voltage-gated I_Na and I_K (Fig. 1r and Supplementary Fig. 8b), and the average peak amplitude of the I_Na in the enTsOrg was −1,004.71 pA, which was significantly greater than that of the cells in the sOrg (Fig. 1r). These results showed that neurons in the enTsOrg exhibited significantly greater excitability than neurons in the sOrg.

Calcium imaging was performed to record the changes in neuronal activity associated with changes in the intracellular Ca²⁺ concentration³ ([Ca²⁺]_i) in the organoids (Fig. 1s). Upon stimulation with KCl (60 mM), neurons in the enTsOrg displayed a significant increase in [Ca²⁺]_i, indicating a robust response to the stimulus. By contrast, neurons in the sOrg exhibited a milder response to the same stimulation. The statistical analysis of the area under the curve (340/380 ratio) further supported these findings, with enTsOrg showing a greater integrated [Ca²⁺]_i response (107.0 ± 5.60) over time than sOrg (95.45 ± 2.96). In addition, an elevated [Ca²⁺]_i was observed in the enTsOrg, further indicating heightened neuronal activity. These results collectively suggested that, compared with sOrg, enTsOrg neurons exhibit greater overall activity and a more pronounced response to external stimuli.

The above results confirmed that enTsOrg displays a distinct quasi-spinal cord architecture characterized by centrally located neuronal cell bodies radiating from peripheral nerve fibres. enTsOrg harboured multiple neurotransmitter neurons and has a developmental fate reminiscent of that of the thoracic dorsal neural tube. The enTsOrg also exhibited robust neural signalling and a strong response to external stimuli. Its importance as a central relay station for intricate neural signal transduction processes is underscored. These features of enTsOrg emphasize its potential as a central relay station for intricate neural signalling processes, offering promising prospects for ameliorating motor impairments resulting from SCI.

enTsOrg transplantation facilitates hind-limb motor function restoration after complete thoracic SCI

Given the resemblance of enTsOrg to the cellular composition and characteristics of the thoracic spinal cord, along with its robust electrophysiological activity, we conducted a study involving implantation of enTsOrg in an immunocompromised animal model of thoracic complete transection SCI to investigate its potential for restoring neurotransmission and motor function following paralysis (Fig. 2a). We determined the optimal developmental stage for organoid transplantation in a SCI model. Organoids cultured for 21 and 42 days were transplanted into mice with SCI, and hind-limb motor functional recovery was assessed. To evaluate graft survival and integration, we examined cellular apoptosis in the acute phase (1, 3, 5 and 10 days after transplantation) and neural development in the chronic phase following transplantation. Terminal deoxynucleotidyl transferase dUTP nick-end labelling assays revealed significantly increased apoptosis in 42 day organoid grafts compared with 21 day grafts during the acute phase after SCI (Extended Data Fig. 3b,c). In addition, the 21 day enTsOrg exhibited a greater abundance of stem cells (SOX2⁺) with self-renewal and differentiation potential on days 1 and 3 after transplantation. After 10 days, this group displayed a significantly enhanced occurrence of neurons (NeuN⁺) (Extended Data Fig. 3b,c). While mice receiving 42 day enTsOrg exhibited substantial, although incomplete, improvement in hind-limb motor function beginning at 4 weeks after transplantation, this was less pronounced than that observed with 21 day organoids (Extended Data Fig. 3a). At 7 weeks after transplantation, histological analysis revealed graft atrophy and impaired development of thoracic MNs in the 42 day organoid group (Extended Data Fig. 3d). Based on these findings, we selected 21 day organoids for subsequent functional studies.

Fig. 2: Transplantation of enTsOrg into thoracic segmental SCI leads to pronounced restoration of motor and neurological function.

a, Schematic diagram of the experimental design. On day 21, the organoids were transplanted into animals for in vivo functional detection within 14 weeks of surgery. b, BMS of SCI mice in each treatment group. Two-way ANOVA followed by uncorrected Fisher’s LSD, n = 20–22 biological replicates per group in 1–7 weeks and n = 4–7 biological replicates in 8–14 weeks. *P ＜ 0.05; **P ＜ 0.01; ***P ＜ 0.001. c, Stick diagrams obtained from lateral images of hind limbs during walking of each group at 14 weeks after the operation. The thighs, calves and feet are represented by blue, grey and red sticks, respectively. d, Representative image of the hind-limb position of mice during walking in the enTsOrg group 14 weeks after the operation. Bar graphs showing the stride length, stance duration, swing-to-stance ratio (the ratio of the swing phase duration to the stance phase duration of the bilateral hind limbs) and gait symmetry of the mice acquired with a digital footprint system. One-way ANOVA followed by Welch’s correction (n = 4 biological replicates). e, Schematic diagram of the horizontal ladder experiment with indication of hind-limb trajectory in sOrg and enTsOrg group while moving. f, Statistical histograms of the percentage of the total miss, slip at contact, slip at lift onset and successful position (correction, partial placement and correct placement) during the horizontal ladder experiment in each group in 14 weeks, two-sided unpaired t-test (sOrg, n = 14 biological replicates; enTsOrg, n = 14 biological replicates; sham, n = 6 biological replicates). g, Heat map of the percentage of failures and successes for each group during the horizontal ladder experiment. h, Fluorescence images of pMNs (Olig2), neuronal axons (MAP2) and MNs (ISL1) in spinal cord cross-sections of each group in 7 weeks, n = 3. The magnified areas are shown in the white boxes. i, Schematic diagram of MEP recording with representative curves and quantitative statistics for the sOrg, enTsOrg and sham groups in 7 weeks. One-way ANOVA followed by Welch’s correction, n = 7 biological replicates. j, Immunostaining results for a neurofilament marker (SMI312) and a postsynaptic dense protein (PSD95) in high-magnification images of spinal cord tissue 7 weeks (left) and 14 weeks (right) after transplantation, with bar graph showing quantification results. Two-sided unpaired t-test, n = 3 biological replicates. The magnified areas are shown in the white boxes. k, Representative immunohistochemistry images of human cytoplasmic marker (STEM121) in the sOrg and enTsOrg in 14 weeks, n = 3. Scale bars, 200 μm in h and k, and 100 μm in j. Data represent the mean ± s.e.m. in b, d, f and j. Data in i are presented as box-and-whisker plots showing the median (centre line), the 25th–75th percentiles (box limits) and the minimum and maximum values (whiskers); all individual measurements are shown as points overlaid on the plot. wpi, weeks post injury.

Source data

In stem cell-based tissue regeneration studies, graft over-proliferation or tumorigenicity are major concerns. Volumetric analysis showed that graft size remained stable from week 1 to week 14 after transplantation, with no significant expansion (Extended Data Fig. 2a,b). Long-term survival and weight gain curves showed the safety of organoid transplantation (Extended Data Fig. 2d,e), with no detectable carcinoembryonic antigen expression in transplanted or adjacent spinal cord tissues (Extended Data Fig. 2c). The biomaterial LDH has been reported to possess exceptional in vivo biocompatibility and safety profiles, establishing it as a promising candidate for applications in stem cell transplantation research²². We simulated the hypoxic microenvironment faced by the organoids after transplantation. Results showed that LDH has protective effect on cell apoptosis under hypoxic conditions (Extended Data Fig. 1f). We monitored cell apoptosis rates in organoids during the acute and long-term periods following transplantation. Initial high levels of apoptosis (weeks 1–2) were attributed to hypoxia and ischaemia, but rates declined thereafter, stabilizing at low levels by week 3. enTsOrg exhibited significantly reduced apoptosis compared with sOrg at 2 weeks after transplantation (Supplementary Fig. 5). The locomotion of model animals is affected by disrupted neural signalling following SCI (Fig. 2b and Supplementary Fig. 6). During a 14 week intervention with enTsOrg in SCI model mice, noteworthy differences in Basso mouse scale (BMS) scores were observed compared with those in the injured group at 4 weeks. Subsequent treatments showed an even more pronounced impact of enTsOrg on improving motor function than sOrg (Fig. 2b). Animals subjected to enTsOrg implantation displayed conspicuous dorsiflexion and protraction of the hind limbs during locomotion, while mice in the sOrg group exhibited a dragged hind-limb movement pattern, indicating impaired mobility (Fig. 2c). Digital footprint data were recorded and analysed to evaluate motor function recovery at 14 weeks after transplantation (Fig. 2d, Supplementary Video 1). After enTsOrg transplantation, stride length and stance duration significantly outperform the sOrg group. In addition, the swing-to-stance ratio of the enTsOrg group exhibited a smaller difference compared with the sham group than the sOrg group, and gait symmetry more closely approximates to 1, indicating enhanced coordination and stability in autonomous motor function²³. We also assessed the hind-limb motor coordination of the animals using the horizontal ladder test (Fig. 2e). In animals subjected to sOrg transplantation, the most frequently deficit observed was foot-slip at contact, followed by total miss and slip at lift onset, while after enTsOrg transplantation, the animals exhibited a significantly reduced frequency of slip at contact and demonstrated a markedly higher incidence of correction and partial placement of the paw on the rung (Fig. 2f). Overall, the animals in the enTsOrg transplantation group demonstrated a higher frequency of successful outcomes (0.36), including corrections, partial placements and accurate placements, compared with those in the sOrg transplantation group (0.13) (Fig. 2g). These findings highlight the superior efficacy of enTsOrg transplantation in enhancing the motor coordination of the animals.

Considerable changes in the expression of neuron-related proteins (NeuN, βIII-TUB and GAP43), especially MN-associated proteins (ChAT and HB9), were observed in the post-transplantation enTsOrg group (Supplementary Fig. 10c,d). These findings provide evidence of enhanced maturity in neural functionality. Cross-sectional analysis of the transplantation region revealed high densities of pMNs (Olig2⁺) and MNs (ILS1⁺), as well as dense neuronal axons labelled with microtubule-associated protein 2 (MAP2) (Fig. 2h). The establishment of functional neural connections between the host and the graft is crucial in the application of organoid transplantation for spinal cord regeneration⁴. Motor-evoked potential (MEP) was used to assess motor conduction integrity within the spinal cord. Following enTsOrg implantation, the amplitude in the enTsOrg group was 26.70 ± 8.46 μV, which was significantly greater than that in the sOrg group (9.62 ± 2.38 μV) (Fig. 2i). Our examination revealed the presence of dense neural fibres within the transplants, while several bundles of nerve fibres were observed to traverse the boundary region of the injury and penetrate the host tissue 7 weeks after transplantation, and no significant boundary was observed 14 weeks after transplantation. These observations not only affirm the robust survival of the transplanted organoids but also establish the formation of neural structural connections with the host. The enTsOrg animals also exhibited significantly elevated levels of postsynaptic density protein 95 (PSD95) compared with those in the sOrg group, both within the transplants and at the host interface (Fig. 2j). To ascertain the contribution of transplanted organoids to the reconstitution of impaired neural circuits, we used stem employing marker for human cytoplasm (STEM121) to track the cytoplasm of human cells (Fig. 2k). After enTsOrg transplantation, clear fibrous morphological fluorescence signal was detected in the host’s posterior spinal cord, suggesting a connection between the graft and the host’s neural tissue. By contrast, no significant continuous signal was detected in the host at the rear end of the graft after sOrg transplantation. To further validate structural and functional connectivity between enTsOrg and the host spinal cord, we performed anterograde and retrograde neural tracing. We used biotinylated dextran amine (BDA) as an anterograde tracer, injected into the motor cortex to map projections to the grafts and host, and pseudorabies virus (PRV) as a retrograde tracer, injected into the tibialis anterior muscle to map projections from the grafts and host. In enTsOrg-transplanted animals, BDA and PRV signals were detected both within the grafts and in the host posterior and anterior spinal cord, respectively. By contrast, sOrg-transplanted animals exhibited PRV signals only within the grafts, with no BDA signal detected. (Extended Data Fig. 4a). These findings provide evidence that enTsOrg serves as a bridge to the injured spinal cord, facilitating neural circuit reconstruction and promoting signal conduction.

In addition to motor function, we also evaluated whether the grafts contributed to sensory function transmission in the SCI animals. The recovery of sensory function in mice was evaluated using the von Frey test and the hot plate test. Compared with the SCI group and the sOrg group, the enTsOrg group exhibited a more sensitive response to mechanical stimuli (Extended Data Fig. 5a), and the latency of response to thermal stimuli was significantly reduced (Extended Data Fig. 5b), indicating the recovery of sensory function following the transplantation of enTsOrg. In addition, ST-seq analysis revealed the high expression levels of precursor markers for sensory neurons (NEUROG1, NEUROG2 and NEUROD1) as well as the sensory neuron marker (ELAVL3)⁶ in enTsOrg compared with sOrg (Extended Data Fig. 5c). Immunofluorescence results revealed the presence of the Mrgprd (Mas-related G protein-coupled receptor D)-expressing nociceptors in the organoids at 7 weeks after transplantation, with a notable increase observed by 14 weeks. In addition, elevated levels of Mrgprd signalling were detected in the enTsOrg compared with sOrg at the same stage of transplantation (Extended Data Fig. 5d).

Enhanced neuron maturation in engrafted organoids

We further investigated the cellular interactions between the transplanted organoids and the host. At 7 weeks after organoid transplantation, we collected spinal cord tissue samples for ST-seq analysis to unravel the cellular diversity and developmental dynamics of enTsOrg and sOrg, while the electrophysiological functionality of the grafts was evaluated at 14 weeks after transplantation (Fig. 3a). During distinct phases after transplantation, we procured samples to evaluate the expression profiles of the neural progenitor cell (NPC) markers Nestin and SOX2, and the mature neuronal marker NeuN by immunostaining (Fig. 3b). The data elucidated that by 3 weeks after transplantation, the cells within enTsOrg showed a pronounced upregulation of Nestin and SOX2 (Fig. 3c). 5-Bromo-2′-deoxyuridine assay results showed that this was due to the rapid proliferation of NPCs expressing Nestin⁺ and SOX2⁺ cells between 1 and 3 weeks (Supplementary Fig. 4). Subsequently, at the 5 week mark, there was a significant surge in the expression of mature neuronal differentiation compared with the sOrg grafts, which was sustained at an elevated expression plateau (Fig. 3b,c). The ST-seq results based on tissue distribution revealed the considerable presence of neuron progenitors in the central region of the graft, distant from the host, in the sOrg group. By contrast, enTsOrg exhibited a greater degree of cellular maturity, with 67.76% of the spots identified as mature neurons and only 32.24% identified as progenitors (Fig. 3d,e). Moreover, the expression of genes related to mature neurons, including those involved in neurite formation (TUBB3), axon formation (MAP2) and synapse development (SYP, SYN1), was notably greater in the enTsOrg population (Fig. 3f). The above results show a greater level of neuronal maturity in the enTsOrg following transplantation. By comparing the grafts to the developmental stages of human spinal cord tissue from Carnegie stages (CS) 12 to 19 (ref. ⁶), we identified that the predominant cell populations in sOrg were similar to those at CS12, which corresponds to the fourth week of human embryonic development. In enTsOrg, the dominant cell clusters resemble those at CS14, equating to the fifth week of human embryonic development (Fig. 3g,h). The aforementioned findings indicate that enTsOrg shows an accelerated developmental progression relative to sOrg in vivo.

Fig. 3: Advanced maturation of neuron in enTsOrg revealed by spatial transcriptomic and in vivo electrophysiological analysis.

a, Schematic diagram of the experimental design. ST-seq at 7 weeks after transplantation and in vivo electrophysiological experiments at 14 weeks after transplantation. b, Representative immunofluorescence timeline of mature neuron marker (NeuN) and NPC marker (SOX2, Nestin) in the transplanted-organoids. c, Statistic analysis of NeuN-positive cells and SOX2 and Nestin-positive cells in sOrg and enTsOrg at different time points after transplantation. The folded line showing the trend of the population of cells. Two-way ANOVA followed by uncorrected Fisher’s LSD (n = 3 biological replicates per group). d, Spot plots of neuron and progenitor distributions in organoid transplant areas by ST-seq at 7 weeks after transplantation. e, Pie diagrams of the neuron and progenitor percentage in d. f, Violin plots of progenitor and neuron-related gene expression levels by ST-seq. g, Distribution of predicted developmental stages CS12, CS14, CS17 and CS19 cell types in the organoid-transplanted area of the spinal cord tissues at 7 weeks, by transcriptome analysis referring to human embryonic spinal cord dataset (GSE171892). h, Bar plots showing the spot numbers with CS12, CS14, CS17 and CS19 identification in transplanted area. i, Distribution of excitatory neurons and inhibitory neurons in the organoid-transplanted area of the spinal cord tissues. Bar plots showing the spot numbers with excitatory and inhibitory neuron identification, n = 2 biological replicates. j, Schematic diagram of in vivo electrophysiologic signal collection. k, Histograms showing distributions, mean firing rate (z-score) of all units in sOrg and enTsOrg at 14 weeks after transplantation; each point indicates a unit. Two-sided Kolmogorov–Smirnov test (record from 6 animals in sOrg and enTsOrg group). l, Neurons distribution based on waveform properties in each group; each symbol represents one unit. m, Representative waveforms of two discrete neuron populations: P-IN and P-PN, according to waveform properties analysis by k-means. n, Proportion of P-PN and P-IN in organoid grafts of sOrg and enTsOrg at 14 weeks (record from 6 animals in sOrg and enTsOrg group). o, Mean firing rate (z-score) quantification of the P-IN and P-PN in the organoid grafts; each point indicates a unit, and black lines indicate the mean value. Kolmogorov–Smirnov test, n > 3. Scale bars, 100 μm in b. Data represent the mean ± s.e.m. in c. Data in k are presented as box-and-whisker plots showing the median (centre line), the 25th–75th percentiles (box limits) and the minimum and maximum values (whiskers); all individual measurements are shown as points overlaid on the plot. CS, Carnegie stages.

Source data

The interaction between excitatory and inhibitory neurons in the spinal cord forms the foundation of neural signal transmission. Their coordinated activity enables the precise control of muscle activity and movement by the nervous system, facilitating balanced, coordinated and refined motor control. In enTsOrg, excitatory (SLC17A6) and inhibitory (GAD1) neurons were widely distributed and extend into the host tissue. Conversely, in sOrg, the number of functionally mature neurons was remarkably fewer, and the distribution was more localized in the centre of the graft. Inhibitory neuron proportions in enTsOrg were more physiologically relevant, suggested more coordinated signalling communication with host neurons (Fig. 3i).

In vivo electrophysiological investigations were conducted at weeks 7 and 14 in organoid graft to evaluate the maturation of cellular electrophysiological functions in engineered enTsOrg within an in vivo environment (Fig. 3j). The results showed that enTsOrg grafts exhibit more active neural firing than sOrg grafts and also showed a greater degree of similarity to the sham group (Fig. 3k and Supplementary Fig. 8c,d,g). Based on the characterization of extracellular spike waveform classification results, engineered organoids were able to exhibit putative excitatory principal neurons (P-PN) and inhibitory interneurons (P-IN), characterized as broad spike waveforms with low-rate regular firing patterns and thin spike waveforms with sustained high-frequency firing, respectively (Fig. 3m–o and Supplementary Fig. 8f)²⁴. Repolarization time and trough peak latency analysis illustrated that the distribution of neurons corresponded to the k-means clustering outcomes, with discrete clustering of the two cell types, P-PN and P-IN (Fig. 3l and Supplementary Fig. 8e). Signal acquisition for the sOrg group was challenging, and the resulting classification also lacked clear narrow-spiking units at both 7 and 14 weeks after transplantation; this was consistent with the classification results obtained from the ST-seq data.

Cellular composition diversification and functional maturation in engrafted organoids

Based on the gene markers for major neurotransmitter classes in the spinal cord²⁰, we classified the cells of the organoids into cholinergic neurons (ChAT, AChE, SLC5A7), GABAergic neurons (GAD1, GAD2, SLC32A1), glutamatergic neurons (SLC17A6, SLC17A7, SLC17A8), dopaminergic neurons (TH, DDC), serotonergic neurons (TPH1, TPH2) and neuron progenitor cells. Uniform manifold approximation and projection (UMAP) analysis revealed that the transplanted organoids developed diverse neurotransmitter-class neurons (Fig. 4a and Supplementary Fig. 9b). The analysis of neurotransmitter classifications revealed a notable disparity between enTsOrg and sOrg. enTsOrg exhibited a remarkable abundance of specialized functional neurotransmitter neurons, whereas sOrg comprised a substantial proportion of immature neuronal precursors (Fig. 4a). After determining the composition of mature neurons, we found that enTsOrg and sOrg showed similar quantities of cholinergic and glutamatergic neurons, both of which are prominent neurotransmitter classes in mature neural populations. Furthermore, compared with sOrg, enTsOrg promoted a greater proportion of GABAergic and dopaminergic neurons. These findings elucidated the distinct neurodevelopmental profiles of enTsOrg and sOrg, underscoring the potential of using enTsOrg for advancing neural regeneration strategies. In addition to the differences in proportions, these neurotransmitter-related genes exhibited notable variations in expression (Fig. 4b). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed that, compared with sOrg versus iPSCs, enTsOrg versus iPSCs exhibited considerably enriched signalling pathways associated with neural development and functional maturation (Supplementary Fig. 10a). These pathways included axon guidance, the sphingolipid signalling pathway, the synaptic vesicle cycle and the generation of several functional neurotransmitter synapses (GABAergic synapses, dopaminergic synapses and glutamatergic synapses).

Fig. 4: enTsOrg grafts show the composition and functional maturation of principal neurons associated with motor function.

a, UMAP analysis of projections of neurotransmitter-classified neurons with pie diagrams showing the subtype proportions by ST-seq at 7 weeks; n = 2 biological replicates. b, Heat map of neurotransmitter-related gene expression in the enTsOrg and sOrg groups at 7 weeks by ST-seq. c, Representative immunofluorescence images of interneurons (Calb2), cholinergic neurons (ChAT), GABAergic neurons (GAD1) and glutaminergic neurons (vGlut2) in the enTsOrg and sOrg groups at 7 weeks. The magnified areas are shown in the white boxes; white arrowheads indicate signal-positive cells. d, Heat map showing the qualification analysis of Calb2⁺, ChAT⁺, GAD1⁺ and vGlut2⁺ cells in organoid grafts of sOrg and enTsOrg by immunostaining. e, Representative immunofluorescence timeline showing the changes of cholinergic neurons (ChAT⁺) number in organoid grafts after in vivo transplantation; n ≥ 3. f, Statistic analysis of ChAT-positive cells in sOrg and enTsOrg at different time points. The folded line showing the trend of the population of cells. Two-way ANOVA followed by uncorrected Fisher’s LSD (3 wpi, n = 3 biological replicates; 5 wpi, n = 4 biological replicates; 7 wpi, n = 6 biological replicates; 14 wpi, n = 5 biological replicates). g, Schematic diagram of the in vivo optogenetics experimental set-up with representative immunofluorescence image showing the co-localization of virus fluorescent signals HNA, and ChAT-positive cholinergic neurons in the transplanted area of enTsOrg group. h, Effect of optogenetic inhibition of labelled ChAT⁺ neurons on the relative firing frequency (spike count/total count) during a 200 s light on (590 nm, 10 mA, 100% duty cycle, yellow shaded) or light off period in the enTsOrg and sOrg groups at 14 weeks (n = 3 biological replicates). i, Peri-event raster plots and histogram of an example unit in a 5 s light on (590 nm, 10 mA). j, Peri-event spectrogram analysis of an example unit in a 200 s light on (590 nm, 10 mA, 50% duty cycle). k,l, Representative MEP waves (k) and amplitude statistics (l) of each group before, during and after optogenetic inhibition (14 weeks); one-way ANOVA followed by Welch’s correction, n = 4 biological replicates. Scale bars, 100 μm in c and e, and 50 μm in g. Data represent the mean ± s.e.m. in f. Data in l are presented as box-and-whisker plots showing the median (centre line), the 25th–75th percentiles (box limits) and the minimum and maximum values (whiskers); all individual measurements are shown as points overlaid on the plot. Calb2, calbindin 2; GAD1, glutamate decarboxylase 1; vGlut2, vesicular glutamate transporter 2; HNA, human nuclear antigen.

Source data

Immunofluorescence labelling was performed to validate the neuronal cell diversity and proportions of different neuronal subtypes in the transplanted sOrgs. Spinal cord tissue harbours a diverse array of neurotransmitter-secreting neuron types within its interneuron population. A substantial population of cholinergic neurons (ChAT⁺ and NeuN⁺) is prominently observed within the graft area (Fig. 4c and Supplementary Fig. 10b). Quantitative analysis revealed that mature cholinergic neurons accounted for 18.28% of the total neuronal population in the sOrg group. The enTsOrg group exhibited a substantially higher proportion, with cholinergic neurons comprising 32.94% of the total neuronal population (Fig. 4d). In addition to cholinergic neurons, a diverse population of neurons, including GABAergic (GAD1⁺), glutamatergic (vGlut2⁺) and dopaminergic (TH⁺) neurons, was observed within the grafts (Fig. 4c and Supplementary Fig. 9c). Given the pivotal role of cholinergic neurons in the coordinated control of motor functions and the transmission of neuronal excitability, we performed immunostaining on integrated spinal cord tissues at 3, 5, 7 and 14 weeks after transplantation to analyse the developmental dynamics of cholinergic neurons. We found that while a relatively enriched population of cholinergic neurons could be cultivated under in vitro conditions, the survival of these neurons was compromised in the immediate period after transplant (within 3 weeks) due to the harsh microenvironment. However, over time, the cholinergic neurons within the graft exhibited a progressive increase, with a significant elevation in numbers at 7 and 14 weeks after transplantation in enTsOrg compared with sOrg (Fig. 4e,f).

We used a combination of in vivo electrophysiological and optogenetic approaches both 7 weeks and 14 weeks after transplantation to verify whether cholinergic neurons in engineered organoids function as spinal MNs in the host (Fig. 4g and Extended Data Fig. 4b). In vivo electrophysiological recordings indicated that those categorized as putative principal neurons were presumed to be MNs in the spinal cord, whereas P-INs are generally considered to be GABAergic neurons. Consequently, we used optogenetic inhibition of ChAT⁺ neurons to observe changes in P-PN firing. The results showed that the virus was a good label for the ChAT-positive cholinergic neurons (Fig. 4g and Extended Data Fig. 4c), and the firing of P-PN in enTsOrg showed a notable decline following 200 s of continuous photoinhibition, whereas no significant change was observed in sOrg grafts (Fig. 4h and Extended Data Fig. 4d). Figure 4i and Extended Data Fig. 4e illustrate the typical P-PN neuron, which exhibited a notable decline in raster firing following 20 rounds of 5 s photoinhibition. Spectrogram analysis further revealed a substantial reduction in frequency (Fig. 4j and Extended Data Fig. 4f). In addition, the MEP showed a considerable decrease in amplitude during photoinhibition (Fig. 4k,l and Extended Data Fig. 4g,h). Trajectory analysis of mouse behaviour in the open field revealed that optogenetic inhibition reduced locomotor activity, with a more pronounced effect in the enTsOrg group (Extended Data Fig. 4g,i). The above effects of optogenetic inhibition correlate with the observed differences in the proportion of cholinergic neurons between the two groups of grafts, as evidenced by the ST-seq results.

Transplanted enTsOrg at the injury site presents corresponding segmental and D–V patterning

We examined the ability of transplanted tissues to replicate region-specific patterning and heterogeneity at the recipient site; the transcriptomic sections of the organoids were clustered based on the developmental characteristics of the neural tube cells. ST-seq analysis showed that 89.35% of spots from enTsOrg exhibited thoracic spinal cord features, notably more than the 57.08% observed in sOrg (Fig. 5a) compared with human embryonic spinal cord data (GSE136719; GSE, Gene Expression Omnibus series accession number)²⁵. Transcriptomic profiling of enTsOrg organoid grafts and two other datasets (GSE171892 (ref. ⁶) and GSE188516 (ref.⁷)) from human embryonic spinal cord revealed a consistent enrichment of thoracic segment-specific spots (Extended Data Fig. 6c). The results revealed that most cells in the transplanted enTsOrg population exhibited distinct labelling characteristics indicative of thoracic segments. The cells in grafts were also clustered according to the anterior–posterior (A–P) axis gene markers²⁶, which included the cervical segment (Hoxa6, Hoxc6), thoracic segment (Hoxb8, Hoxa9, Hoxc9, Hoxd9), lumbar segment (Hoxa10, Hoxc10) and sacral segment (Hoxc12, Hoxa13, Hoxd13), and the genes related to thoracic segment exhibited high expression (Fig. 5b,c), consistent with the in vitro generated features and matching the segments of the injured transplantation site. Reverse transcription (RT)-qPCR confirmed the greater expression of thoracic-specific genes, including Hoxb8, Hoxc8, Hoxa9, Hoxb9, Hoxc9 and Hoxd9, in the enTsOrg group than in the sOrg group (Fig. 5d). The immunostaining results provided confirmation of the elevated abundance of HOXC9⁺ MNs within the enTsOrg graft 5 weeks after transplantation, with expression levels being sustained at a relatively high level thereafter (Fig. 5e).

Fig. 5: The enTsOrg grafts show developmental patterns characterized by thoracic spinal segment and D–V features.

a, Spot plots of the distributions of the different partitions of the A–P axis (including cervical, thoracic and lumbar regions) in organoid transplant areas (7 weeks after transplantation) by transcriptome analysis referring to human embryonic spinal cord dataset (GSE136719), with a pie diagram visualizing the proportions, n = 2. b, Schematic diagram of spinal segment (upper row). UMAP plots showing the expression of A–P axis related Hox genes in the enTsOrg and sOrg (bottom row). c, Heat map of Hox gene expression determined by ST-seq. d, qPCR analysis of thoracic-specific Hox genes in spinal cord tissues at 7 weeks; expression levels were normalized to the sOrg group using the 2^−ΔΔCt method. Two-sided Welch’s t-test, n = 3 or 4 biological replicates. e, Immunofluorescence timeline of thoracic-specific marker (HOXC9), cholinergic neuron marker (ChAT) and MN marker (ISL1) in the transplanted organoids with the qualification analysis of the HOXC9⁺/ChAT⁺ or HOXC9⁺/ISL1⁺ cells proportion in total cells. Two-way ANOVA followed by uncorrected Fisher’s LSD (3 wpi, 5 wpi and 14 wpi, n = 3 biological replicates; 7 wpi, n = 5 biological replicates). f, Schematic diagram of D–V patterning of the neural tube and spinal cord. g, UMAP plots of progenitor and neural domains in D–V patterning of the neural tube and spinal cord, with a pie diagram visualizing the proportion of these cell types by ST-seq at 7 weeks. h, Bubble plots of D–V patterning-specific gene marker expression of sOrg and enTsOrg. The bubble size indicates the mean scaled gene expression level. i, Violin plot of the expression levels of D–V patterning-specific genes in different D–V regions of enTsOrg determined by ST-seq. j–l, Immunofluorescence images of ventral-specific markers (NKX2.2 (j) and FoxA2 (k)) and a dorsal-specific marker (Pax2 (l)) in the enTsOrg and sOrg groups at 7 weeks. Heat maps showing the distribution and expression levels of related markers. The magnified areas are shown in the white boxes; white dotted lines separate the dorsal and ventral sides. Scale bars, 100 μm in e, and 200 μm in j, k and l. Data represent the mean ± s.e.m. in d and e. RP, roof plate; dI2, dorsal interneuron 2; dI4, dorsal interneuron 4; FP, floor plate; HOXC9, homeobox C9; Pax2, paired box 2.

Source data

Clustering based on the D–V axis revealed cell populations exhibiting features of the floor plate, MN, roof plate, dI4 and ventral interneuron (p3) subtypes (Fig. 5f). Among these, the enTsOrg group had the highest abundance of spots with MN features, accounting for approximately 53%, while the sOrg group had only approximately 13% of spots with MN features. However, the nonmature subpopulations, including roof plate, floor plate, p3 and pMN, were found to be predominant in the sOrg group (Fig. 5g). The bubble plots illustrated the expression of progenitor-specific and neuron-specific gene markers along the D–V axis in the spinal cord tissue of the enTsOrg and sOrg groups (Fig. 5h). The violin plots depict the variations in the expression of genes associated with the D–V axis in different regions of enTsOrg (Fig. 5i). These results highlight the elevated expression of gene in the ventral region, particularly the MN region in the enTsOrg. Furthermore, distinct specialization is observed in the dorsal region, including the dorsal progenitor cell type 2 and dI4. These dorsal interneurons typically undergo development in the neural tube and mature after MN²⁷. The high expression of these associated genes shows the presence of simulated spinal cord-like neural diversity within the enTsOrg population and reflected a greater tissue maturity level than that of the sOrg population.

To further investigate the spatial distribution characteristics of the partition-specific genes along the D–V axis after transplantation, we performed immunofluorescence staining on cross-sections of the spinal cord tissue. According to the immunofluorescence results, NK2 homeobox 2 (NKX2.2) was relatively densely expressed in the transplanted sOrgs. It is expressed primarily in the ventral neural tube and is involved in specifying the subtype identity of MNs and interneurons¹². This contributes to their functional diversity and connectivity within the neural circuitry, which is crucial for the proper development and organization of ventral neuronal populations. The expression of NKX2.2 was primarily concentrated on the ventral sides of the grafts, resembling the physiological developmental pattern (Fig. 5j). Similarly, in the enTsOrg grafts, there was greater expression of forkhead box A2 (FoxA2) on the ventral side, and the overall expression level was greater than that in the sOrg grafts (Fig. 5k). FoxA2 is a transcription factor that is involved in neural tube development, is expressed in the floor plate and is involved in regulating the development of different neuronal subtypes, including MNs and interneurons, which are essential for the establishment of proper neural circuitry and motor function¹². In addition to the region-specific expression of genes in the ventral region of the neural tube, the expression of the dorsal-specific gene Pax2 (paired box 2) has been observed in grafts (Fig. 5l). Pax2 is a transcription factor that provides guidance for developing dorsal-specific cell types within the neural tube, including the generation of dorsal interneurons and sensory neurons, which are crucial for coordinating reflexes and sensory processing¹². Pax2 expression in the enTsOrg population was strongly localized to the dorsal region of the graft, whereas Pax2 expression was disorganized in the sOrg population.

Taken together, these findings indicate that the gene expression patterns of enTsOrg were more indicative of thoracic region characteristics following transplantation. Moreover, enTsOrg has acquired mature D–V axis developmental features. In addition, the expression of these D–V axis-specific genes in the enTsOrg group even showed spatial distribution characteristics like those observed in the neural tube.

Integration of enTsOrg transplants promotes the formation of specialized neurons for spinal cord repair

Within the spinal cord tissue, a diverse array of MN subtypes exists, each occupying distinct spinal cord segments and governing the intricate motor activities of various muscles. Specific populations of thoracic MNs, known as MN columns, can be further classified into subgroups, including the preganglionic motor column (PGC), hypaxial motor column (HMC) and median motor column (MMC), which can be further divided into the MMCm and MMCl subgroups²⁸. The PGC subtype is highly expressed in the phrenic nerve of the thoracic spinal cord. The HMC subtype is in the ventrolateral region of the thoracic spinal cord and plays a crucial role in maintaining balance and controlling movement. The MMC subtype controls the movements of back muscles, contributing substantially to maintaining body posture and enabling back movements. These MN column subtypes exhibit region-specific expression, with pronounced expression in the thoracic segment of the spinal cord. Analysis of the specific marker genes associated with these MN column subtypes revealed greater expression in the enTsOrg than in the sOrg. This observation was supported by both the heat map and UMAP results, underscoring the heightened expression of these genes in enTsOrg (Fig. 6a). In addition to the prominent upregulation of thoracic-specific Hox family genes, the expression level of neurotensin (Nts) in enTsOrg was markedly greater than that in sOrg (Supplementary Fig. 9a). Nts emerged as a gene displaying substantially heightened expression, specifically within the thoracic MNs. This finding suggested a greater abundance of thoracic MNs in enTsOrg.

Fig. 6: Transplanted enTsOrg promote the generation and functionalization of critical neurons that restore motor function.

a, Diagram showing the distribution of MN subtypes in the spinal cord thoracic segment with a heat map showing the expression levels of related genes in the enTsOrg and sOrg groups at 7 weeks determined by ST-seq; n = 2 biological replicates. UMAP plots showing the expression of the genes in the MN subtypes (PGC, HMC, MMCm and MNCl). b, Enrichment analysis of the KEGG pathways of the MNs in the sOrg versus enTsOrg. c, Projection of enTsOrg and sOrg onto UMAP clustering of the MN pool, alpha MNs and gamma MNs with heat maps showing associated gene expression levels by ST-seq. d, UMAP plots and heat maps showing the Vsx2 expression level before (upper row) and after (bottom row) organoid transplantation according to ST-seq. e, Immunofluorescence analysis of specific interneurons (Vsx2⁺) in longitudinal and cross sections of spinal cord tissues in 7 and 14 weeks. The magnified areas are shown in the white boxes; white arrowheads indicate Vsx2⁺/vGlut2⁺ cells, n = 3. f, Villon plot of PTCH1, SHH and CYP26A1 expression in organoids before transplantation (21 days) detected by ST-seq (n = 2). g, qPCR analysis of SHH and RA pathway-related genes in enTsOrg and sOrg before transplantation (21 days); expression levels were normalized to the sOrg group using the 2^−ΔΔCt method. Two-sided unpaired t-test, n = 4. h, Western blot results of the protein expression of SHH pathway members with the bar graphs representing quantitative analysis in organoids before transplantation (21 days). i, Representative immunofluorescence images of PTCH1/NKX6.1 and CYP26A1/βIII TUB in organoids before transplantation, with bar graph showing quantification results (right). Two-sided unpaired t-test (PTCH1 and NKX6.1: sOrg group, n = 3; enTsOrg group, n = 4; CYP26A1 and βIII TUB: n = 7). The magnified areas are shown in the white boxes. j, Heat map of SHH pathway-related gene expression determined by ST-seq at 7 weeks after transplantation (n = 2). k, Western blot results of the protein expression of PTCH1, GLI1 and NKX6.1 with the bar graphs representing quantitative analysis in organoids 7 weeks after transplantation. l, Representative immunofluorescence images of CRABP2 and CYP26A1 expression 14 weeks in organoids after transplantation, with bar graph showing quantification results (below). Two-sided unpaired t-test, n = 4. m, Western blot analysis of p-PI3K, p-AKT and p-GSK3β expression in spinal cord at 7 weeks after injury, with the bar graphs representing quantitative analysis. n, Representative immunofluorescence images of p-PI3K, p-AKT and p-GSK3β expression in neurons of spinal cord tissues 14 weeks after transplantation, with quantification results shown below. Two-sided unpaired t-test (p-PI3K, n = 3; p-AKT and p-GSK3β, n = 4). The magnified areas are shown in the white boxes. Scale bars, 200 μm in e, and 100 μm in i, l and n. Data represent the mean ± s.e.m. in g, i, l and n). MMCm, median motor medial subcolumns; PTCH1, Patched 1; PI3K, phosphatidylinositol 3-kinase; AKT, AKT serine/threonine kinase; GSK3β, glycogen synthase kinase 3 beta.

Source data

We performed enrichment analysis of KEGG pathways for the MN population in the sOrg versus enTsOrg groups (Fig. 6b). The results revealed enrichment not only in neuronal differentiation signalling pathways, such as axon guidance, but also in other pathways, including ubiquitin-mediated proteolysis and the Notch signalling pathway. These signalling pathways may play crucial roles in the subpopulation and functional specialization of MNs. In addition, MNs can be classified into distinct pools based on their functionalities, with each pool comprising a group of MNs that control similar muscles and possess specific functional characteristics. Among these, alpha MNs are primarily responsible for the main control of skeletal muscle movement, while gamma MNs function in the internal regulation of muscle fibres to maintain muscle tone and adaptability. These two MN subtypes coordinate muscle movement and ensure proper motor function²⁰. Overall, the gene expression analysis revealed notably greater expression of related genes in the enTsOrg group than in the sOrg group (Fig. 6c). Immunofluorescence validation also showed elevated expression of the MN synaptic vesicle glycoprotein (SV2) in the enTsOrg animals, indicating improved MN function (Extended Data Fig. 7j). These findings provide evidence for diverse subgroups of MNs and enhanced MN functionality.

In addition to the emergence of MN subtypes, we also observed a critical neuron subtype (Vsx2⁺/vGlut2⁺). It has been reported that these specific interneurons can regenerate and reconnect with their natural target area in the lumbar spinal cord. Furthermore, the absence of these neurons impedes the recovery of hind-limb walking ability in SCI mice²⁹. This gene exhibited similar expression levels in both enTsOrg and sOrg before transplantation. However, following transplantation, there was a substantial increase in its expression in enTsOrg, while its expression in sOrg notably decreased (Fig. 6d). Immunofluorescence validation confirmed the survival of this crucial neuronal population (Vsx2⁺/vGlut2⁺) in the enTsOrg population (Fig. 6e).

We investigated the mechanisms underlying LDH engineering in promoting the thoracic-segmental specification and MN differentiation before and after transplantation. Our previous study revealed that LDH can bind to various cell membrane receptors in a time-dependent manner, thereby regulating stem cell fate²². We observed high expression levels of PTCH1, as well as elevated expression of its downstream targets GLI family zinc finger 1 (GLI1), NK6 homeobox 1 (NKX6.1) and sonic hedgehog (SHH) in the enTsOrg before transplantation (Fig. 6f–i). This indicates that LDH facilitates MN differentiation in organoids by modulating the SHH signalling pathway. It was also observed that cellular retinoic acid binding protein 2 (CRABP2) was highly expressed in the nuclei of stem cells (Extended Data Fig. 6b) in the early developmental stage of enTsOrg (21 days in vitro). High expression of cytochrome P450 family 26 subfamily a member 1 (CYP26A1) and retinoic acid receptor alpha (RARα) was also observed in enTsOrg (Fig. 6h,i). Upregulation of the RA-degrading enzyme CYP26A1 by LDH modulates RA signalling, leading to patterned neuronal development along the A–P axis and the acquisition of thoracic segment identity in organoids. Following transplantation, the associated genes of the SHH and RA signalling pathways maintain expression patterns similar to those observed in vitro (Fig. 6j–l). Furthermore, the activation of downstream signalling pathways related to neuroprotection and functional maturation was detected. Western blot analysis revealed a notable increase in phosphorylated PI3K and AKT in the enTsOrg group at 7 weeks after transplantation, along with elevated phosphorylation of the downstream targets GSK3β (Fig. 6m). Immunostaining confirmed these findings (Fig. 6n and Extended Data Fig. 7a). The observed changes validate the increased maturity of motor neuronal functionality. We observed substantial enrichment of ubiquitin-mediated proteolysis KEGG in enTsOrg (Fig. 6b), and this enrichment may be a crucial factor enabling enTsOrg to integrate more effectively into the host and fulfil its functional role³⁰. The results confirmed the significant regulation of neural maturation and ubiquitin-related genes including UCHL1 (Extended Data Fig. 7b,c,e). Previous studies have indicated that its elevated expression can modulate MFN2 expression, thereby regulating neuronal development and axonal extension³¹. Results also revealed high expression of the MFN2 gene in the enTsOrg grafts (Extended Data Fig. 7c,d,f). However, no significant changes in the expression of the dynamin-like protein (Dlp1), which regulates mitochondrial fission, were observed (Extended Data Fig. 7d,g). VDAC1, a mitochondrial marker, exhibited high expression and a regular neuronal fibre-like arrangement in enTsOrg (Extended Data Fig. 7d,g).

The differences observed in the subtypes of neurons mentioned above can be influenced by the microenvironment at the interface with the host. The classification of spots based on the species grouping of cell populations^6,32 revealed that 15.74% of spots were human-mouse cell hybrids in the enTsOrg group compared with approximately 8.4% in the sOrg group (Supplementary Fig. 11a). The results indicated greater cellular integration between the enTsOrg transplant and the host, suggesting increased cell structural connectivity and neural communication. Moreover, a substantial population of mesoderm I and haematopoietic cells was observed in adjacent host cell clusters close to the sOrg transplants (Supplementary Fig. 11b). The presence of this barrier hindered intercellular communication and impeded the extension of nerve fibres between the transplanted cells and the host neurons. At the periphery of the organoid transplantation site, notable upregulation of neuronal populations in the host tissues of enTsOrg was observed, accompanied by the expression of regeneration-associated genes (Atf3, Sox11, Klf6, Bdnf, Adcyap1, Gap43) (Supplementary Fig. 11c). This heightened expression indicated a favourable environment for neuronal survival and developmental maturation. These findings provide further evidence that precisely matched organoids can efficiently restructure neural circuits and faithfully recapitulate the functionality of the corresponding tissues.

The above results show that enTsOrg has developed specialized neurons that match the transplantation region, including specialized MNs and interneurons. These neurons exhibit enrichment of key genes, allowing them to perform enhanced neural functions. The enTsOrg graft can establish favourable feedback regulation with the host spinal cord. Such an improved microenvironment in the host spinal cord may be crucial for better integration of enTsOrg.

Discussion

Recent advances in stem cell transplantation and central nervous system (CNS) organoid technologies offer promising avenues for SCI repair. However, effective integration and functional contribution of transplanted cells require precise control of their regional identity⁵, for example, cortical neurons typically integrate successfully only into corresponding host circuits³³. Recent studies demonstrate that transplanted spinal cord NPCs can self-organize in vitro into structures resembling spinal dorsal horn ganglia and express region-specific neural markers. Critically, these cells retain their regional identity during culture, suggesting the host spinal cord can recognize and guide targeted axonal connections. Further investigation reveals that these NPCs possess pre-determined regional fates, rather than being passively directed by the host environment. For instance, transplantation of progenitors derived from the dorsal spinal cord preferentially promotes the regeneration of host sensory and motor axons⁴. These findings underscore the importance of controlling regional heterogeneity in transplanted cells to optimize graft composition and enhance innervation of host circuits. Furthermore, studies have begun to define the potential and challenges of CNS organoid-mediated SCI repair. Transplantation of brain organoids yielded limited survival of CTIP2⁺ deep-layer neurons, underscoring the challenges of integrating organoids from disparate brain regions³⁴. Directly reprogrammed sOrgs exhibited D–V patterning and segmental Hox gene expression, indicating suitability for region‑specific repair. However, despite observing axonal regeneration and synapse formation following murine transplantation, functional recovery of SCI mice remained limited, suggesting insufficient host integration of mature neuron-based organoids¹⁰. These findings highlight the need for improved region-specific and transplantation strategies to enhance the therapeutic potential of sOrgs.

We investigated whether organoids with regionally matched characteristics recapitulate the effects of predetermining regional differentiation fate in stem cells. By comparing to multiple human transcriptome datasets^6,7,25, we found that enTsOrg exhibited a higher proportion of thoracic segment-specific transcripts compared with sOrg, increasing from 85% to approximately 90% after transplantation, while sOrg decreased from 65% to 55%. This outcome was corroborated by enriched expression of characteristic thoracic segment genes (Figs. 1f and 5a and Extended Data Fig. 6c). In vivo transplantation also amplified differences in D–V axis organization between enTsOrg and sOrg. A decrease in neuronal progenitors observed in enTsOrg after transplantation (Fig. 3d,e) mirrored the developmental decline from gestational weeks 7 to 11 (ref. ³²), coinciding with increased dI4 expression, a marker of dorsal neuron development, and suggesting a progression towards dorsal neuronal maturation (Fig. 5g). This aligns with the established developmental trajectory of the spinal cord, where dorsal neuron development follows earlier ventral specification. Spatial analysis confirmed an organized distribution of neuronal markers within enTsOrg, reflecting progressive maturation. Complete SCI repair requires neural neogenesis and circuit reconstruction at the lesion site. Transplantation of enTsOrg promoted differentiation of stem cells into mature MNs and induced changes in adjacent tissues indicative of neural repair. Compared with sOrg, enTsOrg were associated with reduced numbers of mesodermal and hematopoietic cells in the degenerative area of the mouse tissue adjacent to the injury area (Supplementary Fig. 11b), suggesting enhanced graft-host integration. We observed a more robust recovery of inhibitory interneurons within the spared but activated region slightly distal to the degenerative area of enTsOrg-transplanted mice (Supplementary Fig. 11d), accompanied by increased expression of regeneration-associated genes (Supplementary Fig. 11c), indicated of a positive recovery from SCI¹⁴. Spatial transcriptomics revealed a greater proportion of human-murine chimeric cells in the enTsOrg group (Supplementary Fig. 11a), potentially reflecting improved structural alignment and cellular communication with adjacent tissues. Transplantation of enTsOrg promoted differentiation of stem cells into mature MNs and induced changes in adjacent tissues indicative of neural repair.

Transplantation of region-matched organoids represents a potential therapeutic strategy for CNS injury, yet the optimal timing remains a key consideration. Compared with mature organoids, immature organoids displayed greater plasticity and survival after transplantation by promoting functional neuronal integration and motor recovery, and migrating to the thalamus and hippocampus to enhance synaptic connectivity and neurotrophic support. Mature organoids, however, displayed limited regenerative capacity due to a depleted stem cell population³⁵. Consistent with these findings, Early transplantation of enTsOrg (21 days in vitro) preserved stem cell potential and segmental specificity, resulting in improved motor function compared with both mature-stage (42 days in vitro) enTsOrg and sOrg. Increased neuronal apoptosis was observed in the 42 day groups attributed to depleted stem cell populations to support proliferation and differentiation (Extended Data Fig. 3b,c). Following transplantation, the stem cells within 21 day sOrg grafts failed to differentiate effectively due to a lack of regional matching (Extended Data Fig. 3d). These findings highlight the importance of maintaining stem cell potential and regional heterogeneity in organoid transplantation for maximizing therapeutic benefit.

We hypothesized that in vivo transplantation would support continued organoid differentiation, generating regionally heterogeneous and functionally mature tissues, thereby improving motor function. Transplantation of enTsOrg resulted in sustained improvements in hind-limb motor function, shown by enhanced motor scores and locomotor ability, including reduced hind-limb dragging and improved grip strength. The observed improvements in motor function are primarily attributable to continued differentiation and maturation of MNs following enTsOrg transplantation. To understand the underlying mechanisms, we investigated neuronal differentiation within the organoids, focusing on thoracic spinal cord regional heterogeneity. During development, thoracic MNs differentiate into MMC neurons innervating the dorsal axial muscles that innervate axial muscles along the entire cephalocaudal axis and specific PGC neurons derived from the same precursors as somatic motoneurons innervating the sympathetic ganglion, as well as the HMC neurons that innervate intercostal and abdominal wall muscles³⁶. After transplantation, the enTsOrg grafts expressed more mature MN pool genes that were similar to those expressed at developmental CS14, whereas the sOrg graft expression, which lacks segmental specificity, remained at the level of CS12 (Fig. 3g,h). MNs further diversify into alpha, beta, and gamma subtypes, and enTsOrg grafts exhibited more pronounced expression of both alpha and gamma MN markers than sOrg grafts, indicating advanced differentiation and established muscle innervation¹⁸, corroborated by high expression of the SV2 motor endplate (Extended Data Fig. 7j). Earlier maturation of MN subtypes in enTsOrg grafts explains the superior therapeutic outcomes observed in electrophysiological, optogenetic, and locomotor assessments, despite comparable numbers of cholinergic MNs in both enTsOrg and sOrg grafts. Mature, regionally heterogeneous MNs in enTsOrg are capable of forming functional synaptic connections with host animals, thereby facilitating precise control over a range of motor functions.

Functional recovery following SCI requires not only MN maturation but also the restoration of functional interneuronal circuits. Remodelling generates crucial Vsx2⁺/vGlut2⁺ interneurons in the mid-thoracic spinal cord, can transmit spinal supraspinal commands to the lateral half of the spinal cord, their absence compromises the restoration of hind-limb ambulation ability in paralysed mice³⁷. In vitro, both organoid types exhibited comparable numbers of MNs and Vsx2⁺/vGlut2⁺ interneurons, yet enTsOrg displayed a more functionally advanced profile. Transplantation resulted in a decrease in Vsx2⁺ cells within sOrg and a considerable increase in enTsOrg (Fig. 6d,e). enTsOrg grafts exhibited elevated Nts expression (Supplementary Fig. 9a), consistent with high expression in mid-thoracic spinal neurons, suggesting differentiation towards visceral MN subtypes. Nts expression defines preganglionic MNs and potentially contributing to motor functional recovery by modulating inhibitory interneuron activity and preventing muscle spasms. The enTsOrg grafts exhibited elevated dopaminergic neuron density, suggesting a contribution to functional recovery. We observed expression of TH and DDC in transplanted neurons (Supplementary Fig. 9b,c), coinciding with persistent activation of spinal D1 and D2 receptors following injury, potentially compensating for diminished brain input. This activation correlated with improved motor function, gait stabilization, and locomotor performance. Spinal TH⁺ neurons in reforming bladder reflex circuits following SCI, thereby modulating bladder and sphincter reflex activities. Increased dopamine neuron density following organoid transplantation potentially enhance neuronal excitability and facilitate motor command transmission by modulating spinal circuitry³⁸. After transplantation, enTsOrg expressed diverse MN and interneuron subtypes, with sequencing and electrophysiological data showing a balanced ratio of inhibitory and excitatory neurons (Fig. 4i). Remodelled neural circuits contribute not only to motor recovery but also to sensory reconstruction. The enTsOrg grafts exhibited higher expression of sensory nerve precursors and markers, and superior recovery of mechanical and thermal pain sensation, correlating with increased expression of the pain receptor Mrgprd (Extended Data Fig. 5)³⁹. These results show that organoid transplantation effectively promotes spinal cord repair through neural circuit remodelling and signalling restoration, offering a potential therapeutic strategy for paralysis.

We propose that LDH enhances the survival, differentiation, and maturation of both MNs and interneurons within transplanted organoids, potentially by modulating stem cell fate via biomaterial effects. Recent study demonstrated that MgFe-LDH promotes neural differentiation through SHH signalling by activating PTCH1. Single-cell sequencing revealed that MgFe-LDH-induced NPCs exhibit elevated expression of the MN precursor marker NKX6.1, crucial for SHH-mediated MN progenitor emergence²². Spinal cord development requires coordinated signalling pathways, notably SHH, which establishes a ventral-to-dorsal gradient governing neuronal fate. RA signalling is also critical for neural precursor differentiation and proper spinal cord development⁴⁰, but must be tightly regulated. SHH maintains and reinforces expression of Cyp26 genes, which degrade RA, preventing supraphysiological activation of RA signalling⁴¹. We found that LDH-lacking organoids exhibited increased Hox5 and Hox6 expression in vitro and displayed enhanced brachial and cervical segment characteristics (Fig. 1f). This suggests that LDH modulates RA signalling to control subtype-specific differentiation of thoracic segment MNs via Hox gene expression. Fluctuations in RA concentration can alter MN fate, increasing brachial-specific Hoxc6 expression while suppressing HOXC9⁺ thoracic segmental MN generation⁴². Coordinated regulation of CYP26A1 and CRABP2 reduces RA levels, influencing Hoxc6 and Hoxc9 expression and extending Hox expression along the A–P axis. Sustained RA receptor activation impedes differentiation of dorsal and median column neurons and induces apoptosis of thoracic segment MNs⁴³, highlighting the importance of moderate RA concentrations. Our findings show that LDH-mediated control of RA signalling is essential for establishing a finely structured thoracic neural circuit and promoting directed differentiation of stem cells to thoracic segment MNs (Fig. 6f–l).

We observe that sustained activation of the SHH pathway following transplantation modulates PI3K/GSK3β signalling and regulates CYP26A1 expression in vivo; the regulation of RA levels was sustained, thereby facilitating the optimization of neuronal differentiation and segment patterning. Synergistic activation of SHH and RA enhances GSK3β S9 phosphorylation, promoting neuronal survival and axonal regeneration⁴⁴ (Fig. 6m,n and Extended Data Fig. 7a). This process is mediated, in part, by upregulation of ubiquitin C-terminal hydrolase L1 (UCHL-1), which stimulates protein kinase B (AKT) phosphorylation and regulates mitochondrial function, reducing oxidative stress⁴⁵. Comparative analysis revealed significantly increased UCHL-1 expression in the enTsOrg group, correlating with elevated ubiquitin levels and enhanced axonal regeneration and tissue repair following SCI. UCHL-1 modulates mitochondrial function by regulating MFN2 levels and stabilizing the mitochondrial network, preventing neuronal loss⁴⁶. LDH regulates PI3K/GSK3β and MFN2 expression, synergistically enhancing neuronal survival and neurogenesis to promote functional recovery following SCI (Extended Data Fig. 7b–i).

By using engineering approaches to construct sOrgs, we recapitulated the diversity of neuronal subtypes and thoracic regional heterogeneity, thereby mimicking the organizational patterns of the thoracic spinal cord. Transplantation into SCI models resulted in functional recovery mediated by integration of these neurons into host circuitry, exhibiting enhanced thoracic segment-specific organization. Furthermore, the findings suggest that achieving an exact match with the injury site efficiently restores neural function, resulting in a substantial reorganization of neural connectivity with the host and leading to a comprehensive enhancement of locomotor function in spinal cord-injured animals. Consequently, in the establishment of developmental models, drug screening or transplantation repair, maximizing the advantages of organoids necessitates the construction of organoids with intricate structures tailored to target organs. This precise in vitro–in vivo culture model also represents a valuable strategy for using organoids in nerve repair, thereby providing opportunities for advancing the field.

Methods

Cell culture and generation of sOrgs

Wild-type fibroblast-derived iPSCs were maintained on Matrigel (354277, Corning)-coated plates using mTeSR1 media (85850, StemCell Technologies). Immunostaining for OCT4 and TRA-1-60 was performed to confirm the pluripotency of the iPSCs before organoid generation (Fig. 1a).

The organoids were generated as previously reported with some optimization⁴⁷. Specifically, dissociated iPSCs (30,000 per well) were plated in ultra-low-attachment 96-well round-bottom plates with 10 μM Y-27632 (S1049, Selleckchem) in mTeSR1. After 12 h, the medium was replaced with neuralization induction medium containing LDN-193189 (1 μM, 04-0004-10, Stemgent) and CHIR99021 (3 μM, 4423, Tocris Bioscience). On day 3, 0.1 μM RA (04-0021, Stemgent) was added to the medium to promote caudalization. On day 10, the neurospheres were encapsulated in LDH–Matrigel or Matrigel and further differentiated in medium supplemented with RA and purmorphamine (1 μM, 04-0009, Stemgent). On day 14, organoids were transferred to low-attachment 6-well plates on a shaker (65 rpm). On day 17, 10 ng ml⁻¹ brain-derived neurotrophic factor (450-02-10, PeproTech) and 10 ng ml⁻¹ glia-derived neurotrophic factor (450-10-10, PeproTech) were added to promote maturation of the organoids. The enTsOrg and sOrg were collected routinely at days 14, 21 and 42. iPSC line N12 and line N13a were applied for generation of organoids.

For 2D induction, iPSCs single cells were planted into 24-well plates with Matrigel-coated coverslips. On day 10 of induction, LDH (10 μg ml⁻¹) was added into the culture medium.

Preparation of LDH and LDH-engineered Matrigel

LDH was synthesized hydrothermally as previously reported²²: NaOH (10019718, Sinopharm) was dissolved in CO₂-free double-distilled H₂O, followed by the addition of Mg (NO₃)₂ (80075918, Sinopharm) and Fe (NO₃)₃ (80072718, Sinopharm). The mixture was stirred for 30 min at 60 °C in a nitrogen atmosphere. The resulting suspension was washed twice and then subjected to a hydrothermal reaction in a kettle at 100 °C for 16 h. The products were washed twice and then stored at 4 °C. LDH–Matrigel (10 μg ml⁻¹) was prepared by mixing LDH in PBS/Matrigel (1:9 v/v) until uniform.

Characterization of LDH, LDH-engineered Matrigel and organoids

The physicochemical characterizations of nanoparticles and LDH–Matrigel were performed as previously reported⁴⁸: X-ray diffraction (D2 Phaser, Bruker; 2θ ranging from 10° to 80°, Cu Kα1 radiation) for phase analysis of LDH; X-ray photoelectron spectroscopy (K-Alpha, Thermo Fisher) for surface elements; circular dichroism for protein secondary structure of Matrigel and LDH–Matrigel; SEM (GeminiSEM 300, Zeiss) for LDH, LDH–Matrigel and organoid micromorphology.

In vitro hypoxic microenvironment model

Six-well plates containing 21 day cultured organoids were packed into an anaerobic bag to build a hypoxic microenvironment in vitro and was placed back in the incubator for 2 h. For enTsOrg group, 10 μg ml⁻¹ of LDH was added in the culture medium before the reaction began. After completion of the reaction, groups of organoids were collected for the cell apoptosis detection.

RNA-seq and analysis

The sOrg and enTsOrg (days 14 and 21; n = 3 per group, biological replicates) were collected and transported to the Beijing Genomics Institute (BGI) for total RNA isolation and RNA-seq analysis. Briefly, the samples were subjected to RNA isolation, after which the RNA samples were subjected to library preparation. The messenger RNA was enriched using oligo (dT) magnetic beads, followed by fragmentation. The fragmented RNA was then reverse transcribed into complementary DNA, and cDNA libraries were prepared by adding sequencing adapters and performing PCR amplification. Sequencing was performed on a DNA nanoball sequencing (DNBSEQ) system (DNBSEQ-T7, PE150). Wild-type iPSCs served as controls. KEGG, Gene Ontology and differential gene expression analyses were performed and visualized on the Beijing Genomics Institute website (https://biosys.bgi.com/). P < 0.05 was considered to indicate statistical significance.

Whole-cell membrane clamp electrophysiology

Whole‑cell patch‑clamp recordings were performed on sOrg and enTsOrg in standard artificial cerebral spinal fluid using 4–6 MΩ electrodes under stereo zoom microscopy. The composition of artificial cerebral spinal fluid and the pipette solution are shown in the Supplementary Table 1. Whole-cell membrane clamp recordings were made in current-clamp mode with 10 pA step⁻¹, 500 ms in duration and were elicited every 5 s to record neuronal sAPs and iAPs. Voltage‑clamp recordings used a ramp protocol (−80 to +60 mV, 1 s) and stepwise depolarizations (−70 to +70 mV, 10 mV increments, 300 ms) to measure voltage-gated sodium currents (I_Na) and voltage-gated potassium currents (I_K). Online P/N subtractions were performed by Clampex 10.7 software (Molecular Devices). Data were recorded using an Axopatch 200B amplifier (Molecular Devices), filtered at 2 kHz, and sampled at 10 kHz for voltage clamp mode, or 20 kHz for current clamp mode, respectively. Clampfit 10.7 (Molecular Devices) together with GraphPad Prism 10 were used for the data analysis.

Calcium imaging

Organoids were collected and washed in Hank’s balanced salt solution (HBSS, C0218, Beyotime), then loaded with 5 μM Fura-2/AM (F1201, Thermo Fisher) for 30 min at 37 °C. The Fura2-AM working solution was removed, and the organoids were washed with HBSS then placed in pre-warmed organoid culture medium. The Fura-2-loaded organoids were illuminated with 340 or 380 nm excitation light by a high-speed wavelength switcher (DG4, Sutter), and the respective fluorescence images at each wavelength were collected at 1 fps (frames per second) by a high-speed charge-coupled device (CCD, ORCA-Flash4.0, Hamamatsu). All the data were acquired and analysed by Metafluor software v.7.8.0 (Molecular Devices). The ratio of the two wavelengths (340 nm/380 nm) was considered an indicator of intracellular Ca²⁺ level. ΔF/F₀ was used to normalize the intracellular calcium ion transients in the sOrg and the enTsOrg. ΔF is the transient amplitude, and the mean ratio of the first minute was calculated as the baseline (F₀).

MEA recording

The neural spike activity of the organoids was assessed using a MEA. Organoids (35 days) were placed on Matrigel-coated (1:100 in DMEM/F12, Gibco) MEA plates (M384-tMEA-6B, Axion Biosystems; 64 electrodes with 30 μm diameter and 0.04 MΩ resistance). After 1 week of incubation, neural spike recordings were obtained using the Maestro pro-MEA system and AxIS Software Spontaneous Neural Configuration (Axion Integrated Studio Navigator v.1.5, Axion BioSystems). A recording duration of 20 min was chosen for each measurement, during which the neural activity of the organoids was captured.

Animal model and organoid transplantation

All experimental procedures were approved by the Animal Welfare Committee of Tongji Hospital Affiliated to Tongji University in Shanghai (approval number 2022-DW-SB-44) and conducted in accordance with their standards, under the laboratory animal permit held by Tongji Hospital (license number SYXK (Hu) 2019-0005 and number SYXK (Hu) 2024-0004).

Female NOD-PrkdcscidIl2rgem1/Smoc (M-NSG) mice (20 ± 2 g, 7–8 weeks; Shanghai Model Organisms Center) underwent T9 laminectomy, with complete 2 mm spinal cord transection. The animals were randomly divided into five groups. In the enTsOrg group (n = 25), enTsOrgs with LDH–Matrigel droplets (21 days) were implanted into the lesion areas. In the sOrg group (n = 25), sOrgs with Matrigel droplets (21 days) were implanted into the lesion areas. In the Matrigel group (n = 25), Matrigel in gel form was implanted into the lesion areas. In the SCI group (n = 25), no treatment was administered after the operation. In the sham group (n = 25), laminectomy was applied to expose the dorsal spinal segments. Postoperative analgesia was administered using rodents-specific protocols (ibuprofen 3 mg kg⁻¹). Humane end-points were strictly observed, requiring immediate euthanasia when animals showed ≥20% weight loss or developed irreversible morbidity. The bladders of the mice were massaged two times daily, and food and water were provided ad libitum.

Kinematic assessment

The BMS open-field ratings were recorded weekly to evaluate the recovery of hind-limb movements in the model mice. The assessments were performed on a grid (1 m × 1 m × 0.5 m) for 5 min per animal by two independent observers blinded to the group assignments within 14 weeks. Left and right hind-limb movements were scored separately and averaged. The weight change was recorded weekly to assess the safety of organoid transplantation.

Gait analysis was captured by the DigiGait Imaging system (Mouse Specifics) and was performed 14 weeks after injury (4 mice per group). The target joints were marked with different coloured ink for lateral locomotor data acquisition, and the soles of the feet were marked in red for footprint information. Gait parameters data were analysed and generated by DigiGait analysis software v.16A (Mouse Specifics). The footage images captured by the high-speed camera were analysed frame by frame and simplified into stick diagrams to assess hind-limb coordination and motor function in different groups.

The horizontal ladder task was performed at 7 weeks and 14 weeks. The mice were placed in a transparent walking compartment with evenly spaced rungs. The high-speed camera was positioned on the side of the ladder to record each side of the mice while they were walking back and forth in the compartment. The horizontal ladder data were analysed by a scoring system as described previously⁴⁹.

Sensory behaviour assessment

Mechanical sensitivity was tested using von Frey filaments and a hand-held force transducer (IITC Life Sciences) on wire-mesh flooring after 30 min acclimation. Polypropylene filaments with minimal stimulus strength were applied vertically to the centre of the hind paw, applying pressure until the filaments bent into a ‘C’ or ‘S’ shape. If no positive response was observed from the hind paw, such as rapid withdrawal, retreating, hunching, licking the stimulated area, or escape behaviours, the filament with the higher intensity was used. The up–down method was used to determine the appropriate von Frey filaments for the formal experiment. Each von Frey filament was applied for approximately 5 s. Upon a positive response, the pressure meter automatically recorded the stimulus intensity as the paw withdrawal threshold. Each paw was tested 5 times, with a 20 s interval between tests.

Hot plate (YSL-21, Shanghai Yuyan Instruments) with the temperature of 50 °C (hot stimulus) was applied to assess thermal hyperaesthesia of mice hind paws. The latency period for the hind paw withdrawal in mice was determined by measuring the time from contact with the hot plate to the onset of a positive response. The retraction of the hind paws and the twitching of the hind limbs are considered as positive responses.

MEP recording

Mice were randomly chosen for MEP recording by using Keypoint II dual-channel evoked potential/electromyography (AlphaLab SnR, Alpha Omage). Briefly, animals were placed on a stereotaxic device (RWD Life Science) after anaesthetization; the brain surface was exposed at the hind-limb area of the sensorimotor cortex, and electrodes were attached to the pia mater to provide stimulation at a current of 1.5 mA. The MEP was recorded by a bipolar disk electrode in the sciatic nerve. The raw data files were converted to MATLAB (MathWorks) format using Map File Converter software v.5.1.8. Subsequently, the data were acquired and then plotted in Origin 2021 (OriginLab).

Electrophysiological procedures and optogenetics experiments

Electrophysiological procedures were performed 7 weeks and 14 weeks after the transplantation of organoids. These procedures were carried out in a radiofrequency shielded room to minimize external electrical interference. For 7 week optogenetics, 300 nl rAAV-ChAT-eNpHR3.0-EGFP (AAV2/9, 5.0 × 10¹² viral genomes per ml, BrainCase) was injected into organoids 1 day before transplantation. For 14 week experiments, 300 nl was injected into ventral graft sites. High-density flexible neuralelectrode (NSDS, NeuroXess, 16 electrode sites) with optical fibres (200 μm diameter core, NA 0.22, Nufern) was inserted into the xenotransplantation site and affixed with dental cement. We applied 10 ms pulses of yellow light (590 nm) at 50 Hz (100% and 50% duty cycle) for 200 s for neuronal inhibition. Electrophysiological recordings were performed using an OmniPlex Neural Recording Data Acquisition System (Plexon) with a 40 kHz sampling rate. The acquired raw signal files were pre-processed by Offline Sorter and NeuroExplorer (Plexon). Spikes were sorted via k-means clustering algorithms into excitatory principal neurons and interneurons⁴⁷. During optogenetics experiments, the behavioural changes of animals were video-recorded and analysed with YOLO11 Nano object detection algorithm for motion trajectory extraction.

Neural tract tracing

To observe and detect axonal regeneration after organoid transplantation, as well as the reconstruction of neural networks between newborn neuronal synapses and the host spinal cord, we used BDA and PRV for neural tracing. At 7 weeks after transplantation, mice received BDA (10%, Invitrogen) injections into the unilateral sensorimotor cortex at 4 sites (500 nl per site; A–P = 1.0, 0.5, –0.5, –1.0; medial–lateral = 1.5; depth = 0.6 mm). Two weeks after BDA injection, the PRV-CAG-3Ms virus (2 × 10⁹ p.f.u. ml⁻¹, Brain Case) was injected into the unilateral anterior tibialis at four sites (0.75 μl per site) according to the instructions provided by the manufacturer. Mice were killed 7 days after the PRV virus injection.

Spinal cord tissue collection for immunostaining

To investigate the developmental dynamics of organoids after being transplanted into the host, animals were perfused at 1, 3, 5, 7 and 14 weeks after the organoid transplantation to collect tissue samples. The spine was extracted and post-fixed overnight in 4% paraformaldehyde and then transferred into 15% to 30% sucrose solution for gradient dehydration. After dehydration was completed, the spinal cord tissue was peeled out from the spine; 5-bromo-2′-deoxyuridine (10 mg ml⁻¹ in PBS) was injected intra-peritoneally every 2 days from week 1 to 2 after surgery, and samples were collected 1 week after the injection.

Cell apoptosis analysis

The collected organoids were fixed in 4% paraformaldehyde overnight, followed by dehydration in gradient sucrose solution at 4 °C, and were embedded in the optimal cutting temperature compound (O.C.T., 4583, SAKURA) and frozen at −80 °C. Frozen sections of the samples, each with a thickness of 12 μm by a freezing microtome (CM3050S, Leica), were prepared. To determine cell apoptosis in the organoid sections, a terminal deoxynucleotidyl transferase dUTP nick-end labelling (TUNEL) detection kit (40307, Yeasen) was used following the manufacturer’s instructions.

10× Visium spatial transcriptomics

Organoids and mice in each group were chosen randomly for ST-seq. The sample sections were layered on a spatially barcoded array for in situ 2D RNA-seq spatial transcriptomics. The area of each capture region was 6.5 mm × 6.5 mm, with approximately 5,000 spots. The diameter of each spot was 55 μm, and the distance between the spot centres was 100 μm. Each spot contained unique molecular identifier (UMI) and poly (dT) to capture RNA in the target region to obtain spatial information for reverse transcription, amplification and library sequencing. The libraries were generated by the Visium Spatial Gene Expression Slide & Reagent Kit (10× Genomics), and the libraries were sequenced on an Illumina NextSeq 2000 sequencer. The sequencing results were separated into two fastq files of paired-ended reads after demultiplexing the paired-end files. The clean reads without the poly(A) tail sequence and the template switch oligo sequence were subsequently aligned to the GRCh38 (version 100) genome assembly, and Space Ranger software v.3.0 was used for quantification.

ST-seq data analysis and cell-type annotation

Quantitative gene expression matrices were analysed using the Seurat pipeline. All the data were merged into a single Seurat object after quality control processing and then scaled and normalized using the ‘ScaleData’ function. The top 1,500 variable genes were identified using the ‘vst’ method. Principal component analysis was performed to reduce the dimensionality, and the first 20 principal components were integrated as inputs for UMAP and t-distributed stochastic neighbourhood embedding (t-SNE) analysis. Gene expression analysis and mapping were performed using Loupe browser 8.

After pre-processing and cluster analysis of the obtained transcriptome data, the clusters were manually annotated using prior knowledge and reference datasets, integrating single-cell RNA-seq data from human embryonic spinal cord (GSE171892)⁶ and developing mouse spinal cord (E-MTAB-7320)³² to define the major cell types. Subsequently, we achieved the annotation of marker genes for the corresponding cell types by integrating datasets that included D–V cell types, neurotransmitter types²⁰ and A–P classification²⁶. To elucidate the similarity between sOrgs and human spinal cord segments, particularly the thoracic segment, we used a systematic approach that encompassed human embryonic spinal cord single-cell datasets GSE171892 (ref. ⁶), GSE136719 (ref. ²⁵) and GSE188516 (ref. ⁷). For the annotation of mouse-derived cells, cells were roughly categorized by a list of known marker genes³² and then further delineated into cell populations by a list of progenitor and neuronal markers. After completing the labelling of human and mouse cell types within the samples, the annotated reference dataset was integrated with the acquired data.

RT-PCR analysis of gene expression

Tissue samples (transplantation area ±2 mm margins) were collected in RNAiso Plus (9108, TAKARA) reagent according to the manufacturer’s protocol. NanoDrop ND-2000 spectrophotometer (Thermo Fisher Scientific) was used to detect RNA quality and concentration. Reverse transcription of the RNA was performed using a Primer Script Reverse Transcriptase Kit (2690 A, Takara) following the manufacturer’s instructions. The primers used are listed in Supplementary Table 2. RT-PCR was performed using TB Green Premix Ex Taq (RR820A, Takara). Applied Biosystems software v.1.4.3 and Microsoft Office Excel v.2023 were used for the data analysis.

Immunochemistry

The sections were blocked with 5% donkey serum (017-000-121, Jackson) and 0.3% Triton-X (X100-100ML, Sigma) in PBS for 1 h at room temperature, after which the primary antibodies were diluted in blocking buffer and incubated at 4 °C overnight. The secondary antibodies were diluted in blocking buffer and incubated at room temperature for 1 h. The antibodies used are listed in Supplementary Table 3. DAPI (D9542-5MG, Sigma) was used to stain the cell nuclei. Fluorescence signals were detected by a confocal microscope (LSM880, Zeiss). Fluorescence signal quantification was performed with ImageJ software v.1.5 (NIH).

Western blots

A protein extraction kit was used to isolate proteins according to the manufacturer’s protocol. The protein concentration was determined by a Bicinchoninic Acid Protein Assay Kit (P0012, Beyotime). The proteins were separated via SDS–PAGE and transferred onto PVDF membranes. The membranes were blocked with 5% bovine serum albumin (C0500-3010, CELLGROGEN) in Tris-buffered saline-Tween (TBST, T1087, Solarbio) and incubated with primary antibodies in 1% bovine serum albumin overnight at 4 °C. Then, the membrane was incubated with secondary antibody for 1 h at room temperature. Antibodies used are listed in Supplementary Table 3. The blots were visualized by a chemiluminescence imaging system. Band intensities were quantified by ImageJ software (NIH).

Statistics

All the statistical analysis and graphic plots were generated in GraphPad Prism 10 (GraphPad Software) and Origin 2021 (OriginLab). Welch’s t-test was used for the analysis of statistically significant differences between the means of two groups (fitting Shapiro–Wilk normality test); one-way analysis of variance (ANOVA) was used for more than two groups of data following Dunnett’s T3 multiple comparisons test, and two-way ANOVA was applied for multi-variables analysis. The specific statistical methods are listed in the figure captions. P values are determined by the New England Journal of Medicine (NEJM).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.